Introduction

Ventilator-induced lung injury (VILI) can increase morbidity and mortality beyond that associated with acute lung injury (ALI) and the acute respiratory distress syndrome (ARDS). It has been estimated that up to 35,000 deaths/year could be prevented by implementing lung-protective mechanical ventilation [1]. There are three basic mechanisms involved in the pathogenesis of VILI: (a) volutrauma, (b) recruitment/derecruitment (R/D), and (c) biotrauma. Volutrauma and R/D are mechanical injuries due to overstretching of the pulmonary parenchyma and shear-stress due to the collapse and reopening of alveoli, respectively. These two mechanical injuries cause the release of inflammatory mediators from damaged lung tissue that results in a secondary injury known as biotrauma [2]. The majority of the data defining these three mechanisms of VILI were generated in animal studies, and, of these, volutrauma is the only mechanism confirmed in the clinical setting. Indeed, the mechanism of biotrauma has recently been challenged [3]. However, the rational development of protective mechanical ventilation strategies depends upon an understanding of the alveolar mechanics involved in lung injury.

The mechanical behavior of ventilation at the alveolar level has to date only been inferred by methods that measure aggregate alveolar behavior. The inferences made from these techniques yield distinct models of alveolar behavior. Computer-assisted tomography [4, 5, 6] and whole-lung pressure/volume (P-V) curve analysis [7, 8, 9] have been used to infer alveolar behavior from regional and global aeration in the injured lung. Both approaches have supported a model of alveolar recruitment. The changes in regional lung volume assessed with parenchymal markers [10, 11, 12] have challenged the existence of alveolar R/D as a mechanism of VILI. Their data support a model of lung injury in which alveolar edema rather than collapse is the predominant pathological feature.

In vivo microscopy has been used to directly visualize subpleural alveoli. With this technique we have observed R/D of individual alveoli in a model known to deactivate surfactant, i.e., Tween lavage [13, 14, 15]. In the present study we evaluated the derangement of alveolar mechanics that occurs in lung injuries of distinct cause by in vivo microscopy in the nondependent lung. The models of lung injury selected included our previous model of primary surfactant dysfunction (Tween lavage), two classical experimental models of ARDS (endotoxin and oleic acid infusion), as well as a common model for causing VILI in the naive lung (ventilation with high peak and no end-expiratory pressure). We hypothesized that: (a) there would be an alteration in alveolar mechanics that is correlated with the onset of hypoxemia and (b) specific alterations in alveolar mechanics would depend upon the cause of lung injury. Portions of the data were presented at the American Thoracic Society Meeting in 2004 and was published in abstract form [16].

Methods

Study protocol

After approval by the local animal ethics committee, 30 male Sprague-Dawley rats (300–550 g) were anesthetized (ketamine/xylazine) and mechanically ventilated. A right thoracotomy was performed on all study animals for the placement of the microscopic apparatus. Baseline hemodynamic, ventilatory, and microscopic measurements were obtained 15 min after instrumentation. Animals were separated into five groups (six per group) consisting of four methods of lung injury and a control group. Lung injury was induced with Tween-20, oleic acid (OA), injurious ventilation (VILI), or endotoxin (lipopolysaccharide, LPS) in the lung injury groups. Animals were followed until the development of hypoxemia (PaO2/FIO2 <300 mmHg) at which time all hemodynamic, ventilatory, and microscopic parameters were again measured. All animals were followed until hypoxemia developed (or 4 h postsurgery in the control group) at which time the animals were killed.

Ventilation

After initiation of positive-pressure ventilation (Galileo, Hamilton Medical, Reno, Nev., USA) all animals received an intravenous dose of pancuronium (0.8 mg/kg). Mechanical ventilation was initiated in a pressure-control mode with 50% decelerating flow, a respiratory rate of 35/min, positive end-expiratory pressure (PEEP) 3 cmH2O, and a tidal pressure of 14 cmH2O. The tidal pressure was adjusted to maintain baseline tidal volume (10–12 cc/kg). The inspiratory-to-expiratory ratio (I:E) was 1:2 in all groups. The fraction of inspired O2 was 1.0. The respiratory rate (RR) was adjusted to prevent hypercarbia (PaCO2 >45). Animals who received injurious ventilation deviated from this ventilatory algorithm as described below.

Animal groups

The five groups were defined as follows:

  • Tween (n=6): animal lungs were lavaged (16 cc/kg) three times with 2% Tween-20 (Fisher Scientific, Fair Lawn, N.J., USA) in saline [17].

  • Oleic acid (OA, n=6): 1% oleic acid (0.1 cc/kg, Sigma, St. Louis, Mo., USA) in isotonic saline was given intravenously [18].

  • Ventilator-induced lung injury (VILI, n=6): the tidal pressure was increased to 40 cmH2O and PEEP was decreased to 0 cmH2O (ZEEP).

  • Endotoxin (LPS, n=6): animals received intravenous endotoxin (Escherichia coli lipopolysaccharide O111:B4, Sigma) prepared in isotonic saline at a concentration of 5 mg/cc. The animals received a dose of 12.5 mg/kg every 30 min.

  • Control (n=6): animals subjected to surgical preparation and data collection only for a total of 4 h.

Ventilatory parameters

Airway pressures (Paw) and flow were transduced from the Y of the ventilator tubing and recorded by the ventilator. The ventilator derives inspiratory resistance (Rinsp) and compliance (Cstat) without interruption of ventilation from the least squares fit method applied on a breath-by-breath basis using flow, volume, and airway pressure measurements obtained at 60 Hz. This derivation for compliance is applicable in our study as the animals were paralyzed and plateau pressures were stable [19]. The compliance measurements were reportable to only one decimal place with this method

Hemodynamics and gas exchange

The right internal jugular vein was cannulated with a 0.030-in. inner-diameter Silastic catheter. The left carotid artery was cannulated with a 0.58 mm inner-diameter polyethylene catheter. Transduced arterial pressures (Edwards, Irvine, Calif., USA) were monitored (78532B, Hewlett Packard, Andover, Mass., USA), and the mean arterial pressure (MAP) was recorded. Arterial blood samples were analyzed with hemoximeters (ABL-5 and OSM-3, Radiometer, Copenhagen, Denmark). No vasoactive or inotropic agents were used on animals in this study. Hypotension (MAP <60) was treated with 1-cc aliquots of isotonic saline at a maximum rate of 20 cc kg−1 h−1.

In vivo videomicroscopy

In vivo videomicroscopy of the nondependent lung was performed at baseline and at the time of hypoxemia. The microscope (XDFM, Olympus, Orangeburg, N.Y., USA) accommodates a reflected light, brightfield illuminator (U-LH100UH, Olympus). Four contiguous 1.0 mm2 fields on the anterior pleural surface of the right lower lobe were filmed with the in vivo microscope at baseline and again when hypoxemia developed (PaO22/FIO2 <300 mmHg). Within each frame at inspiration, a 50-µm wide vertical band was defined as the region of interest (ROI). Any discernable object with obvious and continuous borders that intersected the ROI was considered an alveolar structure. The areas of manually outlined alveolar structures at peak-inspiration (AI) and end-expiration (AE) were computed with image analysis software (Image Pro, Media Cybernetics, Carlsbad, Calif., USA; Fig. 1). The proportion decrease in alveolar size from peak-inspiration to end-expiration was calculated as an index of alveolar stability: I−E (%)=[(AI−AE)/AI]×100.

Fig. 1
figure 1

A 50-µm-wide vertical band was considered the region of interest (ROI). All defined alveolar structures that intersect the ROI were traced for analysis (dots)

Full size image

Assessment of lung edema

Lung edema was assessed by measurement of the wet-to-dry lung weight ratio (W/D). At necropsy the left lung of all animals was isolated and weighed immediately. After incubation in a dry oven for a minimum of 72 h the lungs were reweighed, and the ratio was calculated. No data for Tween W/D is given due to the potential artifact of retained Tween.

Pulmonary histology

At necropsy the right lung was excised and fixed with 10% formalin instilled to a pressure of 25 cmH2O. After 24 h the tissue was blocked in paraffin, and serial sections were made for staining with hematoxylin and eosin. The slices for histology were made through the center of the right dependent lobe such that histology at the pleural surface could be analyzed and compared with in vivo subpleural videomicroscopy.

Statistics

Reported values are mean ±SE. The mean values for each animal at baseline and injury were used for statistical analysis. Parametric analysis of variance and Tukey’s “honestly significantly different” test were applied across all groups at baseline and injury. The level of significance was set at p<0.05. These computations were performed with the JMP 5.0.1.2 statistics software (SAS Institute, Cary, N.C., USA).

Results

Ventilatory parameters

Ventilatory data are listed in Table 1. Baseline measurements for all parameters were similar in all groups. Lung injury was associated with a fall in static compliance (Cstat) that was significant in all injury groups as compared to the control group. The mean inspiratory resistance (Rinsp) was higher in the VILI and Tween groups at injury than at baseline. However, the standard deviation of Rinsp was very large likely due to excessive edema fluid in the airway in a few animals rather than airway constriction. Edema fluid was observed from the tracheal tube in the animals with excessively high Rinsp.

Table 1 Ventilatory data (RR respiratory rate, Ppk peak inspiratory pressure, Rinsp inspiratory resistance, Cstat static compliance)
Full size table

Hemodynamics and gas exchange

Hemodynamic and gas exchange data are listed in Table 2. The pH in controls fell with time. This was most likely due to a metabolic rather than a respiratory component. Baseline measurements were equivalent among groups. Acidemia was higher in the LPS group than in controls. There was no significant difference in hemoglobin levels among groups at any time. Animals subjected to Tween injury expressed profound hypotension at the onset of hypoxemia. The development of hypoxemia took longer in the LPS group than in all others (Fig. 2).

Table 2 Gas exchange and hemodynamics (MAP mean arterial pressure)
Full size table
Fig. 2
figure 2

The arterial oxygenation index (PaO2/FIO2) fell below 300 mmHg in the Tween, ventilator (VILI), oleic acid (OA), and endotoxin (LPS) injury groups at different rates. *p<0.05 vs. other injury groups

Full size image

Alveolar mechanics

Numeric data

Alveolar mechanics were studied only in the nondependent lung. At the onset of hypoxemia the number of alveoli identified per ROI significantly decreased in all injury groups except VILI. Fewer alveoli per region of interest indicate total collapse of a large number of alveoli. Tween, VILI, and OA all caused a large and significant increase in alveolar instability at injury compared to baseline (Fig. 3, video clips). LPS caused little change in stability with values similar to those of the controls group even though fewer alveoli were identified (Fig. 4).

Fig. 3
figure 3

Alveolar instability as measured by the proportion decrease in size from peak-inspiration to end-expiration in controls and in Tween, ventilator (VILI), oleic acid (OA), and endotoxin (LPS) induced lung injury. Injury measurement for the control group was 4 h after surgery since the animals did not become hypoxemic. *p<0.05 vs. baseline

Full size image
Fig. 4
figure 4

Visible alveolar density at baseline and injury in controls and in Tween, ventilator (VILI), oleic acid (OA), and endotoxin (LPS) induced lung injury. Injury measurement for the control group was 4 h after surgery since the animals did not become hypoxemic. *p<0.05 vs. baseline

Full size image

Analysis of in vivo photomicrographs

In vivo photomicrographs of the same subpleural field at peak-inspiration and end-expiration of the control group at baseline and at the end of the study are shown in Fig. 5. Alveoli in the control lung filled the entire field without any areas of dark atelectasis (Fig. 5a). These alveoli did not change size appreciably from end-expiration to peak-inspiration (Fig. 5b). Photomicrographs of the remaining groups at peak-inspiration and end-expiration are arranged in Fig. 6. Tween caused marked alveolar instability (dots) and microatelectasis. This same pattern of injury (i.e., microatelectasis and alveolar instability) was also seen in the VILI and OA groups (Fig. 6). Of note were the large aerated structures seen in the Tween images (Fig. 6, b). Alveolar mechanics in the endotoxin-injured lung (LPS) were very different from the other 3 injury models (Fig. 6). Alveoli were very stable with minimal fluctuation in size during ventilation (dots). Additionally, there were large white amorphous areas in which alveolar boundaries could not be clearly seen (Fig. 6, c).

Fig. 5
figure 5

In vivo photomicrographs of subpleural alveoli in control rats at peak-inspiration and end-expiration at low (a) and high (b) power. The field is filled with alveoli that change little in size (dots) during respiration. Bar in a 200 µm; bar in b 100 µm

Full size image
Fig. 6
figure 6

In vivo photomicrographs of 1.0 mm2 sections of subpleural alveoli at peak-inspiration and end-expiration in Tween, ventilator (VILI), oleic acid (OA), and endotoxin (LPS) injured rat lungs. Alveolar size (dots) decreases in the Tween, VILI, and OA groups but remains relatively constant in the LPS group. There are large areas of microatelectasis (a) present in the Tween, VILI, and OA images. There are large aerated structures (b) present in the Tween and OA images not seen in the VILI photomicrograph. There are large areas of amorphous, whitish structures (c) seen in the LPS images. Bar 200 µm

Full size image

Assessment of lung edema

All groups demonstrated significantly more lung water than controls (Fig. 7). The W/D ratio in the Tween group is not shown due to possible artifact by infusing Tween into the lungs.

Fig. 7
figure 7

Lung water in controls and in ventilator (VILI), oleic acid (OA), and endotoxin (LPS) induced lung injuries. *p<0.05 vs. control

Full size image

Pulmonary histology

The subpleural alveoli of control lung (Fig. 8a) are fully distended and no alveolar wall edema or congestion is evident. The prominent proteinaceous alveolar edema seen in endotoxin-injured lung (Fig. 8b) is evident to a lesser degree in oleic acid-injured lung (Fig. 8d) and scarcely identified after injurious ventilation (Fig. 8c) or Tween injury (Fig. 8e). Alveoli and alveolar ducts were more widely expanded in the LPS group than in all other groups. Alveolar wall edema was observed in all injury groups but to a lesser degree in the Tween group.

Fig. 8
figure 8

The subpleural alveoli of control lung (a) are fully distended and no alveolar wall edema or congestion is evident. There is prominent alveolar edema in endotoxin-injured lung (b), evident in widened alveolar walls and patchy eosinophilic proteinaceous material. Congestion is not noted. There is moderate alveolar wall edema but little, if any, intra-alveolar proteinaceous material or evidence of congestion in lung subjected to injurious ventilation (c). Oleic-acid injured lung (d) exhibits alveolar wall edema and intra-alveolar proteinaceous material although less pronounced than endotoxin-injured lung. Moderate congestion is noted. There is minimal wall edema but scattered areas of airspace edema in Tween-injured lung (e). Bar 200 µm

Full size image

Discussion

The most important findings in this study are that at a similar degree of lung injury (defined as a PaO2/FIO2 <300 mmHg) the cause of this injury effects the pathological change that occurs in dynamic alveolar mechanics. Specifically, the impact of different causes of ALI on dynamic alveolar mechanics are complex, leading to either unstable alveoli that collapse and inflate with each breath (VILI, Tween and OA) or to a more complex, heterogeneous injury pattern in which alveoli become poorly defined with a solid white appearance but stable throughout ventilation (LPS). These data demonstrate that the hypoxemia in ALI can be caused by mechanisms other than alveolar collapse and instability. These experiments were designed to determine whether alveolar mechanics at a similar degree of lung injury differ with cause. They were not designed to uncover the mechanism that induced these abnormal alveolar mechanics.

Possible mechanisms of altered alveolar mechanics

Pulmonary surfactant dysfunction is the most likely explanation for the development of the alveolar instability observed in this study. It is well known that normal surfactant function is essential for alveolar stability [20]. Although we did not measure surfactant function in these experiments, each model of lung injury in this study is known to deactivate surfactant by various mechanisms. Tween directly inhibits surfactant function [15, 21]. Alveolar instability has been shown in several studies utilizing Tween lavage as a model of ALI [13, 14]. Injurious ventilation can cause surfactant deactivation by four mechanisms: (a) loss of surfactant from the alveolus into the smaller airways [22], () conversion of surface active large aggregates into less-surface active small aggregates [23], (c) inactivation of surfactant by ventilator-induced high protein pulmonary edema [24], and (d) loss of surfactant into the bloodstream [25].

Oleic acid is directly cytotoxic and rapidly (within minutes) injures vascular endothelial cells [26, 27, 28]. Although the edema induced by oleic acid may inhibit surfactant function, Oyarzun et al. [29] demonstrated that oleic acid injury does not deplete surfactant, as do other acute lung injury models, but rather caused a direct surfactant deactivation. Previous work from our laboratory supports these results, demonstrating that oleic acid (as well as palmitic and stearic acid found in meconium) directly inhibit surfactant function [30].

The dynamic alteration in alveolar mechanics observed in the endotoxin group was radically different from that seen in the other groups. Endotoxin caused only minimal alveolar instability. Although endotoxin can deactivate surfactant [31], the fact that alveoli did not become unstable suggests that either the timing or degree of surfactant dysfunction in the LPS model was distinct from the other injury models. In lieu of alveolar instability, distinctive, rounded alveolar silhouettes [32] lost their clear borders after endotoxin, resulting in an appearance similar to edema-filled subpleural alveoli seen using confocal microscopy [12]. Despite this similarity in appearance the argument that edema is the distinguishing feature of endotoxin injury in this study is challenged by two disparities: (a) the appearance of subpleural alveoli in saline-filled lung (smooth, solid, gray surface) was much different than that caused by LPS (complex conglomerate of thick rough and folded white areas with scattered remnants of alveolar wall and no identifiable alveolar movement during ventilation), and (2) measured lung water (W/D) was not greater than in the other groups.

The histology gives us a few clues as to the pathology observed in the LPS group. Alveoli and alveolar ducts in the LPS group expanded with intra-alveolar fibrin, suggesting the presence of intra-alveolar edema (Fig. 8b). All other injury groups had either normal sized alveoli or areas of atelectasis. The VILI group (Fig. 8c) in particular had a great deal of alveolar collapse including collapsed areas on the pleural surface. This suggests that atelectasis is not the cause of the amorphous thickened white subpleural areas seen following LPS. It is possible that alveoli were expanded and edema-filled but not reflected in a higher W/D ratio.

Limitations of the study

Our protocol has three major limitations. First, these data are observational and not mechanistic. Second, we judged that rather than the selection of an arbitrary time after injury, a similar fall in oxygenation would better standardize the degree of injury. The subsequent difference in latencies to hypoxemia may possibly confound our interpretation. The possibility exists, and is supported by the findings of this study, that oxygenation does not adequately reflect the degree of injury. Third, only nondependent subpleural alveoli were analyzed. It is well know that dependent portions of the lung have the greatest pathology in ARDS patients. It is possible that subpleural alveoli in the dependent lung would have been unstable following LPS. In a separate study that used a VILI model in the same animal alveolar instability in the nondependent lung was observed prior to that of the dependent lung [33]. In spite of these limitations our primary hypothesis was confirmed by these data. Limitations of the microscopic technique are described in the Electronic Supplementary Material.

Clinical implications

Improper mechanical ventilation can exacerbate the lung injury associated with ARDS, leading to a ventilator induced lung injury. It is now known that lowering the tidal volume can reduce the mortality associated with ARDS [34]. Low tidal volumes are believed to reduce VILI by decreasing plateau pressures and/or by stabilizing alveoli, i.e., reducing cyclic collapse and expansion. Our findings suggest that alveolar instability is not present in every patient with ARDS and that oxygenation does not reflect the degree of injury present. This emphasizes the need for additional investigation of the role of abnormal alveolar mechanics in the pathogenesis of VILI.

Conclusions

OA and VILI caused a rapid onset of alveoli instability similar to that caused by Tween, which is known to deactivate surfactant. This suggests that surfactant deactivation is key to the development of alveolar instability. Endotoxin did not cause alveolar instability at a similar degree of lung injury (i.e., similar P/F ratio), suggesting that surfactant deactivation is not a primary component of endotoxin pathophysiology in this model. Our findings demonstrate that alveolar instability is not a universal occurrence in ALI and that the cause of lung injury influences the pathological changes in dynamic alveolar mechanics.