Published ahead of print on December 12, 2002, doi:10.1164/rccm.200208-889OC
© 2003 American Thoracic Society
Ventilator-induced Cell Wounding and Repair in the Intact LungDepartment of Medicine, Division of Pulmonary and Critical Care Medicine, Thoracic Diseases Research Unit; Department of Pathology; and Department of Physiology and Biophysics, Mayo Clinic, Rochester, Minnesota Correspondence and requests for reprints should be addressed to Correspondence and requests for reprints should be addressed to Ognjen Gajic, Stabile 8-62, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. E-mail: gajic.ognjen{at}mayo.edu
We tested the hypothesis that cells of ventilator-injured lungs are subject to reversible plasma membrane stress failure. Rat lungs were perfused with the membrane impermeable fluorescent marker propidium iodide and randomized to one of four ventilation strategies. Subpleural lung regions were imaged with confocal microscopy, and cell injury was quantified as the number of propidium iodide–positive cells per alveolus. The number of injured cells was significantly greater in lungs ventilated with large tidal volumes and zero end-expiratory pressure than in lungs ventilated with small tidal volumes and positive end-expiratory pressure (p < 0.01). Cell injury correlated with lung weight gain, change in dynamic compliance, and histologic injury scores. In a second set of experiments, lungs were mechanically ventilated for 30 minutes at high tidal volume settings, whereas propidium iodide was perfused either during or after injurious ventilation. Labeling after removal of injurious stress revealed significantly fewer injured cells (0.25 ± 0.09 to 0.08 ± 0.08, p < 0.01). We conclude that cells of ventilator-injured lungs are subject to reversible plasma membrane stress failure.
Key Words: ventilator-induced lung injury • plasma membrane resealing • confocal microscopy • isolated perfused rat lung • stress failure Ventilator-induced lung injury (VILI) is characterized by mechanical failure of the blood–gas barrier. As shown with electron microscopy more than 20 years ago, widespread endothelial and epithelial cell injury is one of the hallmarks of the entity and accounts at least in part for the increased microvascular permeability of ventilator-injured lungs (1–5). Plasma membrane blebs and cytoskeletal disruptions occur in association with intercellular and intracellular gaps exposing basement membrane. Whereas electron microscopy defines cellular ultrastructure in intricate detail, the technique is limited by finite sampling and is therefore not well suited for quantifying injury on the scale of whole lungs. Light microscopy, on the other hand, does not have sufficient spatial resolution to define lesions in individual cells. Therefore, much VILI research has focused on the consequences of cellular injury such as edema, inflammation, and tissue remodeling rather than on the determinants of the cellular stress failure itself (6–8). We introduce a new method for grading mechanical lung injury that rests on the assessment of cell membrane integrity and that involves confocal imaging of subpleural lung regions. The lung is perfused with the membrane impermeable label propidium iodide (PI) (Molecular Probe, Eugene, OR). When PI enters a cell through a membrane defect, it intercalates with DNA and emits a red fluorescence upon excitation with blue light. We have previously reported on the use of PI as a marker of plasma membrane stress failure in cultured alveolar epithelial cells and have now adapted this method for the study of mechanically ventilated isolated perfused lungs (9, 10).
Plasma membrane stress failure occurs when the matrix to which a cell adheres undergoes large deformations (1, 11–14). As a result, the cell might be forced to assume a shape with a large surface-to-volume ratio at which the plasma membrane might experience lytic tension. Most breaks are repaired within seconds, usually via a calcium-dependent lipid trafficking response (15–17). Failure to reseal the membrane defect invariably leads to cell death. The ultimate fate of wounded and resealed lung cells is unclear. In many instances, wounded and resealed cells maintain normal function. In certain model systems, wounding causes expression of early response genes associated with the translocation of the nuclear transcription factor nuclear factor- The primary objective of the experiments described in this report was to test the hypothesis that cells of ventilator-injured lungs are subject to plasma membrane stress failure. To this end, we compared a PI uptake–based index of cell wounding with conventional injury markers such as weight gain, change in lung compliance, change in pulmonary vascular resistance, and light microscopy. Our secondary objective was to assess the potential for cell resealing under conditions of injurious stress. We compared the cell injury index of lungs that were PI perfused during high-volume ventilation with that of lungs that were PI perfused after high-volume ventilation. The latter had consistently fewer PI-positive cells. This finding suggests that most wounded cells reseal their membrane defect and that a relatively small fraction of wounded lung cells undergo necrosis.
Animal Preparation Male adult Sprague-Dawley rats weighing 275 to 380 g were anesthetized by intraperitoneal administration of sodium pentobarbital (50 mg/kg). A tracheostomy was performed, and the lungs were mechanically ventilated with a tidal volume (VT) of 6 ml/kg (Harvard Rodent Ventilator, Model 683; South Natick, MA). The thorax was opened, and heparin was administered into the right ventricle. The animal was killed by exsanguination. The pulmonary artery and the left atrium were cannulated and perfused with a red cell–enriched 4% dextran (40KD) in Krebs bicarbonate buffer (37°C) at a flow rate of 6 ml/minute. Fluorescent label PI was added to the perfusate either during or after experimental ventilation. Ventilator stroke volume, inspiratory CO2 concentration, and airway and vascular pressures were recorded, and the ventilator setting was adjusted according to the experimental protocol. All monitored variables were displayed and recorded in digital form with an IBM-PC running in-house–developed, customized software (Labview for Windows; National Instruments Corp., Austin, TX). At the end of the experiment, the lungs and heart were excised en block, weighed, and then submerged in saline for immediate confocal microscopy.
Experimental Protocol
Laser Confocal Microscopy-based Assessment of Cell Injury and Resealing in Intact Lungs
In Vitro Validation of a Method to Characterize Cell Injury and Resealing
Light Microscopy
Data Analysis Interobserver variability was evaluated in preliminary experiments by paired comparisons of two blinded raters using Wilcoxon signed-rank test. Analysis of variance and Kruskal-Wallis test were used to test for differences between the groups, followed by Tukey-Kramer test for multiple comparisons (JMP Version 4.04; SAS Institute, Cary, NC).
High-volume Ventilation Is Associated with Plasma Membrane Wounds High tidal volume ventilation (groups II, III, and IV) caused cellular injury, as evidenced by the presence of PI-positive cells on the fluorescent images (Figure 1 , top). The number of PI-positive cells per alveolus (cell injury index) increased from 0.04 ± 0.01 in group I to 0.10 ± 0.07 in group II, 0.18 ± 0.09 in group III, and 0.35 ± 0.13 in group IV (p < 0.01 analysis of variance; Figure 1, bottom). Although groups III and IV were ventilated with the same tidal volume (40 ml/kg), the application of 3 cm H2O PEEP in group III was associated with a lower number of wounded cells (p = 0.05).
There was excellent interobserver agreement with regard to the number of PI-positive cells per alveolus (median interobserver difference in cell injury index in any one experiment = 0.001, range from -0.11 to 0.04, p = 0.79). However, there was considerable field-to-field variability in the number of wounded cells, suggesting marked heterogeneity and a focal nature of the injury. Mean SDs in cell injury index across 10 random images within a specimen were 0.06 in group I, 0.08 in group II, 0.15 in group III, and 0.32 in group IV.
Cell Injury Index Correlated with Conventional Injury Parameters
Two representative light microscopic images contrast the normal architecture of a lung that was ventilated at group I settings with that of a lung from group IV (Figure 2 , top). Note the diffuse alveolar damage, the widespread hyaline membrane formation, the perivascular hemorrhage, and the small airway epithelial injury in the group IV lung. The mean histologic injury score was 0.5 ± 0.5 in group II and increased to 7.0 ± 2.2 and 8.5 ± 0.6 in groups III and IV, respectively (Figure 2, bottom). The difference in histologic injury score between groups III and IV was not statistically significant. Epithelial necrosis and cell sloughing were observed in distal airways and alveolar ducts. In lungs with the most severe injury, epithelial lesions also involved larger airways. There was no difference in the distribution of injury between subpleural and central lung regions.
The degree of cellular injury in subpleural alveoli, as reflected in the number of wounded cells per alveolus, correlated with traditional injury markers across the four experimental groups (Table 1 and Figure 3) . There was no correlation between cell injury index and conventional injury parameters within groups, however. Furthermore, although the cell injury index of group IV was significantly greater than that of group III, none of the conventional injury markers distinguished between these two groups; that is, they failed to reveal a PEEP effect (Figure 3).
Most Injured Cells Reseal upon Removal of the Injurious Stress Injured lungs that were PI labeled after resumption of a noninjurious ventilator setting had a significantly lower cell injury index than lungs that were PI labeled during injurious mechanical ventilation (Figure 4) . Specifically, the cell injury index fell from 0.25 ± 0.09 to 0.08 ± 0.08 after the injury stimulus had been removed (p < 0.01). In contrast, there was no significant effect of label application timing on conventional injury markers such as change in airway and vascular pressures, weight gain, fall in inflation compliance, and histologic injury score (Table 2) .
To test whether the lower cell injury index of postinjury labeled lungs could be explained by a normalization of vascular leakiness and the consequent exclusion of PI from the alveolar spaces, we harvested edema fluid from post injury labeled lungs and applied it to wounded A549 cells. PI staining of cells at the wound edge could be consistently demonstrated (data not shown). Therefore, we interpret the decrease in the number of PI-positive cells after resumption of noninjurious ventilation as evidence of restored plasma membrane integrity (i.e., wound resealing).
In Vitro Validation of a Method to Characterize Cell Injury and Resealing
We have introduced a new method for quantifying cell wounding in an ex vivo rat model of VILI. The method identifies cells with either transient or permanent plasma membrane stress failure. We have validated the method by demonstrating the expected dependence of cell injury on VT and by correlating our index of cell wounding with conventional lung injury parameters. We have also shown that removal of injurious stress causes a decrease in the number of labeled cells and have attributed the decrease in labeling to a restoration of plasma membrane integrity, that is, wound resealing. Our objective was to show that cells of ventilator-injured lungs undergo reversible plasma membrane stress failure; it was not to establish efficacy of a specific ventilation strategy in a clinically relevant disease model. Therefore, we chose an experimental approach that could be easily referenced and the results compared with data in the published literature (26, 27). Not surprisingly, the choice of VT was a major determinant of injury by any measure (Table 1 and Figures 1 and 2). That alveolar epithelial and endothelial cells are targets of mechanical injury has been appreciated for decades (1–5). Whereas cells themselves are probably not the principle stress-bearing elements of the blood-gas barrier, they undoubtedly sense and transduce deforming stress. We know from in vitro studies of epithelial monolayers that relatively large strains must be applied to the substratum before a significant amount of wounding can be demonstrated (11, 14). A loss of barrier function and a change in the distribution of surface forces undoubtedly promote such conditions in the injured lung. Because there is considerable topographical heterogeneity in both character and extent of mechanical ventilation associated lung lesions, electron microscopy is not a technique with which one can easily quantify injury of an entire lung. In contrast, the PI labeling method gives information on a scale conducive to estimating whole lung injury. However, the method does not distinguish between epithelial, endothelial, and basement membrane lesions, all of which are subject to ventilator-induced stress failure. The relative determinants and susceptibilities of these structures to stress failure remain incompletely understood. There is a complex interplay between capillary pressure and flow, surface tension, lung volume, and the rate of lung expansion on the stress distribution along the blood–gas barrier. It is not known whether these variables interact with sufficient degrees of freedom to produce distinct disease specific patterns of cellular injury. Our light microscopic observations confirm earlier reports that VILI is associated with lesions in small airways and alveolar ducts (26, 28–30). Some of these structures may not be visible with single photon confocal microscopy because it is hard to illuminate tissue more than 50 µm below the pleural surface. Nevertheless, there was good agreement between the PI labeling–based injury index and the histologic injury score. Given our choice of ventilator settings, it should come as no surprise that lungs in groups III and IV revealed distinctly more injury than lungs in groups I and II. Because of this bifurcation in group responses, there was a good correlation between the PI labeling–derived cell injury index and conventional measures of lung injury. There is no single gold standard of injury. Edema, hemorrhage, influx of inflammatory cells, cytokine levels, and gene expression of proteins as well as the number of cells with ultrastructural lesions reflect different dimensions of injury. They ought to be correlated, but there is no reason to believe that indices with different temporal expression profiles will be perfectly correlated at any one instant in time. Nevertheless, the statistically significant differences in cell injury estimates between lungs of groups III and IV deserve closer scrutiny. It appears that the addition of a modest amount of PEEP (3 cm H2O to group III) prevents cellular stress failure even though weight gain, mechanical impedance, vascular resistance, and light microscopic injury scores were similar between the two groups. The heterogeneous nature of the injury and spatial differences in sampling (the conventional markers were measured in the whole lung while the PI uptake was restricted to subpleural alveoli) may have contributed to the observed differences. Another possible explanation for the discordance between the PI labeling–based cell injury estimates and conventional lung injury parameters is to consider the PEEP effect real and the conventional injury parameters too insensitive to reveal it. PEEP is thought to protect the lungs from mechanical injury by minimizing the shear stress associated with the repeated opening and collapse of small airways and alveolar ducts (26, 31–33). This is particularly relevant in a preparation in which the transmural pressure is allowed to fall to zero on every breath. Recently, this mechanism of PEEP derived benefit has been challenged in favor of one that views the displacement of edema from small airways and alveoli to interstitial spaces as the key event (34). Nevertheless, the net effect of either mechanism could be a reduced local wall stress resulting in fewer injured cells (35). It must also be emphasized that inflation compliance and lung weight gain are not nearly as sensitive parameters of vascular permeability as is the filtration coefficient, which we did not measure (36). The latter might well have uncovered beneficial PEEP effects on edema formation in lungs ventilated with high tidal volumes. Finally, one should not assume that cell injury and vascular leakiness are always neatly correlated. A great deal of new information on the role of endothelial cell mechanics and cytoskeletal remodeling in the regulation of vascular permeability has emerged in recent years (37). Plasma membrane stress failure is but one of many stimuli that trigger cell remodeling and thereby mediate changes in fluid flux across the vascular wall (36). We have attributed the difference in labeling between lungs that were perfused with PI during injury and lungs that were perfused with PI after injury to a wound healing response. Before discussing the biologic implications of this finding, we consider alternative interpretations and potential technical limitations of our methods. First, we need to consider alternative mechanisms of PI uptake, such as fluid phase pinocytosis or simple diffusion. PI is a membrane impermeant nuclear label that is routinely used in cell viability assays. An extensive literature search failed to uncover specific documentation of PI uptake by pinocytosis. We consider this mode of entry unlikely based on the following observations: (1) There is virtually no PI uptake in cultured alveolar epithelial cells incubated for 30 minutes with PI in the presence of phorbol ester (as an endocytosis stimulator) unless cells are permeabilized (see Figure E1 in the online supplement). (2) Minimal uptake is observed when PI is administered for 30 minutes at noninjurious ventilator settings (see Figure 1 and Figure E2 of online supplement). One of the central assumptions of the technique is that the local label concentration is high enough for PI to enter the cell and label its nucleus should a plasma membrane break occur. Because we perfused the lungs with PI containing solutions, this assumption is likely satisfied for endothelial cells. However, one cannot a priori assume that a sufficient amount of PI gains access to the alveolar space under conditions during which the filtration coefficient is relatively normal. In other words, vascular permeability could be a confounding variable and explain why fewer lung cells took up PI in experiments in which label perfusion was restricted to the post injury period. Two observations argue strongly against this interpretation. In several instances, we sought to remove label from the vascular space by perfusing the preparation for 5 minutes with label free Krebs solution. We then cut the lung in 50-µm thick coronal slices because we wanted to analyze the distribution of PI-positive cells with respect to small airways and alveolar ducts. These attempts invariably failed because virtually every cell in the specimen appeared labeled. We attributed this observation to our inability to remove label from alveolar edema fluid combined with the cell injury caused by cutting the specimen. In additional experiments, we removed edema fluid from lungs that had been label perfused after injury. The edema fluid was then applied to a monolayer of A549 cells, the monolayer wounded with a scalpel and subsequently imaged for PI fluorescence. Invariably, the cells near the wound edge were PI positive. Therefore, we conclude that even after removal of the injurious insult, sufficient amounts of PI enter the alveolar spaces to label injured pneumocytes. Several lines of evidence suggest that removal of mechanical stress causes a rapid restoration of endothelial structure and vascular barrier function. In a series of experiments on isolated mesenteric frog vessels, Neal and Michel showed that increasing perfusion pressure opens both intraendothelial and interendothelial gaps and that these gaps recover within minutes of lowering pressure (38–40). The groups led by West and Mathieu-Costello and Dreyfuss and colleagues came to a similar conclusion after a detailed electron microscopic examination of ventilation and perfusion-injured lungs (1, 2, 41, 42). Neal and Michel argued that gap formation was an adaptive cellular stress response rather than the consequence of a basement membrane break (43). To our knowledge, we provide the first direct evidence that the VILI lesion is associated with transient loss of plasma membrane integrity. However, we do not know whether reversible plasma membrane stress failure is a necessary stimulus for strain-related remodeling of the blood–gas barrier. In summary, we have demonstrated that ventilation-induced lung injury is associated with both reversible and irreversible forms of plasma membrane stress failure. The consequences of these events on lung inflammation and repair are of obvious clinical and biologic interest. Because the determinants of plasma membrane stress failure may not be intimately linked to the molecular mechanisms that govern membrane repair, lipid trafficking and membrane resealing should also be considered targets in the search for effective pharmacoprotection from VILI.
Supported by a grant from the National Institutes of Health (HL-63178) This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org. Received in original form August 19, 2002; accepted in final form December 9, 2002
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
B (
< 570 nm) on channel 1 and PI (
