Adult male Sprague–Dawley rats weighing 330–600 g were anesthetized with intraperitoneal ketamine (90 mg/kg) and xylazine (10 mg/kg) at the onset of the procedure, and as needed throughout the procedure to maintain surgical anesthesia. A tracheostomy was established with a 2.5 mm pediatric endotracheal tube. Paralysis was then achieved with intravenous pancuronium (0.8 mg/kg) and the rats were placed on pressure control ventilation with 50% oxygen delivered via a Galileo ventilator (Hamilton Medical Inc., Reno, NV, USA). Baseline ventilator settings included a control pressure (P_control, the pressure applied above that of PEEP during the inspiratory phase) of 14 cmH₂O and a PEEP of 3 cmH₂O. The respiratory rate was initially titrated to maintain a PCO₂ of 35–45 mmHg.

Rats were then placed on zero PEEP and a midline sternotomy was performed with removal of the right third through sixth ribs. Lung volume history was standardized by generating a single inflation from zero PEEP to a peak pressure of 25 cmH₂O at a constant rate of 3 cmH₂O/sec (PV Tool™; Hamilton Medical Inc.).

A carotid arterial catheter was placed for blood gas analysis (model ABL5; Radiometer Inc., Copenhagen, Denmark) and for inline measurement of the mean arterial pressure (TruWave™; Baxter Healthcare Corp., Irvine, CA, USA). The internal jugular vein was cannulated for fluid and drug infusion. Fluid resuscitation was performed with a 1 cm³ bolus of warmed lactated Ringer's solution when the mean arterial pressure fell below 60 mmHg.

The protocol was as follows. After surgical instrumentation, the rats remained on MV and were randomly assigned to one of four groups: Group I, dependent + low PEEP (n = 5), P_control = 45 cmH₂O, PEEP = 3 cmH₂O; Group II, nondependent + low PEEP (n = 4), P_control = 45 cmH₂O, PEEP = 3 cmH₂O; Group III, dependent + high PEEP (n = 5), P_control = 45 cmH₂O, PEEP = 10 cmH₂O; and Group IV, nondependent + high PEEP (n = 5), P_control = 45 cmH₂O, PEEP = 10 cmH₂O.

The only difference between the dependent and nondependent groups with similar PEEP was the position of the animal (Figure 1). Animals were placed in the lateral decubitus position, left lung up to measure the nondependent lung alveolar mechanics (Groups II and IV) and left lung down to film the dependent lung alveolar mechanics (Groups I and III) (Figure 1). In vivo microscopy was accomplished in the dependent lung by rotating the microscope 180° so that the objective was pointing up, and the microscope was positioned under the rat and attached to the pleural surface (Figure 1).

Concomitant with the initiation of the injurious ventilator strategy, the respiratory rate was set to 20 breaths/min in all groups. Time zero was designated as the time immediately following initiation of the experimental ventilatory strategy. Hemodynamic data, lung function data, and in vivo microscopic data were recorded at baseline and every 15 minutes after initiation of the experimental protocol. The protocol was terminated after 90 minutes.

A microscopic coverslip mounted on a ring was lowered onto the pleural surface, and the lung was held in place by gentle suction (≤5 cmH₂O) at end inspiration for placement of an in vivo videomicroscope (epi-objective microscope with epi-illumination; Olympus America Inc. Melville, NY USA). At each timepoint, the apparatus was reattached to the lung so that a different cohort of alveoli was sampled every 15 minutes. The microscope objective was moved from the top to the bottom of the coverslip field by field, and each new field was photographed for the measurement of alveolar mechanics (Figure 2). Microscopic images of alveoli were viewed at a final magnification of 130× with a color video camera (model CCD SSC-S20; Sony, Tokyo, Japan) and recorded on Pinnacle Studio Plus software (Pagasus Imaging Corporation Tampa, FL) Each field measured 1.22 × 10⁶ μm² and was filmed throughout five complete tidal ventilations for subsequent analysis of alveolar mechanics.

Frame-by-frame analysis was performed by capturing still images of alveoli at peak inspiration and at end expiration. For each visual field, the subset of alveoli analyzed consisted of those that contacted a vertical line bisecting the visual field and represented approximately 10 alveoli per field, the length of that line measuring approximately 1 mm. Five microscopic fields were analyzed for each animal at each timepoint (Figure 2). The outer walls of individual alveoli were manually traced at both end inspiration and end expiration. The areas of these tracings were computed with image analysis software (Empire Imaging Systems; Image Pro, Syracuse, NY, USA) and are referred to as the area at peak inspiration (I) and the area at end expiration (E). The degree of alveolar stability (%I - EΔ) was defined as the percentage decrease in alveolar size during tidal ventilation:

%I - EΔ = 100 × [(I - E)/I]

For each animal at each timepoint, the mean I and the mean E values were calculated, yielding a single value. These aggregate values were then used in the statistical analysis. Similarly, %I - EΔ was calculated for each individual alveolus, and the mean %I - EΔ value for each animal at each timepoint was then compared using standard statistics (see Statistical analysis).

Arterial blood gases, systemic arterial pressures, and pulmonary parameters (tidal volume) were recorded at baseline and then at 15-minute intervals. Pulmonary parameters were measured inline by the Galileo ventilator (Hamilton Medical Inc.).

The trachea was cannulated and the lung was inflated with 10% formalin by gravity to a pressure of 25 cmH₂O. Each lung was identified as a dependent lung or a nondependent lung and was grouped for histological assessment. After 24 hours, the tissue was blocked in paraffin and serial sections were made for staining with H & E. The slides were reviewed at high magnification for edema (400×).

Alveolar instability was caused in two additional rats by 30 minutes of injurious MV (peak inspiratory pressure (PIP) = 45 cmH₂O, PEEP = 3 cmH₂O), similar to injury in Group I and Group II of this study. This injurious ventilation caused the alveolar mechanics of subpleural alveoli to change from stable (that is, little to no change in size with ventilation) to unstable (that is, very large change is size with tidal ventilation), determined by in vivo microscopy within 60 minutes of application. Once unstable alveoli developed, the animals were sacrificed and the lungs were removed en bloc and perfused through the pulmonary artery with 10% formalin at an intravascular pressure of 20 cmH₂O for 24 hours.

The lungs of one rat were inflated and held constant at an airway pressure of 45 cmH₂O (when subpleural alveoli were observed to be fully inflated with the in vivo microscope), and the lungs of the second rat were fixed at an airway pressure of 3 cmH₂O (when subpleural were observed to be mostly collapsed with the in vivo microscope). Following 24 hours of fixation at constant perfusion and airway pressure, the lungs were blocked, sliced, and stained with H & E. These data were used to identify the potential mechanism of alveolar collapse.

Experiments described in this study were performed in accordance with the National Institutes of Health guidelines for the use of experimental animals in research. The protocol was approved by the Committee for the Humane Use of Animals at our institution.

All results are presented as the mean ± standard error of the mean. An all-pairs, Tukey HSD (honestly significantly different) test was used to compare more than two groups. Student's t test was applied for all pair-wise comparisons. We accepted P < 0.05 as significant. Data were analyzed using JPM software (version 5; SAS Institute Cary, NC, USA).

Normal alveoli before injurious ventilation are very stable, and they did not change size appreciably during tidal ventilation (Additional file 1). Injurious MV caused alveolar instability faster (60 minutes) in the nondependent + low PEEP group (Figure 3 and Additional file 2) as compared with the dependent + low PEEP group (Figure 3 and Additional file 3). By 75 minutes, however, the %I - EΔ was no longer different between these groups although it trended higher in the nondependent + low PEEP group. The addition of 10 cmH₂O PEEP prevented the development of alveolar instability for the entire experiment in both the nondependent and dependent lungs (Figures 3 and 4, and Additional file 4)

At 45 cmH₂O airway pressure (PIP) most alveoli in the in vivo microscopic field are inflated (Figure 5a,c), and at 3 cmH₂O (PEEP) most alveoli collapsed (Figure 5b,d). Alveoli at the PIP are inflated and fill the in vivo microscopic field (Figure 5a, dotted line surrounds representative alveoli), and the alveolar walls are very thin (Figure 5c, arrows). At the PEEP the subpleural alveoli collapse (Figure 5b, dark atelectatic areas identified by arrows), and the alveolar walls are thickened (Figure 5d, arrows). The thickened alveolar walls suggest that alveolar collapse is by folding of the alveolar walls [14].

The arterial PO₂ and PCO₂ were not significantly different in the low PEEP versus the high PEEP groups (Table 1) even though alveoli were unstable only in the low PEEP groups (Figures 3 and 4). There were no significance changes in intra-alveolar edema or in interstitial edema between groups (Table 2).

There was a significantly smaller tidal volume in the PEEP 10 cmH₂O groups compared with the PEEP 3 cmH₂O groups. There was no significant difference in lung compliance or mean arterial pressure at 90 minutes between groups. There were no differences in intravenous fluid resuscitation between groups.

Our data are contrary to the findings of Nishimura and colleagues, who showed that lung injury was not gravity dependent [22]. Using a closed-chest rabbit VILI model they found that lung injury was not uniformly greatest in the dependent portions of the lung. Nishimura and colleagues demonstrated that lung injury was very regional but that the most severe injury always occurred in the dorsal portion of the lung regardless of whether the dorsal lung was in the dependent or nondependent position. In contrast, our study showed that the nondependent lung was the first to develop alveolar instability. Nishimura and colleagues, however, did show that prone position slowed the onset of atelectasis (VILI) [22], which supports our finding that body position affects the rate at which VILI develops.

Both of these studies suggest that VILI is not uniform throughout the lung, but rather occurs preferentially in specific areas; however, there is no consensus whether this specificity of injury is due to the gravitational or anatomical position of the lung. The reason for the discrepancy may involve the species being studied (rat versus rabbit), or the tools used to measure the injury (in vivo microscopy versus computed tomography scan). It is possible that there was more injury in the dorsal portions of the lung in our study, which could be not identified with in vivo microscopy. Likewise, there may have been a gravity-dependent increase in alveolar instability in Nishimura and colleagues' study that was not identified with the computed tomography scan. Finally, our study looked at open-chest rats whereas the Nishimura and colleagues study used closed-chest rabbits. Perhaps the influence of the chest wall resistance to inflation changed the location of injury in the two models.

In addition, the interpretation of the computed tomography scan has recently been called into question. Hubmayr suggests that the increased density seen by computed tomography scan in ARDS patients is caused by open alveoli flooded with edema rather than by atelectasis [23]. Perhaps the dorsal injury seen on the computed tomography scan occurs regardless of whether the animal is in the prone or the supine position because the anatomical shape of the rabbit lung causes increased edema in that dorsal portion of the lung.

The lung can be described as an elastic sponge that is compressed by its own weight, especially when edematous (that is, nondependent lung compresses dependent lung), and by the weight of other organs (that is, the heart). Albert and Hubmayr [24] confirmed by computed tomography scan in humans that the heart compresses a significant amount of lung tissue and that the prone position relieves much of this compression. The weight of the nondependent lung and the heart would cause the dependent lung to become less compliant and would divert a larger percentage of the tidal volume into the more compliant nondependent lung. Veldhuizen and colleagues have previously shown that large tidal volumes cause pulmonary surfactant dysfunction [25,26]. The development of alveolar instability in our VILI model was therefore probably due to a large tidal volume-induced surfactant deactivation. In addition, if a larger tidal volume was being delivered to the more compliant nondependent lung, surfactant deactivation would be exacerbated – which may explain why alveolar instability occurred more rapidly in the nondependent lung.

These findings have clinical significance since the amount of healthy lung tissue is drastically reduced in ARDS [27], and thus a 'normal' tidal volume might direct excessively large volumes into the healthy tissue and cause VILI similar to that in the present study. Indeed, it has been shown that smaller tidal volumes significantly reduce mortality in ARDS patients [1].

In this study, the addition of PEEP prevented repetitive recruitment and derecruitment in both the nondependent and dependent lung regions. Our study used a PEEP of 10 cmH₂O, since it was previously shown in our laboratory by Halter and colleagues that 10 cmH₂O PEEP stabilized alveoli following a recruitment maneuver [13]. These data support those of Dreyfuss and colleagues that PEEP will reduce injury to the normal lung ventilated with high volumes and peak pressures [2,28]. Therefore it appears that it is not the high PIP that causes VILI, but rather the large change in pressure from PIP to the end-expiratory pressure that causes injury that ultimately results in altered alveolar mechanics.

The mechanisms by which PEEP reduces VILI and stabilizes alveoli are twofold: the increase in end-expiratory pressure could prevent alveolar collapse, or the decreased tidal volume when 10 cmH₂O PEEP was applied could prevent alveolar overdistension. Although either mechanism could be responsible for the results in this paper, the literature supports the concept of a large tidal volume-induced deactivation of pulmonary surfactant causing alveolar instability [29]. We therefore conclude that the most probable mechanism of PEEP-induced alveolar stabilization is by prevention of alveolar collapse.

Our results are complex, however, since high PEEP prevented alveolar instability but did not reduce pulmonary edema measured histologically. This suggests that PEEP prevents the onset of mechanical VILI (that is, unstable alveoli) but not inflammatory VILI (that is, injury secondary to sequestered neutrophils). Neutrophil-released proteases and reactive oxygen species could cause an increase in vascular permeability with resultant edema formation without alveolar instability. It is possible that if we had allowed the study to continue past 90 minutes, the combination of mechanical and inflammatory injury in the low PEEP group would have caused more edema than that in the lung with high PEEP and stable alveoli. Another explanation for the increase in edema with high PEEP possibility is that barotrauma occurred in the absence of alveolar instability due to the high peak inflation pressure.

Lung histology was studied at the PIP and at the PEEP to determine a potential mechanism of abnormal alveolar collapse and re-expansion. We used the histological configuration of the collapsed alveoli to speculate on the mechanism of this collapse. Tschumperlin and colleagues found that the alveolar walls were thickened at low airway pressure [14], very similar to those in the present study fixed at 3 cmH₂O (Figure 5c, arrows). Using electron microscopy they demonstrated that the thickened alveolar walls were due to alveolar wall folding, and concluded that alveoli do not change size by balloon-like expansion and contraction but rather by folding and unfolding like a paper bag [14]. We conclude that the probable mechanism by which unstable alveoli collapse and expand in the injured lung is not by balloon-like isotropic expansion, but rather due to the folding of the alveolar walls.

Another interesting finding was that the arterial PO₂ was not significantly reduced (actually it was slightly higher) in the low PEEP group with abnormal, unstable alveolar as compared with that in the high PEEP ventilation group with normal, stable alveoli.

The present study clearly demonstrated that alveoli in the low PEEP group were unstable, and we know from previous studies that alveolar instability leads to VILI if alveoli are unstable for 3–4 hours [5,12]. A normal arterial PO₂ does not therefore necessarily identify a healthy lung with normal alveolar mechanics, and nor does it identify a lung that is not being subjected to mechanical VILI.

We postulate that the arterial PO₂ remained elevated in our study even with unstable alveoli because oxygen was exchanged during the portion of the ventilatory cycle in which the unstable alveoli are inflated. This hypothesis was supported by Pfeiffer and colleagues, who demonstrated a cyclic change in arterial PO₂ utilizing an ultrafast inline PO₂ sensor [11]. The arterial PO₂ in these studies fluctuated with each breath in an animal ARDS model with unstable alveoli. The arterial PO₂ can therefore be maintained if the PIP is high enough to open most of the alveoli during inflation. Forcing collapsed alveoli open to improve the PO₂, however, will greatly increase lung injury since alveolar recruitment/derecruitment is one of the primary mechanisms of VILI. These data can loosely be extrapolated to the bedside, and would suggest that it might be possible to normalize PO₂ by increasing the airway pressure, but at the expense of causing a significant VILI.

Our microscope has a limited depth of field (70 μm), and therefore only allows for alveolar analysis in two dimensions. Also, the subpleural alveolar mechanics might still differ from those within the lung parenchyma. Subpleural alveoli have less structural support since these alveoli are not surrounded on all sides by adjacent alveoli (that is, one wall of a subpleural alveolus is attached to the visceral pleura rather than to another alveolus). This anatomic arrangement may lessen the structural support provided by alveolar interdependence, causing subpleural alveoli to become unstable sooner than those within the lung. A classic paper by Mead and colleagues showed that even if not surrounded by alveoli on all sides, there is still a significant structural interdependence between alveoli [30].

The suction that stabilizes the lung tissue on the cover slip might prevent normal pleural expansion and contraction, and thus may prevent healthy alveoli from changing size normally with ventilation. Although we have not totally eliminated this possibility, we have shown in a previous study that suction slightly but significantly increased both the alveolar size and stability. These changes were very subtle, with an alveolar size change from expiration to inspiration being 1.1% in the suction group increasing to 8.3% in the nonsuction group [21]. This slight change in alveolar size with ventilation even without suction was in stark contrast to the dramatic change in alveolar size (for example, total collapse at end expiration or 100% change in size) that occurred following prolonged exposure to injurious MV. Suction therefore does not seem to artificially stabilize normal alveoli nor does it prevent alveoli from becoming unstable following injury.

Finally, the fact that we must open the chest to obtain our in vivo microscopy may alter the way that normal and injured alveoli behave mechanically.