Mechanosensitive activation of mTORC1 mediates ventilator induced lung injury during the acute respiratory distress syndrome

Acute respiratory distress syndrome (ARDS) is a highly lethal condition that impairs lung function and causes respiratory failure. Mechanical ventilation maintains gas exchange in patients with ARDS, but exposes lung cells to physical forces that exacerbate lung injury. Our data demonstrate that mTOR complex 1 (mTORC1) is a mechanosensor in lung epithelial cells and that activation of this pathway during mechanical ventilation exacerbates lung injury. We found that mTORC1 is activated in lung epithelial cells following volutrauma and atelectrauma in mice and humanized in vitro models of the lung microenvironment. mTORC1 is also activated in lung tissue of mechanically ventilated patients with ARDS. Deletion of Tsc2, a negative regulator of mTORC1, in epithelial cells exacerbates physiologic lung dysfunction during mechanical ventilation. Conversely, treatment with rapamycin at the time mechanical ventilation is initiated prevents physiologic lung injury (i.e. decreased compliance) without altering lung inflammation or barrier permeability. mTORC1 inhibition mitigates physiologic lung injury by preventing surfactant dysfunction during mechanical ventilation. Our data demonstrate that in contrast to canonical mTORC1 activation under favorable growth conditions, activation of mTORC1 during mechanical ventilation exacerbates lung injury and inhibition of this pathway may be a novel therapeutic target to mitigate ventilator induced lung injury during ARDS.


Introduction
The acute respiratory distress syndrome (ARDS) is a devastating condition that affects over 200,000 patients annually in the U.S. with mortality rates up to 45% in its most severe form. (1)(2)(3) The only treatment for patients with ARDS is supportive care with mechanical ventilation (MV), and although life-saving, MV can exacerbate preexisting lung injury and even cause de novo injury, known as ventilator induced lung injury (VILI).(4) Limiting lung distention by using low tidal volume (TV) ventilation decreases mortality in ARDS patients, (5) but factors such as regional heterogeneity lead to persistent injury even with low tidal volume ventilation.(6) VILI arises from three injurious forces including excessive stretch (volutrauma), increased transmural pressure (barotrauma), and mechanical stress from repetitive collapse and re-opening of lung units (atelectrauma). (7) The molecular mechanisms by which these injurious forces cause lung injury remain poorly understood which has limited the development of targeted therapies to prevent lung injury in patients requiring mechanical ventilation and in patients with ARDS.
Mechanotransduction is the process by which physical forces are transduced into biologic responses. The lungs stretch cyclically during spontaneous breathing and positive pressure mechanical ventilation. The precise mechanisms by which the lungs sense and respond to injurious forces during mechanical ventilation are not well understood. Several cell types have been implicated in the transduction of mechanical signals following injurious ventilation including epithelial (8,9) and endothelial cells (10).
One approach to elucidate the molecular mechanisms of lung injury during mechanical ventilation is to identify mechanosensitive signaling pathways that are differentially activated by the various types of VILI. mTOR complex 1 (mTORC1) is a ubiquitously expressed multi-protein complex with serine/threonine kinase activity that is involved in cellular metabolism and response to stress. mTORC1 activation has been also been implicated in mechanotransduction in muscle. (11)(12)(13) In the lung, mTORC1 contributes to the pathogenesis of lung injury in response to cigarette smoke (14) and endotoxin (15).
However the role of mTORC1 activation in response to mechanical forces in the lung and its role in the pathogenesis of VILI has not been examined. In contrast to canonical activation of mTORC1 under favorable growth conditions, we discovered that mTORC1 is pathologically activated by injurious forces during mechanical ventilation. Based on this observation, we investigated whether mTORC1 activation mediates the development of lung injury following injurious mechanical ventilation and whether inhibition of this pathway might represent a novel therapeutic target for patients with ARDS.

Results
Volutrauma and atelectrauma activate mTORC1 in epithelial cells in murine models of injurious mechanical ventilation. We used a model of simultaneous volutrauma (tidal volume (VT, 12 cc/kg) and atelectrauma (positive end expiratory pressure [PEEP] 0 cm H2O) that impaired lung function, induced inflammation (i.e. bronchoalveolar lavage (BAL) neutrophils and IL6 levels), and increased barrier permeability (i.e. BAL protein levels) (Supplemental Figure 1). mTORC1 activation was assessed in lung tissue by immunoblotting for phosphorylated isoforms of the ribosomal S6 protein (16,17) and S6 kinase. (18) Spontaneously breathing control mice and mice ventilated with noninjurious settings (VT 6 cc/kg, PEEP 5 cm H2O) had low levels of mTORC1 activation. In contrast, mice ventilated with injurious high tidal volumes (12 cc/kg) or without PEEP had increased levels of phosphorylated S6 ( Figure 1A). Immunostaining revealed that the primary site of mTORC1 activation was the epithelium, with the most intense activation in the airway epithelium ( Figure 1B-D). Given that critically ill patients frequently develop ARDS and require mechanical ventilation (MV) in the setting of existing lung injury, we also subjected mice to a 2-hit model of acute respiratory distress syndrome (ARDS) in which MV was initiated in the setting of polymicrobial sepsis induced by cecal ligation and puncture (CLP). Twenty-four hours after sub-lethal CLP or sham operation, mice were subjected to injurious MV (CLP/VILI-12 cc/kg, PEEP=2.5) for 4 hours. Mice subjected to CLP/VILI had impaired lung function and increased markers of lung injury compared to uninjured control mice or either injury alone (Supplemental Figure 2) which correlated with increased mTORC1 activation in whole lung tissue ( Figure 1E). As was the case with VILI alone, the most prominent site of mTORC1 activation following CLP/VILI was the epithelium. These data demonstrate that mTORC1 is activated in lung epithelial cells following injurious mechanical ventilation.
Tsc2 deletion in airway epithelial cells exacerbates lung injury during mechanical ventilation. To determine whether mTORC1 activation in lung epithelial cells mediates the development of lung injury or represents a compensatory response that promotes recovery, we used a genetic approach to generate mice with increased mTORC1 in airway epithelial cells. We bred mice with homozygously floxed alleles of Tsc2 (a negative regulator of mTORC1) (19) to mice expressing Cre recombinase under the control of the airway epithelial specific CC10 promoter (20). CC10-Cre mice were bred to mT/mG reporter mice (21) to confirm airway epithelial specific expression of the Cre recombinase prior to breeding with Tsc2 flox mice. (Supplemental Figure 3).   Figure 3). Cre+ mice were subjected to simultaneous volutrauma and atelectrauma (12 cc/kg, PEEP 0 cm H2O) for 4 hours and had increased lung stiffness (i.e. elastance) and decreased inspiratory capacity compared to Crecontrols (Figure 2A-B). Cre+ mice were also more hypoxemic during injurious ventilation ( Figure 2C). Notably, recruitment of inflammatory cells, barrier permeability (BAL protein), and pro-inflammatory BAL cytokine and chemokine levels were not different following VILI or in spontaneously breathing control mice ( Figure 2D-H, Supplemental Figure 4). These data demonstrate that airway epithelial Tsc2 deletion exacerbates physiologic lung injury (i.e. increases lung stiffness, decreases inspiratory capacity, and impairs oxygenation) following injurious ventilation without altering lung inflammation or barrier dysfunction.

Mechanically ventilated patients with diffuse alveolar damage have increased mTORC1
activation in lung tissue. To assess whether mTORC1 is activated in patients with ARDS undergoing mechanical ventilation, we obtained formalin-fixed lung tissue from a pathology tissue bank. Specimens were analyzed from 5 patients with diffuse alveolar damage (DAD), the most common lung pathology in ARDS patients. (23,24) Normal lung tissue adjacent to resected lung tumors from 5 subjects was used as a control. Clinical characteristics of the subjects are shown in Supplemental Table 1 and there were no significant differences between groups. Notably all patients were receiving mechanical ventilation at the time of biopsy. Immunostaining for phosphorylated ribosomal S6 revealed increased mTORC1 activation in lung tissue from patients with DAD compared to normal lung tissue. (Figure 3A-F, Supplemental Figure 5) Although diffuse staining was present throughout the lung, it appeared that the most prominent site of staining was the epithelium including airway epithelial cells. (Figure 3D-F) The intensity of P-S6 staining was quantitated by image analysis and was significantly higher in patients with DAD compared to controls. (Figure 3G) To determine whether mTORC1 activation was due to the underlying lung injury or the use of mechanical ventilation, we obtained additional biopsies from spontaneously breathing patients undergoing transbronchial biopsies during bronchoscopy or from surgical wedge resections from mechanically ventilated patients. Biopsies had a variety of pathologic findings including several with diffuse alveolar damage, several types of interstitial lung disease, airway diseases, and no specific pathology. In general, spontaneously breathing patients had less mTORC1 activation compared to mechanically ventilated patients regardless of the underlying pathology. (Supplemental Figure 6) Patients without lung injury had low levels of mTORC1 activation. (Supplemental Figure 6) These data suggest that injurious mechanical ventilation in the context of underlying lung injury activates mTORC1 in the lung epithelium.
In vitro models of volutrauma and atelectrauma activate mTORC1 in primary human airway epithelial cells. To investigate the molecular mechanisms of mTORC1 activation in epithelial cells following mechanical ventilation we utilized in vitro models of the various forms of ventilator induced lung injury ( Figure 4A-C). To simulate volutrauma, primary human airway epithelial cells were grown on flexible membranes and subjected to cyclic stretch that models overdistention during injurious mechanical ventilation (25,26). Volutrauma increased phosphorylation of S6 kinase and ribosomal S6 within 30 minutes in distal small airway epithelial cells (SAECs) and proximal human bronchial epithelial cells (HBE) ( Figure 4D, Supplemental Figure 7). mTORC1 activation was dose and time dependent and rapidly decreased following cessation of injurious stretch.
(Supplemental Figure 7). Injurious stretch also increased phosphorylation of 4E-BP1, a regulator of protein translation that is phosphorylated when mTORC1 is activated (27) ( Figure 4D). Volutrauma induced mTORC1 activation was prevented by allosteric (i.e. rapamycin) and active site (i.e. Torin 2) mTOR inhibitors (Supplemental Figure 7). To model airway collapse and reopening during atelectrauma, we grew SAECs to confluence on collagen coated glass slides and exposed them to a moving air-liquid interface by propagating a fluid bubble over the monolayer which simulates the movement of edema fluid in ARDS patients undergoing mechanical ventilation. (28) As was seen with volutrauma, atelectrauma rapidly increased P-S6 levels in SAECs ( Figure   4E). Interestingly, we did not see the increase in S6 kinase or 4-EBP1 phosphorylation that was seen following volutrauma. To model barotrauma, SAECs were cultured at an air-liquid interface and cells were subjected to high levels of transmural pressure. (29,30) Although mTORC1 was activated following barotrauma, the degree of activation was much less than that seen with volutrauma and atelectrauma. (Figure 4F). In summary, these data demonstrate that mechanical forces during in vitro volutrauma and atelectrauma activate mTORC1 mirroring the findings from our murine model and patients with ARDS.
Volutrauma activates mTORC1 through reactive oxygen species-dependent activation of the ERK pathway. We sought to determine the molecular mechanisms by which injurious ventilation activates mTORC1. Several canonical signaling pathways integrate biochemical and physiologic cues and activate mTORC1 via a series of phosphorylation events, including the ERK and Akt pathways. To determine whether these pathways mediate mTORC1 activation in our volutrauma model, we treated human primary airway epithelial cells with a specific ERK 1/2 (SCH772984) or Akt (MK-2206) inhibitor immediately prior to injurious stretch. In vitro volutrauma rapidly increased ERK1/2 phosphorylation that was potently inhibited by SCH772984 ( Figure 5A). ERK inhibition completely abrogated the increase in S6 phosphorylation following injurious stretch indicating that stretch induced mTORC1 activation is ERK dependent ( Figure 5A). In  Figure 5H). These data indicate that the mechanism of mTORC1 activation during volutrauma involves the release of cellular ROS that act as second messengers to activate mTORC1 via the canonical ERK pathway.
Rapamycin mitigates physiologic lung dysfunction during VILI. To determine the therapeutic potential of targeting mTORC1 activation to prevent VILI, we treated wildtype mice with the mTOR inhibitor rapamycin (5 mg/kg IP) or vehicle control at the time mechanical ventilation was initiated. Mice were subjected to simultaneous volutrauma and atelectrauma (12 cc/kg, PEEP 0 cm H2O) for 4 hours. Treatment with a single dose of rapamycin dramatically decreased P-S6 levels in whole lung tissue following VILI ( Figure 6A). Following injurious ventilation vehicle treated mice had increased lung elastance (i.e. stiffness) compared to rapamycin treated mice ( Figure 6B). Over the period of ventilation, rapamycin treated mice had significant lung recruitment as evidenced by an increase in inspiratory capacity that was not seen in vehicle treated mice ( Figure 6C). Although rapamycin prevented physiologic lung dysfunction following VILI, it did not affect recruitment of inflammatory cells (Figures 6D-E) or alter BAL IL-6 or KC levels. (Figure 6F-G). BAL protein levels were also not different between rapamycin and vehicle treated mice ( Figure 6G) mirroring our findings in mice with airway epithelial Tsc2 deletion. These data demonstrate that rapamycin administered at the time mechanical ventilation is initiated decreases mTORC1 activation and prevents physiologic lung injury during mechanical ventilation independent of lung inflammation or barrier permeability.

mTORC1 activation exacerbates surfactant dysfunction during injurious ventilation.
Since our data demonstrate that mTOR inhibition ( Figure 6) and Tsc2 deletion (   44,45) We hypothesized that airway epithelial cells also release ATP in response to VILI but that mTORC1 activation in these cells impairs this compensatory response and exacerbates to surfactant dysfunction. To test this hypothesis we subjected SAECs and HBEs to in vitro volutrauma and found a 2-5 fold increase in the concentration of extracellular ATP released into the culture medium ( Figure 8A-B). A similar increase in extracellular ATP was seen when SAECs were subjected to in vitro atelectrauma ( Figure 8C). Interestingly, in vitro barotrauma did not increase extracellular ATP levels in either SAECs or HBEs suggesting that different mechanical forces may have unique effects on the release of extracellular ATP (Supplemental Figure 9). mTORC1 inhibition with rapamycin or Torin 2 significantly increased the release of extracellular ATP following volutrauma ( Figures 8D-E). Given that volutrauma induced mTORC1 activation is ERK dependent, we treated HBE cells with an ERK inhibitor prior to stretch and found that ERK inhibition prior to volutrauma also increased the release of extracellular ATP ( Figure 8F). These data indicate that airway epithelial cells, release ATP into the alveolar space following volutrauma and atelectrauma. However, mTORC1 activation in these cells limits the release of extracellular ATP which can be enhanced by treatment with mTORC1 or ERK inhibitors.
Paracrine release of ATP from airway epithelial cells may be a compensatory mechanism to increase surfactant release from adjacent AT2 cells during injurious ventilation.

Discussion
Classically, favorable growth conditions (e.g. presence of nutrients and growth factors) activate mTORC1, and cellular stress (e.g. starvation) inhibits mTORC1 . (46,47) In contrast to canonical mTORC1 activation in the setting of favorable growth conditions, we found that this pathway was activated in the lung epithelial cells during injurious mechanical ventilation in preclinical animal models, in in vitro models of VILI, Activation of mTORC1 has been reported in mouse models of inflammatory lung injury including endotoxin induced lung injury (15,55,56), hyperoxia (54), and bacterial infection (57). Similar to our findings, mTORC1 activation was also increased in alveolar and airway epithelial cells following LPS-induced lung injury. (55) In genetic mouse models in which mTOR was deleted from alveolar and airway epithelial cells, cytokine levels were decreased and pulmonary edema was increased following LPS-induced lung injury. (55) In contrast with these prior reports, we did not see differences in lung inflammation or barrier permeability with mTOR inhibition in our model of injurious ventilation. (Figures 6). This may be due to inherent differences in the mechanism of lung injury, the timepoints studied, or a combination of these factors. mTORC1 activation has also been examined in inflammatory cells such as neutrophils (15) and Th17 cells (58)  Future studies will be needed to further explore the role of mTORC1 inhibition to prevent lung injury in ICU patients undergoing mechanical ventilation. code was used to analyze the droplet shape in each frame to obtain surface tension and surface area measurements. First, the Sobel method was used to detect the droplet edge in each frame. Next, the differential equations governing force balances for an arbitrary axisymmetric droplet shape (77) were integrated to determine a theoretical droplet shape based on the surface tension, tip radius of curvature and angle of inclination parameters. Finally, a non-linear least-squared regression algorithm was used to vary the surface tension and tip radius of curvature parameters until the theoretical droplet shape matched the measured droplet shape. The droplet shape was also integrated to determine the surface area in each frame. This algorithm results in surface tension vs time and surface tension vs area hysteresis loops that were further analyzed for minimum surface tension.

In vitro volutrauma model
Cells were cultured on pronectin-coated BioFlex culture plates (FlexCell International Corporation) and grown to confluence. Cells were subjected to varying degrees of biaxial cyclic stretch delivered via a Flexcell Strain Unit FX-5000 (FlexCell) for varying amounts of time prior to protein isolation or imaging. Control cells were placed in the stretching device, but not subjected to mechanical stretch.
Polyacrylamide gels or glass slides were coated with collagen type I (Sigma-Aldrich) at a concentration of 5 μg/cm 2 following surface activation of the glass slide with sulfo-SANPAH (Thermo Scientific). Cells were cultured on the gels or glass slides and placed in a microfluidic chamber (Bioptechs Inc.) with a rectangular silicone gasket to create the channel for cyclic propagation of media at a linear velocity of 30 mm/s for varying amounts of time.

In vitro barotrauma model
Oscillatory pressure was applied as described previously. (79) Briefly, cells were cultured on Transwell inserts (0.4 µm pore size, PET; Corning Inc.) and grown to confluence.
Media was removed from the apical chamber of the Transwell and a rubber stopper was inserted into the culture well in order to hermetically seal the apical chamber. Tubing was threaded through the rubber stopper and connected to a manometer and small animal ventilator (Harvard Apparatus). The ventilator was set to 0.2 Hz at a magnitude of 20 cm of H2O of oscillatory pressure.

Extracellular ATP measurements
Cells were subjected to in vitro models of VILI in the presence of an ectonucleotidase inhibitor ARL 67156 (50 µM) to prevent ATP degradation. Media was collected following in vitro models of VILI and extracellular ATP was measured using a luciferase-luciferin assay (Invitrogen).

Measurements of cellular reactive oxygen species
Cells were grown to confluence and changed to basal medium 24 hours prior to in vitro volutrauma. MitoSOX (5 µM) and calcein (4 µM) was added to each well 30 minutes prior to injury. Antimycin A (20 µM) was used as a positive control. Following stretch cells were washed 3 times with PBS, membranes were cut out of the Flexcell plates and inverted into a 60 mm dish containing a small amount of basal medium immediately prior to obtaining fluorescence images. For MitoSOX studies, nine separate 100X images were captured from each membrane. Cells were defined in ImageJ using an ROI detection algorithm on the calcein stained images and MitoSOX intensity was quantitated in ImageJ for each cell within capture. Average intensity per cell was calculated for each capture. For CellROX experiments, average fluorescence per high power field was quantitated using ImageJ.

Immunoblotting
Cells or lung tissue were lysed in RIPA buffer (VWR) supplemented with protease (Roche Diagnostics) and phosphatase inhibitors (Sigma-Aldrich). Cells were mechanically disrupted using a rubber scraper followed by a freeze/thaw cycle. Lung tissue was disrupted using a tissue homogenizer. Protein extracts were centrifuged to pellet insoluble debris, protein concentration was measured by BCA assay (Thermo Scientific), and protein concentrations were equalized prior to boiling in the recommended sample buffers supplemented with 2.5% beta-mercaptoethanol (Bio-Rad). SDS-PAGE electrophoresis was performed using 4-12% gradient bis-tris gels or 16% tricine gels (Invitrogen) and proteins were transferred to nitrocellulose membranes (Bio-Rad). Membranes were incubated in primary antibodies overnight at 4°C following blocking in 5% w/v nonfat dry milk in TBST. After incubation with HRP-conjugated secondary antibodies, proteins were visualized using SignalFire Elite ECL Reagent (Cell Signaling Technology) and imaged using a ChemiDoc XRS+ System (Bio-Rad). Quantitative assessment was performed using Image Lab software (Bio-Rad).

Immunohistochemistry and image analysis
Mouse lungs were perfused free of blood with 10 ml of sterile saline and inflated with 10% formalin to a pressure of 30 cm H2O prior to overnight fixation. Following fixation lung tissue was embedded in paraffin. For immunostaining lung sections from mice and humans were deparaffinized and rehydrated prior to boiling in antigen retrieval buffer (10 mM Na citrate/0.5% Tween 20) for 30 minutes. Tissue was permeabilized with 0.3% Triton X and quenching and blocking was performed using a cell and tissue staining kit (R&D Systems) according to the manufacturer's protocol. Tissue was incubated with anti-phospho-S6 antibody (Ser235/236) in 2% goat serum in PBS overnight at 4°C. Nonimmune serum and no primary antibody controls were performed with each experiment. H&E staining was performed using standard techniques. Blinded review was performed by a veterinary pathologist and a lung injury score was calculated. (80) Stained sections were scanned at The Ohio State Wexner Medical Center Digital Imaging Core using a NanoZoomer 2.0-HT Scan system (Hamamatsu Photonics) to generate digital whole-slide images. Quantification of the digital images was performed using the VIS software suite (Visiopharm, Hoersholm, Denmark). A single ROI was then made using a tissue detection application to measure a mean DAB stain intensity of all tissue including within each tissue section. Non-tissue areas within tissue sections such as airway and vascular lumens were excluded from the mean intensity measurement.

BAL protein and cytokine analysis
Protein concentrations were measured in BAL fluid using a standard BCA assay (Thermo Scientific) and various analytes including IL-6, KC, and VEGF-A were measured by ELISA (R&D Systems) according to the manufacturer's instructions.

Study approval
All animal studies were approved by the Institutional Animal Care and Use Committee at           Wild type mice were treated with rapamycin (rapa) or vehicle (veh) immediately prior to injurious mechanical ventilation (VILI, tidal volume 12 cc/kg, PEEP 0 cm H 2 O) for 4 hrs. A) Following mechanical ventilation, protein was isolated from lung tissue of ventilated (n=6) and spontaneously breathing (SB) control (n=4) mice and immunoblotted for phosphorylated S6 (P-S6). B) The change in lung elastance was measured during the 4 hr period of mechanical ventilation. *indicates that rapamycin is a statistically significant factor, p<0.05, by 2-way ANOVA, data presented as mean + SEM. C) The percent change in inspiratory capacity (IC) was quantitated following 4 hrs of VILI. Data not normally distributed, analyzed by Mann-Whitney test, *p<0.05. Following mechanical, a bronchoalveolar lavage (BAL) was performed and total inflammatory cells (D), neutrophils (E, PMNs, n=14 vehicle, n=18 rapa), and protein levels (F) were measured. KC (G) and IL-6 (H) levels were measured in BAL fluid by ELISA. n=16 for vehicle and n=18 for rapamycin unless otherwise noted.