Does oxygen tune cellular mechanotransduction?

in pioneering studies conducted during the latter part of the 19th century, James Lorraine Smith observed that 50% of mice breathing inspired oxygen concentrations between 70 and 80% died within one week (14). He noted that mice had a remarkable ability to recover from the toxic effects of hyperoxia and speculated that the same was true of humans. He has been proven correct since many patients with the acute respiratory distress syndrome recover with near normal lung function despite prolonged mechanical ventilation with high inspired oxygen tensions. In the interim, the effects of hyperoxia on lung cell morphology and function have been described in exquisite detail (4) and are being enriched by an expanding body of knowledge concerning reactive oxygen species (ROS) enzymology (7). The current view on the pathogenesis of pulmonary oxygen toxicity holds that a spatial imbalance between ROS-generating and ROS-scavenging signaling molecules triggers distinct cell death pathways in alveolus resident cells (16). This is often in conjunction with and perpetuated by a cytokine mediated proinflammatory response (1). Yet attempts to translate this knowledge into efficacious lung-protective interventions continue to be frustrated by the plethora of intracellular proteins with putative oxidation motives such as reactive cysteine residues (19). Moreover, the recent focus on physical stress as the cause of ventilator-associated lung injury (VALI) seems to have relegated clinicians' concerns about oxygen toxicity to one of secondary importance.

In the present issue of this journal, Roan and colleagues (13) proffered the intriguing hypothesis that hyperoxia, by virtue of stiffening alveolus resident cells, may increase their susceptibility to wounding by deforming stress. Using atomic force microscopy, they showed that oxygen exposure increases the elastic modulus (a measure of stiffness) of alveolar epithelial cells (AEC) due to actin stress fiber formation and polymerization of the cortical actin network. In an attempt to relate these observations to biophysical lung injury mechanisms, they show that cyclic stretch causes detachment of hyperoxia-treated AEC monolayers from their substratum, in line with cytopathological changes in epithelial and endothelial cells of ventilator-injured lungs (6).

The proposed interaction between oxygen exposure and biophysical lung injury mechanisms warrants a brief discussion of cellular micromechanics and the topographical distribution of stress and strain in edematous, mechanically ventilated lungs. Whereas most VALI scholars agree that alveolar overdistension and cyclic recruitment and collapse of unstable lung units are prevalent injury mechanisms, there is less certainty how stresses associated with either mechanism are transduced to generate a specific biological response. Viewed through the lens of cellular stress failure, alveolar overdistension is generally thought to be associated with matrix strains that effect lytic tensions at cell-cell and cell-matrix contacts. In contrast, “recruitment and collapse” predispose airway and alveolar lining cells to injurious interfacial stresses as liquid bridges are continually formed, displaced, and destroyed in the small airways and air spaces of edematous lungs (12). Experimental models of “overdistension injury” typically employ cell monolayers that are stretched in liquid culture (17), whereas the injury associated with recruitment and collapse is typically mimicked by advancing gas bubbles or by destroying liquid bridges in cell-coated microchannels (2, 9). Although cytopathological changes in ventilator-injured lungs are widespread and involve both epithelial and endothelial lesions consisting of the loss of basement membrane anchorage, loss of cell-cell contact, plasma membrane blebbing, and frank basement membrane fracture (6), the time scale at which injured cells remodel and repair prevents making conclusions about the nature of the stresses that cause these lesions. The problem is further compounded by lingering uncertainty about breathing-associated parenchymal strains at the scale of interest, i.e., that of the alveolar basement membranes and the cells that decorate them (8).

Several studies using reduced systems have explored whether and how manipulations of cell mechanical properties alter their susceptibility to membrane wounding by deforming stress (5, 11, 15, 18). Although results varied across cell species, interventions, and experimental models, the disruption of endothelial monolayers in response to a contractile stimulus is consistent with Roan's observation that stiffer cells lose substrate adherence when cyclically stretched (10, 13). Yet in seeming contradiction, a hypertonic stimulus, which also increases cell stiffness (20), protected AECs against interfacial stress injury, and was cytoprotective in ex vivo mechanically ventilated rat lungs (11). How can these apparently discrepant observations be reconciled? It can be intuitively understood that the shear stress between a much stiffer elastic membrane and an adherent cell monolayer increases as a function of the imposed deformation and the monolayer's elastic modulus. Roan and colleagues indeed confirmed this prediction using a finite element model (13). It is equally predictable that, the stiffer the cell, the pressure gradient of an advancing air-liquid interface acting on the surface of lining cells in a small airway will cause a smaller shape change. Considering the cell as a poroelastic structure composed of a compressible network of cytoskeletal elements that is bathed in an incompressible liquid, namely the cytosol, Oeckler et al. (11) have argued that the magnitude of the imposed shape change determines the likelihood of plasma membrane stress failure. This is because, in a poroelastic structure, a surface pressure gradient and resulting shape change causes a local displacement of the cytosol relative to the actin cytoskeleton (3), thereby generating a hydrostatic pressure gradient across the plasma membrane that, if large enough, will lead to bleb formation and membrane rupture. Thus, considering the differences in orientation and magnitude of stresses between experimental models of overdistension and models of opening and collapse injury, it is not surprising that interventions that stiffen cells could prove harmful in one setting yet cytoprotective in the other. This begs the question how well the different ex vivo models of cell deformation injury capture the regional micromechanics of the intact lung and what the relative risks of damaging the lung by so-called overinflation and underrecruitment truly are.

Answers to these questions would require 1) a better description of cellular microstrains in hyperinflated aerated alveoli compared with cell strains in partially liquid filled airways and air spaces; 2) a better understanding of resulting inter- and intracellular stress distributions; and 3) a computational model of the cell that is sensitive to the organization of and interactions between stress bearing elements, their respective lytic tensions and one that incorporates rate constants for structural remodeling. Unfortunately, available imaging techniques do not have sufficient spatial or temporal resolution to provide regional strain data at the scale of interest. For the same reason, the debate about the relative risks of injuring the lung by overdistension vs. underrecruitment suffers from a lack of valid surrogate end points on the basis of which the efficacy of injury mode specific interventions may be judged.

In conclusion, Roan and colleagues' observations point out a previously unappreciated link between hyperoxia, ROS signaling, cytoskeletal remodeling, and mechanotransduction. Taking their lead, I have chosen to explore certain implications of such a link in the context of cellular stress failure as one effector in the pathogenesis of mechanical VALI. Time will tell whether targeting the biophysical properties of lung cells will result in novel biological insights and/or efficacious lung protective treatments. In the interim we should acknowledge that, in complex biological systems, cause and effect are rarely transparent.

GRANTS

This work was funded by National Heart, Lung, and Blood Institute Grant R01 HL63178.

DISCLOSURES

No conflicts of interest, financial or otherwise are declared by the author(s).