Anesthetized Yorkshire pigs weighing 25–35 kg were pretreated with glycopyrrolate (0.01 mg/kg, intramuscular) 10–15 min before intubation and were pre-anesthetized with telazol (5 mg/kg, intramuscular) and xylazine (2 mg/kg, intramuscular). Sodium pentobarbital (6 mg/kg per hour) was delivered intravenously via a Harvard infusion pump (model 907; Harvard Apparatus, Holliston, MA, USA) to achieve continuous anesthesia. Animals were ventilated using a Galileo™ ventilator (Hamilton Medical, Reno, NV, USA) with baseline ventilation (Vt 12 cc/kg, PEEP 5 cmH₂O, and fractional inspired oxygen 100%) at a rate of 15 breaths/minute, adjusted to maintain arterial carbon dioxide tension at 35–45 cmH₂O.

A left carotid artery cutdown was performed to gain access for blood gas measurements (Model ABL 2; Radiometer Inc., Copenhagen, Denmark), blood oxygen content analysis (Model OSM 3; Radiometer Inc.), and systemic arterial blood pressure monitoring. A thermodilution pulmonary artery catheter was inserted through the right femoral vein for mixed venous blood gas and oxygen content sampling, along with cardiac output and lung function determinations (Baxter Explorer™ Baxter Healthcare Corp., Irvine, CA, USA). A triple lumen catheter was placed into the right internal jugular vein for fluid, anesthesia, and drug infusion. Pressures were measured using transducers (Argon™ Model 049-992-000A, CB Sciences Inc., Dover, NH, USA) leveled with the right atrium and recorded on a 16 channel Powerlab/16s (AD Instruments Pty Ltd, Milford, MA, USA) with a computer interface.

Surfactant deactivation was achieved by endotracheal instillation with Tween-20 surfactant detergent as previously described [22,23]. Briefly, pigs were placed in the right lateral decubitus position and a 0.75 cc/kg 10% solution of Tween-20 in saline was instilled into the right, dependent lung beyond the tracheal bifurcation. Following lavage, the endotracheal tube was reconnected to the ventilator for three breaths and the lungs were then inflated with a Collins supersyringe to twice the baseline Vt for one breath in order to enhance Tween distribution. The endotracheal tube was suctioned, rendering it free from residual Tween and the previous mechanical ventilation regimen was resumed for several minutes. The animal was then rotated to the left lateral decubitus position, and the Tween lavage procedure was repeated in the left lung.

A right thoracotomy was performed with removal of ribs five to seven to expose the lung for in vivo microscopy. The in vivo microscope (epiobjective, epillumination) provides real-time images of subpleural alveoli. Our technique for in vivo microscopy is described in detail elsewhere [24] (video footage illustrating the technique is available on the internet [25]). Briefly, the microscope uses a coverslip suction head apparatus. The apparatus is positioned on the visceral pleural surface of the diaphragmatic lobe of the exposed right lung, and gentle suction is applied (5 cmH₂O) at end-inspiration to affix the lung in place. Suction was minimal to limit motion artifact with respiration, without altering alveolar mechanics [22-24]. The microscopic images were viewed using a video camera (CCD SSC-S20; Sony), recorded using a Super VHS video recorder (SVO-9500 MD; Sony, Tokyo, Japan), and analyzed using a computerized image analysis system (Image Pro™; Media Cybernetics, Carlsbad, CA, USA). Still images of alveoli were extracted from video at peak inspiration and end-expiration, and alveolar areas were measured using computer image analysis (Figure 1). Alveolar stability was expressed as the dynamic change in alveolar area between inspiration and expiration (I-EΔ), with higher values of I-EΔ representative of greater alveolar instability. I-E% was calculated by dividing I-EΔ by the alveolar area at end-expiration.

Following surgical preparation, continuous filming of subpleural alveoli was performed before surfactant deactivation to serve as controls. Video was recorded during ventilation with all possible permutations of three experimental levels of Vt (6, 12, and 15 cc/kg) and three experimental PEEP levels (5, 10, and 20 cmH₂O), generating a total of nine experimental groups (Table 1). We chose these tidal volumes because 6 and 12 cc/kg were used in the ARDSnet trial and 15 cc/kg is still used in some hospitals. We felt that the PEEP levels covered the gambit between low, medium, and high PEEP used in current clinical practice. In addition, we chose not to conduct a recruitment maneuver before applying PEEP for two reasons: although recruitment maneuvers are used by many clinicians, they are not currently the standard of care; and it is possible that the recruitment maneuver itself, with a high airway pressure for an extended period of time, could damage the lung [26] and obscure our primary goal of determining the role of multiple ventilator strategies (combination of Vt and PEEP) on alveolar stability and VILI.

The order of the nine combinations was randomized. Ventilation was maintained at each combination for 5 min to acquire video in order to assess alveolar mechanics before changing ventilation. After all nine Vt/PEEP combinations in healthy lung, Tween instillation was performed as described above. The in vivo microscope was again placed on the visceral pleural surface and video was recorded for all nine combinations of Vt and PEEP in the surfactant-deactivated lung in a similar manner. It is important to note that the same alveoli were filmed for each Vt/PEEP combination. In the event that alveoli moved out of our field of view for any of the Vt/PEEP combinations, they were excluded from the data analysis. Thus, our data represent the effect of each Vt/PEEP combination on the same individual alveoli in the normal and surfactant-deactivated lung.

The phase I protocol was designed to determine which combination of Vt and PEEP was most effective at stabilizing alveoli. In the subsequent phase II protocol, we tested the hypothesis that the combination of Vt and PEEP determined in the initial phase that resulted in the most stable alveoli would produce the least lung injury, and that the combination that resulted in the most unstable alveoli would result in more severe lung injury. In phase I, we found that a Vt/PEEP combination of 5 cmH₂O PEEP and 15 cc/kg Vt caused the most alveolar instability (highest I-EΔ and I-E%), and a combination of 20 cmH₂O PEEP with 6 cc/kg Vt caused the least alveolar instability (lowest I-EΔ and I-E%). Thus, these were the two Vt/PEEP combinations that were tested in phase II.

Following surgical preparation, the in vivo microscope was placed on the visceral pleural surface of healthy swine lung and subpleural alveoli were recorded before Tween instillation to serve as controls. Lavage was then performed with Tween as described above. The in vivo microscope was again placed on the visceral pleural surface and animals were divided into two groups: animals in the high Vt/low PEEP group (least alveolar stability) were ventilated with Vt 15 cc/kg and PEEP 5 cmH₂O; and those in the low Vt/high PEEP group (most alveolar stability) were ventilated with Vt 6 cc/kg and PEEP 20 cmH₂O. Alveolar size at expiration, inspiration, and the number of alveoli per field were measured at each time point. Five minutes of in vivo microscopic footage was recorded every 30 min for three hours. It should be noted that the same four microscopic fields were recorded at each time point to standardize the data collected.

At necropsy the lungs were inflated to 25 cmH₂O pressure and held at this pressure for 60 s to normalize lung volume history. The lungs were than allowed to deflate to atmospheric pressure and the samples were taken immediately as described below. A 3 × 3 × 3 cm cubic section of the right lung taken directly beneath the in vivo microscope viewing field and was fixed in 10% formalin. The fixed tissue contained the alveoli that were being observed with the in vivo microscope. The tissue was blocked in paraffin and serial sections were made for staining with hematoxylin and eosin.

A blinded observer evaluated lung tissue; details of this scoring methodology are published elsewhere [6]. Briefly, the slides were reviewed at low magnification to exclude areas containing bronchi, connective tissue, large blood vessels, and areas of confluent atelectasis, such that histologic data was from parenchymal tissue. These parenchymal areas were assessed at high magnification (400×) in the following manner. Five high power fields (HPFs) were randomly sampled. Features including alveolar wall thickening, intra-alveolar edema fluid, and number of neutrophils were assessed in each of the five HPFs. Specifically, alveolar wall thickening, defined as greater than two cell layers thick, was graded as '0' (absent) or '1' (present) in each field. Intra-alveolar edema fluid, defined as homogenous or fibrillar proteinaceous staining within the alveoli, was graded as '0' (absent) or '1' (present) in each field. A total score/five HPFs for alveolar wall thickening and intra-alveolar edema fluid was recorded for each animal. The total number of neutrophils was counted in each of the five HPFs and expressed as the total number/five HPFs for each animal. All data are expressed as mean ± standard error.

Serum and bronchoalveolar lavage (BAL) fluid were obtained at baseline and when the animals were killed. Serum and BAL levels (ng/ml) of IL-1, IL-6, IL-8, IL-10, and tumor necrosis factor (TNF)-α were determined by enzyme-linked immunosorbent assay (Endogen, Woburn, MA, USA).

Neutrophil elastase activity was determined in serum drawn both at baseline and at the end of the experiment, and in BAL fluid obtained at necropsy. Specifically, elastase activity was determined by incubating either 100 μl serum or BAL fluid and 400 μl of 1.25 mmol/l methoxy succinyl-ala-pro-val-p-nitroanilide (specific synthetic elastase substrate) in a 96-well enzyme-linked immunosorbent assay plate at 37°C for 18 hours. After incubation, the optical density was read at 405 nm. Data are expressed as nanomoles elastase substrate degraded per milligram of protein per 18 hours (nmol/l per 18 hours per mg).

Matrix metalloproteinase (MMP)-2 and MMP-9 activities were measured using a type I gelatin zymography technique. A volume of 20 μl BAL fluid or 2.5 μl serum was electrophoresed (30 mA) for two hours at 4°C. The slab gels were then incubated for one hour with 2.5% Triton X-100 at 22°C and the gels washed with water, then incubated at 37°C in TRIS/NaC/CaCl₂ buffer overnight. The gels were stained with Coomasie blue, destained with 20% methanol/5% acetic acid (22°C), and the molecular weights of the gelatinolytic zones were compared with standard MMP-2 and MMP-9. The concentrations of MMP-2 and MMP-9 were calculated by scanning of the gels using an image densitometric system (Kodak Image Analysis System; Kodak, Rochester, NY, USA). MMP-2 and MMP-9 concentrations are expressed in densitometric units.

A 2 × 2 × 2 cm section of lung directly adjacent to each histologic section was used for wet-to-dry weight ratio determination. The samples were placed in a dish and weighed, dried in an oven at 65°C for 24 hours, and weighed again. This was repeated until there was no weight change over a 24-hour period, at which time the samples were deemed to be dry. Lung water is expressed as a wet to dry weight ratio.

The experiments described in this study were performed in adherence with the US 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 values are reported as mean ± standard error. Differences between groups were determined using one-way analysis of variance, and differences within groups were determined using repeated measures analysis of variance. Whenever the F ratio indicated significance, a Newman-Keul test was used to identify individual differences. P < 0.05 was considered statistically significant.

As expected, control alveoli before Tween endotracheal instillation were very stable during ventilation, with no significant differences for any of the alveolar mechanics parameters (alveolar area at peak inspiration, alveolar area at end-expiration, I-EΔ, and I-E%) regardless of Vt/PEEP combination (Table 1 and Figure 2a; also see Additional file 1). Following Tween endotracheal instillation, significant alveolar instability (high I-EΔ and I-E%) was observed in several Vt/PEEP groups, the most dramatic being the combination of the lowest PEEP (5 cmH₂O) with the highest tidal volume (15 cc/kg; Table 1 and Figure 2b; also see Additional file 2). For any given tidal volume following Tween instillation, higher levels of PEEP were directly related to alveolar stabilization (lower I-EΔ and I-E%). Furthermore, for any given PEEP setting, progressive increases in Vt produced a progressive trend toward increased alveolar instability (Table 1 and Figure 2b).

At baseline before Tween endotracheal instillation, as expected there were no significant differences for any of the alveolar mechanics parameters (alveolar area at peak inspiration, alveolar area at end-expiration, I-EΔ, and I-E%) for either the low Vt/high PEEP or the high Vt/low PEEP group (Figure 3b; also see Additional file 1).

Immediately following Tween instillation alveolar instability increased dramatically, with significantly higher values for I-E% observed for both groups (Figure 3b; also see Additional file 2). In the low Vt/high PEEP group, alveoli were stabilized from the 30 min time point throughout the duration of the three hour study period, with little change in I-EΔ and I-E% values, which were similar to baseline levels (Figure 3b). In contrast, in the high Vt/low PEEP group, alveolar instability persisted as late as two hours into the protocol, with significantly elevated I-E% values compared with baseline levels (Figure 3b).

There were also significant differences in the number of alveoli present in the microscopic field, which we used as a measure of alveolar microatelectasis (Figure 3a). Although both groups demonstrated similar numbers of alveoli at baseline and immediately after Tween instillation, ventilation with the high Vt/low PEEP combination resulted in progressive microatelectasis, because alveoli continually collapsed during the three hour protocol (Figure 3a). In the low Vt/high PEEP group, however, the number of open alveoli remained constant throughout the three hour duration of the study (Figure 3a).

There were no significant differences between the two Vt/PEEP combinations in terms of peak or mean airway pressures, static compliance, and alveolar-arterial gradient at any time point during the 3-hour study (Table 2). Mean airway pressures were statistically higher in the low Vt/high PEEP group. Despite attempts to normalize partial carbon dioxide tension (PCO₂) with increases in respiratory rate (maximum rate allowed by protocol was 35 breaths/min), hypercapnia in the low Vt/high PEEP group was substantial, resulting in significant respiratory acidosis at all time points compared with the high Vt/low PEEP group. With the exception of oxygen saturation at the 60 and 120 min time points, the low Vt/high PEEP strategy produced superior arterial oxygen tension and oxygen saturation throughout the study (Table 2).

Alveolar instability and microatelectasis were associated with a significant lung injury, as measured histologically. High Vt/low PEEP caused alveolar septal thickening, intra-alveolar proteinaceous edema, and neutrophil infiltration. This injury was ameliorated in the low Vt/high PEEP group (Figure 4). There was no difference in lung wet:dry ratio between the two groups (Figure 4). Although there was a significant increase in intra-alveolar edema histologically, the increase was small (Figure 4). Our injury scale is from 0 to 5; the low Vt/high PEEP group scored 1.00 ± 0.15 and the high Vt/low PEEP group scored 2.6 ± 0.33. It is likely that wet:dry ratio was unable to detect such a small difference in intra-alveolar edema.

Levels of cytokines, MMP-2, MMP-9, and neutrophil elastase for both serum and BAL fluid are reported in Table 3. No significance was identified between the groups in IL-1, IL-6, IL-8, IL-10, TNF-α, MMP-2, or MMP-9 level in either serum or BAL fluid.

Reduced lung injury with low Vt ventilation has been the subject of much investigation, and this strategy has become the standard-of-care for ARDS patients [1,18]. A study by Frank and coworkers [31] demonstrated reduced atelectasis and alveolar epithelial injury when Vt was reduced from 12 to 6 cc/kg. In a clinical trial involving 44 ARDS patients [10], reduction in mean tidal volumes (11.1 versus 7.6 cc/kg) produced a marked reduction in BAL fluid levels of TNF-α, IL-1, IL-6 and IL-8, suggesting that lower Vts may reduce biotrauma-induced VILI.

Although low Vt ventilation has become the standard-of-care for ARDS patients, it may exacerbate lung injury if insufficient PEEP is applied to prevent end-expiratory alveolar collapse [32]. One of the aims of the present study was to show the relative value of lowering Vt versus raising PEEP in reducing alveolar stability. We demonstrated that increasing PEEP from 5 to 10 cmH₂O with a Vt of 6 cc/kg provided much greater alveolar stability (53.7% decrease in I-EΔ) than reducing Vt from 15 to 6 cc/kg at either 5 cmH₂O (46.7% decrease) or 10 cmH₂O (31.2% decrease) PEEP. If these results can be extrapolated to clinical treatment of acute lung injury/ARDS, then there is certainly a clear benefit from low Vt ventilation, but there is a potentially greater benefit from even modest increases in PEEP.

Richard and coworkers [20] demonstrated that alveolar derecruitment is more a function of reduced plateau pressures than of low Vt. In addition, they showed that increased levels of PEEP could prevent derecruitment. These findings are consistent with the results of the present study. In addition, low Vt/high PEEP ventilation – similar to that used in our low Vt/high PEEP group – has yielded improvements in oxygenation [33-35] and reduces both intra-alveolar protein levels [33] and lung injury [34,35], supporting the hypothesis that decreased Vt and increased PEEP work synergistly to reduce alveolar instability and reduce VILI.

It has been suggested that injurious mechanical ventilation, such as high Vt and/or low PEEP levels, produces lung injury through biotrauma. Stretch imposed on alveolar epithelial cells has demonstrated dramatic increases in IL-8 release as well as IL-8 gene transcription in vitro [13]. A clinical study involving 44 ARDS patients [12] identified a significant reduction in IL-6, IL-8, and TNF in those patients ventilated with a low Vt in combination with elevated PEEP. In the present study histologic injury was significantly worse in the high Vt/low PEEP group, but the levels of inflammatory mediators were not significantly increased by this strategy in either serum or BAL fluid. Furthermore, neutrophil elastase actually declined over time, regardless of ventilation strategy. These data suggest that mechanical trauma (shear stress from unstable alveoli) rather than biotrauma is the initial mechanism of VILI. If this study had been conducted for a longer time, then we hypothesize that inflammatory mediators would have increased in the high Vt/low PEEP group. In our previous study [27] we did identify increases in IL-6 and IL-8 when we extended the study by an additional hour, although proteases were not increased, similar to the present study. There was a significant increase in the number of polymorphonuclear leukocytes in lung tissue in the high Vt/low PEEP group compared with the low Vt/high PEEP group, even though there was no difference in the measured inflammatory mediators. It is known that cytokines are not free floating in the plasma but can be bound to cells. This suggests that there was an increase in the tissue-specific cytokines in lung in the high Vt/low PEEP group that resulted in increased polymorphonuclear leukocyte sequestration. We previously showed in a similar animal model that there can be an increase in tissue bound cytokines (TNF and IL-6) [27].

Detailed critiques of this in vivo microscopic technique have previously been reported [6,22-24,27,36,37]. This in vivo microscopic technique allows measurements of alveoli in only two dimensions, and thus we measured alveolar cross-sectional area at inspiration and expiration and these data were used to calculate changes in alveolar size with ventilation (I-EΔ). Although this technique only measures alveolar mechanics in two dimensions, the mechanics of alveoli in the normal and surfactant-deactivated lung are profoundly different. Therefore, our hypothesis that alveolar instability is injurious to the lung appears valid, despite our inability to measure precise changes in alveolar volume. Additionally, our in vivo microscopic technique does not provide us with a global measure of alveolar mechanics, but rather we are restricted to the subpleural alveoli in our microscopic field. We have recently demonstrated that subpleural alveoli do not over-distend even at very high airway pressure (60 cmH₂O; see the data repository by DiRocco and coworkers [37]), and so we did not expect to observe alveolar over-distension in the PEEP 20/Vt 15 group.

Although not ideal, this technique provides a bridges between purely physiologic approaches to assessment of alveolar mechanics (such as pressure-volume curve analysis) and purely anatomic approaches (such as computed tomography scanning). The short duration of the study might not have been sufficient time to allow a change in inflammatory mediators to take place. Ventilation with low Vt resulted in a significant increase in PCO₂, which could not be normalized by increasing respiratory rate. It has been shown that high PCO₂ can protect against VILI [38], and so it is possible that the reduction in tissue injury in the low Vt/high PEEP group could have been due to high PCO₂ rather than stabilization of alveoli. Finally, we did not use a recruitment maneuver before setting PEEP and Vt, and it is possible that the results of the experiment would have been altered if a recruitment maneuver had been performed.

Although we used a small number of animals in each group (n = 3/group), the facts that the data were very tight (low standard error) and that we achieved statistical significance in our primary end-point (alveolar stability) suggest that the study had sufficient power to address the the issue considered in the present study. No bias was introduced by a single animal; otherwise, the data would have had a very high standard error.

In this study we considered whether a combination of Vt and PEEP that resulted in alveolar instability cause lung injury. To address this issue, we directly observed subpleural alveoli for stability and, at necropsy, removed the lung tissue that had been observed using the in vivo microscope for histologic analysis. This methodology allowed used to correlate alveolar instability with lung injury and test our hypothesis. However, there were several confounding factors that do not allow us to extrapolate these results readily to the ARDS patient. First, our alveolar sample size was very small and included subpleural alveoli only, so we do not know whether the area of lung sampled is representative of the entire lung. Second, our samples were from nondependent lung areas and our results might have been different in the dependent lung. Finally, we must open the chest to attach the in vivo microscope, and so we do not know whether our findings would have been different if we had been able to obtain the same information with a closed chest.

Tween causes serious lung injury, regardless of the type of mechanical ventilation that the Tween-injured lung is subjected to. Static compliance fell significantly in both groups following Tween instillation and did not significantly recover with time. This suggests that the static compliance was at a nadir following Tween and could not be further reduced by VILI. However, we did observe a significant improvement in partial arterial oxygen tension in the low Vt/high PEEP group, suggesting protection of the lung from VILI.

Cytokines were not significantly increased in this study, which is not consistent with many other experiments demonstrating that high Vt/low PEEP ventilation strategies increase plasma and BAL fluid cytokine levels. This could be for two reasons. Tween is a unique injury model, and other studies demonstrating cytokine increase have used other lung injury models. Also, this experiment was very short, and if we had extended the diuration of the study we might have identified significantly increased cytokine levels. Finally, not all studies have demonstrated that cytokines are released with injurious ventilation [39], and our findings support this hypothesis.