The relative balance of GM-CSF and TGF-β1 regulates lung epithelial barrier function

Lung barrier dysfunction is a cardinal feature of the acute respiratory distress syndrome (ARDS). Alcohol abuse, which increases the risk of ARDS two-to fourfold, induces transforming growth factor (TGF)-beta 1, which increases epithelial permeability and impairs granulocyte/macrophage colony-stimulating factor (GM-CSF)-dependent barrier integrity in experimental models. We hypothesized that the relative balance of GM-CSF and TGF-beta 1 signaling regulates lung epithelial barrier function. GM-CSF and TGF-beta 1 were tested separately and simultaneously for their effects on lung epithelial cell barrier function in vitro. TGF-beta 1 alone caused an similar to 25% decrease in transepithelial resistance (TER), increased paracellular flux, and was associated with projections perpendicular to tight junctions (spikes) containing claudin-18 that colocalized with F-actin. In contrast, GM-CSF treatment induced an similar to 20% increase in TER, decreased paracellular flux, and showed decreased colocalization of spike-associated claudin-18 with F-actin. When simultaneously administered to lung epithelial cells, GM-CSF antagonized the effects of TGF-beta 1 on epithelial barrier function in cultured cells. Given this, GM-CSF and TGF-beta 1 levels were measured in bronchoalveolar lavage (BAL) fluid from patients with ventilator-associated pneumonia and correlated with markers for pulmonary edema and patient outcome. In patient BAL fluid, protein markers of lung barrier dysfunction, serum alpha 2-macro-globulin, and IgM levels were increased at lower ratios of GM-CSF/TGF-beta 1. Critically, patients who survived had significantly higher GM-CSF/TGF-beta 1 ratios than nonsurviving patients. This study provides experimental and clinical evidence that the relative balance between GM-CSF and TGF-beta 1 signaling is a key regulator of lung epithelial barrier function. The GM-CSF/TGF-beta 1 ratio in BAL fluid may provide a concentration-independent biomarker that can predict patient outcomes in ARDS.