Different Ventilation Strategies Affect Lung Function but Do Not Increase Tumor Necrosis Factor-α and Prostacyclin Production in Lavaged Rat Lungs In Vivo 

The study was approved by the local animal committee of the Erasmus University (Rotterdam, The Netherlands). Care and handling of the animals were performed in accordance with the European Community Guidelines (86/609/EC). The studies were performed in male Sprague-Dawley rats (body weight, 250–330 g; Harlan CPB, Zeist, The Netherlands). An overview of the different experimental groups is presented in table 1.

One group of animals served as spontaneously breathing controls (control group, n = 10). Six additional animals served as nonventilated, serum tumor necrosis factor-α(TNF-α)-positive controls. They were injected with 15 mg/kg LPS intraperitoneally (5 ml Salmonella  abortus equi S. form; Metalon GmbH, Wusterhausen, Germany) and exposed for 90 (n = 3) and 240 (n = 3) min, respectively, to LPS (group LPS^ip). This group was used as a positive control for serum TNF-α.

Six animals served as nonventilated, bronchoalveolar lavage TNF-α–positive controls. They received 5 mg/kg LPS dissolved in 2 ml saline intratracheally as previously described 12and were allowed to breathe spontaneously for 90 (n = 3) and 240 (n = 3) min, respectively (group LPS^it). This group was used as a positive control for bronchoalveolar lavage TNF-α. To exclude an effect of intratracheally administered saline on bronchoalveolar lavage TNF-α, six animals were given 2 ml saline intratracheally and were allowed to breathe spontaneously for 90 (n = 3) and 240 (n = 3) min, respectively (groups Sal^it).

In all animals, after induction of anesthesia with 2% enflurane in 65% nitrous oxide in oxygen, a sterile polyethylene catheter (outer diameter, 0.8 mm) was inserted into a carotid artery for drawing arterial blood samples. Before tracheotomy, the animals received 60 mg/kg pentobarbital sodium administered intraperitoneally (Nembutal; Algin, Maassluis, The Netherlands), and the enflurane concentration was decreased to 0.5–1.0%. Thereafter, a sterile metal cannula was inserted into the trachea.

In the nonventilated animals, a bronchoalveolar lavage was performed with saline (32 ml/kg heated to 37°C), and 4 ml heparinized blood was taken from the arterial line. The animals were then killed by an overdose of pentobarbital sodium through the penile vena.

After cannulation of the trachea, muscle relaxation was induced in all other animals by intramuscular administration of 2 mg/kg pancuronium bromide (Pavulon; Organon Teknika, Boxtel, The Netherlands) followed by immediate connection to a ventilator. Body temperature was kept within normal range with a heating pad. The animals were mechanically ventilated in parallel (with more than one animal connected to the ventilator) with a Servo Ventilator 300 (Siemens Elema, Solna, Sweden) in a pressure-constant time-cycled mode with the following settings: frequency, 30 breaths/min; PIP, 13 cm H²O; PEEP, 3 cm H²O; inspiratory/expiratory ratio, 1:2; and 100% oxygen. Initially, PIP was increased to 20 cm H²O for half a minute to recruit atelectatic areas (open-up procedure). Thereafter, the ventilator was returned to its previous settings, and blood gases were recorded (Instrumentation Laboratory, Synthesis 25, Milan, Italy). Anesthesia was maintained with hourly injections of pentobarbital sodium (60 mg/kg intraperitoneally), and muscle relaxation was maintained with hourly injections of pancuronium bromide (2 mg/kg intramuscularly).

One group of nonlavaged (healthy [H]) ventilated animals served as controls, and ventilator settings were not changed. Control ventilation (PIP, 13 cm H²O; PEEP, 3 cm H²O; inspiratory/expiratory ratio, 1:2; frequency, 30 breaths/min; fraction of inspired oxygen [FI^O2], 1.0) was continued for 35 + 90 (n = 13) and 35 + 240 (n = 13) min (group 13/3H). The 35-min period was included to compensate for a period during which the animals in other ventilated groups were being lavaged (see following section).

To investigate whether washout of alveolar macrophages by the lavage procedure might have affected TNF-α release, 40 healthy animals were not lavaged (H) but underwent mechanical ventilation only. After the open-up procedure, they were ventilated for 35 min at control settings (PIP, 13 cm H²O; PEEP, 3 cm H²O; I/E ratio, 1:2; frequency, 30 breaths/min; FI^O2, 1.0) to compensate for the 35-min period in which the animals in the other experimental groups were being lavaged and ventilated. After this period, the animals were exposed to the following ventilator settings (frequency, I/E ratio, and FI^O2were not changed):(1) a PIP of 32 cm H²O with 6 cm H²O of PEEP for 90 (n = 10) and 240 (n = 10) min (group 32/6H);(2) a PIP of 32 cm H²O without PEEP for 90 (n = 10) and 240 (n = 10) min (group 32/0H). In this group, dead space was increased to keep arterial carbon dioxide partial pressure > 20 mmHg.

In 100 animals that received mechanical ventilation, acute lung injury was induced after the open-up procedure by repeated bronchoalveolar lavage according to Lachmann et al.  14Each lavage was performed with saline (32 ml/kg) heated to 37°C. Just before the first lavage, PIP and PEEP were elevated to 26 and 6 cm H²O, respectively. Lung lavage was performed over a 35-min period and repeated four to five times to achieve an arterial oxygen partial pressure ≤ 85 mmHg. FI^O2, I/E ratio, and frequency were not changed unless stated otherwise. Immediately after lavage, these animals were randomized to be treated with

• Partial liquid ventilation (PLV) at a dose of 15 ml perfluorocarbon per kilogram body weight (Perflubron; Alliance Corporation, San Diego, CA) for 90 (n = 10) and 240 (n = 10) min (PIP, 26 cm H²O; PEEP, 6 cm H²O).

• Exogenous surfactant at a dose of 120 mg/kg for 90 (n = 10) and 240 (n = 10) min. The surfactant used was isolated from minced pig lungs that were processed as previously described. 15The freeze-dried material was suspended in warm saline to a concentration of 40 mg/ml and administered intratracheally, for which the animals were disconnected from the ventilator. The surfactant suspension was administered as a bolus dose followed by a bolus dose of air (12 ml/kg) directly into the endotracheal tube via  a syringe, immediately followed by reconnection to the ventilator (PIP, 26 cm H²O; PEEP, 6 cm H²O).

• An increase in PEEP and PIP of 2 cm H²O for 90 (n = 10) and 240 (n = 10) min, resulting in a PEEP of 8 cm H²O and a PIP of 28 cm H²O (group 28/8).

• An increase in PIP of 6 cm H²O for 90 (n = 10) and 240 (n = 10) min, resulting in a PIP of 32 cm H²O and a PEEP of 6 cm H²O (group 32/6).

• High PEEP for 90 (n = 10) and 240 (n = 10) min (open-lung concept [OLC]). In these groups, the lungs

were opened by increasing PIP to 40 cm H²O, PEEP to 20 cm H²O, and I/E ratio to 1:1; frequency was set at 100 breaths/min after lung lavage. After 2–3 min, PIP was decreased to 32 cm H²O, and PEEP was set at 16 cm H²O. These settings were shown not to result in auto-PEEP because end-expiratory flow on a ServoScreen (Siemens) connected to the ventilator was zero. The high ventilatory frequency and I/E ratio were chosen to maintain normocapnia.

In all ventilated animals, blood gases were recorded at 5, 30, 60, and 90 min (in the animals ventilated for 90 min) and at 5, 30, 60, 120, 180, and 240 min (in the animals ventilated for 240 min) after starting the experimental mode of mechanical ventilation. At the end of the study period, bronchoalveolar lavage was performed in the ventilated animals with saline (32 ml/kg heated to 37°C), and 4 ml heparinized blood was taken from the arterial line, identical to the method described for nonventilated animals. The animals were then killed by an overdose of pentobarbital sodium through the penile vena.

The blood and the bronchoalveolar lavage fluid (BALF) of all animals were centrifuged at 4°C at 400 ×g  for 10 min to remove cells and cellular debris. Supernatant of both blood and BALF were taken and snap-frozen on liquid nitrogen and stored at −80°C until further analysis.

The protein concentration of the BALF supernatant was determined with a photospectrometer (Beckman DU 7400; Beckman Instruments Inc., Fullerton, CA) at 595 nm using the Bradford method (Bio-Rad protein assay; Bio-Rad Laboratories, Munich, Germany) with bovine serum albumin (Sigma, St. Louis, MO) as a standard. 16 

Prostacyclin (PGI²) was assessed as the stable metabolite 6-keto-prostaglandin (PG) F^1aand was measured by enzyme immunoassay (Cayman, Ann Arbor, MI). The cross-reactivity of the detecting antibody was 2,3-dinor-6-keto PGF^1α8.7%; PGF^2α2.1%; and PGE², PGF1^α, PGD², and thromboxane B²all < 1%. Rat TNF-^αwas assessed by rat-specific enzyme-linked immunosorbent assay (Genzyme, Cambridge, MA). The TNF test shows no cross-reactivity with any other rat (i.e. , interleukin [IL]-1β, IL-2, IL-4, interferon-γ, monocyte chemoattractant protein-1) or murine (i.e. , IL-1α, IL-3, IL-5, etc. ) cytokine tested up to concentrations of 1 μg/ml. The lower limits of detection were 10 pg/ml for 6-keto-PGF1^αand 50 pg/ml for TNF. The intra-assay and interassay coefficient of variation was < 10% for both tests.

Intragroup comparisons for arterial oxygen and carbon dioxide partial pressures at the time points before and after lavage and at 5 and 240 min were analyzed with a repeated-measures analysis of variance. Intergroup comparisons for protein, mediator levels, and arterial oxygen and carbon dioxide partial pressures were analyzed with analysis of variance. If analysis of variance resulted in a P  value ≤ 0.05, a Tukey-Kramer post-test was performed. Statistical significance was accepted at P ≤ 0.05. All data are reported as mean values ± SD. The statistical power of the cytokine measurements was calculated by assuming a two-sided t  test with an α level that was adjusted by the Bonferroni method for the number of comparisons made.

The recovery percentage of the BALF fluid in the different groups is presented in table 1. Arterial oxygenation and carbon dioxide levels over time with statistically significant differences are given in tables 2 and 3, respectively. In groups 28/8 and 32/6, oxygenation levels did not significantly recover from postlavage values. In the OLC, PLV, and surfactant groups, oxygenation levels were restored to prelavage values after 60 min, although in the PLV group they gradually decreased over the next 3 h. In group 32/0H, oxygenation levels decreased over time as a result of mechanical ventilation only, whereas they remained stable in group 32/6H.

The protein concentration was increased in the BALF of all lavaged animals after 90 and 240 min compared with controls (0.19 ± 0.05 mg/ml;fig. 1). However, treatment with OLC (90 min, 0.94 ± 0.34 mg/ml; 240 min, 2.38 ± 0.53 mg/ml), surfactant (90 min, 0.77 ± 0.38 mg/ml; 240 min, 1.96 ± 0.7 mg/ml), or PLV (90 min, 0.80 ± 0.33 mg/ml; 240 min, 1.83 ± 0.93 mg/ml) partially reduced the BALF protein concentrations compared with groups 28/8 (90 min, 1.80 ± 1.12 mg/ml; 240 min, 3.82 ± 1.54 mg/ml) and 32/6 (90 min, 3.66 ± 0.67 mg/ml; 240 min, 4.28 ± 1.42 mg/ml).

Because previous studies had shown that ventilation alone may be sufficient to cause release of the eicosanoid PGI 2,3as well as the important proinflammatory cytokine TNF 2,3from isolated lungs in vitro , in the present study we focused on these two mediators. Figures 2 and 3depict data on BALF and serum concentration of TNF-α and PGI², respectively. There was no significant increase in serum concentrations of these mediators caused by the ventilation procedures. Increased serum TNF levels were found only in the serum of the intraperitoneal LPS-treated animals (no serum TNF-α levels were detectable in Sal^it90 and 240 and LPS^it90 and 240). TNF levels in the BALF were significantly increased in groups LPS^it90 and 240 when compared with all other groups (BALF TNF-α levels in groups LPS^ip90 and 240 were 59 ± 55 pg/ml and 88 ± 78 pg/ml, respectively, and did not significantly vary from controls).

With respect to PGI², significantly increased concentrations of 6-keto-PGF^1αin BALF were observed in the OLC group after 240 min only (320.9 ± 179.4 pg/ml; controls, 105 ± 149 pg/ml). In serum, PGF^1αconcentrations within each ventilated group were higher at 90 min than at 240 min. Within the lavaged-animal groups, all values at 240 min were lower than those at 90 min in all groups except for PLV 90 versus  32/6 240 and OLC 240. Because we found no evidence for a ventilation-induced release of cytokines into either the circulation or the alveolar space, we performed a statistical power analysis to calculate the difference in mediator levels that, under our conditions, would have been detected with a power of 80%. According to this analysis, we had an 80% chance to detect an increase in serum TNF by 170 pg/ml, in lavage TNF by 350 pg/ml, in serum PGI²by 170 pg/ml, and in lavage PGI²by 175 pg/ml.

No animal died during the experiment, nor did any develop pneumothorax.

This study demonstrates that surfactant therapy, PLV, and ventilation with high PEEP reduce protein accumulation in the bronchoalveolar lavage–accessible spaces in this rat lung lavage model of acute lung injury. TNF-α concentrations of BALF were increased only in the groups treated with LPS intratracheally and showed no differences between all other groups. Thus, our in vivo  mediator results did not confirm the results of previous studies in isolated healthy and LPS-exposed lungs subjected to different modes of hyperinflation, 1–3which showed ventilation-induced inflammatory mediator release and reduced release with ventilation strategies using higher levels of end-expiratory pressure. Moreover, of note, the large increase in serum TNF induced by LPS intraperitoneally shows that under our conditions, changes in serum TNF would have been noticed. The power analysis showed that we had a power of 80% to detect an increase of ≥ 170 pg/ml. The reasons for the lack of agreement between our in vivo  studies and the aforementioned in vitro  studies 1–3are speculative and multiple and may be a result of the limitations of isolated lung preparations. In our study, high PEEP at high levels of mean airway pressure (OLC group) resulted in a significant release of PGI²in the bronchoalveolar lavage–accessible space.

Previous studies were conducted in intact, healthy rats at tidal volumes comparable to the ones used by Tremblay et al.  2in their ex vivo  isolated and nonperfused rat lungs (tidal volume ≥ 40 ml/kg). 9–12Such studies have shown that the permeability changes as well as the changes in lung mechanics and oxygenation associated with this type of mechanical ventilation can (at least partially) be prevented by application of PEEP or administration of exogenous surfactant before mechanical ventilation. 9–12 

The application of ventilator settings resulting in tidal volumes ≥ 40 ml/kg is lethal in intact rats within 1 h, likely because of respiratory failure. 17However, this period is within the time frame of first assessment of inflammatory mediators in the in vitro  studies reported previously. 1–3For this reason, we investigated the effect of different ventilation strategies on inflammatory mediator release in an in vivo  rat lung lavage model of acute lung injury at more clinically relevant airway pressures.

Bronchoalveolar lavage increases surface tension of the alveolar lining fluid and decreases lung–thorax compliance 18; bronchoalveolar lavage primarily affects the surfactant system, and the lavage procedure itself does not alter elastic properties of the pulmonary parenchyma. 19Treatment procedures to improve oxygenation are aimed at the following:

1. Counterbalancing the increased retractive forces by applying pressure-controlled ventilation that recruits collapsed lung areas by applying an inspiratory pressure that overcomes the opening pressure of collapsed but recruitable lung units. After recruitment, ventilation pressures are reduced, and PEEP is set just above the critical closing pressure of these lung units to prevent end-expiratory collapse. The pressure amplitude is set as small as necessary to maintain normocapnia (OLC). 20 

2. Decreasing alveolar surface tension toward prelavage levels by application of surface active material (exogenous surfactant therapy). 21 

3. PLV, in which ventilation is superimposed on lungs that are filled with perfluorocarbons (which are capable of dissolving high amounts of oxygen and carbon dioxide), thus preventing expiratory alveolar collapse and maintaining gas exchange. 22 

All of these strategies initially improved oxygenation index (arterial oxygen partial pressure/FI^O2) > 500 mmHg compared with ≤ 85 mmHg after bronchoalveolar lavage; however, oxygenation in the PLV group decreased over time. Results from previous studies have suggested that with PLV, oxygenation decreases over time as a result of evaporation of perfluorocarbon because of its low vapor pressure, with subsequent derecruitment of alveoli over time. 23In our study, groups in which PEEP was increased by 2 cm H²O (group 28/8) or in which PIP was increased by 6 cm H²O (group 32/6), oxygenation did not significantly improve from postlavage values (PEEP, 6 cm H²O; PIP, 26 cm H²O). These data indicate that after surfactant impairment, simply increasing PEEP or PIP does not result in mechanical ventilation with higher mean lung volumes and improved oxygenation, but that an active recruitment procedure should be performed or alveolar air–liquid tension should be decreased by surfactant or perfluorocarbon to improve oxygenation. In the present study, oxygenation in healthy animals ventilated with a PIP of 32 cm H²O without PEEP (group 32/0H) decreased from > 500 mmHg to < 85 mmHg, whereas oxygenation in animals ventilated with 6 cm H²O (group 32/6H) was preserved. These data confirm previous findings on the beneficial effect of PEEP on oxygenation during mechanical ventilation with moderately high to high PIP. 11,17Arterial carbon dioxide was maintained within normal limits in all groups except for the animals in group 28/8, which became hypercapnic, and those in group PLV, which became hypocapnic. Differences in arterial carbon dioxide tension are well explained by the interaction between differences in compliance, ventilation pressure, and dead space (group 32/0H), which will result in different tidal volumes and rates of arterial carbon dioxide elimination.

Intra-alveolar protein levels in the PLV, surfactant, and OLC groups were lower than those in groups 28/8 and 32/6. These findings can be explained by differences in lung parenchymal stretch and epithelial stretch, in particular with widening of intracellular junctions. 24Increased protein infiltration may also be caused by alveolar collapse due to surfactant impairment, which will increase suctioning from the capillary into the direction of the interstitial spaces and alveolus. 12Therefore, reducing surface tension by exogenous surfactant (surfactant group) or perflubron (PLV group) and preventing alveolar collapse by high PEEP levels (OLC group) are important explanations for the reduced protein concentrations in these animal groups compared with groups 28/8 and 32/6. The lower intra-alveolar protein concentration of group 32/6H compared with those of group 32/0H confirms previous findings on the beneficial effect of PEEP on the permeability of the alveolo-capillary barrier to protein. 9,11The fact that there was no difference in the protein concentration of groups LPS^itand Sal^itindicates that TNF-α did not play an important role in promoting intra-alveolar protein infiltration.

Mechanical ventilation with high PEEP levels at high levels of mean airway pressure resulted in an increase in the PGI²level of BALF after 240 min of mechanical ventilation (OLC group). It may be speculated that PGI²release as a result of mechanical ventilation will protect the lung from reduced capillary perfusion as a result of compression associated with mechanical ventilation at high mean airway pressures. Because the barrier function of the alveolo-capillary membrane in lavaged lungs is lost even to large molecules, 25PGI²may freely diffuse over the alveolo-capillary barrier, which makes its origin unclear. Stretching of both cultured rat lung cells 26and cultured endothelial cells 27has been shown to result in PGI²production. We do not have an explanation for the consistent finding of a decrease in serum PGI²concentration over time. It may be the result of decreased flow speeds in the lung vasculature 28or depletion of a PGI²pool, altered PGI²metabolism, or a naturally occurring physiologic vascular adaptation to changes in lung perfusion as a result of mechanical ventilation.

The BALF control levels of TNF-α were on the same order of magnitude as in isolated rat lungs. 2Our data showed no statistically significant effect of lavage or ventilation on BALF TNF levels in vivo . Alveolar macrophages are primary candidates for mediator release 29and TNF-α in particular, 30which may be induced by mechanical stretch. 6The failure to observe ventilation-dependent TNF release in vivo  in this study could, in theory, have been a result of the washout of alveolar macrophages from the alveolar spaces. To exclude this possibility, we included two groups of healthy, nonlavaged rats exposed to comparable PIP levels but different PEEP levels. There was no increase in either serum or lavage TNF-α levels in healthy nonlavaged animals as a result of mechanical ventilation compared with nonventilated controls, and the values were on the same order of magnitude as those in the ventilated and lavaged animals. Therefore, it is unlikely that bronchoalveolar washout of macrophages affected bronchoalveolar TNF-α levels.

The absence of any effect of ventilation on TNF or PGI²release raises the question as to the minimum difference in mediator levels that the present study would have detected. Because we used as many as 10 animals per group, the power of the present study was sufficient to note a moderate increase of either PGI or TNF levels in serum (< 170 pg/ml) or BALF (PGI, 175 pg/ml; TNF, 350 pg/ml). Thus the increase that was detectable was much less than that after LPS treatment, which was only used as an internal control. The power of the present study to detect an increase in BALF TNF of 1,000 pg/ml as reported after high-volume zero PEEP ventilation of nonperfused isolated rat lungs 2was > 99%. These considerations show that the present study had sufficient power to show even moderate changes in mediator levels. However, it should be noted that the absence of increased TNF or PGI levels in mechanically ventilated healthy animals or animals with homogeneous lung injury as induced by the lavage procedure, does not allow prediction of what would happen in inhomogeneously injured lungs, such as those typical for acute respiratory distress syndrome.

In conclusion, different ventilation strategies had a profound effect on lung protein permeability and oxygenation in lungs from lavaged rats in vivo . However, except for an increase in the level of PGI²in the BALF by mechanical ventilation with high levels of PEEP (OLC), we could not demonstrate any increase in TNF-α or PGI²levels in serum or BALF as a result of these different ventilatory strategies in lavaged rats in vivo . These results are in contrast to previous findings in isolated perfused lungs. 1–3We therefore stress that caution is required in the extrapolation of data on ventilation-induced inflammatory mediator expression in isolated lung preparations to in vivo  preparations and the clinical situation.

The authors thank Laraine Visser-Isles for English-language editing.