Endothelial cell signaling and ventilator-induced lung injury: molecular mechanisms, genomic analyses, and therapeutic targets

Abstract

Mechanical ventilation is a life-saving intervention in critically ill patients with respiratory failure due to acute respiratory distress syndrome (ARDS). Paradoxically, mechanical ventilation also creates excessive mechanical stress that directly augments lung injury, a syndrome known as ventilator-induced lung injury (VILI). The pathobiology of VILI and ARDS shares many inflammatory features including increases in lung vascular permeability due to loss of endothelial cell barrier integrity resulting in alveolar flooding. While there have been advances in the understanding of certain elements of VILI and ARDS pathobiology, such as defining the importance of lung inflammatory leukocyte infiltration and highly induced cytokine expression, a deep understanding of the initiating and regulatory pathways involved in these inflammatory responses remains poorly understood. Prevailing evidence indicates that loss of endothelial barrier function plays a primary role in the development of VILI and ARDS. Thus this review will focus on the latest knowledge related to 1) the key role of the endothelium in the pathogenesis of VILI; 2) the transcription factors that relay the effects of excessive mechanical stress in the endothelium; 3) the mechanical stress-induced posttranslational modifications that influence key signaling pathways involved in VILI responses in the endothelium; 4) the genetic and epigenetic regulation of key target genes in the endothelium that are involved in VILI responses; and 5) the need for novel therapeutic strategies for VILI that can preserve endothelial barrier function.

ventilator-induced lung injury (VILI) is defined as acute lung injury (ALI) induced by mechanical ventilation (11). VILI is considered indistinguishable morphologically, physiologically, and radiologically from the diffuse alveolar damage of ALI (11). VILI is the most common complication arising from the use of mechanical ventilation to treat acute respiratory distress syndrome (ARDS) (11), although mechanical ventilation can also be injurious to the lungs and other organ systems in patients without ALI or ARDS (75, 105, 106). The incidence of VILI in mechanically ventilated ARDS patients has been estimated at 86% (258). Furthermore, when ARDS cases are stratified according to disease severity, the incidence of VILI was found to be 48.8% in the entire patient population, 87% in late ARDS, 46% in intermediate ARDS, and 30% in early ARDS (114). In addition to aggravating the course of disease in patients with ALI/ARDS, mechanical ventilation can also lead to the development of ALI/ARDS. Studies of medical-surgical patients receiving mechanical ventilation without ALI at the time of intubation found that between 6.2% (106) and 24% (105) developed ALI. However, it should be noted that the advent of protective ventilation strategies (described below) has significantly reduced the development of ARDS (73).

VILI: CLINICAL PROBLEM AND STATUS

Mechanical ventilation with excessive inflation pressures, tidal volumes, and flow rates can produce a wide array of local and systemic adverse effects. These pathophysiological changes occur during mechanical ventilation due to four types of injury to the lung: 1) damage caused by lung overdistension or volutrauma, 2) a direct effect of high pressure on the lung, i.e., barotrauma, 3) shear stress from repetitive opening and closing of alveoli, i.e., atelectotrauma, and 4) the generation of cytokines and inflammatory signaling i.e., biotrauma (279). In ARDS, the damage to the lung is heterogeneous with alveoli ranging from essentially normal to flooded alveoli that cannot engage in gas exchange and atelectatic or partially flooded alveoli that can be inflated and recruited to participate in gas exchange with mechanical ventilation (279). Ultimately, ALI/ARDS induced by VILI is caused by impaired gas exchange due to alveolar flooding, or the complete collapse of the alveoli, resulting in respiratory failure in critically ill patients (338). The injury to the lung parenchyma leads to the damage of the endothelial cell layer of pulmonary capillaries and the type I and type II epithelial cells of the alveoli (26). Subsequently, the loss of the barrier between the alveolar epithelial and pulmonary capillary endothelial cell (EC) allows fluid from the capillaries to leak into the interstitium and the alveoli, causing pulmonary edema and atelectasis (26). Once activated in ALI, pulmonary ECs express specific markers and proteins that can increase vessel tone, blood coagulation, permeability, leukocyte recruitment, and apoptosis (338) (Fig. 1). Since the pulmonary edema from increased endothelial permeability is the hallmark of VILI and ARDS, increasing our understanding of how the endothelial layer responds to mechanical challenege is vital to the development of effective endothelial-targeted treatments. Thus this review will focus on endothelial damage in the pathogenesis of VILI.

Fig. 1.

Development of acute lung injury (ALI). Under physiological conditions the cell-cell communications between endothelial and epithelial cells maintains a tight barrier preventing fluid buildup or inflammatory cell infiltration into the alveolus. This maintains efficient gas exchange. Under various pathological conditions (hypoxia, mechanical stress, bacterial infections, etc.) the epithelial layers are injured. This allows protein-rich fluids and inflammatory cells to accumulate in the alveolus. This leads to an increase in oxidative stress and the release of inflammatory cytokines and the eventual compromise of gas exchange. This leads to the development of ALI and potentially acute respiratory distress syndrome (ARDS).

Both animal and human studies show that mechanical ventilation can induce and exacerbate lung injury. Thus the current standard of care for patients with ALI/ARDS is the use of protective lung ventilation strategies (77, 278, 341). These ventilator strategies are based on a landmark 2000 study in which 861 patients with ALI/ARDS were assigned to a low tidal volume strategy of 6 ml/kg of predicted body weight (PBW) with a plateau pressure less than 30 cmH₂O or 12 ml/kg with a plateau pressure less than 50 cmH₂O. There was a reduction in mortality from 39.8 to 31% in the lower tidal volume group (5). These results shaped the current tidal volume guideline recommendations of 6 ml/kg of PBW for the management of patients with ALI or ARDS (74). In addition to the reduction of tidal volume, increasing the level of positive end-expiratory pressure (PEEP) is now considered an integral part of protective ventilation. PEEP is defined as air pressure in the lungs that is maintained above atmospheric pressure at the end of the expiratory cycle by a ventilator and is used to prevent the collapse of bronchioles and alveoli at the end of expiration. A meta-analysis of three clinical trials (44, 201, 205) examining the efficacy of PEEP coupled with a low tidal volume ventilation strategy found that higher levels of PEEP were associated with improved survival among ARDS but not ALI patients (43). However, it must be noted that these strategies are supportive and not therapeutic. Thus there is intense interest in understanding the molecular mechanisms by which VILI leads to the development of ARDS with the hope that this will lead to more proactive strategies to reduce the significant morbidity and mortality associated with this disease. The following sections will describe our up-to-date understanding of how VILI occurs and also highlight where our knowledge is lacking.

MECHANICAL FORCES AND THE ALVEOLAR-CAPILLARY UNIT

Mechanical, or hemodynamic, forces associated with blood flow affect the entire vasculature and can have a profound impact on physiological and pathological vascular responses. There are three types of hemodynamic forces: shear stress, compressive stress, and tensile stress (71). Flowing blood constantly exerts hemodynamic stress on the endothelium lining the blood vessels once the heart begins to produce a fetal circulation (71). Shear stress is a frictional force arising from blood flow and acts tangentially to the vessel luminal surface (71). Blood pressure acts perpendicularly to the cell surface, creating a compressive stress within the vessel (71). Tensile stress arises from the pulse pressure being transmitted circumferentially to cells via their contacts with the extracellular matrix causing strain (71). Cell deformation in response to applied stress is described as strain and depends on the mechanical and structural properties of a cell (71). Shear strain is characterized by angular distortion in response to shear stress (111). Compressive (shortening) and tensile (stretch) strain represent a change in length per unit (original) length (111).

The endothelium is located between flowing blood and the vascular wall (71). Cells lining the arterial circulation are exposed to fluid forces of a much greater magnitude than those experienced by other tissues (71). Therefore, EC are capable of altering their structure and mechanical properties, resulting in the generation of internal cellular stresses that equalize the external forces (71). However, pulmonary vascular ECs are exposed to different hemodynamics and additional mechanical forces resulting from respiratory cycles (35). Compared with the pulmonary circulation, the vascular endothelium lining the systemic circulation experiences higher hydrostatic pressure (mean arterial pressure, 70–110 mmHg), shear rates of 15–70 dyn/cm² (54), and a 5–10% stretch associated with vessel distension caused by pulse pressure (35). In comparison, the pulmonary circulation has lower hydrostatic pressure (mean pulmonary artery pressure, 10–20 mmHg) and shear stress (pulmonary artery, 20 dyn/cm² (305). In addition to cyclic stretch (CS) derived from the intraluminal blood pressure, lung capillary strain is also associated with respiratory cycles (35). One study shows that the surface area of the alveolar epithelium increases by 35% as lung volume is increased from 40 to 100% of total lung capacity (314), and the surrounding capillary endothelium is thought to stretch to a similar extent (35).

Mechanical Stress in the Ventilated Lung: Effect on the Lung Vasculature

Mechanical loads on pulmonary vasculature are well tolerated. However, lung overdistension caused by mechanical ventilation at high tidal volumes transmits pathological mechanical stress to alveolar epithelium and pulmonary vasculature resulting in wall stress. The alveolar wall can be regarded as a string of pulmonary capillaries, with the result that increased tension in the alveolar wall caused by high states of lung inflation is transmitted to the capillary wall (344). Thus wall stress in the pulmonary capillaries can be increased by raising the transmural pressure (Ptm) of the capillaries and by increasing the longitudinal tension in the alveolar wall (344). The three principal forces acting on the capillary wall are circumferential tension due to the Ptm of the capillaries; surface tension of the alveolar lining layer due to the bulging of capillaries into the alveolar space, and longitudinal tension associated with lung inflation (Fig. 2A). In seminal studies, the West group showed that the exposure of rat pulmonary capillaries to a Ptm of 40 mmHg induces ultrastructural changes in the alveolar wall involving both the alveolar epithelium and capillary endothelium (317, 345). Similarly, increasing lung inflation while maintaining a constant capillary Ptm also disrupts the capillary endothelial and alveolar epithelial layers (99), the result being increased permeability of pulmonary capillaries, loss of vascular barrier integrity, and increased vascular leak (58, 89, 161, 179, 371) (Fig. 2A). The term “stress failure” was coined to explain these effects (345). It should be noted that the capillary leakage induced by elevated vascular Ptm and/or increased lung distension increase longitudinal tension within the alveolar-capillary unit and this is thought to enhance, rather than produce, stress failure (99). During high tidal volume ventilation, hemodynamic forces within the capillaries should actually decrease secondary to a decrease in both pulmonary flow and Ptm, but the substantial increase in longitudinal tension induces direct endothelial injury that can be observed at an ultrastructural level (76). More recent in vivo data show that even a short exposure to high tidal volume ventilation (<15 min) activates inflammatory signaling (357). Available data suggest that pulmonary EC are the most sensitive cell type to mechanical stretch (357). Furthermore, the pattern of inflammatory pathway activation seems to be specific to VILI and differs from lipopolysaccharide (LPS) (357). This suggests that VILI rapidy and directly induces pulmonary vascular inflammation as pathological mechanical forces are sensed by the lung (357). These data also shed light on the observed phenomenon by which VILI causes injury to nonpulmonary organs through lung-derived signaling agents (156). The recruitment of Gr-1^high monocytes to the lung has been shown to be involved in the development of VILI (352) by enhancing systemic cytokine production (333). This has led to the idea that, similar to the “decompartmentalization” hypothesis, ECs respond to pathological mechanical stretch via passage of signals to inflammatory cells within the capillaries allowing amplification of the inflammatory signal into the systemic circulation (357).

Fig. 2.

Pathological lung overdistension caused by mechanical ventilation at high tidal volumes transmits pathological mechanical stress to alveolar epithelium and the pulmonary vasculature altering pulmonary gene expression profiles. A: cartoon depicts the 3 forces acting on the alveolar capillary unit. Circumferential tension caused by the capillary transmural pressure, longitudinal tension in the alveolar wall due to lung inflation, and surface tension of the alveolar lining layer, which is an inward-acting force to support the capillary as it bulges into the alveolar space when capillary transmural pressures are high. P_air, pressure of air inside the lung alveoli; P_blood, capillary blood pressure. The cartoon is adapted from West et al. (345) and Birukov et al. (34; by permission) and depicts the alveolar wall as a string of pulmonary capillaries. Increased alveolar wall tension by mechanical ventilation-mediated lung distention is transmitted to the capillary wall leading to “stress failure” and the loss of vascular barrier integrity and increased vascular leak. In vitro, high tidal mechanical ventilation-mediated cyclic stretch would be reflected by exposure to 18% cyclic stretch (34, 315, 355). B: regional heterogeneity in gene expression in VILI (275). NAMPT (circled) is one of the genes upregulated by high tidal volume ventilation in the base portion of the injured lung. Ctrl, control group; Vent, group with high tidal volume ventilation.

The increased tension in the alveolar wall produced by mechanical ventilation also appears to be location specific and significant regional heterogeneity exists, as reflected by gene expression profiling of different lung segments (Fig. 2B) (275). Measurements of mechanical stress in the mechanically ventilated lung are technically challenging due to the complexity of local distension patterns in the lung parenchyma; however, calculations based on animal models suggest that if the lung volume increases from 40 to 100% of the total lung capacity, the alveolar epithelial cell basal surface area increases by 34–35% (314, 315, 355). These studies translate in vitro to indicate that a 25% surface area increase, as produced by exposure to 5% CS, reflects the physiological levels of mechanical stress in alveolar epithelium produced by spontaneous breathing (34, 315, 355). In contrast, high tidal volume mechanical ventilation-mediated mechanical stress results in a 40–50% surface area increase as would be reflected in vitro by 18% CS (34, 315). Consistent with these biophysical indexes, we, and others, have shown that exposure to 5% CS (physiological stress) has vascular-protective effects with enhanced EC barrier integrity via Rac GTPase activation and gene expression profiles that are anti-inflammatory (36, 39, 107, 355). In contrast, EC exposure to 18% CS, reflecting the pathological level of mechanical stress produced by mechanical ventilation, significantly increases EC permeability via RhoA GTPase activation and EC paracellular gap formation (Fig. 1) in association with inflammatory cytokine gene expression (2, 37, 125, 379). Preconditioning of lung EC to either 5 or 18% CS results in differential responses to edemagenic agents with 5% CS attenuating and 18% CS exaggerating edemagenic agonist-mediated declines in transendothelial cell electrical resistance, even after cessation of CS stimulation (37–39).

Influence of Mechanical Stress Regulation of Vascular Barrier Integrity

To study the effects of circumferential CS on EC function, ECs are grown on a flexible membrane that is precisely deformed by a microprocessor-controlled vacuum, providing equibiaxial tension (22). This allows the cells to be subjected to defined levels of cyclic strain in a variety of wave patterns. In this way, ECs can be exposed to various levels of cyclic strain that are either physiological or pathologically elevated, and the effect on specific EC markers can be determined. Many of the mechanoreceptors and downstream signaling pathways that allow ECs to respond to shear stress are also stimulated by tensile stress (35). Physiological levels of CS are associated with cell quiescence, induction of intercellular adhesion genes, and increased EC barrier integrity and recovery after challenge with edematous agents, such as thrombin (35). However, pathologically elevated amplitudes of CS occur under conditions such as hypertension or alveolar overdistention and result in barrier dysfunction, increased apoptosis, increased synthesis of extracellular matrix proteins, increased inflammatory cell adhesion, activation, and the release of inflammatory cytokines (35). Such conditions occur in the setting of complications arising from excessive mechanical stress generated by mechanical ventilation, i.e., VILI (35). Cyclic strain regulates many processes by altering the expression and/or activation of the following proteins in ECs. For example, cyclic strain regulates vessel tone by modulating endothelial nitric oxide synthase (eNOS) (16), cyclooxygenase-2 (COX-II) (383), and endothelin 1 (ET-1) (190); proliferation by modulating platelet-derived growth factor (PDGF) (291), tyrosine kinase receptors: Flk-1, Tie-2, and Tie-1 (387), and vascular endothelial growth factor (VEGF) (298); migration by modulating urokinase plasminogen activator (uPA) (Flores C, Hysi P, Ma SF, Christie J, Theisen-Toupal JC, Dudek SM, Garcia JG, unpublished observations) (321), tissue plasminogen activator (tPA) (146, 321), plasminogen activator inhibitor-1 (PAI-1) (321), monocyte chemoattractant protein-1 (MCP-1) (359), matrix metalloprotease (MMP)-2, and MMP-9 (66); angiogenesis by modulating VEGF (298), MMP-2, MMP-9 (66), uPA (321), and RGD-dependent integrins (331); and cell-cell adhesion by modulating ICAM-1 (24), VCAM-1 (24), zonula occludens (ZO)-1 (62), occludin (62), and VE-cadherin (28). However, little is known about the transcriptional mechanisms that regulate the expression of junctional proteins and ultimately barrier function in pulmonary ECs exposed to CS.

Animal models of VILI consist of mechanical ventilation with high tidal volume, low PEEP (0–3 cmH₂O), or a combination of ventilation with high tidal volume and low PEEP (354). Animal models that induce VILI by purely overdistending the lungs utilize tidal volumes ranging from 15 to 45 ml/kg (353) and demonstrate differences in lung physiology compared with noninjurious tidal volumes of 6–10 ml/kg (304). However, one study found that only exposure to very high tidal volumes (30–40 ml/kg) induces nonrecruitable changes in respiratory mechanics, edema, and inflammation in mice (353). In contrast, moderately high tidal volume (10–20 ml/kg) strategies induced similar pathologies that were completely reversible with recruitment, suggesting these changes were not due to overstretching the lungs but were the result of atelectasis (353). In addition, it has been suggested that the differences in lung physiology between animals ventilated with moderately high tidal volumes (15–20 ml/kg) and those ventilated with lower tidal volumes (6–10 ml/kg) may be the result of differences in respiratory rate or minute volume, which are altered to manage arterial CO₂ (353). Higher respiratory rates during VILI can exacerbate pulmonary edema and cytokine production (255); therefore, altering respiratory frequency introduces an other variable. The intratracheal delivery of bacteria or bacterial products, such as LPS, followed by exposure to mechanical ventilation with moderately high tidal volumes (20 ml/kg) have been used to create a “two-hit” model of VILI that is presumed to more accurately reflect clinically ARDS/VILI (30, 253). The preinjured lungs are more sensitive to lung injury induced by stretch although it is difficult to distinguish the contribution of stretch vs. other insults (173). The majority of VILI studies have been conducted in adult rodents, although an increasing number of studies are also being performed in newborn animals (63). In adult mice (134, 351) and rats (63), high tidal volume ventilation results in lung injury and decreased lung compliance, and similar results are found in newborn rats (63). Adult rats were found to be more susceptible to VILI than newborn rats (63). Supplemental oxygen is frequently used to improve survival during prolonged periods of mechanical ventilation (84, 351). The proinflammatory properties of inhaling a high fraction of inspired oxygen are well known and may contribute to additional lung injury (3, 380) and confound interpretation of the results. The combination of high tidal volume ventilation and/or low PEEP and hyperoxia resulted in lung injury, decreased lung compliance, and increased inflammatory response in adult (84, 220, 245) and infant rodents (63).

GENE REGULATION IN ARDS AND VILI

VILI-associated mechanical stress exacerbates lung inflammation and triggers alterations in lung gene expression. Multiple investigators have attempted to link the pathobiology of VILI with genomics (140, 165, 188) with the goal of identifying biomarkers for diagnosis/prognosis, as well as discovery of therapeutic targets to maximize the utility of low tidal volume ventilation in the intensive care unit (ICU). As it is difficult to separate the role of VILI in patients developing ALI/ARDS, the majority of “omic” studies of VILI have been carried out in human cells (lung microvascular ECs or other lung cells) (96, 357), or in tissues from preclinical models of VILI (without ALI) (188). However, despite increased utilization of systems biology-like understanding of ALI (genes to populations) (283), there remains a lack of understanding of the heterogeneity of response in patients with ALI and VILI. On a very positive note, candidate gene genomic approaches have led to successes in identifying VILI-associated genes and signaling pathways (e.g., IL-6) (188), and novel biomarkers (NAMPT/PBEF) (382) and genes involved in the coagulation pathways (e.g., APC) (89).

Identification of Novel Genes Dysregulated During VILI

To identify the most reliable pathways, and differentially regulated genes, associated with VILI, studies have successfully used an in silico ortholog gene approach for “most conserved VILI genes” to identify VILI candidate genes (120). The thematic underpinning of this approach is the hypothesis that patients with VILI and preclinical animal models of VILI will exhibit commonality in expression of evolutionarily conserved genes across species. Using profiling results from preclinical models of VILI (rat, mouse, dog) and mechanically stressed human lung endothelium, 3,077 genes have been identified whose expression is altered across all four species in response to high tidal volume mechanical stress. Filtering these results for unidirectional change in gene expression (with >1.3-fold change) refined the list to 69 genes, reflecting specific ALI-associated gene categories (modules/ontologies): coagulation, inflammation, chemotaxis/cell motility, and immune response (120). This approach has identified multiple genes already recognized as likely to play a role in ALI (IL-6, AQP-1, PAI-1). The identification of conserved gene pathways suggests that mechanical stress-mediated lung injury has many similarities with other types of ALI (such as endotoxin-mediated ALI), and diagnostic/therapeutic approaches should exhibit additional similarities. In addition, these profiling studies have also identified a number of novel genes not previously known to be mechanistically involved in ALI (120) (Table 1).

Table 1.

Genes showing significant changes in expression throughout all biological systems tested

Gene Symbol	Fold Change	P Value Ventilation vs. Control
CXCR4	1.62	0.007
GJA-1	1.33	0.026
IL1R2	1.88	0.026
GADD45A	1.71	0.004
BTG-1	1.38	0.007
TFF-2	−1.32	0.045

From Ref. 123.

CXCR4 and IL1R2 are known ALI markers. IL1R2 and CXCR4 are EC receptors that mediate acute inflammatory responses to facilitate leukocyte infiltration and cytokine release. BTG1 is a member of an antiproliferative gene family that regulates cell growth and differentiation. Expression of this gene is highest in the G₀/G₁ phases and reduced through G₁. BTG1 also interacts with several nuclear receptors, functions as a coactivator of cell differentiation (177), and is a regulator of apoptosis. However, its role in the development of VILI has not been well characterized. GJA1 encodes connexin 43, a member of the connexin gene family and a component of gap junctions that provide intercellular channels for the diffusion of low molecular weight molecules from cell to cell. Recent studies have suggested that GJA1 is upregulated in ALI and may be involved in the EC barrier disruption (164, 225). GADD45a is a multifunctional DNA repair gene with mice lacking GADD45α gene (GADD45α^−/−) exhibiting significantly increased VILI susceptibility (208). GADD45α depletion results in increased ubiquitination of Akt and results in both increased proteasomal degradation of Akt and decreased Akt phosphorylation in response to mechanical stress (208, 212). TFF2 encodes trefoil factor 2, a mucin binding protein with high abundance in mucus (191). Recently it has been shown that TFF2 accelerates gastric epithelial restitution through the activation of calcium-dependent signaling and ERK1/2 phosphorylation (79). The identification of the involvement of the CXCR4 receptor is considered to be the first validation of a TFF receptor (79). Validation of its role in ARDS was shown by TFF2 knockout mice having exacerbated VILI (149). Although the mechanism of action is still unresolved, TFF2 is considered a protective factor in VILI. Of these VILI biomarkers, CXCR4, IL1R2, and GJA1 are considered as deleterious contributors, and GADD45a and TFF2 are protective factors to VILI.

Epigenetics and VILI

Epigenetics refers to heritable changes in gene expression not involving changes to the underlying DNA sequence, i.e., a heritable change in phenotype without altering the genotype. Epigenetic changes modulate how cells read their genes. Epigenetics allows the regulation of protein expression by switching genes on or off and can be inherited stably during cell differentiation and development. Epigenetics is involved in many normal cellular processes (243). At least three systems underlie epigenetic control in cells: DNA methylation, histone modification, and micro (mi)-RNA-associated silencing. All of those play a crucial role in the pathogenesis of inflammatory lung diseases. There has been a significant focus recently on DNA methylation and miRNA-associated epigenetic regulation in VILI with a paucity of studies on histone acetylation associated with the development of VILI (160, 388).

Epigenetic regulation via DNA methylation mainly occurs in 5′ flanking areas designated CpG islands, in which high concentrations of CpG dinucleotides are found. DNA methylation is a common epigenetic strategy that cells adopt to turn off gene expression (142, 382). Recent data have provided strong evidence for an important role of DNA methylation in VILI. DNA methylation is altered in the UCHL1 promoter resulting in reduced Akt1 activation via increase Akt1 ubiquitination at K48 (212). In addition, nicotinamide ribosyltranserase (NAMPT), also known as pre-B cell colony-enhancing factor (PBEF) (PBEF/NAMPT) expression is induced by mechanical stretch, a mimic of VILI in cultured cells, via a DNA demethylation event that results in increased STAT5 binding activity (292). These results suggest that DNA methylation changes is a promising epigenetic biomarker for VILI.

MicroRNAs (miRNAs), small noncoding RNA sequences, modulate gene expression by binding to 3′-untranslated regions (3′-UTR) of mRNA’s interfering with mRNA stability and/or translation. Recent studies have demonstrated that miRNAs are a crucial regulator of gene expression in lung and systemic inflammation, especially VILI (6, 326, 336, 388). In a murine model of VILI, miRNA expression is induced rapidly by high tidal volume (326). Furthermore, mechanical stretch mediated changes in miR-374a and miR-568 target the 3′-UTR sequence of the PBEF/NAMPT mRNA leading to enhanced mRNA levels (6). Thus, although the analysis of the role of miRNAs in ARDS/VILI remains in its infancy, the data obtained so far suggest that specifically targeting individual miRNA is a potentially promising therapeutic approach for ARDS/VILI.

Transcription Factor Regulation by Mechanical Stress

As discussed above it is becoming clear that both genetic and epigenetic regulation of target genes in response to excessive mechanical stress is involved in the development of VILI. However, the genetic/epigenetic regulation of gene expression during ARDS/VILI is also dependent on the activation or inhibition of transcription factors (TFs). An inventory of the current available literature identifies nine transcription factors that appear to be involved in gene regulation in response to mechanical stress: STAT5 (18, 370), STAT3 (356), SOX18 (122), Sp1 (272), Nrf2 (236, 306), KLF2 (Sun XG, Sammani S, Mathew B, Jacobson JR, Garcia JG, unpublished observations), NFAT (132), IRF7 (213), EGFR (236). Several of these TFs are well established as being activated by mechanical stimuli, such as STAT family members, Sp1, and Nrf2; and other TFs have only been recently identified such as SOX18. Unfortunately, the knowledge of the relationship between these TFs and the development of VILI is very poorly understood and more studies are required to understand their role. Key questions that remain unanswered include 1) How does mechanical stress activate/inhibit these TFs? 2) How do different types of mechanical stress regulate the activity of these TFs? 3) What are the key downstream target genes essential for mechanical stress-associated signaling/defense regulated by these TFs?

Health Disparities and Genetic Contributions to ARDS and VILI

Chronic lung diseases, such as asthma or chronic obstructive lung disease, even lung cancer, have been clearly marked with uneven population susceptibility, an important category of health disparity, in multiple dimensions including race, ethnicity, sex, age, socioeconomic status, and geographic location (113). Health disparities have also been observed in ARDS (113). The burden of ARDS morbidity and mortality affects African Americans (AA) disproportionately and AA ARDS patients have the highest mortality rate and AA men have an even higher mortality rate than AA women (217). Hispanics also have a higher susceptibility for ARDS and a higher mortality rate (260). Although there is a lack of extant family pedigrees for ARDS, a total of 34 genes have been reported to influence ARDS susceptibility (206), the majority of which are considered as candidate genes based on the current pathophysiological understanding of ARDS. These genes belong to pathways linked to inflammation, coagulation, EC function, reactive oxygen radical generation, and apoptosis. These candidate genes were identified by novel genetic approaches, including genome-wide association studies and human peripheral blood gene expression data. The genetic risk for ARDS seems to vary both by ancestry and by the subtype of ARDS (206), suggesting that both factors may be valid considerations in clinical trial design. For example, the Fas pathway regulates apoptosis, inflammation, and epithelial/endothelial cell injury. A few common genetic variants in Fas are associated with susceptibility to developing clinical lung injury (116, 198). NAMPT/PBEF is a known sepsis and ARDS prognostic biomarker (18, 382), associated with poor clinical outcomes in ventilated patients with sepsis and sepsis-induced ARDS. Several NAMPT genetic variants are significantly associated with ARDS risk and increased mortality (292). In addition, a promoter polymorphism (rs2814778) in the Duffy antigen/receptor for chemokines (DARC) gene is associated with a 17% increase in 60-day mortality in AA ARDS patients (166). We previously reported the GADD45a promoter single nucleotide polymorphism (SNP) rs581000 to be associated with ALI (213). Specifically, the C allele reduced the risk of both sepsis and ALI in AAs from a Chicago cohort (P = 0.009, adjusted P = 0.05 under a dominant model) and in a Spanish cohort (P = 4.20E−06, adjusted P value = 0.00003 under a dominant model, FDR <1%). However, the association did not reach significance in European Americans (Chicago cohort). Together, the prevailing data support the hypothesis that AA ARDS patients have a higher risk of death (166, 198). However, caution is also warranted as putative causal genes require more validation in independent study populations, to avoid false positive outcomes by “overfitting.” Indeed, genome-wide association and sequencing studies for susceptibility and outcomes in multiple populations are already in progress. In addition, not all studies have found this link between ethnicity and ARDS susceptibility (45). However, overall epidemiological observations suggest that, although rate of incidence might not be affected by ethnic genetic variations, mortality or disease severity are influenced by the ethnic background. This raises a novel aspect to understand ARDS disparity by identifying ethnic genetic variations and evaluating their severity on ARDS outcome, independent of their role in susceptibility to ARDS itself. Thus it is imperative that a comprehensive genetic dissection of ARDS/VILI be undertaken to allow a complete understanding of how racial disparity affects ARDS/VILI outcomes that may allow for the development of more personalized therapies for ARDS/VILI.

Potential Challenges to Current Investigative Approaches

The “big data” era has generated genomic- and genetic-intensive data sets leading to increased knowledge of ARDS and VILI. These data provide novel approaches for personalized medicine or precision medical care for ARDS and VILI, while potential challenges still exist at four levels: first, the validity of data. As ARDS and VILI lung biopsy samples are relatively more difficult to acquire than other chronic lung disease, which might require lung transplantation, most of the lung gene expression data are from cell models (e.g., mechanically stressed endothelial cells), or animal models (e.g., VILI mice lungs). The genomic information collected from these models (especially gene expression) need to be extensively validated in human disease platforms. Second, there are inherent limitations with the specific assay platforms used. For example, the microarray analysis has a more limited signal window than RNA sequencing, while genome-wide association study chips only examines a small portion of known SNPs. Thus these assays will only yield data relating to the transcription level (relative mRNA quantity) of a gene or the DNA variants (SNPs) of a gene respectivey. To extend these data sets comprehensive and integrative analyses using genetics (DNA), genomics (mRNA), proteomics (protein and posttranslational modification), and metabolomics (protein function) are desperately required. Third, the potential for misinterpretation of data exists, due to the complexity and unique characterization of these omics studies. More validation studies are required to differentiate biomarkers from therapeutic targets, epiphenomena from pertinent regulatory mechanisms, and susceptibility genetic variants from causative genetic variants. Fourth, the “data blizzard” from these studies generate multiple redundant data sets. Thus novel analytical methods are now required perhaps even more than are new genetic/genomic data sets.

CELLULAR MECHANISMS OF MECHANICAL STRESS-INDUCED INJURY

Reactive Oxygen Species and Endothelial Injury

The generation of reactive oxygen species (ROS) is a result of living in an oxygen-rich environment and organisms have evolved elaborate systems to detoxify these species. However, these protective pathways can be overwhelmed, leading to “oxidative stress.” In the lungs multiple cell types, including endothelial cells, neutrophils, eosinophils, alveolar macrophages, and alveolar epithelial cells, are major ROS generators. Within the cell, several enzymes are involved in generating ROS. These include NADPH oxidase (NOX), uncoupled NOS, dysfunctional mitochondria, and xanthine oxidase. There is evidence that all these systems may be involved in the oxidative stress associated with ARDS/VILI. Antioxidants reduce the severity of ARDS/VILI in response to mechanical ventilation (49, 69, 252) as well as other ALI/ARDS models such as LPS (7, 56, 143, 297), influenza A (369), hyperoxia (145), toxic gas (273), ischemia-reperfusion (I/R) (386), sepsis (144, 184), acid aspiration (372), and burn and smoke inhalation (365). The link between increased oxidative stress and lung injury to such a diverse group of pathological stimuli as well as the success of antioxidant therapy in preclinical models makes targeting ROS generation an attractive target in ARDS. However, complicating the development of more targeted therapies is that fact that the source of the ROS is far from clear. For example, ROS derived from xanthine oxidase are stimulated by mechanical stretch (1, 378) as well as other lung injury models including I/R (268, 342) and LPS (133). However, NOX-derived ROS have also been shown to be increased during the lung injury associated with mechanical ventilation (124) as well as I/R in the perfused mouse lung (178), during hyperoxia (235), infuenza virus infection (309), and acid aspiration (70, 367). In addition, NOX-derived ROS can be derived from different isoforms with increased ROS during LPS exposure dependent on NOX1 in macrophages (193) and NOX2 in the lung (266, 348, 358), demonstrating the complexity of designing targeted therapeutic interventions against ARDS/VILI. Mitochondrial-derived ROS (9, 10) have also been implicated in the injury induced by mechanical ventilation (147). LPS (78), I/R (119, 274), and sepsis (52) also stimulate mitochondrial ROS generation in the injured lung (8).

Uncoupled eNOS also appears to be an important source of ROS in lung injury induced by high tidal volume mechanical ventilation (325). NOS uncoupling is also associated with lung injury induced by both gram-positive (55) and gram-negative bacteria (123) exposure, as well as smoke inhalation and burn injury (219). However, the mechanism by which eNOS becomes uncoupled can be different. Thus in response to mechanical ventilation the uncoupling of NOS is associated with the oxidation of tetrahydrobiopterin (BH₄) (325). Increases in the BH₄ oxidation product dihydrobiopterin have been shown to cause NOS uncoupling (121). The downstream effector of NOS uncoupling is likely peroxynitrite, formed from the interaction of NO with superoxide. Peroxynitrite levels in the lung have been shown to increase in response to high tidal volume mechanical ventilation (195, 196). Peroxynitrite leads to tyrosine nitration, a covalent modification that adds a nitro group (-NO₂) to one ortho carbon of tyrosine’s phenolic ring to form 3-nitrotyrosine. Protein tyrosine nitration introduces a net negative charge to the nitrated tyrosine at physiological pH, thus altering structural properties and catalytic activity of the protein (126). Little is known regarding the role of protein nitration in VILI. However, in other forms of ALI, proteomic analysis been able to identify the nitration of proteins that may be important for the development of ALI/ARDS; these include sphingosine 1-phosphate lysase 1 (SIP lysase 1) (381), Rho-GTPase-activating protein 5 (RHOGAP5) (381), and RhoA itself (247). However, only in the case of RhoA has the nitration been validated as being important for the development of ALI/ARDS (123, 381).

Mechanical Signal Transduction and EC Cytoskeleton Reorganization

Mechanical signal transduction is mediated via the extracellular matrix and is transferred to intracellular signaling molecules via a complex network that receives and responds to the mechanical signals. These signals include ROS-related or non-ROS related mechanical signals. The non-ROS-related mechanical stress signals through the cytoskeleton network, including transmembrane components for cell-cell attachments [adherens junction (AJ) and tight junction (TJ)] as well as cell-matrix junctions (focal adhesions); intracellular signaling protein complexes such as tyrosine (focal adhesion kinase, pp60^Src) kinases (110), serine (Erk, JNK, and p38 MAP kinases) protein kinases, inositol lipid kinases (phospholipase C), and some growth factor receptors (VEGF and PDGF receptors) and the cytoskeleton that links the membrane junction to intracellular protein complexes. The EC cytoskeleton plays a key role in mechanical signal transduction, as CS on EC cytoskeleton triggers multiple cellular signal cascades with ion channels, and G proteins (Gi and Gq), and protein kinases (protein kinase C, MAP kinases, nonreceptor protein tyrosine kinases) (reviewed in Ref. 35). Protein kinase-mediated phosphorylation of specific cytoskeletal proteins (such as nmMLCK) and signaling molecule induce cytoskeletal rearrangement. Low-magnitude CS (5%, mimicking physiological breathing) selectively activates the small GTPase Rac1 that promotes peripheral translocation of actin polymerization proteins (including nmMLCK and cortactin) and enhancement of cortical actin cytoskeleton. In contrast, high-magnitude CS (18%, mimicking pathological high tidal volume mechanical ventilation) stimulates the small GTPase RhoA and potentiates stress fiber formation and barrier dysfunction induced by edemagenic agonists (36).

nmMLCK.

MYLK, the gene encoding the critical cytoskeletal effector myosin light chain kinase (MLCK), plays a prominent role in the inflammatory lung injuries observed in ARDS and VILI (60, 108, 211). There are two major types of MLCK isoforms encoded by MYLK, via two distinct promoters: the nonmuscle MLCK isoform (nmMLCK, 210 kDa) and smooth-muscle MLCK isoform (smMLCK, 135 kDa). smMLCK is identical to the COOH-terminal half of nmMLCK, containing the key enzymatic activity-related domains (catalytic domain, CaM binding domain). nmMLCK is highly expressed in endothelial and epithelial cells and regulates the EC cytosketon (actin filaments) in a highly dynamic and spatially dependent fashion. nmMLCK, unlike many cytoskeleton molecules, has a dual role in EC barrier regulation with enzymatic inhibition attenuating capillary leak in diverse models of lung injury, whereas, paradoxically, nmMLCK is a key participant in barrier restoration via spatially specific actomyosin contractile events in the peripheral actin ring and in lamellipodia (46, 80, 81, 83, 112). This complex dual role is due to the differential activation of kinases in the process of VILI and ARDS. During the period when inflammation is developing, pp60^Src is activated (by edemagenic mediators such as thrombin or LPS) to phosphorylate nmMLCK at Y464/Y471, and thereby stimulates stress fiber formation in the central portion of the cell leading to cellular contraction and gap formation. Conversely, during the resolution of inflammation, nmMLCK is phosphorylated (on the same sites by endogenous barrier enhancers such as S1P or hepatocyte growth factor, HGF) and binds cAbl kinase, and the complex is translocated to cell periphery where it facilitates the formation of cortical actin (81, 112, 175, 240, 267, 323, 327) thereby strengthening the barrier and promoting vascular integrity. Studies in genetically engineered nmMLCK^−/− mice confirmed nmMLCK as an ARDS and VILI candidate target in vivo (211, 335) with nmMLCK gene/protein expression and enzymatic activities each contributing to risk and severity in preclinical ARDS and VILI models and humans (108, 207, 211, 212). These studies also suggest that nmMLCK could be a therapeutic target for VILI, although this may be limited due to the dual role in both driving EC barrier disruption as well as barrier restoration.

Besides the studies in cell biology and preclinical VILI models, the genetic contribution of MYLK to ARDS is also well studied. We have shown that MYLK SNPs with high minor allelic frequencies in African descent individuals, a vulnerable population at risk for ARDS and reduced ARDS survival (45), confer increased ARDS susceptibility and mortality (60) and increased risk of developing severe asthma (4, 92, 109). These findings have confirmed MYLK as a causative contributing factor to ARDS and VILI.

Sphingolipid receptors.

Sphingolipids are critical structural components of cell membranes and bioactive lipids that regulate diverse signaling pathways and contribute to a variety of pathologies that underlie cancer, inflammation, injury, edema, and infections (222). Of the several hundred sphingoid bases described to date, at least six, namely, sphingomyelin (SM), sphingosine (Sph), sphingosine 1-phosphate (S1P), ceramide, ceramide 1-phosphate (Cer1P), and sphingosylphosphorylcholine (lyso-SM), are considered key signaling and regulatory bioactive lipids (222). Of these six lipids, S1P is the best studied in EC cytoskeleton function regulation. S1P elicits its cellular effects through five known G protein-coupled S1P receptors (S1PR1 to S1PR5), formerly known as endothelial differentiation gene (EDG) receptors, which are expressed in various cell types with S1PR1 and S1PR3 the major subtypes in lung EC. Although S1PR1 and S1PR3 share significant homology they exert distinct biological functions and spatial distributions, coupling to different G proteins (S1PR1 couples to Gi, while S1PR3 couples to Gq). Differences in receptor-complex internalization and recycling may also provide specificity for each of the S1P receptors (200). The deletion of S1PR1 in mice is embryonically lethal due to vascular leakage (185), whereas S1PR3 deletion appears to exert no discernible effect. We were the first to show that S1PR1 activation by S1P enhances EC barrier (118) and ameliorates endotoxin- or high tidal volume ventilation-mediated lung injury in animal models (264, 287). Consistent with the barrier-protective role of S1PR1, the pretreatment of wild-type mice with an S1PR1 antagonist SB-649146, or the use of S1PR1^+/− mice, reduced S1PR1 agonist (S1P or SEW2871)-induced barrier protection after LPS challenge (264). These evidences suggest S1PR1 plays a key positive role in endothelial barrier regulation, while S1PR3 is a negative contributor.

Although S1PR3 was identified as a receptor with opposite effects of S1PR1 (277), our recent evidence further confirmed plasma S1PR3 as a potential biomarker in murine and human ALI (295). S1PR3 is nitrated and secreted upon ALI/VILI, both in vivo (patient and murine model) and in vitro (cultured human ECs) (295). The biological activity of nitrated and secreted S1PR3 (in microparticles) has been confirmed to mediate endothelial barrier disruption via as yet unknown molecular mechanisms (295). Multiple ALI/VILI-targeted strategies have employed manipulation of S1PR1 and S1PR3 function. However, many agonists such as SEW2871 and AUY954 for S1PR1 exhibit poor water solubility, thereby limiting their use in animal models of lung inflammation and injury. A systematic study of S1P receptor-mediated signaling and the study of intracellular targets that regulate S1P concentrations in vascular cells may provide further insights into the mechanisms underlying the barrier function disruption seen in ALI and VILI. Similar to MYLK, S1P receptor SNPs are significantly associated with ARDS. S1PR3 promoter SNPs rs7022797 and rs11137480 are linked with altered promoter activity and ARDS susceptibility (293). S1PR1 SNP rs59317557 is also significantly associated with severe asthma, a respiratory disease sharing similar endothelial pathology (294). This compelling evidence highly suggest that, like MYLK, S1P receptors can be potential effective therapeutic targets in ARDS and VILI.

Mechanical Stress Signaling Through Epidermal Growth Factor

EGFR-mediated signaling has been implicated in various cellular physiological and pathological processes. EGFR-activated signaling regulates many downstream signaling pathways, including MAP kinase and PI3K/Akt signaling. As mechanical stretch stimulates EGFR-mediated signaling, this pathway may be an important mechanotransducer (64, 312, 313, 316). The inhibition of EGFR signaling decreases mechanical ventilation-induced lung vascular leak, alveolar permeability and neutrophil accumulation in the bronchoalveolar lavage fluid and attenuates the expression of mechanical ventilation-responsive genes (32). Interestingly, intratracheal instillation of EGF failed to induce lung injury, suggesting that EGFR signaling is necessary, but may not be sufficient, to cause ALI (32). Exposure of the mouse lung to EGF enhances neutrophil accumulation, suggesting EGF involvement in enhancing ALI inflammation (319, 320). However, other studies suggest EGF is protective against certain forms of ALI including endotoxin (366), burn injury (29, 186), and toxic industrial compounds (183). However, it should be noted that we have recently shown that EGFR signaling induced by oxidative stress mediated EGFR subunit dimerization (249). This suggests there could be differences between physiological signaling, induced by EGF, and pathological signaling that is redox dependent. Because oxidative stress is a hallmark of multiple forms of ALI, including VILI (50, 237, 239, 252, 270, 288, 299, 306, 325, 334), this could lead to sustained EGFR signaling and extend the inflammatory phase.

Mechanical Stress Signaling Through TRPV4

TRPV4 is a member of the transient receptor potential cation (TRP) channel subfamily V (vanilloid) type 4 (117). TRPV4 channels are highly expressed in the lung endothelium (374). TRPV4 activation induces swelling, blebbing, and detachment of both the epithelium and capillary endothelium resulting in alveolar flooding (328). Seminal work from the Townsley group has shown that in the lung, TRPV4 is a rapidly activated mechanosensitive receptor (129). The activation of TRPV4 in lung EC by, for example, hydrostatic stress leads to an increase in intracellular Ca²⁺ and this is involved in disrupting the EC barrier, leading to increased pulmonary edema (158). There are also effects at the alveolar level as the Ca²⁺-mediated stimulation of NO generation from eNOS inhibits Na⁺ channels in the epithelium reducing alveolar fluid clearance (162). Increases in TRPV4 expression and activation have also been demonstrated in the lung as a consequence of heart failure (308). The orally active TRPV4 inhibitor GSK2193874 reduces calcium flux in cultured cells and attenuates pulmonary edema in reponse to the elevated venous pressure occurring during heart failure (308). With respect to VILI, work from the Townsley and Kuebler groups have shown that the elevation of lung microvascular pressure increases endothelial Ca²⁺ via TRPV4 activation (373) and that TRPV4 inhibitors reduce pulmonary edema in isolated perfused mouse lungs (129). TRPV4 inhibitors also attenuate the macrophage activation associated with VILI (128). Recent studies have demonstrated that TRPV4^−/− or GSK2193874 treated mice are protected against acid-induced ALI (375). Interestingly TRPV4 inhibition was only protective if given in a preventative manner (375). This suggests that the signaling pathways activated by TRPV4 are early in the development of ALI. Furthermore, the lack of a therapeutic window suggests that the downstream targets of TRPV4 may be more viable targets for therapy. In addition, there appear to be feedback pathways associated with actin cytoskeleton remodeling that regulate TRPV4 signaling as MLCK activation attenuates mechanosensitive calcium fluxes secondary to a reduction in the cell surface expression of TRPV4 channels (238).

Akt1 Signaling and Mechanical Injury

AKT (PKB) kinase is part of a critical signal transduction pathway that regulates cell survival, growth, proliferation, immunity, and permeability (192). The expression, activation, degradation, and localization of AKT is tightly regulated in normal cells (192). AKT1 is subjected to multiple forms of posttranslational modifications (PTM). Multiple studies have established the upstream signaling cascades and events that lead to the phosphorylation events at Thr308 and Ser473 (48, 95) that induce the full activation of AKT. The phosphatidylinositol-3-OH kinase (PI3K), a lipid kinase, is a canonical regulator of AKT activation in response to a variety of ligands, while multiple other non-PI3K stimuli can contribute to the activation (246, 284, 286). Emerging work has identified a role for AKT1 in the context of endothelial barrier function and lung injury. Seminal work in the past has shown the role of AKT in the cross talk between TJs and AJs (300). In this process, endothelial VE-cadherin in AJs upregulates the gene encoding the TJ adhesive protein, claudin-5 via the phosphorylation of forkhead box factor (FoxO1) through AKT activation and limiting the translocation of Tcf-4-β-catenin to the nucleus. This effect requires the release of the inhibitory activity of FoxO1 and β-catenin transcriptional repressor complex and reveal a critical role for AKT in the link between AJs and TJs and in permeability (115, 300).

Although the regulation of AKT signaling by phosphorylation is well established over the last decade, other types of PTMs of AKT have been identified including ubiquitination and nitration. Although ubiquitin conjugation is generally considered a marker for protein degradation, recent studies have demonstrated the pleiotropic roles of ubiquitination including protein processing, membrane trafficking facilitating protein activation, and transcriptional regulation (51). Recent studies have described AKT ubiquitination (51) via several enzymes including STIP1 homology and U-box containing protein 1, E3 ubiquitin protein ligase (CHIP), breast cancer 1 (BRCA1) (360), tetratricopeptide repeat domain 3 (TTC3) (290), and mitochondrial E3 ubiquitin protein ligase 1 (MUL1) (17). Family members of the AKT pathway including PI3K also have been shown to undergo ubiquitination by Cbl-b ubiquitin E3 ligase, the role of which has not yet been defined (51).

Recently a role of AKT deubiquitination has been described as a protective event against VILI. This is regulated via stress-induced growth arrest and DNA damage-inducible a (Gadd45a)/ubiquitin carboxyl terminal hydrolase 1 (UCHL1) (208, 212, 213). GADD45a^−/− mice have increased susceptibility to VILI and this is associated with reduced AKT1, but not mRNA. Overexpression of a constitutively active AKT1 transgene significantly reduces lung injury in VILI challenged GADD45a^−/− mice (208, 212). In vitro, exposing ECs to mechanical stress induces AKT1 phosphorylation and its trafficking and activation at the membrane (212). GADD45a is essential to this process as both GADD45a silenced ECs and GADD45a^−/− mice have increased AKT1 ubiquitination at K48 and proteasomal degradation. These events also correlated with the loss of UCHL1 that normally removes K48 polyubiquitin chains bound to AKT1. Loss of GADD45a significantly reduces UCHL1 expression due an increase in UCHL1 promoter methylation (212).

In addition to PTM by phosphorylation and ubiquitination, AKT1 can also be activated by nitration in pulmonary ECs (248). The endogenous NOS uncoupler, asymmetric dimethylarginine (ADMA) eNOS phosphorylation at the AKT1-dependent phosphorylation sites Ser(S)617 and S1179. Furthermore, ADMA enhanced AKT1 nitration and increased its activity. Mass spectrometry identified a single nitration site in Akt1 located at the tyrosine (Y) residue 350 located within the client-binding domain. Replacement of Y350 with phenylalanine abolished the ability of peroxynitrite-mediated nitration to induce eNOS translocation to the mitochondria (248). It is unclear whether a similar nitration-mediated activation mechanism occurs during the development of ARDS and whether it plays a significant role in attenuating the EC barrier disruption that drives the disease. Furthermore, it is possible that new therapeutics that modulate AKT PTM (phosphorylation, ubquitination, nitration) could be of clinical value. However, more work will be required to elucidate how these novel modifiers are regulated during ARDS to better define their mechanistic roles in regulating AKT function. In summary, VILI-related stimuli modulate AKT PTMs, which leads to EC injury and remodeling. Furthermore, modulating AKT PTMs could be a potential therapeutic target in VILI.

Cell-Cell Interactions in Endothelial Cells

Interactions between cells are necessary for maintaining a tight endothelial barrier in the lung. These fall into three types of intercellular functional complexes: AJs, gap junctions, and TJs, all of which are regulated by mechanical stress (Fig. 3). Each will be discussed below in some detail and the effect of mechanical stress discussed. Epithelial cells also contain another junctional complex, the desmosome. Desmosomes link the intermediate filamentous network between adjacent cells and like AJs, are comprised of cadherins, albeit desmosomal cadherins. By anchoring the intermediate filaments to the plasma membrane, they form a supracellular network that strengthens tissues, protecting them against mechanical damage. Endothelial cells appear to lack desomsomes (72). In addition, endothelial cells also contain focal adhesion complexes that allow the cell to interact with the extracellular matrix (ECM) via integrin receptors (Fig. 3). Modulation of the focal adhesion complex has been shown to regulate both endothelial shape and endothelial barrier integrity (97, 98, 204). Extensive experimental evidence have confirmed that the mechanical stress associated with VILI significantly impacts the expression, signaling, and regulation of EC junctions, which leads to injurious remodeling and EC barrier disruption (41, 59, 82, 203).

Fig. 3.

Mechanical stress effects on endothelial junctional integrity. The endothelial barrier is maintained by 3 types of intercellular functional complexes: adherens junctions, gap junctions, and tight junctions. The tight junction is maintained through the interactions of claudins, occludins, and zonula occludens-1 (ZO-1). In endothelial cells, adherens junctions are formed through the interactions of the extracellular domains of vascular endothelial cadherin (VE-Cad) between adjacent cells. The cytosolic portion of VE-Cad is bound to p120-catenin, α-catenin (α-Cat), β-catenin (β-Cat), and γ-catenin (γ-Cat). Gap junctions are direct channels between the cytoplasm of adjacent cells formed by the connexin family of proteins. Endothelial cells also form focal adhesion complexes that allow mechanical and regulatory signals to be transduced from the ECM, via integrin receptors, to inside the cell. Focal adhesions are large macromolecular complexes that require interactions between Paxillin (PAX), Focal adhesion kinase (FAK), G protein-coupled receptor kinase interactor-1 (GIT1), Talin, and Vinculin. The cartoon adapted from Wang and Dudek (376). VILI (or mechanical stress) have been shown to have effects on the proteins that regulate EC barrier function: ZO-1 is upregulated (62) and dissembled from TJ (4, 37, 125, 379) by VILI. Occludin is degraded and dissembled from TJ by VILI (384, 385). VE-Cad is phosphorylated and ubiquitinated by VILI (231). nmMLCK is upregulated (6) and activated (60, 108, 211) by VILI. β-Cat is upregulated by VILI (329). VILI promotes FAK phosphorylation at Y397 and Y576 (271). Integrin β4 is phosphorylated upon VILI (57). +, upregulated; D, degradation; P, phosphorylated; U, ubiquitination; DIS, dissembled; A, activated.

Adherens junctions.

Zonula adherens or AJs are located below TJs in the lateral plasma membrane and are formed by the cadherin family (171). The extracellular domains of cadherin bind with other cadherins within the plasma membranes of adjacent cells (171) while the cytosolic portions are bound to p120-catenin, β-catenin, α-catenin, and vinculin, linking cadherins to the actin cytoskeleton (171). In general, cadherins allow the formation of strong mechanical attachments between adjacent cells. VE-cadherin in particular plays a major role in regulation of vascular paracellular permeability and leukocyte transmigration (118). VE-cadherin has also been shown to regulate claudin-5 expression (see claudins below) (300), perhaps explaining why AJs form earlier than TJs in ECs (171). As AJs assemble, cadherin–catenin clusters actively transform the actin cytoskeleton (172). These clusters in turn regulate the activity of the small GTPases with Rac or Rho being involved in regulating AJ assembly (33, 42, 368), and Cdc42 regulating AJ maintenance (131). The cell must actively balance Rac and Rho signaling to ensure optimal AJ assembly, and cross talk is required for the actin reconfiguration that occurs during AJ assembly (131). During cell contact formation, Rac inhibits Rho (265) through the recruitment of the Rho GTPase activating protein, p190RhoGAP, to p120 catenin (349). After contacts have formed, Rho, via the activation of Rho kinase, is then required for actomyosin contractility (362). Another regulator of AJ assembly is IQGAP1, a multidomain protein with regions containing high sequence homology to Ras-GAPS (339). Its name is derived from the four IQ motifs that bind calmodulin and a COOH-terminal GAP-related domain (GRD) initially predicted to act as a traditional GTPase Activating Protein (GAP) (343) although it is now known that IQGAP1 does not act either as a either a guanine exchange factor (GEF) or as GAP since its RasGAP domain is inactive (135). Instead, IQGAP1 regulates the activity of small GTPases by acting as a scaffold protein to either stabilize the active or inactive GTPase forms and/or through the recruitment of multiple GTPase activity modulators (47, 153–155). Thus IQGAP1 binding is able to regulate the activity of a number of the small GTPases (154). IQGAP1 binding can both activate and inhibit small GTPase activity both positively and negatively (47, 346). IQGAP1 binding to the activated forms of Rac1 and Cdc42 is able to inhibit GTP hydrolysis, prolonging catalytic activities that are critical for reestablishment of the endothelial cell barrier following injury (154). In addition to binding activated Rac1/Cdc42 (174), this region of IQGAP1 interacts with catenin and cadherin AJ proteins (221), as well as microtubule tip-binding protein CLIP-170 (127). Similar to other ARDS-associated stimuli, the mechanical stress associated with VILI negatively regulates AJs, including Src-dependent phosphorylation of VE-cadherin, leading to its internalization and ubiquitination (231).

Gap junctions.

Gap junctions are direct channels between the cytoplasms of adjacent cells, which allow ions and small signaling molecules to diffuse freely between neighboring cells. Gap junctions allow the metabolic and electrical activities of neighboring cells to be linked. In ECs, gap junctions play an important role in triggering blood coagulation (226), inflammatory cytokine production (228), tube formation (227), migration (227), and regulating the permeability of the endothelial monolayer (225). Gap junctions are composed of transmembrane proteins of the connexin family. Four to six connexins assemble in the plasma membrane to form a connexon, which then aligns with a connexon of an adjacent cell. An open channel is then formed between the two cytoplasmic compartments. Recent studies have shown that the key endothelial connexin, Cx43, is upregulated by VILI-associated CS (225). This facilitates CS-mediated EC barrier disruption although the mechanism is still unresolved. These findings suggest that connexins play an essential role in CS-induced EC barrier regulation and are possibly novel therapeutic targets.

Tight junctions.

TJs perform two distinctive functions: first, they maintain cell polarity by restricting the diffusion of ions and small molecules between the apical and basolateral plasma membranes, the fence function (324); second, they form a seal between adjacent ECs, preventing uncontrolled transendothelial flux into the tissue parenchyma, barrier function) (88). The strands of the TJs completely encircle the apical region of the cell (281). Each strand in the network is composed of transmembrane proteins of the claudin, occludin, or junctional adhesion molecule (JAM) families (see below). These transmembrane proteins bind to similar type proteins on adjacent cells, thereby forming a seal between the two plasma membranes. Similar to other EC junctions, the TJ is affected by VILI-associated mechanical stress. For example, cyclic stress upregulates ZO-1 levels in vascular ECs (62). This results in TJ instability and dissembly, leading to a persistent disruption of the EC barrier (189).

Proteins of the tight junction.

occludin.

Occludin, a 60- to 65-kDa protein was the first integral TJ protein discovered (101). Occludin has four transmembrane segments divided into five separate domains, A–E (65). Domain A is the NH₂-terminal intracellular domain; domain B is the first extracellular loop; domain C is a short intervening intracellular loop; domain D is the second extracellular loop; and domain E is the COOH-terminal intracellular domain containing an important coiled-coil motif (65). The two extracellular loops form homophilic interactions with occludin from adjacent ECs (65) while domain D is important for membrane localization (202). The coiled-coil motif in the cytoplasmic COOH-terminal domain E allows an interaction to occur with the GUK domain within ZO-1 (102, 182) and is necessary for membrane trafficking (102, 182) and dimerization (40). The expression of occludin correlates with the degree of EC barrier function in different vascular beds. Occludin can be phosphorylated at multiple sites and these phosphorylation events play an important role in TJ assembly and stability as well as EC barrier function. Improved EC barrier function is associated with both increased phosphorylation of occludin on serine and threonine residues (61, 263) and tyrosine dephosphorylation (251, 303, 332). Occludin expression is also heavily regulated. Physiological cyclic strain (62) increases occludin expression in ECs while cytokines such as HGF decrease occludin mRNA and EC barrier function (159). Thus occludin is heavily regulated and plays an important role in regulating TJ permeability. VILI promotes occludin degradation via the activation of pp60^Src, leading to persistent TJ dysfunction and lung injury (384, 385). This suggests that occludin is one of the mediators of VILI-induced endothelial barrier dysfunction and may be a possible therapeutic target.

zonula occludens.

The first TJ protein to be identified was ZO-1 (285), while ZO-2 and ZO-3 were discovered later and found to interact with ZO-1 (157). ZO-1 (210–225 kDa), ZO-2 (180 kDa), and ZO-3 (130 kDa) are cytosolic membrane-associated components of tight and AJs in both epithelial and ECs. The ZO proteins are members of the membrane-associated guanylate kinase proteins (MAGUK) family, which have conserved PDZ, Src homology region 3 (SH3), and inactive guanylate kinase (GUK) domains that link different proteins (350). ZO-1 contains three sequential NH₂-terminal PDZ domains: PDZ1 binds to the COOH-terminal regions of claudins-1 to -8 (151), PDZ2 binds with ZO-2 (152) and ZO-3 (151) through DZ domain dimerization, while both PDZ2 and PDZ3 interact with JAMs (85). The SH3 domain interacts with signaling proteins, such as Gα₁₂ (209) and ZO-1-associated kinase (ZAK) (19). The GUK domain of ZO-1 also binds to occludin (87). The COOH-terminal proline-rich domain of ZO-1 binds the cytoskeletal proteins actin (87), AF6 (364), and cingulin (68). In addition to TJ proteins, ZO-1 also binds to the gap junction protein connexin-43 (361), and the AJ proteins, α- (218), β-, γ-catenins (250), armadillo repeat member of the catenin family (AVCRF) (168), and afadin (229). Because AJs form before TJs, ZO proteins are thought to recruit TJ proteins then break away to form apical TJal complexes (12). This possibility is supported by the fact that the depletion of ZO-1 and ZO-2 completely abrogates TJ assembly (322). In addition, ZO-1 deletion in mice results in embryonic lethality, increased vascular permeability, and disrupted cell junctions (167). ZO proteins are highly regulated at the transcriptional and posttranslational level. Tyrosine phosphorylation of ZO-1 induced by the tyrosine kinase, pp60^Src (302) or tyrosine phosphatase inhibitors (280) is associated with increased EC permeability. Serine/threonine phosphorylation of ZO-1 is more controversial. ZO-1 serine/threonine phosphorylation has been associated with both increased (276) and decreased permeability (20, 62). Our knowledge regarding the effects of VILI on ZO-1 is limited; however, studies have demonstrated that the VILI-activation of RhoA reduces the association of ZO-1 with occluding, leading to dissembled and unfunctional TJs (2, 37, 125, 379).

junctional adhesion molecules.

JAMs are 30- to 40-kDa, single-pass transmembrane proteins found in epithelial and endothelial TJs (25). To date, six JAM family members have been identified: JAM A–C, ESAM, CAR, and A33 (25). Structurally, JAMs consist of an NH₂-terminal extracellular domain, a transmembrane segment, and a short COOH-terminal cytoplasmic domain (25). The role of JAMs in regulating TJ permeability is complex. JAM-A reduces permeability (194) while JAM-C has been shown to increase EC permeability (181). JAMs also interact with ZO-1 and may mediate the recruitment of occludin to TJs (85). The effects of VILI on JAMs is unresolved. Hypothetically, mechanical stress would impose a negative effect on JAM proteins by negatively regulating other key TJ proteins such as occludin and ZO-1.

claudins.

The claudin family of proteins is vital for the assembly of TJal complexes. So far 27 claudins have been identified in mammals with molecular mass ranging from 20 to 27 kDa (210). The important role of claudins in regulating paracellular barrier function (148) is derived from studies in occludin knockout mice that show that TJs still form (262). Subsequent experiments established claudin-5 as being key for TJ formation in ECs (215). Occludin recruitment to TJs depends on the presence of claudins, suggesting that occludin may be more of an accessory protein in TJ formation (103). Like occludin, claudins have cytoplasmic NH₂- and COOH-terminal tails and four transmembrane domains (214) with two extracellular loops (EL1 and EL2) in between that are separated by a short, 20-residue, intervening intracellular loop. The COOH-terminal cytoplasmic juxtamembrane region mediates membrane localization and TJ recruitment (259). Claudins can associate through both homomeric and heteromeric interactions (104) although heterodimerization only occurs between specific claudins. Studies examining claudin-5 show that conserved residues within EL2 mediate the homo- and/or heterophilic trans-interaction between the classic claudins 1–10, 14, 15, 17, and 19 (241, 242). Variations in EL1 dictate the size and charge selectivity of each claudin. Evidence also suggests that claudins form paracellular pores with a defined size cutoffs (340). For example, claudin-5 knockout mice are selectively permeable to small molecules (< 800 kDa), suggesting that the remaining claudins form slightly larger pores (224). The prevalence of pore-forming vs. non-pore-forming claudins ultimately determines the degree of barrier tightness within the TJs of a given tissue (100, 163). Claudin-5 is largely considered to be EC specific (215), although claudin-5 is also expressed in other cell types (318). Claudin-5 upregulation improves microvascular endothelial (169) and blood-brain barrier (138) function. Claudin-5 is also dynamically regulated. In particular, changes in the phosphorylation status of claudin-5 drastically alter the barrier properties at the TJ complex. For example, the secondary messenger cAMP induces the phosphorylation of claudin-5 on threonine residues via the activation of protein kinase A, enhancing its membrane distribution and increasing endothelial barrier function (150). While claudin-5 serine phosphorylation by protein kinase C isoforms, α and ζ induces the disappearance of claudin-5 from the cell membrane, increasing endothelial permeability (282). Similarly, MLCK activation decreases claudin-5 expression and membrane localization, increases claudin-5 phosphorylation, and disrupts EC barrier function (130). Furthermore, the ROCK-mediated Thr²⁰⁷ phosphorylation of claudin-5 increases EC permeability (363). At the transcriptional level, claudin-5 is regulated by several TFs. For example, membrane bound VE-cadherin activates the PI3K/Akt pathway that causes the subsequent phosphorylation and inhibition of the forkhead repressor TF, FoxO1, in the cytosol increasing claudin-5 expression. In the absence of VE-cadherin, nonphosphorylated FoxO1 translocates to the nucleus and represses claudin-5 expression via two FoxO1 response elements (300). The expression of claudin-5 has also been shown to be increased by Krüppel-like factor 4 (KLF4) (187) and ETS-related gene (ERG) (216, 377) and downregulated by VEGF (13) and TNF-α/NF-κB (15). Claudin-5 expression and endothelial barrier function can also be regulated by the TF, Sox18 (93). Thus the regulation of claudin-5 in the vascular endothelium can greatly impact TJ formation and ultimately endothelial barrier function. However, its role in regulating mechanical stress related EC barrier dysfunction has not been studied.

THERAPEUTICS AND VILI

Design and delivery of targeted therapeutics are both challenging in patients with ARDS/VILI. Frequently, an ARDS diagnosis is recognized late in the clinical course after significant lung (among other multiorgan) damage has taken place and after excluding other conditions that may also appear to manifest similarly (such as heart failure with pulmonary edema). This late diagnosis poses a second challenge of studying a homogenous patient population in clinical trials, for both biomarker discovery (389) and therapeutic testing. Furthermore, during these late stages, the optimal mode of delivery (intratracheal, intravenous) is unclear. Adding more complexity, the diagnosis of VILI is made more difficult by underlying lung injury from a spectrum of causes (sepsis, trauma, postoperative state), unequal anatomical distribution of injury in all parts of the lungs, and varying severity of injury based on differences in ventilator settings. In fact, the optimal use of a ventilator, is paradoxically life saving in many of these instances. Nonetheless, with mortality rates in excess of 50% in the ICU, treating clinical forms of excessive VILI is a priority that has been underrecognized (5, 27, 31). To date, based on the ARDSnet trial, low tidal volume ventilation and adequately adjusted PEEP appear to be the only proven preventive and therapeutic measures of mechanical ventilation (5). Extracorporeal life support such as extracorporeal membrane oxygenation or extracorporeal CO₂ removal can provide adequate gas exchange in patients with ARDS (27). With emerging research into the pathophysiology and mechanisms of ARDS/VILI, more targeted therapeutics have surfaced that have the potential to reduce the clinical burden.

Anti-inflammatory Therapy

Based on animal and clinical studies, inflammation plays a significant role in ALI (141, 176) and VILI (67, 223, 310, 311). Increased levels of several inflammatory mediators (including TNFα, NAMPT/PBEF, IL-6, and IL-10) are found in ex vivo and in vivo animal models subjected to injurious mechanical ventilation (301) as well as in humans with ARDS. Inhibition of these cytokines and proinflammatory mediators through multiple agents including pharmacological agents, receptor inhibitors, antibodies, siRNA, and gene therapy (94, 136, 137, 220) has started the pipeline for the development of novel therapeutics in preclinical animal models. However, they have not yet translated to clinical trial success in critically ill patients (86, 230). Nonetheless, reduced plasma levels of NAMPT/PBEF, IL-6, IL-8, IL-1β, and TNF-α have been associated with reductions in VILI burden in preclinical clinical studies as well as outcomes (354). While the exact mechanism of how these mediators contribute to VILI are not fully characterized, their measurement as biomarkers may identify VILI-susceptible patients and further risk stratify those who would better respond to protective ventilation strategies. Indeed, given the heterogeneity of the disease, it is likely that personalized medicine may help to address this gap. Several of these new potential therapies may be nonspecific by acting to enhance EC barrier function in response to various stimuli. Specifically targeting systemic inflammation or augmenting tight, adherens, and gap junctions in this common pathway may reduce vascular leak and inflammatory lung injury. A more recent anti-inflammatory therapy being proposed takes advantage of cellular reponses to stress, the heat shock response (HSR). The HSR is characterized by the rapid expression of a conserved group of proteins (heat shock proteins, hsp) that are induced in response to cellular stressors. One isoform, hsp90, is highly conserved from prokaryotes to eukaryotes and constitutes 1–3% of total cellular protein in a cell. Hsp90 is involved in the conformational regulation of over 200 cellular proteins. Hsp90 exists as a homodimer that exists in two conformations: open and closed. The closed conformation enhances protein folding stability, while the open conformation leads to protein degradation. Recent data have shown that one of the major cellular proteins chaperoned by hsp90 is pp60^Src. pp60^Src is an early-activated protein during multiple models of ALI including VILI (384, 385). Studies show that hsp90 inhibitors protect against sepsis-induced ARDS, at least in part by attenuating pp60^Src activity (23, 53, 307). Since hsp90 inhibitors have completed Phase I and II trials as cancer therapeutics, they represent a possible new therapy for the treatment of ARDS and VILI and other inflammatory diseases. However, clinical trials with hsp90 inhibitors to treat ARDS/VILI will need to proceed cautiously as over 50 clinical trials involving 15 different hsp90 inhibitors have validated hsp90 as a clinical target (21, 139, 232–234, 257, 261, 269) but have proved disappointing as therapeutic (337). This failure is directly related to the fact that they all induce a HSR, which upregulates both hsp90 and hsp70 (139, 244, 347). Thus targeting hsp90 activity in ARDS/VILI may need to wait for the development of molecules that act via alternative mechanisms to inhibit hsp90 function without triggering the HSR. This is an active area of research and such molecules have been developed but have not been adequately tested in preclinical animal models (14, 170, 199).

New Therapeutic Targets in VILI and ARDS

Recently, preclinical studies in mice and rats have modeled signaling pathways and targeted therapeutics specific to excessive mechanical stress and VILI. Additionally, unlike proinflammatory pathways that worsen lung injury, several pathways have been reported to promote recovery from injury including the role of acute stress response pathways such as heat shock proteins, Gadd45α, and AKT, leading to new considerations for therapeutic targeting (23, 197, 208, 212, 213, 254, 289). Other studies have therapeutically targeted ROS pathways (90) and have shown that increasing the activity of NRF2, the major cellular regulator of antioxidant gene expression, via bixin, a canonical NRF2 inducer (306), or modulation of ATF3 expression (270), attenuated inflammatory response, and preserved EC barrier function in mice exposed to mechanical ventilation. Furthermore, recent studies have tested the efficacy of repurposing existing FDA-approved therapies for multiple ARDS-specific signaling pathways. For example, imatinib, an effective cancer therapy, inhibits inflammation and MAPK cascade and attenuates a preclinical “two-hit” VILI model that recapitulates human ARDS (180, 256, 371). These pathways have ushered in new concepts such as peptide therapeutics that are designed to specifically target a region of the candidate protein thought to contribute to VILI (247). Unfortunately, despite the number of therapies that improve outcomes in preclinical animal models of ALI/ARDS, translation into successful human clinical trials has been largely nonexistent. Thus it is vital that studies continue to increase our fundamental understanding of the complex processes involved in the mechanical ventilation-exposed injured lung. In addition, as we move away from the one-size-fits-all mentality of treatment and begin to utilize the power of next-generation sequencing, it may be possible to truly achieve a personalized medicine approach to the treatment of VILI/ARDS. This will be vital as ultimately, as with other complex diseases, the universal delivery of a single drug to all patients who suffer from VILI and ARDS is unlikely to succeed, especially given the observed health disparities that exists in VILI and ARDS.

CONCLUSION

Despite nearly five decades of study into ARDS, there are still no specific therapies beyond low tidal mechanical ventilation to reduce the mechanical stress placed on the injured lung. However, even with the advent of protective ventilation ARDS mortality has not significantly declined below 40% (330). Thus VILI still requires the development of new therapies that go beyond reducing the pressures and/or volumes to which the injured lung is exposed. VILI appears to be a progressive disease and, as yet, we do not understand the complex pathways that regulate its temporal progression. As discussed in this review, the complexity of the underlying changes in the pulmonary endothelium that precede the development of VILI development, including changes in gene expression, epigenetic factors, PTM of proteins, changes in junctional components, increased oxidative stress, as well as health disparities and contributions from an individual’s genetic makeup, mean that even therapies that show great promise in preclinical animal models, e.g., anti-inflammatory therapy, have failed to deliver significant reductions in human mortality. These failures are likely due to most studies being too focused on individual components within the cell and failing to take into account that the endothelial barrier is disrupted due to changes in multiple signaling pathways. Thus we appeal to the investigators in the field to take a more holistic approach to studying how the endothelium responds to mechanical stress. In this manner, we believe that we will be able to gain a more fundamental understanding of how the injured lung responds to mechanical stress. It is likely that only such a multifactorial approach will lead to the development of viable therapeutic targets to ameliorate the development of ARDS and VILI.

GRANTS

This research was supported in part by HL60190 (S. M. Black), HL67841 (S. M. Black), HL101902 (S. M. Black), HL126609 (J. G. N. Garcia), HL115014 (J. X.-J. Yuan), and HL096887 (J. R. Jacobson), all from the National Institutes of Health.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

T.W. and S.M.B. prepared figures; T.W., C.G., A.A.D., E.Z., X.W., A.N.G., J.R.J., J.X.-J.Y., J.G.G., and S.M.B. drafted manuscript; T.W., C.G., A.A.D., E.Z., X.W., J.R.J., J.X.-J.Y., J.G.G., and S.M.B. edited and revised manuscript; T.W., C.G., A.A.D., E.Z., X.W., A.N.G., J.R.J., J.X.-J.Y., J.G.G., and S.M.B. approved final version of manuscript.