Protein kinase R-like endoplasmic reticulum kinase is a mediator of stretch in ventilator-induced lung injury

Lung tissue samples (N = 9) were received from the Prospective Registry of Outcomes in Patients Electing Lung Transplantation (PROPEL) bio specimen repository and clinical database (Biobank). This study approved by the IRB of the University of Pennsylvania under protocol number 813685. The Biobank regularly receives lung tissue from donors and recipients of lung transplants. Many donated lungs are declined for transplantation for clinical reasons and these tissues become available for research. We amended protocol 13643/00, Cellular injury mediators of acute respiratory distress syndrome in human lung tissue, to PROPEL, to receive lung tissue samples and limited clinical information on lung tissue samples used for our analysis. Lung tissue samples were collected by the investigators of the current study immediately after the lungs have arrived to the University of Pennsylvania. Lungs were stored on +4C during transportation and they have become available to us within 24 h after harvest. After visual inspection, lung tissue samples were resected from right upper and middle lobes. Samples were used for total protein extraction. Available lung histology report and limited clinical information including age, gender, date of collection, date of death, last available arterial partial oxygen pressure (paO2 in mmHg) within 24 h, inspired percent of oxygen via mechanical ventilation and admission diagnosis was obtained for all samples (Table 1).

We used a combination of available clinical, radiological and histological parameters to diagnose ARDS. Only samples obtained from endotracheally intubated patients were eligible to allow precise measurement of fractional inspired oxygen (FiO₂). The presence of acute bilateral lung infiltrate on imaging within 1 week prior to death, a ratio of arterial oxygen partial pressure in mmHg to FiO₂ less than 300 and a pathological evidence neutrophilic infiltrates or hyaline membranes in the alveolar space with or without septal edema. Histological analysis of samples was performed on H&E stained slides by a pathologist blinded to the clinical history of the patient. The presence of diffuse alveolar damage (DAD) on pathology was not a strict criterion for ARDS as its absence does not definitely rule out the disease [19, 20]. Samples not meeting this criteria were used as non-ARDS controls (Table 1).

Pig heart-lung complexes (N = 12) were obtained from euthanized animals used as untreated controls in Preclinical Therapy Trials for Pediatric TBI study. All animals are housed in accordance with guidelines from the American Association for Laboratory Animal Care and Research. The protocol was approved by the Institutional Animal Care Use Committee at the University of Pennsylvania under protocol number 803401. Heart-lung complexes including the trachea were harvested en block from Yorkshire piglets (age 4 weeks, weight 6–10 kgs, N = 12) post mortem. The pulmonary circulation was flushed via the right ventricle using 1 l sterile normal saline. The heart was carefully removed and the pulmonary trunk was preserved. A 16 G angiocatheter was sutured in the pulmonary artery and connected to a peristaltic surgical pump via tubing. A 3 mm diameter endotracheal tube was inserted in the trachea and sutured in place. The complex was placed in a 37 C heated and humidified chamber hanging by the endotracheal tube. The lungs were perfused with 0.3 L/min MEM supplemented with 3% FBS as explained for the human perfusion-ventilation model. The lung complexes were subsequently randomized in three groups: no stretch (NS), low-stretch (LS) 6 ml/kg tidal volume (Vt) positive pressure ventilation and overstretch (OS) ventilation (Vt = 12 ml/kg) for 4 h using volume control ventilation (Puritan Bennett 7200 ventilator) with 5 cmH2O applied positive end expiratory pressure (PEEP) and FiO2 of 50%. We used 50% FiO2 to match the experimental conditions used in our in vivo rat model. At the end of the experiment the right lower lobe was perfused with 10% formalin and used for histology and the left lower lobe was used for total protein extraction and western blotting as described in lung tissue protein studies.

Tissue pieces (1 cm by 1 cm) were submerged in at least 5-fold volume of cold RIPA buffer and homogenized with Dounce homogenizer. Samples were centrifuged for 20 min at 13,000 RPM on tabletop centrifuge in 1.0 ml aliquots (Beckman Microfuge R, Backman Instruments, Palo Alto, CA). The supernatant was used to determine total protein concentration using colorimetric BioRad Protein DC assay (BioRad, Hercules, CA). Sample protein contents were compared to standard protein curves using KC Junior spectrophotometer plate reader (BioTek Instruments Winooski, VT) at 750 nm wavelength. Equal protein was loaded on precast 4–12% Bis-Tris gels (Invitrogen NuPage Novex mini gels, Life Technologies, Carlsbad, CA). Standard western blotting was performed for Binding immunoglobulin Protein (BiP), phosphorylated (p) and total (t) forms of Protein Kinase R-like Endoplasmic Reticulum Kinase (PERK) and eukaryotic Initiation Factor (eIF)-2α, Activating Transcription Factor (ATF)-4 and 6, CCAAT/Enhancer-binding Protein Homologous Protein (CHOP), Growth Arrest and DNA Damage-inducible Protein (GADD)-34 and X-box Binding Protein (XBP)-1. We used total (t)-forms of PERK and eIF2α to normalize band densities of phospho (p)-forms. For all other proteins β-actin was used as loading control (Santa Cruz Biotechnologies, Santa Cruz, CA). The following primary antibodies were used: BiP (A-10) mouse monoclonal antibody (Santa Cruz), p-PERK (Abcam, Cambridge, MA) and t-PERK (Cell Signaling Technology, Danvers, MA), p-eIF2α (Ser 51) and t-eIF2α rabbit monoclonal antibody (Cell Signaling), ATF4 rabbit polyclonal antibody (Sigma-Aldrich Company, Saint Louis, MO), ATF6 (N-terminal) rabbit polyclonal antibody (Aviva Systems Biology, San Diego, CA), CHOP (F-168) mouse polyclonal antibody (Santa Cruz), GADD34 rabbit polyclonal antibody (Proteintech Group Inc., Rosemont, IL), XBP-1 rabbit monoclonal antibody (Abcam, Cambridge, MA) and B-actin mouse polyclonal antibody (Santa Cruz). In pig tissue studies, the protein levels of N = 4 samples from no stretch, low-stretch and overstretch conditions were compared with western blotting. In human tissue studies, the protein levels of N = 4 of controls and N = 5 of ARDS samples were compared with western blotting. Blots were rinsed with 0.1% Tween 20 containing tris buffered saline (TBST) and incubated with secondary antibodies against rabbit or mouse diluted in 5% milk solution. Blots were rinsed again with TBST treated with high-sensitivity western blot substrate (Pierce ECL Plus, ThermoFisher Scientific, Waltham, MA). High sensitivity autoradiographic film was used to detect chemiluminescence. Immunoblots were scanned and bend densities were analyzed using FIJI software [21].

For immunofluorescence (IF) analysis antibodies p-PERK (#192591, Abcam) and occludin (Invitrogen # 33–1500, ThermoFisher Scientific) were used on formalin-fixed, paraffin–embedded pig lung tissue. Paraffin was cleared with xylene, and slides were rehydrated through descending concentrations of ethanol. Slides were then treated with 3% H2O2/methanol for 30 min. Slides were pretreated in a pressure cooker (Biocare Medical, LLC, and Concord, CA) in Antigen Unmasking solution (H3300, Vector Laboratories, Burlingame, CA). After cooling, slides were blocked in Sudan Black (199664-25G, Sigma-Aldrich) for 20 min at RT. Slide were then rinsed in 0.1 M Tris Buffer, then blocked with 2% fetal bovine serum for 15 min. Slides were then incubated with p-PERK at 1:200 dilution for 1 h at room temperature. Slides were again rinsed, then incubated with anti-rabbit polymer secondary prediluted (K4003, DAKO, Carpenteria, CA) for 30 min at room temp. After rinsing, slides were then incubated with the TSA biotin complex (NEL7490B001, PerkinElmer, Waltham, MA) at 1:50 for 10 mins at room temp. Slides were then rinsed and incubated with Alexa 488 Streptavidin (green) secondary (A21370, Life Technologies, Eugene, OR) 1:200 for 30 min at room temp. After rinsing, slides were then treated in preheated 5% SDS (CS-5585-28, Denville Scientific, Metuchen, NJ) for 7 min at 55C. Slides were then rinsed and blocked with 2% fetal bovine serum again before incubating with occludin at 1:100 for 1 h at room temp. After rinsing slides were incubated for 1 h in Alexa 594 (red) anti-rabbit secondary (A11012, Life Technologies, Eugene, OR). Slides were then rinsed, counterstained with DAPI, and rinsed again before cover slipping with Prolong Gold (P36930, Life Technologies, Eugene, OR). Fluorescence was detected with Zeiss LSM 880 laser scanning confocal microscope using Plan Apo 20X and 40X objective with NA = 0.8. Pictures were taken of 5 independent areas. Alexa 488 positive cells (p-PERK) were count in 100 Alexa 594 positive cells (occludin) per field and the number of dual positive cells was compared between non-ventilated and ventilated lung tissue. Images were analyzed using Fiji imaging software [21] .

To study the relationship between alveolar stretch and ER stress signaling we used ex vivo perfused and ventilated pig lungs. Heart and lung complexes were removed from healthy pigs and were perfused with 3% albumin containing tissue culture media to eliminate circulating immune cells. Subsequently animals were randomized to be exposed to no ventilation, Vt = 6 ml/kg (low stretch) and Vt = 12 ml/kg (overstretch) mechanical ventilation for 4 h with 5 cmH₂O positive end expiratory pressure (PEEP) and 50% FiO2. These stretch settings were chosen to model lung protective (low-stretch) and injurious (overstretch) mechanical ventilation. This model allows the observation of the same physiological changes that take place in the alveoli during in vivo mechanical ventilation, but it has the key advantage of selectively studying parenchymal lung cell contribution to injury signaling. To comprehensively measure ER stress activation we performed western blotting for BiP, XBP-1, ATF6 and PERK in whole lung protein extracts (Fig. 1a). XBP-1 was chosen to study the downstream effects of the IRE-1α signaling activation. We observed increased PERK phosphorylation in response to overstretch but not to low-stretch. There was no significant change in BiP and XBP-1 expression in either stretch conditions. Both the total (p90) and the cleaved (p50) form of ATF6α, when compared to controls, showed significant decrease with overstretch (Fig. 1a). We also detected the total form of the ATF6β (p110) but not its cleaved from (p60). We found significant ISR activation by overstretch which was marked by the increase in eIF2α, ATF4, CHOP and GADD34. However low stretch did not alter ISR signaling. Quantified data is shown in Fig. 1b-i. To confirm the lack of inflammatory cells in the alveolar space hematoxylin-eosin stained lung histology slides are shown in Fig. 2. In the absence of infiltrating cells in the model, we concluded from these experiments that the lung parenchyma can exert ER stress signals independent of infiltrating inflammatory cells. Our data also show that the activation of ER stress signals is dependent on the magnitude of lung stretch.

We next evaluated PERK activation in alveolar epithelial cells in response to overstretch. We exposed pig lungs to overstretch or treated them as unstretched controls. Lung tissue was incubated with green fluorescent-labeled p-PERK antibody and we used red fluorescent conjugated occludin specific antibody to identify alveolar epithelial cells (Fig. 3a). Overstretch significantly increased the number of p-PERK positive cells and we quantified the number of p-PERK positive epithelial cells (Fig. 3b). Epithelial p-PERK localization is shown with high resolution imaging in Fig. 3c. These data suggest that overstretch activates PERK in the alveolar epithelium.

To confirm that PERK activation results in VILI, groups of rats were pretreated with 3, 10 and 30 mg/kg PERK inhibitor (PI) or its vehicle via gavage. Animals were subsequently ventilated with large tidal volume (Vt = 20 ml/kg) for 4 h to maximize injury or let to breathe spontaneously. Pulmonary edema formation was assessed with FITC-labeled albumin extravasation and inflammatory cell recruitment was measured in the BAL. PERK inhibition with moderate to high doses of PI (10-30 mg/kg) significantly reduced pulmonary edema formation (Fig. 4a), total BAL cell count (Fig. 4b) and neutrophil count (Fig. 4c). Low dose PI (3 mg/kg) was sufficient to reduce alveolo-capillary permeability and neutrophil recruitment to the alveoli (Fig. 4a, c). These results show that PERK inhibition improves VILI, and PI can be dosed to selectively study cellular contribution to alveolar injury.

We next explored a mechanistic link between epithelial PERK activation and ER Ca²⁺ signaling. We exposed RAEC-I monolayers to 25% biaxial stretch for 1,3 and 6 h. Unstretched monolayers served as controls. At the end of the experiment monolayers were incubated with green fluorescent labeled p-PERK antibody. The increase in p-PERK was detected with IF at each time point with stretch (Fig. 5a, b). The effect of ER Ca²⁺ release on PERK phosphorylation was studied by the pretreatment of monolayers with Ryanodine (Fig. 5c, d) at the 6-h time point. We have previously shown that 6 h mechanical stretch significantly increase epithelial barrier dysfunction and activates ISR signaling [13]. Ryanodine, at the 20 μM dose, blocks ER Ca²⁺ release. In the absence of Ca²⁺ efflux, PERK activation was significantly reduced, though not completely inactive (Fig. 5c, d). To evaluate if Ca²⁺ signal inhibition affects ISR activation downstream of PERK, we measured eIF2α phosphorylation in the presence and absence of stretch (Fig. 5e, f). Ryanodine treatment without stretch did not change p-eIF2α when compared to vehicle-treated controls. Mechanical stretch significantly increased p-eIF2α, which was reduced by Ryanodine (Fig. 5e, f). These studies provide new evidence that ER Ca²⁺ signals contribute to PERK and ISR activation in the alveolar epithelium.

To evaluate if increased ER stress signaling can also be observed in human ARDS, we used lung tissue from donors whose lungs were declined for transplantation (Fig. 6a). ARDS diagnosis was made post mortem by the presence of acute bilateral lung infiltrate on imaging within 1 week prior to death, a ratio of arterial oxygen partial pressure in mmHg to FiO₂ less than 300 and a pathological evidence neutrophilic infiltrates or hyaline membranes in the alveolar space with or without septal edema. The clinical, radiological and histological characteristics of the patients are listed in Table 1. When compared to non-ARDS controls (samples 1–4), ARDS (samples 5–9) did not change the expression of BiP, XBP-1 and ATF6α (Fig. 6b). ATF6β was not detectable in the human lung tissue. We observed the activation of PERK, eIF2α and the increased expression of ATF4 and CHOP. We did not find significant expression change in GADD34 suggesting impaired deactivation of eIF2α in ARDS (Fig. 6a). Quantified data is shown in Fig. 5b-i. Based on these results, we concluded that PERK activation is a marker of lung injury in ARDS and PERK-mediated ISR signaling may contribute to the pathology of the disease.