Regulation of pulmonary endothelial barrier function by kinases

Abstract

The pulmonary endothelium is the target of continuous physiological and pathological stimuli that affect its crucial barrier function. The regulation, defense, and repair of endothelial barrier function require complex biochemical processes. This review examines the role of endothelial phosphorylating enzymes, kinases, a class with profound, interdigitating influences on endothelial permeability and lung function.

the endothelium forms a barrier that regulates the exchange of blood fluid, electrolytes, and proteins across the vascular wall and is constantly subjected to dynamic remodeling (187). Deviations of normal endothelial barrier function can lead to, or be caused by, various internal or external stresses, including inflammation, sepsis, diabetes, and atherosclerosis (51). Notably, acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) are clinical syndromes of noncardiogenic pulmonary edema caused primarily by increased permeability to proteins across the endothelial and epithelial barriers of the lung (21, 135). Microvascular permeability is increased by inflammatory factors, such as histamine, bradykinin, and platelet-activating factor, growth factors, glycation products, cytokines, reactive oxygen species (ROS), and activated leukocytes (150). These mediators induce a plethora of intracellular signaling events that compromise the barrier integrity by modulating the expression of junction proteins, adhesion molecules, or reorganization of the cytoskeleton and focal adhesion complexes (238). Permeability-altering responses are frequently initiated by the binding of agonists on endothelial surface receptors and the consequent activation of signaling molecules, such as kinases, phosphatases, and GTPases, that control protein expression and degradation (224). These molecular cascades act as modulators of cytoskeletal integrity and contractility, leading to fluctuations of endothelial barrier function (137). In this review we focus on the role of kinases in endothelial paracellular permeability.

Protein kinases catalyze the transfer of inorganic phosphate supplied by ATP onto serine, threonine, or tyrosine residues of substrate enzymes. The specificity of protein phosphorylation is determined by the local amino acid sequence that surrounds the residues to be phosphorylated (8).

Protein phosphorylation is opposed by protein dephosphorylation executed by phosphatases. Many protein phosphatases are promiscuous, in that they do not require amino acid-specific sequences to identify substrates for dephosphorylation. However, the specificity of protein phosphatases is often determined by their regulatory subunits, which target phosphatase to the specific substrate (213). Kinases exist in the active or inactive state (142, 270). Their activation occurs by phosphorylation or release of kinase autoinhibition. Upon activation, a kinase can bind to ATP and transfer phosphate. Kinase activity is regulated by Ca²⁺, Mg²⁺, and binding to negative regulators or to molecular chaperones, such as heat shock protein (Hsp)-90 (66, 210). An emerging body of evidence suggests that kinase-induced protein phosphorylation is an important regulator of endothelial permeability (Fig. 1).

Fig. 1.

Interaction among major kinase pathways in the pathophysiological control of endothelial barrier function. FAK, focal adhesion kinase; GAP, GTPase-activating protein; AMPK, AMP-activated protein kinase; PAK, p21-activated kinase; ROCK, RhoA-associated protein kinase; LIMK, LIM domain kinase; MLCK, myosin light chain kinase; pMLCK, phosphorylated MLCK; pcofilin, phosphorylated cofilin; MAPK, MAP kinase.

AMP-Activated Protein Kinase

AMP-activated protein kinase (AMPK) was discovered in 1973 to phosphorylate and activate two key enzymes of lipid biosynthesis, acetyl-CoA carboxylase (35) and 3-hydroxy-3-methylglutaryl-CoA reductase (19). Acetyl-CoA carboxylase kinase activity was shown to be activated by 5-AMP (264), whereas 3-hydroxy-3-methylglutaryl-CoA reductase activity was induced by phosphorylation via an upstream kinase (84). It was initially assumed that these were independent events, but, subsequently, Carling et al. discovered that they were functions of a single protein kinase activated by AMP and phosphorylation (34). Since it became clear that this was a kinase with multiple substrates, it was renamed AMP-activated protein kinase after its key activating ligand (164). Mammalian AMPK (28, 119) exists in heterotrimeric complexes consisting of a single catalytic subunit, α, and two accessory subunits, β and γ (49). There are multiple isoforms of each subunit (AMPKα₁ and -α₂, AMPKβ₁ and -β₂, and AMPKγ₁, -γ₂, and -γ₃) (85). AMPK regulates metabolic, as well as anti-inflammatory, functions (181, 212).

There is strong evidence that AMPK is involved in the regulation of vascular permeability (109) (Fig. 1) via multiple mechanisms. Activation of AMPK with metformin resealed wounds in LPS-exposed rat pulmonary microvascular endothelial cells; metformin failed to induce barrier repair in AMPK-depleted endothelial cells (151). It was recently reported that N-cadherin coordinates AMPK-mediated lung vascular repair, since disruption of N-cadherin's intracellular domain caused translocation of AMPK away from the membrane and attenuated the AMPK-mediated restoration of barrier function in LPS-treated pulmonary endothelium. AMPK activation reversed the LPS-induced increase in vascular permeability, whereas N-cadherin inhibition inhibited the AMPK-mediated repair (110). Additionally, AMPK inhibits RhoA and RhoA-associated protein kinase (ROCK) activation (247).

AMPK was also shown to exhibit anti-inflammatory activities, at least through inhibition of the Toll-like receptor-4-dependent activation of the transcription factor NF-κB (197, 226). Furthermore, AMPK contributes to the resolution of inflammation by facilitating phagocytosis and clearance of apoptotic cells (182, 274). Additionally, the AMPK-dependent phosphorylation and degradation of FoxO1a attenuate angiopoietin-2 expression and protect against the proinflammatory actions of tumor necrosis factor (TNF)-α (52). It was also reported that AMPK activation reduces vascular permeability and airway inflammation by regulating the hypoxia-inducible factor/vascular endothelial growth factor (VEGF)-A pathway (183).

Myosin Light-Chain Kinase

Myosin light-chain (MLC) kinase (MLCK) is a family of Ca²⁺/calmodulin (CaM)-dependent protein kinases that phosphorylate the regulatory MLC (MLC2) (210). MLC posttranslational modification is a key molecular cascade that regulates endothelial permeability and barrier function (210) (Fig. 1). MLCK-mediated phosphorylation of MLC2 (5, 227) at Ser¹⁹ and, subsequently, at Thr¹⁸ induces ATP-dependent actomyosin contraction (105, 127, 169, 210, 242), which increases capillary permeability. Similar to smooth muscle, in vascular endothelium, MLC phosphorylation triggers contraction, resulting in endothelial cell membrane retraction, intercellular gap formation, and barrier compromise (55, 193, 210).

There are distinct isoforms of MLCK (36, 127, 210, 249). The skeletal muscle MLCK is the only tissue-specific MLCK, inasmuch as it is expressed in adult skeletal muscle, but not in cardiac, smooth, or nonmuscular tissues. In contrast, the 130-kDa smooth muscle MLCK is ubiquitous in all adult tissues, including skeletal and cardiac muscle (95). Smooth muscle and nonmuscle MLCK isoforms (smMLCK and nmMLCK, respectively) have wide tissue distribution, and both are expressed in microvascular endothelial cells. They are derived from the same gene (mylk1) on human chromosome 3 and exhibit identical amino acid sequences and protein structures at the COOH terminus, which includes actin-binding, catalytic, inhibitory, CaM-binding, and kinase-related protein domains (64, 127). However, in contrast to smMLCK, nmMLCK also contains a unique 922-amino acid NH₂-terminal fragment containing multiple sites for protein-protein interaction (SH2- and SH3-binding domains), as well as potential regulatory phosphorylation sites for important kinases such as protein kinase C (PKC), protein kinase A (PKA), and MAP kinases (MAPKs) (55, 64).

Endothelial (nonmuscle) MLCK is represented by four isoforms, with MLCK1 and MLCK2 being the most abundant (127). Despite structural similarity, these isoforms reveal some differences in their regulation (23, 56). For example, MLCK1 and MLCK3a are phosphorylated by the Src-family kinases (SFKs) at Tyr⁴⁶⁴ and Tyr⁴⁷¹, which are not present in other MLCK isoforms (210). Importantly, Src-dependent MLCK phosphorylation activates MLCK (23), representing an additional modality involved in MLCK regulation. For example, thrombin increases endothelial permeability in a Src/MLCK-dependent manner via MLC-mediated contractile mechanisms in human lung microvascular endothelial cells (HLMVECs) (112). Additionally, in human umbilical vein endothelial cells (HUVECs), the MLCK-dependent MLC phosphorylation has been shown to regulate the permeability of intact isolated postcapillary venules and pulmonary microvessels in response to histamine and thrombin (245, 246, 272). Mylk^−/− mice are protected against increased lung vascular permeability and neutrophil adhesion induced by intraperitoneal LPS (246, 263). nmMLCK exerts an important role in VEGF-induced permeability. Activation of the VEGF signaling pathway in bovine pulmonary artery endothelial cells (BPAECs) increases MYLK gene and protein levels. Consequent increased nmMLCK mRNA and protein expression is a result of increased nmMYLK promoter activity, regulated in part by binding of the Sp1 transcription factor. This suggests that MYLK may be an important ARDS candidate gene and a therapeutic target that is highly influenced by excessive VEGF concentrations in the inflamed lung (216).

CaM is a major regulator of MLCK activity. MLCK contains a CaM-binding domain that maintains MLCK in an inactive state when CaM is not bound. In human embryonic kidney (HEK-293) cells, the binding of CaM results in MLCK activation (25). In kidney endothelial cells, CaM binding to MLCK in the presence of Ca²⁺ is crucial for MLCK activation (64, 243).

Treatment with the protein phosphatase (PP) inhibitor calyculin A also increases MLC phosphorylation in 3T3 cells (225) and coronary venular endothelial cells (130), suggesting a constitutive MLCK activity. Therefore, it is believed that basal endothelial barrier function is maintained by tonic MLCK activity (193).

MLCK activity is inhibited pharmacologically by compounds such as ML-7 and ML-9 in HEK-293 cells (211, 272). ML-7 improves endothelial dysfunction via tight junction regulation in the aorta of a rabbit model of atherosclerosis (41) and attenuates endothelial cell permeability in association with a downregulation of junction proteins via mechanisms involving MLCK-mediated MLC phosphorylation (276).

Whereas MLC phosphorylation depends on the activation of a MLCK, MLC dephosphorylation is executed by MLC phosphatase (MLCP), which consists of the catalytic subunit CS1β (formerly CS1δ), a M20 subunit with unknown function, and the myosin-targeting regulatory subunit MYPT1 (46). MYPT1 is phosphorylated by Rho kinase at Thr⁴⁹⁶, leading to MLCP inhibition and increased MLC phosphorylation (111). In smooth muscle cells (SMCs), Rho kinase can increase the level of MLC phosphorylation by direct phosphate incorporation at the same residue as does MLCK (11). Several other kinases are suspected in the modulation of MLC phosphorylation (see below), including PKC (28), which phosphorylates not only MLC, but also the specific MLCP inhibitor CPI-17 at Thr³⁸, leading to increased inhibitory potency toward MLCP and PKA, which reduces the phosphorylation of MLC and opposes agonist-induced hyperpermeability in mouse lung endothelial cells (185). That effect is possible due to MLCP association with myosin and MLC dephosphorylation (22).

Nonreceptor Tyrosine Kinases

SFKs are nonreceptor tyrosine kinases that are involved in multiple signaling functions. The prototypical Src isoform viral Src kinase (v-Src) was originally isolated from the cancer-inducing Rous sarcoma virus (30, 186). SFKs include ≥11 isoforms, Blk, Brk, Fgr, Frk, Fyn, Hck, Lck, Lyn, Src, Srm, and Yes (154, 228), with molecular weights ranging from 52 to 62 kDa (99). The isoforms Src, Fyn, and Yes are expressed in various mammalian tissues and are involved in endothelial signaling (178).

SFKs are composed of eight distinct functional regions. From the NH₂ to the COOH terminus, these regions include a myristylated site, the SH4 Src homology domain, a unique region, a SH3 domain, a SH2 domain, a linker, the SH1 kinase/catalytic domain, and the regulatory domain. SH3 and SH2 are protein-protein interaction domains. The SH2 domain binds phosphotyrosine motifs in an inter- or intramolecular manner. The SH1 domain is the site of tyrosine kinase activity (98).

The main phosphorylation sites on Src are Tyr⁴¹⁶ (SH1 domain) and Tyr⁵²⁷ (regulatory domain). Tyr⁴¹⁶ is able to undergo autophosphorylation (194). Tyr⁵²⁷ is phosphorylated and dephosphorylated by many proteins, such as Csk (COOH-terminal Src kinase), SHP-1 (Src-homology 2 domain-containing phosphatase 1), SHP-2, or the protein tyrosine phosphatase PTP1 (102, 275). Both phosphorylation sites play a key role in regulating the activity of SFKs (99).

SFKs mediate microvascular endothelial hyperpermeability through phosphorylation of tyrosine residues on regulatory proteins, including those involved in endothelial cell-cell and cell-matrix adhesions (98, 99, 178, 270). Thus, Src kinase is required for the endothelial hyperpermeability in response to VEGF (58), TNFα (176), or ROS in pulmonary cells (114). Src kinase is also involved in the enhanced hyperpermeability induced by activated neutrophils in human dermal microvascular cells (229). In HLMVECs, Src phosphorylates and activates MLCK, which in turn leads to cytoskeletal actomyosin contraction, actin stress fiber formation, and barrier compromise (23, 99) (Fig. 1).

Hsp90 governs the regulation of v-Src (261). It shifts the conformation of v-Src toward high-activity states that would otherwise be unstable (26). We recently reported that the Hsp90 inhibitor 17-N-allylamino-17-demethoxygeldanamycin (17-AAG) blocks LPS-induced tyrosine phosphorylation of Hsp90 in BPAECs, as well as HLMVECs (17). LPS induces concentration- and time-dependent Src-mediated tyrosine phosphorylation of Hsp90α and Hsp90β in HLMVECs. Mass spectrometry identified Tyr³⁰⁹ as a major site of tyrosine phosphorylation on Hsp90α (Tyr³⁰⁰ of Hsp90β). LPS-induced Hsp90 phosphorylation was prevented by the Hsp90 inhibitor 17-AAG in vitro, as well as in lungs from LPS-treated mice in vivo. Furthermore, 17-AAG prevented LPS-induced Src activation. LPS-induced Hsp90 phosphorylation was also prevented by the Src inhibitor PP2 (17).

In HLMVECs, TNFα causes a SFK-dependent phosphorylation and redistribution of VE-cadherin and gap formation, leading to increased paracellular permeability (9, 99, 144, 176, 257). Furthermore, both focal adhesion kinase (FAK) and paxillin, located in focal adhesion complexes, are Src substrates, and their activities are mainly regulated through phosphorylation by SFKs (190, 228, 262). Both Fyn and Src are associated with FAK and mediate signals initiated by integrin-extracellular matrix (ECM) binding (136). In HUVECs and HEK-293 cells, Src kinases phosphorylate FAK, leading to its binding to α_vβ₅-integrin and causing an increase in endothelial permeability due to disruption of integrin binding to the ECM (59).

In BPAECs, albumin binding to the gp60 receptor triggers Src activation by phosphorylation (98, 99). Once activated, Src causes phosphorylation of caveolin-1 (at Tyr¹⁴) and dynamin-2 (at Tyr⁵⁹⁷), which in turn results in endocytosis and transcytosis of plasma components (208).

H₂O₂ increases the activity of Src and other SFKs, including Lck (18, 86, 167), via oxidation at two cysteine residues and through the dephosphorylation of Tyr⁵²⁷ (72, 191). Exposure of pulmonary endothelial cells to H₂O₂ increases Src activity, leading to endothelial hyperpermeability (114). SFK inhibitors, herbimycin A and PP1, prolong the onset and attenuate the magnitude of the H₂O₂-induced increase in pulmonary endothelial permeability (114).

VEGF-induced increased vascular permeability requires SFKs (58, 252). Animals that do not express Src or Yes do not elicit VEGF-mediated vascular hyperpermeability (58). VEGF receptor (VEGFR)-1, VE-cadherin, and β-catenin form a complex that exerts a key role in endothelial barrier integrity (124, 230) and is dissociated upon VEGF stimulation (69, 251). Activated Src is recruited to Flk-1 after VEGF stimulation (43), suggesting that Src contributes to the VEGF-induced tyrosine phosphorylation of VE-cadherin and β-catenin in HUVECs (60). VEGF stimulation results in the endocytosis of VE-cadherin and the consequent disruption of the endothelial junctions. In axillary lymph node endothelial cells of mice, β-arrestin aids VE-cadherin endocytosis due to its ability to interact with Src (138) and facilitates the pulmonary Src-dependent phosphorylation of cadherin-catenin complexes (180, 251). Furthermore, in cultured embryonic endothelial cells, Src kinase regulates the VEGF-induced abundance of the FAK-α_vβ₅-integrin complexes, which are reduced in cells derived from Src knockout (KO) mice (137). These findings indicate that the Src-dependent and VEGF-induced FAK-α_vβ₅-integrin complex formation is an important regulatory mechanism of vascular permeability (83).

Besides SFKs, pharmacological inhibition of the Abl family of nonreceptor tyrosine kinases by imatinib achieves protective effects in animal models of inflammation. Mice, as well as HLMVECs lacking Abl, exhibit activated Rac1 and Rap1 and attenuated agonist-induced Ca²⁺ mobilization and actomyosin contraction (42).

Protein Kinase C

PKC is a family of Ser/Thr kinases consisting of classical (α, β_I, β_II, and γ), novel (δ, ε, η, and θ), and atypical (ι/λ and ζ) isoforms (116). Classical PKC isoforms are activated by binding to Ca²⁺ and diacylglycerol (DAG), novel PKC isoforms are activated by DAG, and atypical PKC isoforms are not activated by either Ca²⁺ or DAG. Activated PKC phosphorylates target proteins at Ser/Thr residues (8).

Most studies report that PKCs are mediators of endothelial hyperpermeability (270) through the activation of Ras-GTP, Raf, MAP/ERK kinases (MEKs), ERK1/2, c-Jun NH₂-terminal kinases (JNK), and p38 MAPK (Fig. 1). In human and mouse pulmonary endothelial cells, activated PKCs increase actin polymerization (63, 220), induce disassembly of adherens junctions (10, 200), and cause rearrangement of focal adhesions (1).

PKCα mediates endothelial hyperpermeability in response to inflammatory stimulation, as well as during hyperglycemia, ischemia, and angiogenesis (94). PKCβ_II mediates hyperpermeability in the diabetic kidney and retina (70, 250). In HLMVECs, PKCα phosphorylates p120 catenin at Ser⁸⁷⁹, an effect that causes VE-cadherin internalization and a consequent junction destabilization (239). In mice and HLMVECs, inhibition of PKCβ and PKCζ decreases endothelial permeability by decreasing Ca²⁺ levels and impairing MLC-triggered contraction (82, 88, 100, 193). However, it has also been reported that PKCα exerts a protective effect on barrier integrity via RhoA suppression (80, 81), and PKCδ and PKCε may also be barrier-protective (70, 238). Recently, PKCδ KO mice and PKC inhibitors were used to show the involvement of PKCδ in vascular hyperpermeability in ALI (7) in pulmonary cells. The LPS-induced lung injury was more severe in PKCδ KO mice than the corresponding controls. PKCδ inhibition promoted barrier dysfunction in vitro, and administration of a PKCδ-specific inhibitor significantly increased steady-state vascular permeability. PKCδ inhibition increased neutrophil transmigration, indicating that PKCδ inhibition induces structural changes in endothelial cells, allowing extravasation of proteins and neutrophils (7). PKCδ induction has also been shown to decrease endothelial permeability by increasing focal adhesion contacts (88, 224). Furthermore, PKCε may exert a protective role on the heart and brain endothelium (159).

Nitric oxide (NO) has been identified as a potential downstream target of PKC. Endothelial cells exposed to PKC activators exhibit increased levels of NO synthase (NOS) phosphorylation and NO production (57). It has also been suggested that the barrier-weakening effect of PKC is mediated through the endothelial production of NO, suggesting an interactive role of PKC and NOS in the modulation of microvascular barrier function (101). In HLMVECs, PKC-dependent phosphorylation of endothelial NOS at Thr⁴⁹⁵ regulates endothelial NOS coupling and endothelial barrier function in response to G⁺ toxins (37).

Focal Adhesion Kinase

FAK is a cytoplasmic protein-tyrosine kinase that facilitates cytoskeletal remodeling, formation, and disassembly of cell adhesion structures and regulates the Rho-family GTPases (231). It facilitates the localization and cyclic activation of guanine nucleotide exchange factors and GTPase-activating proteins (GAPs) (232). Activated FAK recruits c-Src at focal adhesion sites to form a FAK-Src signaling complex. This signaling complex phosphorylates other focal adhesion signaling proteins such as paxillin and p130Cas, thereby activating diverse signaling pathways important in the regulation of cell migration (267).

FAK forms a complex with p120RasGAP and p190RhoGAP, resulting in the localization of p190RhoGAP at nascent focal adhesions. FAK-induced phosphorylation of p190RhoGAP is associated with the inhibition of RhoA at leading-edge focal adhesions. As Rho and Rac activity are antagonistic, Rho inhibition can also indirectly facilitate Rac activation (232).

The NH₂-terminal domain of FAK binds to the COOH-terminal domain of integrins or growth factor receptors (184, 259). The COOH terminus of FAK is known as FAK-related nonkinase and is involved in the recruitment of FAK to focal adhesions at the cell membrane in response to phosphorylation (157).

FAK activity is mainly regulated through phosphorylation by SFKs (228). Inhibition of FAK tyrosine phosphorylation prevents focal adhesion formation and cell contraction (201, 202). Shear stress and growth factors activate focal adhesions through FAK tyrosine phosphorylation (3, 20, 79, 217). Overexpression of dominant-negative FAK in endothelial cells inhibits cell contraction induced by FAK activation (203). FAKs contain tyrosine phosphorylation sites that are important for focal adhesion assembly regulation (259). Phosphorylation of Tyr³⁹⁷ permits FAK binding to proteins containing the SH2 domain. FAK recruits SFKs to the focal adhesion complex, which in turn regulates FAK (270). Basal FAK phosphorylation levels are important for stabilizing focal adhesions (45) and help maintain normal physiological pulmonary endothelial permeability (184).

FAK-dependent strengthening of focal adhesions enhances endothelial barrier function under normal physiological conditions (259). In human pulmonary artery endothelial cells, FAK translocation to the focal adhesion complex attenuates the decrease in transendothelial electrical resistance induced by inflammatory agents (152) and is required for enhanced barrier function under hyperosmotic conditions (149). In Chinese hamster ovary cells, FAK binds to p120-catenin and suppresses the activity of the Rho family of GTPases (173, 269). It also prevents oxidant-induced barrier dysfunction by regulating VE-cadherin phosphorylation (224). In mouse pulmonary endothelial cells, conditional deletion of FAK disrupts lung vascular barrier function due to destabilization of RhoA and Rac1 activities (204).

However, expression of dominant-negative FAK prevents adherens junction disruption in response to VEGF or oxidants, indicating that FAK may also disrupt barrier function. In HUVECs, VEGF increases FAK tyrosine phosphorylation and recruitment to focal contacts associated with stress fibers (128). Genetic or pharmacological FAK inhibition prevents VEGF-stimulated permeability downstream of the VEGFR, both in vivo and in vitro. Similarly, in pulmonary cells, FAK phosphorylated at Tyr³⁹⁷, Tyr⁵⁷⁶, and Tyr⁹²⁵ contributes to thrombin-induced hyperpermeability, whereas FAK phosphorylated at only Tyr⁵⁷⁶ is associated with cortical actin focal adhesion and barrier strengthening (214, 236).

MAP Kinases

MAPKs mediate various cellular responses to growth factors and cytokines (121). In mammalian cells, the most abundant MAPKs are the extracellular signal-related kinases (ERK1/2), JNK, and p38 protein kinase. MAPK activation modulates the permeability of isolated rat epithelial cell monolayers following cyclic stretch (44). In HUVECs, inhibition of ERK1/2 with PD98059 or inhibition of p38 with SB203580 blocks VEGF-induced hyperpermeability (122, 258). Moreover, in pig coronary venules, ERK1/2 regulates the baseline barrier function and cGMP-induced hyperpermeability in endothelial cells (240, 268).

Many inflammatory agents cause ERK1/2 phosphorylation/activation (61, 115, 244, 254) and induce endothelial cytoskeletal contraction through a p38-related mechanism (75, 122). Advanced glycosylation end products induce ERK1/2 phosphorylation and leakage of retinal microvascular endothelial cells (221). Activation of the ERK1/2 or the p38 pathway can disrupt endothelial barrier function (77) and trigger the H₂O₂-induced loss of endothelial tight junctions (128). Conversely, ERK1/2 suppression is associated with Rho inhibition (177). Furthermore, there is a direct link among p38 MAPK activation, remodeling of the microtubule network, and endothelial barrier regulation: p38 MAPK inhibition attenuates the nocodazole-induced microtubule depolymerization, actin remodeling, and lung endothelial barrier dysfunction (24).

cAMP- and cGMP-Dependent Protein Kinases

PKA is a cAMP-dependent protein kinase that strengthens pulmonary endothelial barrier integrity (54, 92, 160, 171, 206, 209). It causes MLC dephosphorylation (163), dissociation of F-actin from myosin (125), stabilization of cytoskeletal filaments (91), and strengthening of cell-matrix adhesions (123). In rat aortas, PKA can also inhibit leukocyte adhesion and platelet aggregation (80, 129). In HUVECs, cAMP potentiates VE-cadherin-mediated cell-cell contact to enhance endothelial barrier function through an Epac-Rap1 signaling pathway (62). Basal levels of cAMP are critical in maintaining normal barrier function under resting conditions, and increased levels of cAMP prevent increases in microvascular permeability (53, 54, 93). In HUVECs, prostaglandin D₂-D-type prostanoid (DP) signaling promotes endothelial barrier function via the cAMP/PKA/Tiam1/Rac1 pathway (120) and exendin-4 promotes endothelial barrier enhancement via PKA- and Epac1-dependent Rac1 activation (133).

Selective inhibition of NOS, guanylate cyclase, or PKG suppresses shear stress-induced (271) and VEGF-induced (258) hyperpermeability. Most studies suggest that increases in endothelial cGMP are barrier-protective and prevent oxidant-mediated endothelial barrier dysfunction, possibly via PKG1 (158). However, there is also evidence in pulmonary endothelial cells that activation of the cGMP-dependent pathway is a necessary step in eliciting the hyperpermeability response to various inflammatory mediators (93, 153).

Phosphodiesterases (PDEs) catalyze the breakdown of cAMP and cGMP. Three PDE isoforms, cGMP-stimulated PDE II and IV and cGMP-inhibited PDE III, have been shown to affect cAMP-associated alteration of endothelial permeability (50, 53, 54, 171).

TESKs, LIMKs, and Cofilin

In 1984, it was observed that, when phosphorylated, actin-depolymerizing factor/cofilin (hereafter called cofilin) did not bind to actin, suggesting that this phosphorylation may be the crucial posttranslational modification that dictated the interaction between cofilin and actin. Phosphorylated and dephosphorylated cofilin exist, but the phosphorylated species represents a regulated, inactive form of cofilin (148) that is reactivated by dephosphorylation (147). The single cofilin phosphopeptide was sequenced, and Ser³ was identified as the phosphorylation site.

Cofilin has emerged as one of the protein families playing an essential role in actin dynamics at the plasma membrane during cell protrusion (141). Cofilin is a small (19-kDa) ubiquitous protein that binds to both G- and F-actin; it has a higher affinity for ADP-bound subunits and enhances the rate of monomer dissociation from the pointed end of actin filaments (33, 140). In addition, cofilin can also sever actin filaments and, thus, directly generate free actin barbed ends (51). The depolymerization and severing activities of cofilin are presumably due to its ability to bind cooperatively to F-actin and cause a twist in the actin filament, which promotes the destabilization of actin-actin interactions and, thus, fragmentation of the filament (147). Cofilin functions to depolymerize filaments, so that actin can be recycled for another round of polymerization (14, 97) and can induce polymerization of actin by virtue of its severing activity (71, 162).

The severing and depolymerization activity of cofilin in human and mouse pulmonary endothelial cells can be inhibited by phosphorylation on Ser³, which abolishes its actin-binding activity (6, 161). Four different kinases that appear to be downstream of the Rho-family GTPases phosphorylate cofilin: LIM domain kinases LIMK1 and LIMK2 and testis-specific kinases TESK1 and TESK2 (14, 47, 195, 233). Phosphatases, including type 1, type 2A (12), type 2B (148), type 2C (273), and a novel cofilin phosphatase, slingshot (170), have been implicated in the reactivation of cofilin by dephosphorylation in a variety of cell types. LIMK1 induction and cofilin phosphorylation promote endothelial barrier disruption (78).

p21-Activated Kinase

The p21-activated kinase (PAK) family of Ser/Thr kinases is involved in cytoskeletal reorganization, MAPK signaling, apoptotic signaling, control of phagocyte NADPH oxidase (Nox), and growth factor-induced neurite outgrowth (48, 119). Several mechanisms that induce PAK activity have been reported. Binding of Rac/cdc42 to the CRIB (or PBD) domain near the NH₂ terminus of PAK causes autophosphorylation and conformational changes in PAK (119). In rat vascular SMCs, phosphorylation of PAK1 at Thr⁴²³ by 3-phosphoinositide-dependent kinase-1 induces activation of PAK1 (118). Several autophosphorylation sites, including Ser¹⁹⁹ and Ser²⁰⁴ of PAK1 and Ser¹⁹² and Ser¹⁹⁷ of PAK2, have been identified (143).

The effects of PAK on endothelial barrier function remain controversial. In HUVECs, PAK is phosphorylated on Ser¹⁴¹ during its activation downstream of the small GTPase Rac. The phosphorylated subfraction is reported to translocate to endothelial cell-cell junctions in response to serum, VEGF, basic fibroblast growth factor, TNFα, histamine, and thrombin (223), and blocking PAK activation in HLMVECs prevented the increase in monolayer permeability in response to these factors. Similarly, inhibition of PAK, in vivo, reduces permeability in atherosclerosis-prone regions, suggesting that matrix-specific PAK activation mediates elevated vascular permeability in atherogenesis (179). In BPAECs, direct phosphorylation of MLC by PAK leads to increased contractility and increased endothelial permeability (222). In rat fat endothelial cells, PAK disrupts endothelial cell-cell junctions by direct phosphorylation and subsequent internalization and degradation of VE-cadherin (67).

On the other hand, in HLMVECs and mouse lungs, activated PAK phosphorylates and inhibits guanine nucleotide exchange factor H1, which leads to decreased ROCK-MLC activity, decreased contractility, and decreased endothelial permeability (237). Furthermore, PAK protects against an increase in permeability in a hypoxia-induced pulmonary hypertension model (189). It was recently shown that endothelial PAK2 deletion leads to increased vascular permeability (188).

ROCK

ROCK is a Ser/Thr kinase that is known as the downstream target of the small GTP-binding protein RhoA (132). ROCK mediates RhoA-induced actin-cytoskeletal changes via the phosphorylation of MYPT1 at Thr⁴⁹⁶ (81, 117, 218).

Two ROCK isoforms have been identified and have been associated with increases in vascular permeability (Fig. 1). The ROCK1 gene is located on chromosome 18 and encodes a 1,354-amino acid protein (107, 131). The ROCK2 gene is located on chromosome 12 and encodes a 1,388-amino acid protein (145). ROCK1 and ROCK2 share an overall 65% homology in amino acid sequence and 92% homology in their kinase domains (134). ROCK1 is widely expressed in most tissues, except brain and muscle, whereas ROCK2 is most highly expressed in muscle, brain, heart, lung, and placenta (165). Both ROCK1 and ROCK2 are expressed in vascular endothelial cells and SMCs (156, 175, 255).

Angiotensin II and IL-1β upregulate both isoforms of ROCK in human coronary vascular SMCs. This is mediated by PKC and NF-κB (96). In vascular SMCs, silencing of either ROCK isoform leads to an increased protein expression of the other isoform (174).

Although ROCK1 and ROCK2 are ubiquitously expressed and highly homologous, several mechanisms have been reported to differentially regulate ROCK isoform activities. Overexpression of ROCK1 or ROCK2 in mouse lungs and HLMVECs can increase MLC phosphorylation, an effect that is associated with increased pulmonary vascular permeability (265). ROCK2 binds to the MYPT1 subunit of MLCP and plays a predominant role in vascular SMC contractility (248). ROCK2 drives the lysophosphatidic acid-mediated activation of NF-κB (215), and ROCK1 KO mice exert a reduced vascular inflammation and neointima formation after flow cessation-induced vascular injury in the ligated carotid artery (172).

LIMK1 and LIMK2 are well-characterized ROCK substrates, which in turn phosphorylate cofilin in mouse and human lungs. In endothelial cells, it was demonstrated that shear stress, VEGF, and thrombin induce ROCK-dependent LIMK activation and subsequent cofilin phosphorylation, leading to sterol regulatory element-binding protein 2 activation, rearrangements of the actin cytoskeleton, and a decrease in cofilin oligomerization (90).

Hsp90 and Kinases

Hsp90 is required for maintaining the stability and activity of a diverse group of client proteins, including protein kinases, transcription factors, and hormone receptors, that are involved in cell signaling, proliferation, survival, oncogenesis, and cancer progression. Inhibition of Hsp90 alters the Hsp90-client protein complex, leading to reduced activity, misfolding, ubiquitination, and proteasomal degradation of the client proteins, including numerous kinases (31) (Fig. 1). Hsp90 inhibitors are now being actively developed, and >17 agents have entered clinical trials (168). In addition, several laboratories focus on the anti-inflammatory activity of Hsp90 inhibitors in the lung (15, 16).

It was recently demonstrated that Hsp90 regulates Nox activity and is necessary for superoxide production by Nox1, Nox2, Nox3, and Nox5 (38). Hsp90 binds to and regulates Nox protein stability. These actions are opposed by the Hsp90 inhibitors Hsp70 and COOH terminus of heat shock conjugate 70-interacting protein (CHIP), which promote the ubiquitination and degradation of Nox proteins and, thus, reduce ROS production (39). Induction of ROS has been associated with increases in vascular permeability (103). Moreover, the activin receptor-like ALK1 and ALK5 were found to interact with Hsp90, and Hsp90 inhibition suppresses the constitutively active ALK5-induced endothelial hypoxanthine-guanine phosphoribosyltransferase (hyprt) permeability (13). Hsp90 inhibition also prevents and restores LPS-mediated hyperpermeability in HLMVECs. These effects appear to be mediated by inhibition of RhoA signaling, which is an important effector of vascular permeability (89). Furthermore, LPS induces pp60c-Src-mediated tyrosine phosphorylation of Hsp90 in lung vascular endothelial cells and mouse lung, and this effect was prevented in vivo and in vitro by Hsp90 inhibition (17). More recently, it was shown that the transcription factor p53 is involved in mediation of the barrier-protective effects of Hsp90 inhibitors in HLMVECs and mouse lungs. Hsp90 inhibition induces the upregulation of the wild-type p53, an effect that restores endothelial barrier function by disrupting the LPS-induced RhoA/MLC2 pathway (16).

Redundancy and Interplay Within the Kinase Network

The vascular endothelium is the first barrier confronting the fluid or the inflammatory cells migrating from the lumen to the alveoli. Endothelial barrier dysfunction causes major fluid leak, a hallmark of ALI/ARDS. Effective lung repair requires correction of two critical abnormalities: the hyperfiltration across the endothelial barrier and the edema accumulated in the alveolar spaces. In ALI, failure of endothelial cells to sieve proteins is responsible for the hyperpermeability and the resulting pulmonary edema (21).

Kinases are crucial regulators of vascular barrier integrity, since they are able to translate environmental and pathological stimuli into vascular barrier responses. There is intense cross talk between the different kinases, which in turn forges a plethora of intracellular pathways. For example, Src activates MLC by tyrosine phosphorylation in a Rho GTPase-dependent manner (65). Rho kinase-induced MLC phosphorylation is exerted at the same residue as MLCK (104, 234). In HLMVECs, VEGF-induced hyperpermeability involves partial activation of the transcription factor Sp1, which induces MLCK expression (216). Similarly, the thrombin-induced hyperpermeability in HUVECs is Src/MLCK-dependent (27, 74). In mouse lungs, MLC can be directly phosphorylated by PKC, in contrast to PKA, which reduces MLC phosphorylation (77).

PKA exerts its barrier-strengthening effect not only by dephosphorylating MLC (77, 163, 209), but also via Rac1 activation (133) and Src suppression. In HEK-293, mouse NIH-3T3, and NIH-3T3-A14 cells, G protein-coupled receptors inhibit Src signaling by activating Csk in a cAMP-PKA-dependent manner (2).

In rat epididymal microvascular endothelial cells and HUVECs, Src is crucial for the TNFα-induced endothelial barrier disruption, through VE-cadherin phosphorylation (76, 87) and direct phosphorylation of FAK, leading to disruption of integrin binding to the ECM (253). In HUVECs, Src is also involved in VEGF-induced hyperpermeability via tyrosine phosphorylation of VE-cadherin and β-catenin (4).

In HUVECs, VEGF-induced hyperpermeability is MAPK-dependent, as it is abolished by inhibition of ERK1/2 and p38 (155, 205). However, VEGF stimulation also promotes the rapid endocytosis of VE-cadherin by a PAK-dependent mechanism. PAK-mediated phosphorylation of a highly conserved motif within the intracellular tail of VE-cadherin results in the disassembly of intercellular junctions (69). PAK2 has been shown to phosphorylate MLCK (73).

Thrombin induces ERK activation via PKC-dependent and -independent pathways. The PKC-independent pathway requires the involvement of the Src kinases, while the PKC-dependent mechanism depends on PKCε (146). In HEK-293 cells, PKC regulation of the JNK pathway is triggered via PKCε (29), and in rat primary microglia cultures, PKCα is required in the LPS-induced activation of the p38 MAPK (166). PKC is involved in the MLC2 activation, since in BPAECs, MLC phosphorylation occurs with a stoichiometry similar to that mediated by MLCK (241). Furthermore, in HLMVECs, MLCK is inhibited by PAK (199), while PKC upregulates ROCK, which in turn induces MLC phosphorylation (68).

In mouse lungs, FAK activity can be regulated by SFKs and can synergistically regulate VE-cadherin and vascular permeability (207). Src-mediated phosphorylation of FAK is a pivotal regulator of actin and adhesion dynamics, since it controls cell migration and anchorage-independent growth (253). Moreover, FAK deletion in mouse lungs has been known to disrupt the barrier integrity by RhoA destabilization (204). However, in a rabbit model of atherosclerosis, genetic or pharmacological FAK inhibition in endothelial cells prevented the VEGF-stimulated permeability downstream of the VEGFR or Src tyrosine kinase activation (40). A similar effect, namely, disruption of the inflammatory RhoA/MLC2 pathway, can be exerted by a PAK-dependent mechanism in mouse lungs and coronary vessels (199, 237, 256). PAK was found to be activated by AMPK induction in human pulmonary artery endothelial cells (260). LIMKs are downstream of Rho-family GTPases, PAK and ROCK, so that PAK1 and ROCK phosphorylate LIMKs, increasing LIMK activity. In HeLa and COS-7 cells, LIMK inhibits the actin depolymerization activity of cofilin by phosphorylating the Ser³ residue of cofilin (51). Phosphorylation of cofilin at Tyr⁶⁸ by Src leads to its degradation through the ubiquitin-proteasome pathway via a redundant pathway (266). Indeed, Rho kinase is reported to possess an intrinsic barrier-protective activity at the cell margins, apart from the well-established barrier-disruptive activity at contractile F-actin stress fibers (235).

Cell shape is a result of dynamic interactions among the cytoskeleton, the cell membrane, the adhesion complexes, and the environment. Certain shapes are required for particular cell behaviors (198). Motility is generated by formation of filopodial and lamellipodial protrusions at the leading edge (LE); at the trailing edge (TE), a propulsive force is induced by high contractility and disassembly of adhesion (126). Rho-family GTPases control cytoskeletal organization to regulate cell shape (108) and then promote localized changes in cell morphology and coordination of shape changes over the entire cell (192).

In cells migrating using cycles of protrusion, adhesion, and retraction, activation of RhoA promotes adhesion disassembly at the TE as well as cell-wide tension that is critical to restriction of protrusion at the LE (113). In some migrating cells, RhoA is activated at the LE together with cell advancement and is spatially segregated from a proximally localized wave of Rac1 and Cdc42 activity (139, 196). Compartmentalized regulation of Rho-family GTPase activity is critical for generation of particular shapes, since it has been shown that migration is inhibited when Rac activation is uniformly distributed throughout the cell (106, 219).

Activation of the Src, MLC, PKC, MAPK, PKC, ROCK, cGMP, and FAK kinases results in increased endothelial barrier dysfunction. On the other hand, activation of PKA, cAMP, TESKs, and LIMKs is associated with strengthening of endothelial barrier integrity. These effects vary among different cell types. The manner by which kinases operate in the vasculature and regulate endothelial permeability to proteins is further complicated, since these short-lived molecules orchestrate multiple interactions and mediate the cross talk among both them and downstream effectors. Moreover, the spectrum of this complexity becomes even greater, if we consider the lack of consensus over the exact subcellular localization of many of these molecules during inflammatory insults or cell movement (32).

Conclusions

Control of pulmonary endothelial barrier function is governed by complex, interdigitating, and occasionally redundant pathways that reflect the crucial importance of the endothelial barrier in normal homeostasis and the critical role of endothelial barrier dysfunction in numerous pathologies. Kinases represent major players within these pathways and exert profound pathophysiological influences on the function of the vascular barrier. Overexpression of many of these regulatory enzymes is induced by inflammatory stimuli and severely affects normal vascular function, leading to physiological abnormalities. An emerging body of studies suggest that inhibition of the activity of certain of these proteins may have important therapeutic effects in clinical practice.

GRANTS

This research was supported by National Heart, Lung, and Blood Institute Grant HL-101902.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

N.B. prepared the figure; N.B. and A.D.V. drafted the manuscript; N.B., A.D.V., and J.D.C. edited and revised the manuscript; N.B., A.D.V., and J.D.C. approved the final version of the manuscript.