Atelectasis Causes Alveolar Injury in Nonatelectatic Lung Regions

Lung aeration has a heterogeneous distribution in the setting of acute injury, with a continuum ranging from some degree of atelectasis to some degree of overdistention (1-3). Two contrasting compartments may be considered in the acutely injured lung, with a range of intermediate states. One such compartment is atelectatic (i.e., deaerated, compressed, or fluid filled), less ventilated, and characteristically distributed in the dependent lung regions; the other is overdistended and more ventilated, and is termed the "baby lung" (4). Ventilation of the "baby lung" is based on the principle that smaller tidal volumes will prevent undue tidal overdistension of the small volume of available aerated lung (4, 5). The study by the Acute Respiratory Distress Syndrome (ARDS) Network investigators validated this concept beyond the issue of lung injury by demonstrating a relative mortality benefit associated with a lower tidal volume compared with a higher one where levels of positive end-expiratory pressure (PEEP) were similar (6). Our understanding of the atelectatic part of the lung is limited. Ventilator-induced lung injury is worsened by atelectasis, and injury caused by repetitive opening and closing of distal airways in atelectatic areas has been suggested to explain this phenomenon (7-11). However, the coexistence of regional atelectasis and local lung injury has not been established.

To directly address the distribution of injury associated with atelectasis, we used an in vivo model of surfactant depletion and allocated animals to one of two ventilator strategies: (7) high tidal volume and low PEEP, which causes ventilator-induced lung injury and regional (i.e., dependent) atelectasis; and (2) low tidal volume and higher PEEP, which does not cause ventilator-induced lung injury or atelectasis. We used this model to test the hypothesis that in the presence of extensive atelectasis, airway and alveolar injury associated with high tidal volumes would be predominantly localized to the atelectatic regions.

METHODS

After institutional ethics approval (conforming to the guidelines of the Canadian Council for Animal Care), male Sprague-Dawley rats (300-400 g) were used in all experiments. For complete details of the experimental protocol, see the online supplement. A tracheostomy was performed, arterial and venous catheters inserted, arterial blood pressure continuously monitored, and anesthesia maintained with infusion of intravenous ketamine and xylazine.

Experimental Outline

Surfactant depletion was induced by repetitive saline lavage. Warmed saline was instilled into the lungs and gently retrieved; this was repeated until Pa^sub O2^, fell below 125 mm Hg with 100% inspired oxygen. After surfactant depletion, the animals were allocated to either a noninjurious or an injurious ventilation strategy, and mechanically ventilated for 90 min. In the noninjurious group, lungs were recruited and maintained with tidal volume 8 ml/kg and PEEP of 14 cm H2O throughout. In the injurious group, ventilation was with tidal volume of 25 ml/kg and PEEP of 4-7 cm H2O. Five series of experiments were performed (see Figure E1 of the online supplement) as follows.

Series I-Histopathology, Stereology, and Pathophysiology

Histopathologic evaluation was performed on the dependent and non-dependent tissues of animals in both groups (n = 8 in each group), as previously described (12). All histopathologic examinations were conducted by investigators blinded as to experimental group allocation and anatomic site of the tissue sample. The same lung tissues embedded in paraffin were utilized for the stereologic analysis, and mean alveolar volume of each tissue was calculated (Figure E2) (13, 14).

Airspace and airway injury were based on a quantitative method as previously described (8, 12). Airspace injury was expressed in terms of Hyaline Membrane Score: the total number of alveoli, alveolar ducts, and alveolar sacs with (any) hyaline membrane present in eight randomly selected fields per lung section, divided by the total number of alveoli, alveolar ducts, and alveolar sacs in the same fields, then multiplied by 100. All airways, which include intrapulmonary bronchi and terminal bronchioles, were evaluated and scored in each lung section. Each airway was evaluated by the total range of epithelial desquamative injury (bubble formation, cleft formation, sloughing, and denudation), scored from 0-3, summed per lung section. Entirely flattened epithelium without desquamative injury was scored as 1. Airway Injury Score was calculated by dividing this sum by the number of all airways examined. A composite lung injury score (comprising airway and alveolar injury) was calculated. Scores were normalized as described in the complete methodology (see online supplement).

Finally, a semiquantitative analysis of inflammatory cell infiltration was assessed by the following three parameters: presence of inflammatory cells (1) in interalveolar septa and (2) within alveoli, both representing airspace inflammation; and (3) peribronchiolar infiltration of inflammatory cells, representing airway inflammation. Each parameter was evaluated semiquantitatively, using a five-grade scale: normal = 0, questionable change = 1, minimal change = 2, moderate change = 3, and marked change = 4.

Series II-Gravimetric and Volumetric Analysis and Real-Time Polymerase Chain Reaction

Animals were allocated to nonlavaged control, lavaged control, noninjurious strategy, or injurious strategy groups as before (n = 6 in each group). At the conclusion of the protocol, the trachea was occluded at end-expiratory lung volume and the animals exsanguinated. The functional residual capacity (FRC), degree of atelectasis, and lung wet-to-dry weight ratio were measured (15, 16). The dependent and nondependent tissues of the left lung were assessed for cytokine mRNA expression using real-time polymerase chain reaction (PCR) (17).

Series III-Computed Tomography Scan

To exhibit the representative computed tomography (CT) scan images of lung injury, animals were ventilated with either noninjurious or injurious strategy as before (n = 2 in each group). The trachea was occluded at end-expiratory lung volume, and CT scanning performed (18).

Series IV-Cytokine mRNA and Myeloperoxidase Protein Expression

To localize the expression of cytokine mRNA and myeloperoxidase protein, in situ hybridization and immunohistochemistry was performed as previously described (19) on lungs from animals ventilated with either noninjurious or injurious strategy (n = 1 in each group).

Series V-Pressure-Volume Characteristics

The pressure-volume (P-V) curves were constructed on animals after ventilation with either a noninjurious or an injurious strategy (n = 4 in each group).

Statistics

Statistical analysis was as previously recommended (20). Data are expressed as mean ± SD, or median ± quartiles (nonparametric data). Student's t test or analysis of variance (ANOVA) followed by Student-Newman-Keuls testing was used. ANOVA on ranks was used for the semiquantitative analysis of inflammatory cell infiltration. Significance was set at p < 0.05.

RESULTS

Series I-Histopathology, Stereology, and Pathophyslology

Baseline data. The entire protocol was completed on eight animals in each group (noninjurious ventilation vs. injurious ventilation for 90 min). Baseline parameters including animal weight, arterial blood pressure, airway pressure, arterial pH, Pa^sub O2^, Pa^sub CO2^, HCO^sub 3^^sup -^, and base excess (Table 1), and static respiratory system compliance (Table 2) were comparable in both groups before and after saline lavage.

Ventilation strategy. Following group allocation, Pa^sub O2^ was below 125 mm Hg with injurious ventilation, and was significantly greater-above 350 mm Hg-with noninjurious ventilation (p < 0.05; Table 1). Arterial oxygenation was more impaired with injurious than with noninjurious ventilation (p < 0.05; Table 1). Pa^sub CO2^ was elevated, but similar, in both groups.

Functional impact of ventilation strategy. The static compliance of the respiratory system did not change with noninjurious ventilation but decreased after injurious ventilation (p < 0.05; Table 2). In addition, there was global impairment in overall physiologic parameters associated with injurious ventilation, but not with noninjurious ventilation (Table 1).

Development of atelectasis. Animals ventilated with injurious ventilation demonstrated a greater alveolar density (i.e., higher number of alveoli per unit volume) in tissue sampled from dependent (74.1 ± 18.8 × 10^sup 5^/cm^sup 3^) versus nondependent (51.9 ± 12.8 × 10^sup 5^/cm^sup 3^) regions (p < 0.05; Figure 1A). There were no regional differences (i.e., dependent vs. nondependent) in alveolar density in animals ventilated noninjuriously.

Additional stereologic data, including Cavalieri volume (13), and the proportion of lung tissue composed of intraacinar tissue (i.e., airspace, alveolar wall) and extraacinar structures are presented (Table 3). Quantitative stereologic analysis showed that lungs ventilated with injurious ventilation demonstrated a greater volume density of alveolar wall and a smaller volume density of airspace compared with lungs of the noninjurious group (p < 0.05, two-way ANOVA; Table 3), lacking regional differences in either group. Because we believed that the alveolar volume would directly reflect the regional differences in alveolar recruitment, we compared the mean alveolar volume between the dependent and nondependent regions. The estimated mean regional alveolar volume was less in the dependent versus nondependent regions of the injuriously ventilated lungs (p < 0.05; Figure 1B), but regional differences in alveolar volume were not observed in the noninjuriously ventilated lungs (Figure 1B). There was a significant interaction between ventilatory strategy and vertical region, indicating that ventilation strategy (noninjurious vs. injurious) with position in the lung (dependent vs. nondependent) determined mean regional alveolar volume (p < 0.05, two-way ANOVA; Figure 2).

Analysis of lung injury. The composite lung injury histopathologic score was greater with injurious than with noninjurious ventilation (6.55 ± 3.06 vs. 0.56 ± 0.41, p < 0.05). There were also significant regional differences in the composite lung injury score after injurious (but not noninjurious) ventilation, with the following rank order: injurious/nondependent > injurious/dependent > noninjurious/dependent [asymptotically =] noninjurious/nondependent (p < 0.05; Figure 3A).

Alveolar injury (i.e., hyaline membrane score) was minimal-and without regional differences-in noninjurious ventilation (Figure 3B, Figures E3A and E3B). In contrast, injurious ventilation caused severe alveolar-associated injury, which was greatest in the nondependent regions (p < 0.05; Figure 3B, Figures E3C and E3D). In addition, alveolar injury in both dependent and nondependent regions was significantly greater after injurious ventilation than in either region after noninjurious ventilation (p < 0.05; Figure 3B).

Airway-associated injury (i.e., epithelial lesions) was also minimal after noninjurious ventilation (Figure 3C, Figures E3E and E3F). However, there was marked airway epithelial injury in both dependent and nondependent regions after injurious ventilation that was greater than in either region after noninjurious ventilation (p < 0.05; Figure 3C, Figures E3G and E3H). Furthermore, there were no differences in the degree of airway epithelial injury between dependent and nondependent regions after injurious ventilation (Figure 3C).

Finally, a semiquantitative analysis of inflammatory cell infiltration paralleled the above findings (Table 4). Alveolar infiltration of inflammatory cells (airspace inflammation) was greater in the nondependent than in the dependent regions after the injurious ventilation, and the rank order of inflammatory cells in the interalveolar septa, as well as within alveoli, was as follows: injurious/nondependent > injurious/dependent > noninjurious/ dependent [asymptotically =] noninjurious/nondependent (Table 4). In contrast, peribronchiolar infiltration of inflammatory cells (airway inflammation) did not show regional differences in either group, with the following rank order: injurious/nondependent [asymptotically =] injurious/dependent > noninjurious/dependent [asymptotically =] noninjurious/nondependent (Table 4).

Series II-Gravimetric and Volumetric Analysis and Real-Time PCR

Baseline comparisons. The entire protocol was completed on six animals in each group. All parameters were comparable among the groups at baseline, and changes in Pa^sub O2^ and static compliance reflected the initial experimental series (above).

FRC and atelectasis. The effects of the two ventilation strategies on atelectasis formation (i.e., percent atelectasis) and the FRC were measured at end-expiration. The FRC and calculated percent atelectasis were similar in the lavaged control and lavaged noninjurious groups, but the FRC was significantly less (p < 0.05; Figure 4A), and atelectasis significantly greater (p < 0.05; Figure 5) in the lavaged injurious group.

Global and regional lung water. The rank order of overall wet-to-dry weight ratio was as follows: injurious > noninjurious [asymptotically =] lavaged control > nonlavaged control (p < 0.05; Figure 4B). The rank order of the whole-lung wet weight corrected for body weight was the same (Table 5). However, there are two sources of lung water in the context of the current experiment: exogenous (i.e., derived from the saline lavage process) and endogenous (i.e., derived from pulmonary edema resulting from acute lung injury). The quantity of retained water (i.e., saline) was similar in each group (Table 5), but the injury-associated edema was as follows: injurious > noninjurious [asymptotically =] lavaged control (p < 0.05; Table 5).

The regional wet-to-dry weight ratio was the same in all four groups: nonlavaged control, lavaged control, lavaged noninjurious, and lavaged injurious (Figure 4B). The ratio was not different-in total or by region-between the lavaged control and lavaged noninjurious, but the ratio was significantly greater in the lavaged injurious than in the other three groups, regardless of the lung region (p < 0.05; Figure 4B).

Pulmonary cytokine mRNA expression. The regional cytokine mRNA expression was examined in the noninjurious and injurious groups, as well as in an additional control group (lavaged, not ventilated), and was expressed as fold-change relative to the normal control group (not lavaged, not ventilated). The profile of interleukin (IL)-1ß, IL-6, and macrophage-inflammatory protein (MIP)-2 was similar, with minimal mRNA expression in the lungs of lavaged control (nonventilated), and the following rank order: injurious/nondependent > injurious/dependent [asymptotically =] noninjurious/ nondependent [asymptotically =] noninjurious/dependent > control/nondependent [asymptotically =] control/dependent (p < 0.05; Figures 6A-6C). The tumor necrosis factor-a mRNA expression was minimal in the lavaged control, and was greater after noninjurious or injurious ventilation (p < 0.05), without regional differences in either group (Figure 6D).

Series III-CT Scan

Six animals (i.e., two animals in each of three groups) were studied and CT performed at end-expiration. In the nonlavaged control group, the lungs were homogenously aerated and no significant densities noted (Figures 7A and 7B). The lungs ventilated by the noninjurious strategy after saline lavage are similar to the normal control lungs, and were homogenously aerated with minimal densities (Figures 7C and 7D). The lungs ventilated by the injurious strategy after saline lavage showed heterogeneous distribution of densities that were preferentially located in the dependent lung regions, with aeration predominantly in the nondependent regions (Figures 7E and 7F).

Series IV-Cytokine mRNA and Myeloperoxidase Protein Expression

In situ hybridization demonstrated that IL-6 (Figures 8A and 8B) and IL-1ß (Figures E4A and E4B) mRNA followed the same pattern of expression, which depended on the ventilatory strategy and the vertical topography. The expression of IL-6 (Figure 8A) and IL-1ß (Figure E4A) mRNA was weak in the noninjuriously ventilated lung, with no differences between dependent and nondependent regions. In the injuriously ventilated lung, strong positive signals for IL-6 and IL-1ß mRNA were detected predominantly at the alveolar epithelium in the nondependent versus dependent regions (Figure 8B, Figure E4B). Strong signals for IL-6 and IL-1ß mRNA were observed to similar degree in both dependent and nondependent airways (Figure 8B, Figure E4B). Specificity of the mRNA hybridization for both IL-6 and IL-1ß antisense probes was confirmed by the absence of positive signals using corresponding sense probes (Figures 8A and 8B, and Figure E4, left panels).

Immunohistochemistry demonstrated increased numbers of myeloperoxidase (MPO)-positive neutrophils and pulmonary alveolar macrophages (Figure E5) in the intraalveolar septa and alveolar space from the lavaged injurious (Figure E5B) compared with lavaged noninjurious lungs (Figure E5A). MPO expression was greater in the nondependent regions than in the dependent regions (Figure E5B) of the injurious, but not of the noninjurious, lungs (Figure E5A).

Series V-P-V Characteristics

The static P-V curves were constructed from the experimental level of PEEP up to an airway pressure of 25 cm H2O above PEEP. The lung P-V curves showed that the static compliance of lung was more impaired in the lavaged injurious than in the lavaged noninjurious group (Figure 9).

Among all the experimental series (I-V), 18 animals did not survive the protocol-all in the injuriously ventilated group.

DISCUSSION

These data suggest that although distal airway injury occurs in lungs with extensive regional atelectasis, such distal airway injury is generalized throughout the lung and is not localized to the areas that are atelectatic. In contrast, the alveolar injury found in such lungs does not occur in atelectatic regions, but occurs instead in remote nonatelectatic alveoli.

Ventilator-induced Lung Injury and Atelectasis

The current model demonstrated several important features of ventilator-induced lung injury, including the finding that a combination of high tidal volume and lack of recruitment caused overall lung injury, and that such injury was characterized by consistent alterations in histology, lung mechanics, cytokine alterations, and gas exchange. In addition, the effect of ventilation strategy on the distribution of atelectasis was confirmed by histology, stereology, and CT data.

This first demonstration of regional atelectasis in an animal model of this size is consistent with several well-established concepts. The vertical gradient of transpulmonary pressure is known to increase as animal size decreases (21); lung edema formation increases the vertical gradient of pleural pressure, albeit in a large animal (i.e., porcine) model (22); and regional alveolar expansion is directly proportional to the local transpulmonary pressure, as recorded by surface videomicroscopy (23). The current data thus confirm the overall concept that the distribution of regional lung volume is dependent on the vertical position along the superior-inferior axis, and not on the lung weight. The possible determinants for the regional lung volume are not only the gravitational effects of the mediastinum (24, 25) and abdomen (26) but also the force-balance relationship required for the lung to conform to the shape of the thorax (27). Regardless of the exact mechanisms underlying the regional distribution of atelectasis, our model allowed us to examine the precise spatial relationship between tissue injury and alveolar collapse.

Atelectasis and Alveolar Injury-Local Protection, Remote Injury

A striking finding in the current study is the relative absence of alveolar injury in the dependent (i.e., atelectatic) regions in the injuriously ventilated lungs and the presence of injury in the nondependent (i.e., nonatelectatic) regions. This is consistent with the "baby lung" concept proposed by Gattinoni and colleagues (4, 5). The pattern of the injury is schematically illustrated in Figure 10.

Distribution of lung injury has been described in different settings. Extremely high tidal volumes in a canine model resulted in dependent injury (28, 29) whether the lungs were preinjured (28) or not (29). In clinical ARDS, overinflation occurs in aerated areas and bronchial distention in nonaerated regions (30), and similar patterns were produced in a porcine model of multifocal pneumonia (31). The contribution of atelectasis may extend beyond ventilator-induced lung injury as recent work from our group suggests that, without supplemental oxygen, atelectasis per se can cause pulmonary vascular injury (16, 18).

In the current study, the key markers in the evaluation of lung injury-and of its distribution-were histology, cytokine mRNAs, and MPO expression. Hyaline membranes and the infiltration of polymorphonuclear cells are well-established indicators of acute lung injury (32, 33). Because rodents have very short or absent respiratory bronchioles, formation of hyaline membranes confined to the alveolar ducts, sacs, and alveoli is therefore representative of the alveolar injury. MPO reflects neutrophil and monocyte extravasation (34), and together with the increased polymorphonuclear cell infiltration, the increased number of MPO-positive inflammatory cells confirmed increased inflammatory cell activity in the nondependent versus dependent regions after injurious ventilation.

Proinflammatory cytokine mRNA expression (IL-1ß, IL-6, and MIP-2) evaluated by real-time PCR also supports the concept of regional distribution of lung injury. Maximal expression for these proinflammatory cytokines was detected in the nondependent regions of the injuriously ventilated lungs. In situ hybridization provides strong evidence for the regional distribution of lung injury. Both IL-6 and IL-1ß mRNAs followed the same pattern of expression depending on the ventilatory strategies and vertical topography. The expression of IL-6 and IL-1ß mRNAs was weak in the noninjuriously ventilated lung, with no differences between dependent and nondependent regions. In the injuriously ventilated lung, the strongest positive signals for IL-1ß and IL-6 mRNA were detected predominantly at the alveolar epithelium in the nondependent versus dependent regions.

To our knowledge, this is the first study quantitatively demonstrating the regional differences in the expression of cytokine mRNA in an in vivo model. In the injuriously ventilated lungs, the tidal volume would be shifted by the dependent atelectasis toward the nondependent regions, in which alveolar overdistention would occur, resulting in increased mRNA expression of proinflammatory cytokines.

Atelectasis and Distal Airway Injury

The classic paradigm of atelectasis causing local repetitive opening and closing of the airways was initially proposed more than 20 yr ago (7) and has been supported by several elegant studies (8, 9), but not directly proved (35). This hypothesis predicts that both airway and alveolar injury will be most severe in the atelectatic regions of the lung. However, the diffuse airway injury demonstrated in the current study argues against repetitive opening, closing, and injury of the distal airways being confined exclusively to regions that are atelectatic. Indeed, the in situ hybridization showing marked staining for IL-1ß and IL-6 in both dependent and nondependent airways supports the histologically apparent airway injury (i.e., airway epithelial lesions). Our group previously suggested that airway epithelium may be a primary, albeit not the sole, focus in the pathogenesis of ventilator-induced lung injury (19). Taken together with the current data this suggests that ventilation with high tidal volume induces significant airway injury through the entire lung. Such a concept does not refute repetitive airway open and closing as a pathologic process associated with atelectasis, but strongly suggests that associated airway injury is generalized and not localized to atelectatic regions.

The potential mechanisms whereby atelectasis accentuates injury induced by high tidal volume are illustrated in Figure 10. In the current experiments, airway injury was generalized in all lung regions (i.e., atelectatic and inflated regions), and although the mechanism of such injury is unclear, it likely involves local shear stress that might be related to the movement of foam (i.e., air and fluid) with tidal ventilation (36).

Study Limitations

There are several limitations to extrapolation of the current data. First, the atelectasis clinically develops secondarily to lung injury over several days, and the occurrence of atelectasis and lung injury in our model was not identical to the clinical experiences. However, we believe that the surfactant depleted model is useful for our purpose because we focused on the role of atelectasis in the ongoing lung injury process. Second, we considered testing low tidal volume and low PEEP to evaluate the effects of derecruitment per se, but animals were not able to tolerate such a strategy in the current pilot studies, or in other models (16, 18, 37). Third, we could not actually observe the opening and closing of distal airways, and although dynamic CT could have been illustrative as in larger animal models (38-40), it has not been reported in small rodents. Although histology may be a gold standard for demonstration of lung injury, epithelial lesions are optimally seen with electron microscopy.

Fixation, dehydration, and embedding will obviously alter alveolar dimensions from those that existed in vivo due to a number of reasons. For example, lungs were fixed at an airway pressure that produces inflation to total lung capacity in a normal lung and served as a useful standard across all experimental groups, allowing comparison under uniform inflation conditions. The alveolar volumes in any region thus reflected regional compliance. However, this standard fixation pressure was clearly different from the continuously varying pressures that occurred throughout the respiratory cycle in vivo. The further steps in the preparation of histologic sections including dehydration and embedding also alter tissue volumes. Other approaches have been used previously to try to avoid artifacts due to section preparation including frozen sections. However, these are also subject to problems such as the marked changes in gas and fluid volumes that can take place during freezing. Although no method of slide preparation is without artifact, the method of fixation that we used ensured that the postfixation alveolar volume was determined by local regional compliance and thus reflected differences throughout the lung. This view is corroborated by the CT images, which were obtained at the values of PEEP used in vivo. Thus, the absence of regional variability on CT parallels the absence of regional variability in alveolar volume on histology in the noninjuriously ventilated lungs. Conversely, the heterogeneity of regional injury observed in the CT images of the injuriously ventilated lungs parallels that observed in the histology of these lungs and strongly supports our finding that the injurious ventilation strategy produced both regional atelectasis and hyperinflation in the injured lung.

Finally, it is important to consider the saline lavage process. The relative protection of the dependent lung tissue could have occurred as a result of retained saline perhaps forming an impenetrable layer of fluid that protected the alveoli from injurious inflation pressure. It is unlikely that such protection represents an artifact of the experimental model because the retained volume of saline from the lavage process was not different in any of the experimental groups (noninjurious vs. injurious ventilation), and in context of injurious ventilation (where the dependent protection was observed) the retained saline was far exceeded by lung water produced endogenously from the ongoing lung injury process.

CONCLUSIONS

We demonstrate that atelectasis in this model is distributed predominantly in the dependent regions. The alveolar injury was maximal in the nondependent (i.e., nonatelectatic) regions of the injuriously ventilated lungs, consistent with redistribution of ventilation from atelectatic to nonatelectatic areas resulting in overinflation injury. In contrast, distal airway injury was homogeneously distributed. Although we could not actually observe the repetitive opening and closing of distal airways, the current data suggest that the hypothesis of distal airway injury associated with the cyclical phenomenon occurring exclusively in atelectatic regions should be broadened to include distal airways throughout the entire lung. Taken together, these findings suggest that we might revise our current views on the role of atelectasis in the pathogenesis of ventilator-induced lung injury.

Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript.

Acknowledgment: The authors thank Dr. Yoanna Skrobik and Dr. A. Charles Bryan for critically reviewing the manuscript, and Guila Ben David and Nancy Parfield for their expertise in obtaining the computed tomography radiographs. Dr. Bryan died in August, 2005.