Using a Novel Microfabricated Model of the Alveolar-Capillary Barrier to Investigate the Effect of Matrix Structure on Atelectrauma

The alveolar-capillary barrier is composed of epithelial and endothelial cells interacting across a fibrous extracelluar matrix (ECM). Although remodeling of the ECM occurs during several lung disorders, it is not known how fiber structure and mechanics influences cell injury during cyclic airway reopening as occurs during mechanical ventilation (atelectrauma). We have developed a novel in vitro platform that mimics the micro/nano-scale architecture of the alveolar microenvironment and have used this system to investigate how ECM microstructural properties influence epithelial cell injury during airway reopening. In addition to epithelial-endothelial interactions, our platform accounts for the fibrous topography of the basal membrane and allows for easy modulation of fiber size/diameter, density and stiffness. Results indicate that fiber stiffness and topography significantly influence epithelial/endothelial barrier function where increased fiber stiffness/density resulted in altered cytoskeletal structure, increased tight junction (TJ) formation and reduced barrier permeability. However, cells on rigid/dense fibers were also more susceptible to injury during airway reopening. These results indicate that changes in the mechanics and architecture of the lung microenvironment can significantly alter cell function and injury and demonstrate the importance of implementing in vitro models that more closely resemble the natural conditions of the lung microenvironment.


Results
Micro-nano device fabrication. A schematic representation of the micro-nano biomimetic system used in this study is shown in Fig. 1A and the dimensions of the device are shown in Fig. 1B,C. The microfluidic chambers patterned on the Si wafers had a thickness of 200 µm (Fig. 1C). In this study, four different fiber matrices were made using different ratios of polycaprolactone (PCL) and gelatin and three of these matrices (100% PCL, 75/25 PCL/gelatin and 50/50 PCL/gelatin) were subjected to uniaxial mechanical testing as shown in Fig. 2. Nanofiber stiffness was successfully modulated by varying the polymer blend composition where tensile tests revealed that adding gelatin to the polymer blend significantly modulates the stress-strain relationships for the fiber mesh ( Fig. 2A). Specifically, increasing amounts of gelatin reduces the linear elastic modulus of the fiber meshes, with median modulus values of 0.36, 2.10, and 7.20 MPa for pre-wetted fiber meshes fabricated with blends containing 50/50 (PCL/gelatin), 75/25 (PCL/gelatin) and pure PCL, respectively (p ≤ 0.009) ( Fig. 2A,B). As shown in Fig. 3A, fiber size increased significantly (p < 0.001), proportional to the gelatin content, with median fiber sizes of 5.9, 1.12, 0.88 and 0.57 µm for 25/75 (PCL/gelatin), 50/50 (PCL/gelatin), 75/25 (PCL/gelatin) and PCL fibers, respectively. As shown in Fig. 3B, fiber morphology also changed depending on the gelatin concentration, where the use of increased gelatin concentrations yield fibers with a ribbon-like structure while increasing PCL concentration leads to a more dense filament-like structure.
Cell culture. Using this system we were able to successfully seed lung epithelial and endothelial cells in a co-culture configuration on ~20 μm thick PCL-gelatin fiber meshes. Figure 4A shows photographs of the micro-nano biomimetic system used in this study, and a diagram illustrating the distribution of the different cell types within the system. Figure 4B shows confocal fluorescence images of the cells (epithelial and endothelial) cultured on the PCL mesh and demonstrate a high degree of confluence. We also show scanning electron microscopy (SEM) images of the micro-nano fibrous substrate used for cell culture in Fig. 4B. Fig. 5, cells cultured on fiber meshes with different stiffness and structure exhibited different patterns of actin cytoskeletal distribution. Before reaching a confluent state, epithelial cells cultured on softer, less dense fibers (i.e. 50/50 PCL/gelatin fibers), had a more spread morphology in comparison with those on a more rigid, dense fiber mesh (i.e. 100% PCL fibers) (Fig. 5A). Endothelial cells presented spread morphology on both the soft and rigid fiber meshes with increased formation of actin stress fibers when cultured on 50/50 (PCL/gelatin) fibers vs. PCL fibers (Fig. 5B). As shown in Fig. 6A, cells cultured on fiber meshes with different stiffness/structure also exhibited different amounts of epithelial and endothelial tight junctions (ZO-1 and occludin) after 3-days of co-culture. For cells co-cultured on the softer, less-dense 50/50 fibers, we observed little to no ZO-1 or occludin formation in epithelial cells and minimal ZO-1 and occludin formation in endothelial cells. Conversely, for cells co-cultured on the stiffer more-dense PCL fibers, we observed significant ZO-1 and occludin formation in both epithelial and endothelial cells. However, endothelial cells exhibited more defined tight junctions compared to epithelial cells (Fig. 6A). This increase in TJ protein expression on the more rigid/dense PCL fibers was confirmed via western blot analysis of endothelial and epithelial cells cultured on the different fiber substrates (Fig. 6B).

Evaluation of cellular responses. Actin cytoskeleton and Tight junction formation. As shown in
Epithelial/endothelial barrier function. A decreasing trend in dextran transport was observed across epithelial/ endothelial co-cultures on PCL and 75/25 (PCL/gelatin) fiber meshes in comparison with the 50/50 (PCL/gelatin) fibers (Fig. 6C). Specifically, compared to control conditions with no cells, the apparent permeability coefficient decreased significantly (p < 0.029) by 49% and 67% for cells co-cultured on 75/25 (PCL/gelatin) and 100% PCL fibers. In contrast, cells cultured on the 50/50 (PCL/gelatin) fibers did not provide significant barrier function since these cells exhibited an apparent permeability that was not statistically different than the control, no-cell conditions (p = 0.485).
Fluid-filled airway reopening simulations. As shown in Fig. 7, simulating 5 airway reopening events resulted in a significantly higher percentage of cellular death for cells cultured on PCL fibers in comparison with those on gelatin-containing fibers (p ≤ 0.018 for cells on PCL vs. 50/50 fibers). After 5 airway reopening events epithelial cell injury values of 11.4%, 14.0%, and 16.9% were observed for A549 and HUVECs co-cultured on matrigel-coated fibers containing 50/50, 75/25, and 100/0 (PCL/gelatin), respectively (Fig. 7A,B). These results were validated by repeating these airway reopening experiments using co-cultures of primary human small airway cells (HSAECs) and HUVECs. For these studies, 5 airway reopening events also resulted in significantly higher cell death for cells culture on PCL fibers (p < 0.02 for cells on PCL vs 50/50 and for cell on PCL vs 75/25 fibers). The mean injury after airway reopening for primary cells cultured on matrigel-coated fibers containing 50/50, 75/25, and 100/0 (PCL/gelatin) were 12.3%, 10.6%, and 16.2%respectively (Fig. 7A,B).

Discussion
The alveolar-capillary microenvironment is a complex structure that normally provides both efficient gas-exchange and a tight barrier that prevents fluid from entering airspaces within the lung 18 . Disruption of this barrier during acute lung injury leads to pulmonary edema and severe hypoxia and these patients often require mechanical ventilation (MV). Although life-saving, MV generates biophysical forces that can exacerbate lung injury and barrier disruption. Although most investigators have focused on how over-distension causes lung injury 19,21 , recent studies indicate that the surface tension forces generated during the reopening of atelectatic airways/alveoli is the main source of cellular injury 11,12 . However, most previous studies have only investigated how airway reopening causes injury to epithelial cells cultured on rigid glass or plastic substrates [13][14][15][16] and therefore did not account for the complex nature of the lung ECM that consists of a compliant fibrous matrix. Recently, our group developed an atelectrauma model that accounted for one aspect of the lung microenvironment where cells were cultured on soft polyacrylamide gels to account for the complaint nature of lung tissue 28 . That study demonstrated that increases in substrate stiffness leads to a change in cell morphology which in turn causes increased cell injury during airway reopening. Interestingly, increased substrate stiffness also led to increased focal adhesion protein expression (i.e. increased expression of phosphorylated-PAX) and this resulted in less cell detachment during airway reopening. Although that study provided novel insights into the effect of lung compliance on atelectrauma, it did not account for the fibrous nature of the lung ECM and could not be used to evaluate epithelial-endothelial cell interactions and barrier permeability. Therefore, in this study we have used electrospinning techniques to model the fibrous nature of the lung ECM and have developed a novel co-culture model of epithelial-endothelial cell interaction that mimics the in vivo physiology. We have incorporated this co-culture model into a microfluidic system that can simulate the surface tension forces responsible for cell injury during airway reopening and have used this system to evaluate how ECM structure alters barrier permeability and cell injury.
In this study, electrospinning of PCL/gelatin polymer blends was used to model the ECM since electrospun fibers have been shown to simulate the natural topography of the basal membrane [29][30][31][32][33] . Electrospinning allowed for the precise modulation of fiber mesh properties such as fiber diameter, stiffness and density and as a result, we were able to simulate different micro-architectures relevant to normal and abnormal lung tissue. For example, during pulmonary fibrosis increased collagen deposition and remodeling leads to a dense ECM and a significant increase in local tissue stiffness 34 . On the other hand, emphysema involves destruction of the lung ECM architecture and thus a less dense fiber network. Therefore, as shown in Figs 2 and 3B, in this study the 100% PCL fibers can be considered representative of a fibrotic state while the 50/50 PCL/gelatin fibers can be considered representative of the emphysematic state. Of course the clinical pathology often involves other structural changes such as enlarged airspaces during emphysema. The implementation of soft lithography during the fabrication of the current device allows for precise control over channel morphology and dimensions, allowing for easy modulation of the channel volume and length, and therefore the simulation of airways with different dimensions. Although the current study focused solely on how changes in fiber diameter and density influence barrier function and cell injury, future studies could use this system to investigate how other parameters that change during disease, such as airway dimensions, alter barrier function and atelectrauma.
The use of 100% PCL and 75/25 PCL/gelatin mixture lead to dense fiber meshes and increased stiffness as measured by the linear elastic modulus (Figs 2 and 3). Epithelial and endothelial cells cultured on this rigid dense fiber matrix formed tighter monolayers with higher tight junction formation and more effective barrier properties. In contrast, the polymer blends containing 50/50 PCL/gelatin formed softer, low density fiber meshes, which lead to looser monolayers with lower barrier properties, as measured by FITC-labeled dextran (70 KDa) transport across epithelial/endothelial cell co-cultures (Fig. 6C). This response in terms of better barrier properties for cells   cultured on the rigid/dense PCL fibers could be explained by the observed changes in actin cytoskeletal distribution and TJ formation for epithelial and endothelial cells. Endothelial cells on soft 50/50 fibers presented higher formation of actin stress fibers and lower expression of TJs (Figs 5 and 6). As described by Wysolmerski and Lagunoff 35 , and subsequently by Garcia et al. 36 , these type of changes in the actin cytoskeleton of lung endothelial cells play a key role in the regulation of endothelial retraction and barrier dysfunction through a pathway dependent on myosin light chain kinase (MLCK) activation and myosin light chain phosphorylation [35][36][37] . Interestingly, epithelial cells on the soft 50/50 fibers presented diffuse actin staining while epithelial cells on rigid 100% PCL fibers exhibited peri-nuclear condensation of actin. Given the lack of actin stress fibers in epithelial cells, it is unlikely that epithelial retraction plays a significant role in barrier dysfunction. However, epithelial cells cultured on the rigid 100% PCL fibers did exhibit modest TJ formation and this result is consistent with previous studies where A549 cells cultured at an air-liquid did express ZO-1 and occludin in culture. However, epithelial cells cultured on the soft 50/50 fibers did not exhibit TJ formation and this could also partially explain why the co-culture of epithelial/endothelial cells on these soft fibers did not exhibit decreased permeability.
Although the epithelial-endothelial co-culture presented better barrier properties on rigid PCL fiber meshes, airway reopening simulations indicate that epithelial cell death was significantly higher after airway reopening for cells co-cultured on the highly dense/rigid PCL nanofibers. In contrast, the degree of epithelial cells death was lower for cells co-cultured on the less dense/rigid 75/25 and 50/50 PCL/gelatin meshes (Fig. 7A,B). This result is in agreement with previous work by our group, where lung epithelial cells cultured on a stiffer polyacrylamide gels were more susceptible to plasma membrane rupture during airway reopening in comparison with cells on softer gels 28 . In that study we demonstrated that this differential cellular response to substrate stiffness could be explained by significant changes in cell morphology where cells on softer substrates exhibited a flatter topography. As described by Jacob and Gaver 38 , this flatter morphology dissipates the hydrodynamic forces applied by the moving air-liquid interface and is therefore consistent with less cell injury during reopening. In line with these previous studies, in this study epithelial cells on softer 50/50 fibers presented a more spread morphology in comparison with the rounded cells observed on the rigid fibers (Fig. 5A). In addition to responding to matrix stiffness, in this case we also hypothesized decreased spreading on the 100% PCL fibers could be attributed, in part, to a higher density per unit area of submicron-scale fiber structures (characteristic of the rigid PCL fiber meshes) with which the cells interact strongly and would limit their spreading. However, we acknowledge that there are other potential explanations for higher cell injury on the stiffer fibers. First, rigid fiber meshes will sustain less deformation during reopening and as a result, the forces applied by the moving air-liquid interface may be "absorbed" by the epithelial cells resulting in more cell deformation and a higher chance of plasma membrane rupture. An alternative hypothesis is that higher TJ expression in epithelial cells on the rigid PCL fibers may alter the way hydrodynamic stresses are distributed within the cell. In fact, a study by Gaver's group indicates that higher expression of the cytoplasmic TJ multiprotein ZO-1, which directly interact with actin filaments, provides enhanced structural support to cells, dampening out the tangential component of the pressure gradient imposed during airway reopening 39 . It is also important to consider the possible effect of TJ formation in the response of epithelial cells during airway reopening, since the higher expression of cytoplasmic TJs (cell-cell adhesion points) for epithelial cells on rigid PCL fibers may alter stress distribution within the monolayer and thus the strains experienced by the plasma membrane. However, future computational studies that simulate cell deformation during airway reopening may be needed to identify the role of TJ on cell injury 40,41 .
Although we have developed a novel system that simulates important biophysical and physiological features of the lung microenvironment, including ECM fiber architecture, endothelial-epithelial interactions and the fluid mechanical forces responsible for atelectasis, future studies could incorporate additional features to more accurately mimic the lung microenvironment. First, this study utilized non-biological polymers due to the ease and controllability of these systems. However, future studies could incorporate electrospun biopolymers such as elastin and collagen 42 and could also introduce interstitial cells such as lung fibroblasts which could remodel the extracellular matrix. In this study, we used our system to investigate how one type of mechanical force, surface tension forces during airway reopening (i.e. atelectrauma), influences cellular responses. However, other mechanical forces such as mechanical stretch, may also occur in mechanically ventilated lungs. Although the current PCL system is not particularly amenable to simulating mechanical stretching during volutrama due to the very stiff nature of this polymer, future studies could use other more compliant polymers which would allow for the simultaneous application of stretching and surface tension forces. Although the current study indicates that ECM mechanics may be an important determinate of ventilation induced lung injury, future studies should test predictions from the in vitro system using in vivo models. For example, future studies could use mouse models of emphysema and fibrosis to directly test the hypothesis that ventilation of rigid lungs during fibrosis would lead to more cell injury while ventilation of more compliant lungs would lead to more barrier disruption. Finally, although the current device accounts for epithelial-endothelial interactions, future studies could incorporate other cells that play an important role in the alveolar microenvironment such as alveolar macrophages.

Conclusions
In summary, we have developed a micro-nano biomimetic device that more closely replicates the alveolar-capillary barrier and have used this system to investigate how ECM mechanics and structure alters barrier permeability and cell injury during airway reopening. This system accounted for epithelial-endothelial interactions present at the alveolar level, the natural fibrous topography of the basal membrane, and allowed for easy modulation of extracellular matrix stiffness/structure. Using this device, we found that epithelial and endothelial cells are responsive to changes in fiber stiffness and architecture, where epithelial cells developed a more spread morphology and exhibit lower TJ formation on softer and less dense fiber meshes. Endothelial cells presented a similar response in terms of TJ formation but had increased formation of actin stress fibers on softer fibers. Endothelial cells also presented a fairly constant spread morphology on both soft and rigid fiber meshes. These responses lead to decreased barrier permeability for cells co-cultured on rigid and denser fiber meshes. Changes in substrate stiffness and architecture also influenced the degree of cell injury during airway reopening, where cells co-cultured on dense, rigid fiber meshes were more susceptible to injury presumably due to changes in cell morphology observed for epithelial cells cultured on rigid vs. soft fiber meshes. These results demonstrate the importance of implementing in vitro models that more closely resemble the natural conditions of the lung microenvironment. Continued application of this system could lead to a cost-effective, controlled, and systematic analysis of the biomechanical mechanisms involved in ventilator induced lung injury as well as several other pulmonary pathologies.

Materials and Methods
Micro-nano device fabrication. The system developed in this study consists of a three-dimensional multilayered structure (Fig. 1A) that contains two microfluidic chambers which model the airway space and capillary lumen. These chambers are interfaced with a porous mesh of polymeric fibers, which mimics the basal lamina in the lung and is the substrate for co-culturing lung epithelial and endothelial cells. The top and bottom chambers of the device were fabricated using soft lithography techniques and are made of polydimethylsiloxane (PDMS), while the fiber mesh was obtained via electrospinning of a polymer blend containing polycaprolactone -(PCL) and gelatin (porcine skin gelatin type A).
Microfluidic chambers. The layout of the two microfluidic chambers was created using the SolidWorks 3D CAD Design Software (Dassault Systèmes SolidWorks Corp., Waltham, MA) and then printed on a transparency for subsequent use as a mask during the photolithography process. Both chambers were fabricated in PDMS (Dow Corning, Midland, MI) and consist of a hexagonal channel with inlet and outlet connections at the ends of the channel (20mm length, 10mm width, and 200 µm depth) (Fig. 1B,C). The thickness/depth of the microchannel was selected to resemble the diameter of distal pulmonary airways. PDMS was selected due to its excellent stability under physiological-like conditions, biocompatibility, and transparency which allowed imaging inside of the microchannels using regular light and fluorescence microscopy.
The PDMS microfluidic chambers were molded from a silicon master fabricated via standard UV photolithography. Briefly, before spin coating the Si wafers were pre-cleaned with acetone, methanol, and ethanol, and then treated with oxygen plasma (oxygen at 10 sccm, 200 watts, and < 100 mTorr, for 30 seconds), in order to improve the adhesion of the photoresist to the substrate. To further improve the stability of the SU8-2075 layer, an additional thin layer (5 µm) of SU8-2005 was spin coated (at 3000 RPM for 30 seconds) into the Si wafer and flood exposed prior to the deposition of the negative tone photoresist SU-8 2075 (Microchem Corp, USA). The photoresist SU-8 2075 was then spin coated on a 4 inch silicon (Si) wafer at 1000 rpm for 30 seconds to obtain a ~200-250 µm thick layer. The coated wafer was subsequently pre-baked following the directions provided by the manufacturer, and then exposed to ultraviolet light through the photomask previously developed in SolidWorks, which contains the layout of the hexagonal channel. Finally, the patterned wafer was post-processed following the parameters suggested by the manufacturer.
The PDMS microfluidic chambers were obtained by casting PDMS solutions (1:10 ratio of curing agent to precursor) on top of the patterned silicon wafer. The PDMS solutions were allowed to cure for 24 hours at 60 °C, and then peeled from the silicon master. The thickness of the PDMS microfluidic chambers was confirmed via scanning electron microscopy (SEM) (Fig. 1C).
Nanofiber mesh fabrication. Nanofiber meshes were fabricated using a standard electrospinning apparatus which consists of a syringe pump, a high voltage power supply, and a collector. Polymeric nanofibers with different stiffness were obtained by varying the ratio of polycaprolactone -PCL (Mw 14,000, Sigma-Aldrich, USA)/gelatin (porcine skin gelatin type A, Sigma-Aldrich, USA) blends. The polymer blends were produced by mixing PCL and gelatin in hexafluoro-2-propanol -HFP (Sigma-Aldrich, USA) for 24 hours at room temperature. Four PCL/gelatin weight ratios were used in this study: 100/0, 75/25, 50/50 and 25/75. These polymer blends were electrospun at 26 KV at an extrusion rate of 16 mL/h for 2 minutes with a 25 cm distance between the needle tip and collector. These electrospinning parameters were selected in order to obtain ~20 µm thick nanofiber meshes, which resembles the thickness of the lungs basement membrane 43 . The fiber morphology was characterized via SEM and representative SEM images were used to measure the fiber cross-sectional size using the image analysis software Image J (NIH, USA). Tensile test were conducted on dog-bone shaped samples previously immersed in phosphate buffered saline (PBS) for 1 hour, with a cross-head speed of 5 mm/min until failure. The Young's modulus was calculated as the slope of the most linear region of the stress vs. strain curves prior to the yield point.
Assembly of the nano-micro biomimetic system. The three dimensional sealed configuration was obtained by sandwiching the fiber mesh between both microfluidic layers. The binding between microfluidic layers was achieved via oxygen plasma treatment using a TECHNICS Benchtop reactive ion etching at 50 W and 10 cm 3 / min at standard conditions (sccm) O 2 for 30 seconds. The assembled devices were subsequently heated at 70 °C for 10 min to facilitate PDMS-to-PDMS bonding. After device assembly and prior to cell seeding, all fiber meshes were sterilized using UV light and subsequently coated on both sides using a 0.33 mg/mL growth factor reduced basement membrane matrix solution (BD Biosciences, San Jose, CA) for 1 hour at room temperature. This solution was in serum free Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F-12) (Gibco -Life Technologies, Grand Island, NY).
For co-culture experiments, HUVEC cells were injected into the bottom microfluidic chamber at a cell density of 1.8 × 10 5 cells/cm 2 , and the device was flipped over to favor gravity-mediated cell settling and subsequent adhesion. After 24 hours, the device was flipped over and a suspension of A549 or HSAEC cells (1.6 × 10 5 cells/cm 2 ) was injected into the top microfluidic chamber. The co-culture was maintained for 48 hours to reach confluence of both cell types. Culture media was changed every 24 hours.
Evaluation for cellular responses. Actin cytoskeleton. Actin distribution was evaluated on confluent cultures via fluorescence and confocal fluorescence microscopy. Briefly, cells were fixed with 10% Neutral Buffered Formalin (Thermo Scientific, Rockford, IL) for 15 minutes and permeabilized with a 0.1% Triton X-100 (Sigma-Aldrich, St. Louis, MO) solution in PBS for 5 minutes. Actin filaments were stained using Alexa Fluor 488 Phalloidin (Invitrogen, Grand Island, NY) (0.1 μM in PBS for 45 minutes). Cell nuclei were stained using DAPI (Sigma-Aldrich, St. Louis, MO) (0.2 μg/mL in PBS for 5 minutes).
TJ were further characterized via western blot analysis. For this analysis, after treatment, the samples were place on ice and the cells were harvested by immersing the fiber meshes into 50 µL of lysis buffer (10 mM HEPES, pH 7.9; 10 mM KCl; 0.1 mM EDTA; 0.1 mM EGTA; 1 mM DTT; 0.5 mM PMSF; 12.8% protease inhibitor cocktail; and 10 µL of 10% Igepal), vortexed for 15 min at 4 °C and placed overnight at −20 °C. The samples were subsequently homogenized using a pellet pestle motor (Fisherbrand, USA) and centrifuged at 17 g for 15 min at 4 °C. After centrifugation, the supernatant was collected as cytosolic lysate.
Epithelial/endothelial barrier function. Changes in barrier function for co-cultured cells on the different fiber meshes were evaluated using a previously described permeability assay 46 . In this technique the diffusion rate of fluorescein isothiocyanate (FITC)-labeled dextran (70 KDa) (Sigma-Aldrich, St. Louis MO) from the top to the bottom microfluidic chamber was measured. First, a 1 mg/mL solution of FITC-dextran was injected into the top airway chamber and a spectro-fluorimeter (BioTek Synergy HT; 485 nm excitation and 528 nm emission) was used to record changes in fluorescence units in the bottom chamber after 60 minutes. An apparent permeability coefficient (P app ) was calculated according to the following equation: P app = ΔQ/(A*Δt*C 0 ), where ΔQ is the change in concentration in the bottom chamber, A is the surface area of the monolayer, Δt is the time of 60 min and C 0 is the initial concentration in the upper chamber. Note that ΔQ was determined using a calibration curve that relates fluorescent intensity with concentration.
Fluid-filled airway reopening simulations. As described in our previous studies 15,16,22,28 , airway reopening was simulated by propagating air bubbles over the epithelial monolayer in the top chamber. Briefly, the top microfluidic channel was connected to a programmable PHD 2000 syringe pump (Harvard Apparatus, Holliston, MA) and PBS was injected into the top microfluidic channel, to simulate a surfactant-deficient airway fluid. The fluid was then retracted at a rate of 0.3 mm/s to form an air bubble inside the channel that was propagated across the epithelial cell monolayer (Fig. 1D). Cell injury after 5 reopening simulations was evaluated using a fluorescent live/dead cell viability assay (Life Technologies, Grand Island, NY). The image analysis software Image J (NIH, USA) was used to quantify total number of dead and live cells, and the percentage of cell death was computed as the ratio of live to dead cells after propagation of the air bubbles.
Statistical analysis. All data were tested for normality using a Shapiro-Wilk test. Data that followed a normal distribution were analyzed using a one-way analysis of variance (ANOVA) with post-hoc least significant difference test and were reported as mean ± standard error. Data that were not normally distributed were analyzed using non-parametric statistical analysis, i.e. ANOVA on ranks with post-hoc Dunn's test and were reported as median and graphed as median with 75 th percentile.