Reactive Oxygen Species in Inflammation and Tissue Injury

NADPH oxidases were first identified in phagocytes for their role in inducing respiratory burst and bacterial killing (19, 376). So far, there are seven described homologs of NADPH oxidase (NOX1–NOX5 and Duox1 and 2) (Fig. 2). NADPH oxidase homologs differ in their structure, expression levels in different tissue, and their activation mechanisms (227, 233). The catalytic core of classical phagocytic NADPH oxidase 2 (NOX2) consists of two membrane-bound subunits, gp91^phox and p22^phox, which form the flavocytochrome b558 complex. The gp91^phox subunit consists of six transmembrane domains, and its C-terminal region contains the binding sites for flavin-adenine dinucleotide (FAD) and electron donor NADPH. All NADPH oxidases, with the exception of NOX5 and Duox1 and 2, share a similar topological structure of the catalytic core of gp91^phox. NOX5 carries an additional intracellular N-terminal calcium-binding domain. However, Duox1 and 2, in addition to the catalytic NOX5 structure, carry another N-terminus transmembrane α-helix, which possesses a peroxidase homology domain (Fig. 2) (233). NOX1, NOX2, and NOX4 are the major isoforms of NADPH oxidase that are expressed in the vascular system, and each is strongly implicated in inflammation-induced vascular injury (340).

O₂ ^•− production by gp91^phox (current designation NOX2) in leukocytes is essential for killing engulfed microbes within phagolysosomes, the amalgam of phagosomes and lysosomes (155). There have been at least 410 mutations identified in different constituents of the oxidase complex, gp91^phox, p22^phox, p67^phox, p47^phox, or Rac2. These mutations produce the genetic disorder chronic granulomatous disease, an immunodeficiency that is characterized by greatly increased susceptibility to bacterial and fungal infections due to defective oxidase assembly and production of O₂ ^{− •} (177). O₂ ^{− •}generated by NOX2 is induced by the translocation of cytosolic subunits p47^phox, p40^phox, p67^phox, and Rac1 to the plasma membrane and the formation of a complex with cytochrome b558 (Fig. 2a). The assembly of cytosolic subunits with membrane-bound cyt b558 complex results in the transfer of electrons from cellular NADPH to molecular oxygen and the formation of O₂ ^− •. Once activated, neutrophils produce ∼10 nmol/min O₂ ^{− •} per million neutrophils during the oxidative burst (19). The phosphorylation of p47^phox and the activation of Rac1 plays an essential role in the assembly of the oxidase complex. Pro-inflammatory cytokines (TNF-α, GM-CSF, and G-CSF), LPS, phorbol-12-myristate-13-acetate (PMA), and N-formylmethionyl leucyl phenylalanine (fMLP) are well-known kinase inducers of p47^phox phosphorylation (81, 83, 88, 93, 241). p47^phox is phosphorylated at multiple serine residues at C-terminus and is targeted by multiple pathways, such as protein kinase C-δ, -β and -ζ, protein kinase A, Akt, ERK1/2, and p38 mitogen-activated protein kinase (62, 82, 83, 87, 94, 106). The phosphorylation of p47^phox relieves it from auto-inhibitory conformation, which subsequently leads to unmasking of N-terminal SH3 domain and enabling binding to the proline-rich target in the C-terminus of p22^phox. Phosphorylation of p47^phox thereby promotes the recruitment of p67^phox by serving as an adapter and bridging p67^phox with the cyt b558 complex (80).

p47^phox has been documented to be important in the progression of atherosclerosis and pulmonary fibrosis (26, 268). p47^phox null mice were protected against formation of atherosclerotic lesions and development of pulmonary fibrosis (26, 268). Mice deficient in p47^phox subunit or NOX2 were also protected against TNF-α-induced lung inflammation or sepsis-induced lung microvascular injury (126 469). However, there are also contradictory findings. Zhang et al. (468) reported no difference in LPS-induced acute inflammatory responses in NOX2^−/− and p47^phox−/− mice compared with the control mice. On the contrary, they observed enhanced gene expression of inflammatory mediators and increased neutrophil recruitment to the lung and heart, resulting in impaired resolution. This discrepancy in the results was attributed to LPS-mediated ROS generation via NADPH oxidases, which also instead contributed to the resolution of inflammatory response (468). The NOX2 expressed in immune cells such as T cells has also been implicated in angiotensin II (Ang II)-induced hypertension (158). Guzik et al. have shown on the basis of adoptive transfer experiments that T cells lacking the AT1-receptor or p47^phox subunit resulted in decreased aortic O₂ ^•− production and blunted Ang II-dependent hypertension (158). Moreover, Ang II infusion induced the expression of NOX2, p47^phox, and p22^phox in the T cells and stimulated superoxide production (158). In a similar study, Wenzel et al. have shown that Ang II infusion stimulated the accumulation of both macrophages and neutrophils in the mouse aorta (450). The depletion of myelomonocytic cells by inducible diphtheria toxin receptor (LysM^iDTR mice) attenuated Ang II-induced blood pressure increase and reduced vascular O₂ ^•− formation (450). Moreover, adoptive transfer of wild-type monocytes into depleted LysM^iDTR mice re-established the Ang II-induced oxidative stress, and arterial hypertension, whereas adoptive transfer of neutrophils or monocytes lacking NOX2 did not (129). The NOX2 activity has also been implicated in cardiac inflammation and fibrosis. The mice deficient in NOX2 were protected against cardiac remodeling and contractile dysfunction induced by coronary artery ligation (252), aortic banding (153), Ang II infusion (36), or Doxorubicin treatment (471) compared with the wild-type mice. Moreover, in aortic-banding model, the treatment of wild-type mice with N-acetylcysteine resulted in recovery of contractile dysfunction (153). Altogether, these findings highlight the important role of NOX2-derived ROS in cardiac remodeling, which could be prevented by antioxidant treatment.

Rac1 is another important cytosolic subunit that is required for activation of NOX2. Rac1 activation is accompanied by the replacement of GDP residue with GTP by guanine nucleotide exchange factor (GEF), resulting in conformational change of Rac protein by relieving inhibition from Rho GDP-dissociation inhibitor. Rac1 activation promotes binding of p67^phox with cyt b558 complex (79). Rac1 activation is induced by a variety of inflammatory stimuli such as TNF-α (453), interleukin-1β (IL-1β), thrombin, VEGF (299), histamine (321), Ang II (384), ischemia-reperfusion injury (216), and shear stress (463). The expression of the active Rac1 mutant V12Rac1 resulted in enhanced ROS production and loss of endothelial barrier integrity (427). Rac1 activation has also been demonstrated in trafficking of inflammatory cells across the endothelial barrier (428). Cross-linking of vascular cell adhesion molecule-1 (VCAM-1) on endothelial cell surface induced Rac1 activation and ROS generation, and resulted in loss of endothelial cell–cell adhesion and transendothelial leukocyte migration (428).

In contrast to NOX2, NOX1 is activated by homologues of p47^phox and p67^phox, known as NOXO1 (NOX organizer 1) and NOXA1 proteins (NOX activator 1), respectively (23). NOXO1 does not contain an autoinhibitory region and is constitutively active in the cells (410). Co-expression of NOXO1 and NOXA1 with NOX1 in HEK 293 cells was sufficient to generate ROS independent of any stimulus (23, 410). High expression of NOX1 was observed in colon epithelium, and LPS induced the expression of NOX1 and NOXO1 in gastric mucosal cells (208, 410). NOX1 activation was dependent on Rac1 activation, and Rac inhibitor LY-294002 (PI3-K inhibitor) blocked O₂ ^{− •} generation (208). Mice deficient in NOX1 were protected against hyperoxia-induced acute lung injury (54). Surprisingly, NOX2^−/− mice were not protected against hyperoxia-induced lung injury in the same study (54), suggesting that hyperoxia may induce lung injury secondary to NOX1 activation, or NOX1 may compensate for the loss of NOX2. In addition, NOX1-deficient mice were less susceptible to Ang II-induced aortic dissection and aneurysm formation (131). NOX1 was also important in angiogenesis and tumor formation. Mice deficient in NOX1 exhibited impaired angiogenesis and tumor growth (129). Activity of NOX1 was increased in endothelial cells on angiogenic stimulation. However, angiogenesis remained unaffected in NOX2 or NOX4 knockout mice in this study.

NOX4 was first characterized in the kidney as an “oxygen sensor” regulating oxygen-dependent expression of erythropoietin and development of inflammatory processes in the kidney (132). NOX4 has also been termed renal NADPH oxidase (Renox). Activity of NOX4 is largely controlled by expression levels of NOX4 and is independent of the Rac activation or the presence of p47^phox/p67^phox or NOXO1/NOXA1 proteins (12, 132, 275). However, NOX4 activity required the presence of p22^phox (275). The type of ROS generated by NOX4-expressing cells is primarily H₂O_2, whereas O₂ ^•− was almost undetectable in these cells (383). Recently, another mechanism of NOX4 activation has been reported. Lyle et al. have shown that polymerase (DNA-directed) delta-interacting protein 2 (Poldip2) associates with p22^phox to activate NOX4, which regulates cytoskeletal reorganization and cellular migration in an Rho A-dependent manner (257). NOX4 is the predominant homolog expressed in human lung microvascular and human pulmonary arterial endothelial cells compared with NOX1, NOX2, NOX3, or NOX5 (338). Inflammatory stimuli LPS, TNF α, hyperoxia, TGF-β, and hypoxia were demonstrated to enhance ROS generation via NOX4 (29, 191, 242, 295, 335, 338). The expression of intercellular adhesion molecule-1 (ICAM-1), IL-8, and MCP-1 in endothelial cells in response to LPS was demonstrated to be dependent on NOX4 activation (335). Treatment of endothelial cells with NOX4 siRNA decreased LPS induced migration and adhesion of monocytes by 36% and 52%, respectively (335). Increased NOX4 activity is also a reported risk factor in the progression of type-2 diabetic nephropathy, stroke, pulmonary fibrosis, and atherosclerotic lesions (11, 145, 219, 240, 331, 382). However, the contradictory reports in the literature suggest that NOX4 is a vascular protective enzyme rather than a destructive one (415). NOX4 overexpression in the endothelium was found to enhance vasodilation and reduced basal blood pressure, which was related to increased H₂O₂ production and reduced NO inactivation (415). Moreover, in tamoxifen-inducible NOX4 knockout mice, Ang II-mediated aortic inflammation, vascular remodeling, and endothelial dysfunction were exaggerated, which correlated to a decrease in endothelial nitric oxide synthase (eNOS) expression (415). In addition, NOX4 overexpression was found to increase eNOS protein and promote recovery of ischemic tissue by enhanced angiogenesis and aortic sprouting (415). It is not easy to reconcile all these contradictory reports, and the part of the reason behind these discrepancies may be related to the tissue-specific abundance of different NADPH oxidase homologs and differential regulation of NOX4 in different models.

The expression of NOX5 has been reported in human vascular endothelial cells and smooth muscle cells. NOX5 is not expressed in rodents. Four variants of NOX5 have been identified so far, including NOX5α, NOX5β, NOX5γ, and NOX5δ (34). ROS generation by NOX5 is induced by calcium binding on its N-terminal domain, which has four Ca²⁺-binding sites (referred to as EF hands). The binding of Ca²⁺ to NOX5 induced conformational change that enhances ROS generation by NOX5. Unlike other NOX homologs, O₂ ^•− production by NOX5 does not require p22^phox. NOX5 has been reported to be induced by thrombin (34), Ang II (300), and platelet-derived growth factor (197). An increased expression level of NOX5 has been implicated in the coronary artery disease patients, which correlates with the oxidative damage observed in atherosclerosis (157). NOX5 has also been reported to be essential for PDGF-stimulated proliferation of vascular smooth muscle cells and proliferation of endothelial cells and angiogenesis (34, 197). Thus, while NOX5 may be important, it is still relatively understudied.

Mitochondria generate high-energy phosphate bonds of ATP by electrochemical proton gradient created by the transfer of electrons through a series of electron carriers embedded in the mitochondrial membrane. There are four electron transport carriers that are spatially organized in order of their increasing redox potential, complex I (NADH-ubiquinone oxidoreductase), complex II (succinate-ubiquinone oxidoreductase), complex III (ubiquinol-cytochrome c reductase), and complex IV (cytochrome c oxidase) (Fig. 3) (163). Transfer of electrons to molecular O₂ is a tightly controlled process, and only 1%–2% of electrons that leaked out in this process react with O₂, resulting in O₂ ^•− (163). The main sites of O₂ ^•− production in the ETC are complex I and III, with complex I being more predominant in skeletal muscle and neural cells; whereas complex III is more predominant for endothelial cells (Fig. 3) (323, 416). O₂ ^•− generated by mitochondria reacts with manganese SOD (MnSOD) in the mitochondrial matrix to generate H₂O₂, which can cross the mitochondrial outer membrane to access cytosolic targets. This can lead to multiple functional outcomes such as activation of redox-sensitive transcription factors (such as HIF-1α and NF-κB) (60, 269, 435), activation of pro-inflammatory cytokines, and activation of inflammasomes (185, 308).

The genetic mutations in mitochondrial respiratory chain can result in a variety of neurological disorders, including Leigh's syndrome, leukodystrophy, paragangliomas, and pheochromocytomas (97). Mitochondrial-derived oxidative stress has also been implicated in chronic inflammation, cancer progression (206), diabetes mellitus (148, 152, 317 –319, 378), and atherosclerosis (21, 258). Mitochondrial-derived ROS (MtROS) also contributes to LPS-mediated production of pro-inflammatory cytokines IL-1β, IL-6, and TNF-α (48). Notably, MtROS have been implicated in ectodomain shedding of cytokine receptor TNF receptor–1 (TNFR1) in endothelial cells, which is important for regulation of inflammatory progression (364). TNF-α-converting enzyme, which mediates cleavage of TNFR1, is activated by ROS (440). Ectodomain shedding of cytokine receptors facilitates diffusion of soluble receptors into the extracellular space, which dampens the inflammatory response by binding and neutralizing the cytokine ligands. Missense mutations in the extracellular domain of the gene encoding TNFR1 lead to inheritable autosomal dominant disorder known as tumor necrosis factor receptor-associated periodic syndrome (TRAPS) (48). TRAPS is an auto-inflammatory disease that is associated with recurrent fevers, peritonitis, migratory rash, myalgia, and arthralgia. Monocytes and neutrophils isolated from TRAPS patients showed elevated baseline levels of mitochondrial ROS compared with healthy donors and also had constitutive and prolonged activation of JNK and p38 MAPK on LPS challenge (48). Consequently, these cells showed enhanced production of cytokines IL-6 and TNF-α on LPS challenge. Inhibition of mitochondrial ROS with MitoQ suppressed p38MAPK activation and production of IL-6 and TNF-α (48). Although the basis by which TNFR1 mutations leads to increased mitochondrial ROS is not clear, possible mechanisms include retention of the mutated protein in the endoplasmic reticulum, which can induce unfolded protein response, triggering calcium release from ER and depolarizing mitochondria by disruption of the mitochondrial ETC (467).

Another important role of MtROS has been recognized in the regulation of inflammasome, the high-molecular-weight complexes that activate inflammatory caspases (caspase-1 and -12) and cytokines (IL-1β and IL-18) in macrophages (274). Three different prototypes of inflammasomes have been recognized: NALP1, NALP3 (also known as NLRP3), and IPAF. NLRP3 inflammasome, the most well-characterized form, is redox sensitive (99, 379). The key components of NALP3 are NLRP3, apoptosis-associated speck-like protein (ASC), and caspase-1. NLRP3 is a cytoplasmic receptor that interacts with ASC and recruits procaspase-1 (379). The NLRP3 inflammasome has been shown to interact with redox-sensitive protein thioredoxin (Trx)-binding protein-2 (TBP-2, also known as vitamin D3 up-regulated protein 1 [VDUP1]). Increased intracellular ROS generation mediates dissociation of TBP-2 from Trx, enabling association with NLRP3 inflammasome and resulting in its activation (Fig. 4) (473). Mice deficient in NLRP3 or ASC on LPS treatment have reduced serum levels of IL-1β and IL-18 and are resistant to LPS-induced lethality (403). ASC-deficient mice were also protected against LPS challenge (403). On activation by ROS, NLRP3 recruited ASC and procaspase-1, which mediated proteolytic cleavage, leading to activation of caspase-1 and IL-1β and IL-18 (473). The MtROS is important in the activation of NLRP3 inflammasome. Inhibition of mitochondrial function by depleting VDAC with shRNA impaired the activation of NLRP3 and cleavage of IL-1β and caspase-1 (474). In addition to direct activation by ROS, the activity of NLRP3 inflammasome is negatively regulated by autophagy/mitophagy (310, 367). Autophagy/mitophagy indirectly regulates intracellular oxidative stress by clearance of dysfunctional mitochondria and damaged proteins (170). Macrophages isolated from mice deficient in the autophagy proteins LC3B and ATG16L1 have enhanced NLRP3 inflammasome activation and produced higher IL-1β and IL-18 on LPS challenge (310, 367). MtROS has also been implicated in increasing oxidative stress through cross-talk with nitric oxide synthases (NOSs) (293). Arginase II, which activates NOS, was shown to promote the macrophage inflammatory response that contributes to insulin resistance and atherogenesis (293).

Recent reports have highlighted an interesting concept of cross-talk between MtROS and NADPH oxidase, which can drive both feed-forward and feedback regulations of NADPH oxidases (73, 96). Ang II-stimulated mitochondrial H₂O₂ was blocked by NADPH oxidase inhibitor apocynin and protein kinase C inhibitor chelerythrine (96). Moreover, depletion of p22^phox with siRNA also inhibited Ang II-mediated MtROS production (100). In contrast, ROS produced by the mitochondria on serum withdrawal has been shown to trigger NOX1 activation (96). Moreover, Ang II-mediated mitochondrial dysfunction has been shown to increase the expression and activity of NOX1 and NOX4 in vascular smooth muscle cells (96). Block et al. have reported the existence of NOX4 in the mitochondria of rat kidney cortex that was induced in rat model of diabetes (42). In an independent study, Ago et al. have similarly reported the localization of NOX4 in cardiac mitochondria with F₁F₀-ATP synthase as well as the p22^phox subunit of NADPH oxidases (3). However, no such localization of NOX4 or any other NADPH oxidase was observed in mitochondria in other reports (100), whereas Gorlach group localized NOX4 to endoplasmic reticulum in endothelial cells and was shown to be essential for endothelial ROS production and proliferation (346). Thus, further investigations are needed to identify the localization of NOX4 and the potential cross-talk between mitochondria and NADPH oxidase. A dysregulation in this relationship may drive the vicious feed-forward cycle of ROS accumulation that can enhance inflammatory response in different diseases (96, 380).

NOSs are a family of enzymes that catalyze the production of nitric oxide (NO) from L-arginine. There are three different isoforms of NOS, including neuronal NOS (nNOS), inducible NOS (iNOS), and eNOS. Of these isoforms, only eNOS is the membrane-associated protein and is the predominant source of NO in vascular endothelial cells (121). iNOS and nNOS are soluble isoforms and are present in the cytosol. eNOS exists as a dimer consisting of a c-terminus reductase domain of one monomer connected with the oxygenase domain of the other monomer at N-terminus (Fig. 5a). The reductase domain binds to NADPH, FMN, and FAD, and the oxygenase domain binds to prosthetic heme group as well as the cofactor tetrahydrobiopterin (BH₄) and molecular oxygen. The prosthetic heme group connects the two monomers. The electrons are transferred from the bound NADPH in the reductase domain to the heme prosthetic group on oxygenase domain of eNOS. NO is produced by eNOS in two successive mono-oxygenation reactions of arginine, leading to formation of L-citruline (Fig. 5a) (121). The electron transfer in eNOS is a tightly regulated process; however, eNOS uncoupling may lead to transfer of electrons to molecular oxygen rather than arginine, resulting in O₂ ^•− production (121). The superoxide generated by eNOS uncoupling has been implicated in a variety of inflammatory conditions, including acute lung injury (385), diabetes mellitus (178), and Ang II-induced hypertension (298). NO possesses strong anti-inflammatory properties. Dal et al. have shown that iNOS^−/− mice exhibit enhanced neutrophil migration compared with the wild-type mice on LPS challenge (75). Moreover, the mice that have been treated with NOS inhibitors such as NG-nitro-l-arginine, or with a soluble guanylate cyclase inhibitor, ODQ exhibited enhanced LPS-induced neutrophil migration that was accompanied by enhanced expression of ICAM-1 on endothelial cells (74). The anti-inflammatory effects of NO are mediated by suppressing LPS-induced increase in ICAM-1 expression and decreasing the rolling and adhesion of the neutrophils on the endothelium (74).

The cofactor BH₄ plays a crucial role in maintaining the integrity of the eNOS system by stabilizing eNOS dimers (10). BH4 donates the second electron to the first mono-oxygenation reaction of L-arginine, generating N-hydroxy-arginine as the intermediate The deficiency of BH4 leads to eNOS uncoupling, as the intermediate reacts with molecular oxygen, resulting in O₂ ^•− production (Fig. 5b) (10). The reduced bio-availability of BH4 has been reported to be due to oxidation by peroxynitrite, resulting in the formation of inactive BH2 (237). The oxidative inactivation of BH4 is sufficient to produce eNOS uncoupling. The mice deficient in eNOS did not exhibit an increase in vascular superoxide production when compared with the wild-type mice in the mouse model of desoxycorticosterone acetate (DOCA)-salt hypertension, suggesting that eNOS uncoupling is the major source of endothelial-generated O₂ ^•− (235). In addition, p47^phox knockout mice were relatively protected from BH4 oxidation and eNOS uncoupling, suggesting that NADPH oxidase-mediated O₂ ^•− production is an important contributor to oxidative loss of BH4 and eNOS uncoupling (330). Interestingly, the administration of BH4 in eNOS over-expressing mice was found to significantly decrease O₂ ^•− production and reduce the formation of atherosclerotic plaques by preventing eNOS uncoupling (330). Thus, the stoichiometric ration of BH4 to eNOS is critical in eNOS uncoupling.

In addition to BH4, increased levels of asymmetrical dimethyl arginines (ADMA), an endogenous inhibitor of eNOS, reportedly lead to eNOS uncoupling and enhanced O₂ ^•− production (404). Increased levels of ADMA have been reported in endothelial cultures treated with low-density lipoprotein, mouse lung exposed to endotoxin LPS, and septic shock patients (44, 322, 385). ADMA are generated by S-adenosylmethionine–dependent protein arginine methyltransferases that themselves are induced by enhanced oxidative stress (404).

Xanthine oxidoreductase is a soluble high-molecular-weight (270 kD) enzyme that catalyzes the oxidation of hypoxanthine to xanthine and uric acid. It exists in two forms: xanthine dehydrogenase (XDH) and XO (332). XDH reduces NAD⁺, whereas the XO is the main superoxide-producing enzyme. An irreversible proteolytic conversion of XDH to XO can be triggered by ischemic conditions (282). The activity of XDH/XO is specifically induced by IFN-γ in lung microvascular endothelial cells and the pulmonary artery endothelial cells. In contrast, IFN-α/β, TNF-α, IL-1 or -6, LPS, and PMA have no effect on the activity of XDH/XO (104). The activity of XO is up-regulated in the airway inflammatory disorders, ischemia reperfusion injury, atherosclerosis, diabetes, and autoimmune disorders such as rheumatoid arthritis (332). The serum samples from rheumatic patients may have approximately 50-fold higher levels of XO compared with healthy donors (289). Moreover, in diabetic rats, the administration of XO inhibitor allopurinol suppressed the NF-κB activation, neutrophil infiltration, and expression of inflammatory cytokines in the liver, suggesting a key role of XO-derived ROS in type I diabetes (363). XO-derived ROS has also been implicated in pulmonary vascular remodeling induced by chronic hypoxia in neonatal rats (196). The serum activity of XO is up-regulated by hypoxia, and administration of allopurinol limited the hypoxia-induced oxidative stress in the lung and pulmonary vascular remodeling; highlighting the critical role of XO-derived ROS in the chronic hypoxia induced pulmonary hypertension (196). Altogether, the XO-derived ROS seems to be important in a variety of inflammatory disorders; however, the therapeutic advantage of XO inhibitors such as allopurinol for other inflammatory diseases and other forms of organ injury remains to be investigated.

To prevent the damaging effects of oxidants, vertebrate cells have evolved an array of antioxidant defense systems that functions to remove ROS. The antioxidant enzymes SOD (dismutates O₂ ^•− to H₂O₂), catalase, glutathione peroxidase (GPx) (converts H₂O₂ to H₂O), peroxidredoxins, and Trxs are classified as ROS scavengers (Fig. 1) (163). Thus, cells experience an oxidative stress when the capacity of antioxidant enzymes is overcome by enhanced oxidant production. There are three different isoforms of SOD expressed in mammalian cells: SOD1, also known as CuZn SOD, a 32 kD homodimer expressed in the cytosol and nucleus; SOD2 or MnSOD, a 93 kD homotetramer expressed in mitochondrial matrix and extracellular (EC-SOD); and a plasma membrane-associated enzyme, a tetramer with a molecular weight of 135 kD, which has four Cu atoms per molecule. The EC-SOD isoform, although plasma-membrane associated, is a secreted glycoprotein that scavenges extracellular O₂ ^•− (163). All SODs are highly expressed in lung tissue, vessels, and airways (272). However, when their relative distribution is compared, the activities of CuZnSOD and MnSOD are lower in the lung compared with the other organs such as the liver, kidney, heart, and brain; whereas the EC-SOD activity has been reported to be much higher in the lung (217). The knockout mice deficient in CuZn SOD and EC-SOD are viable and show no abnormality associated with oxidative damage, indicating that other antioxidant enzymes compensate for their deficiency under normal physiological conditions (51, 475). However, the knockout mice deficient in MnSOD die in the neonatal stage from dilated cardiomyopathy and impaired neural development (238, 245). The expression of CuZn SOD has been reported to be induced by shear stress and hyperoxia in cultured endothelial cells (188, 223); EC-SOD, in contrast, is induced by a variety of inflammatory cytokines, including TNFα, IFNγ, and IL4 in arterial smooth muscle cells (47, 399). Similar to EC-SOD, the expression and activity of MnSOD is induced by TNF-α and IL-1 as well as by LPS, ROS, and VEGF (1, 201, 292, 431). However, SOD1 knockout mice showed no enhancement of LPS-induced hepatotoxicity or hyperoxia-induced lung injury (180, 475). In contrast, EC-SOD knockout mice exhibited enhanced sensitivity to LPS-induced neutrophilic lung inflammation and hyperoxia-induced lung injury (46, 51). All these observations suggest that tissue-specific distribution of the SOD isoforms and their relative amounts in different tissues is important in determining their role in inflammation.

Catalase is a cytoplasmic 240 kD homotetrameric protein and is an important intracellular antioxidant enzyme detoxifying the H₂O₂ to oxygen and water. The expression of catalase has been reported in alveolar type II cells and macrophages, and highest expression has been reported in the liver and erythrocytes (355). Arita et al. have shown that targeting of catalase directly to the mitochondria in lung epithelial cells protects them from H₂O₂-induced apoptosis (14). Despite this important physiological link, no increase in the activity of catalase was reported after hyperoxia in endothelial cells (202) and bronchial epithelial cells, which made them more susceptible to hyperoxia-induced injury (110). Moreover, the LPS treatment decreased the expression and activity of catalase in mouse lung, a response preceding NF-κB activation (67). The congenital deficiency of catalase known as acatalasia is benign; however, in certain conditions, increased susceptibility to diabetes has been reported (355). Whether catalase is important in the pathogenesis of inflammation remains an open question.

The family of GPx enzymes serves the similar function of detoxifying H₂O₂ as catalase. There are four selenium-dependent GPx enzymes (GPx1–4) in mammalian tissue with a wide tissue distribution (270). GPx enzymes are tetrameric 85 kDa protein and carry four atoms of selenium bound in their catalytic core. These enzymes detoxify H₂O₂ by oxidizing monomeric glutathione (GSH) into dimeric glutathione disulfide (GSSG). Oxidized GSSG is converted back to its monomeric GSH form by glutathione reductase. The expression and activity of GPx is induced by hyperoxia in endothelial cells (202). However, GPx knockout mice showed no hypersensitivity to hyperoxia-induced lung injury, indicating the compensation from other antioxidant enzymes (180). Reduced levels of GSH have been reported in a variety of inflammatory conditions (356), indicative of its important role in the inflammatory response through enhancing oxidative stress.

Peroxiredoxins are a group of related antioxidant enzymes that catalyze the degradation of H₂O₂ to water. There are six different (Prx1–6) identified so far with molecular weights ranging between 17—and 31 kD (190). All subtypes of Peroxiredoxins are expressed in lung tissue (190). The Prx6 knockout mice exhibit enhanced hypersensitivity to hyperoxia-induced lung injury (439), whereas Prx6 overexpressed mice were resistant to such damage (441). Prx1 knockout mice were also reported to be prone to bleomycin-induced lung inflammation (214) and allergic airway inflammation (187). Further, administration by N-acetyl-cysteine in Prx1 knockout mice was found to protect them against bleomycin-induced acute lung injury, indicating a protective role of Prx1 against oxidative damage in inflammation (214).

Trxs are 10–12 kDa redox-sensitive antioxidant enzymes that maintain the proteins in their reduced state by catalyzing the reduction of proteins disulfides to their corresponding sulfhydryls utilizing the reducing equivalents NADPH (311). Three Trx enzymes have been identified thus far (Trx1, 2, and 3). Oxidized Trx is recycled back to its reduced state by Trx reductase. The protective effects of Trxs as an antioxidant enzyme was reported in a variety of oxidative stress-related diseases such as ishchemia reperfusion injury, inflammation resulting from activation of neutrohils, blemomycin-induced lung injury, and inflammation induced by pro-inflammatory cytokines (311). There are two Trxs interacting proteins discovered so far; TBP-1 or p40^phox and TBP-2, also known as VDUP1 (311). The interaction of Trx with p40^phox subunit of NADPH indicates that it may also regulate the ROS generation via NADPH oxidases.

The induction of several antioxidant and cytoprotective genes is mainly regulated by the transcription factor NF-E2-related factor 2 (Nrf2), which is a cap'n’collar basic leucine zipper protein. The Kelch-like ECH-associated protein 1 (Keap1) retains Nrf2 in the cytoplasm and promotes its proteosomal degradation under basal condition (192, 193). In response to oxidative stress, Nrf2 dissociates from Keap1 and translocates into the nucleus, where it dimerizes mainly with the MAF (Maf-G, Maf-F, and Maf-K), JUN (c-Jun, Jun-B, and Jun-D), and ATF (ATF-4) families of bZIP proteins and transactivates a network of genes encoding cytoprotective and antioxidative enzymes containing the antioxidant response element (ARE), 5′-TGAG/CnnnGC-3′ in their promoters such as Gpx, NAD(P)H:quinone oxidoreductase (NQO1), and heme oxygenase-1 (HO1) (Fig. 6). The deficiency of Nrf2 enhances the susceptibility to experimental acute lung injury and impairs the resolution of lung inflammation in mice. Nrf2 activators such as triterpenoids (CDDO-Im) that target cysteine residues of Keap1 have been used to specifically disrupt Keap1:Nrf2 interactions, thereby promoting Nrf2 nuclear accumulation and leading to Nrf2-target gene induction (247) in Nrf2-sufficient but not in Nrf2-deficient cells in vitro and in vivo, suggesting that specific targeting of Nrf2-ARE signaling may provide a novel therapeutic strategy for treating human diseases. We and others have shown reduced levels of acute lung injury and inflammation (360) and emphysema in mice treated with CDDO-Im (402). Recently, triterpenoid analogue, bardoxolone methyl, showed improved renal function in early-stage chronic kidney disease in type 2 diabetes; however, Phase III clinical trial with this compound for very severe-stage patients was halted due to undisclosed safety issues (Reata Phamaceuticals) (ClinicalTrials.gov; NCT01351675). Moreover, the recent study by Zoja et al. has shown that bardoxolone analogues are ineffective in curing diabetic nephropathy in Zucker diabetic fatty rats but instead, the rats receiving such analogues worsen the outcome of the disease (476). While the CDDO-Im potently activates Nrf2 target genes in multiple tissues, proteomic analysis recently revealed that this CDDO-Im interacts with ∼600 different proteins, including many different transcription factors (464). Since Nrf2 confers protection against oxidation-related pathologies and Nrf-2 activation may be a useful antioxidant strategy, it is likely that the unwarranted effects of chronic CDDO treatment may be related to nonspecific off-target effects. Alternatively, it is possible that prolonged activation of Nrf2 signaling may be more complicated because of the possibility of activation compensatory pro-oxidative stress pathways.

They are membrane glycoproteins that are expressed in both endothelial cells and leukocytes and are of three different types: P- (expressed in platelets and endothelial cells), L- (expressed in all leukocytes), and E- selectins (expressed in endothelial cells) (27, 466). The initial rolling of neutrophils on endothelial cells in response to inflammatory stimuli is mediated by selectins (Fig. 7b). Selectins are characterized by the presence of a carbohydrate recognition domain known as lectin that allows a low-affinity binding to sialylated carbohydrate moieties of mucin-like CAMs on leukocytes. These residues, known as sialyl Lewis-x (SLex) antigen, play an important role in leukocyte tethering and rolling (466). Inherited defects in intracellular fucose transport impairs the formation of Lewis-x, resulting in leukocyte adhesion deficiency type II (LAD II) that is characterized by recurrent bacterial infection and impaired ability of neutrophils to bind endothelial cells expressing E-selectin (113, 255, 348). Neutrophil rolling on vascular endothelium is the result of sequential formation and the detachment of selectin-mediated bonds between neutrophils and endothelial cells (466). Selectin mediated rolling of leukocytes on endothelial cells greatly reduced velocity of leukocytes from hydrodynamic velocity of 200 to <5 μm/s (137, 466).

In endothelial cells and platelets, P-selectin (also known as granule membrane protein-140 [GMP-140]) is stored within Weibel–Palade bodies from which it is rapidly recruited to the cell surface during neutrophil migration; whereas E-selectin is up-regulated due to protein synthesis induced by activation of E-selectin gene in response to inflammatory stimuli such as IL-1β, LPS, or TNF-α (154, 166). P-selectin glycoprotein ligand-1 (PSGL-1) is constitutively expressed in neutrophils and T lymphocytes and binds to both P and E-selectin; however, affinity of E-selectin for PSGL-1 is 50 times lower compared with P-selectin (301). E-selectin binds to ligand E-selectin ligand-1 (ESL-1) that is expressed in neutrophils and plays a key role in neutrophil rolling (395, 466). The ligands for L-selectin include PSGL1, P-selectin, and E-selectin as well as Glycam-1, CD34, and MAdCAM-1, which are highly expressed in high endothelial venules (HEVs) and help in the homing of T lymphocytes in lymph nodes (466). On leukocyte activation, L-selectin is rapidly shed from the cell surface via a protease-dependent mechanism, which is important in the mechanism of transmigration and has also been implicated in mediating inflammation (159).

They are heterodimeric proteins, consisting of alpha and beta chains, and are expressed in leukocytes and many other cell types (114). They facilitate firm adherence to the vascular endothelium. β₂-integrins are the primary adhesion molecules that are expressed in leukocytes, and they bind to Ig superfamily member ICAM-1 which is expressed on the endothelial cell surface (114). β₂-integrins are involved in firm adhesion strengthening and transendothelial migration of leukocytes (114). In contrast to selectin-mediated adhesion, which leads to the rolling of neutrophils, adhesion of leukocytes by β₂ integrins is firm and is regarded an essential step for trans-endothelial migration of inflammatory cells. β₂-integrins are activated by chemokines such as CXCL1 (stored in small cytoplasmic granules) and CXCL8 (stored in Weibel–Palade bodies), which are secreted by endothelial cells (456, 465). The binding of chemokines to chemokine receptors (CXCR1 and CXCR2) changes the conformation of integrins from bent to fully extended high-affinity configuration, enabling neutrophil arrest by binding to ICAM-1 that is expressed in endothelial cells (Fig. 7b) (465).

β₂ integrins are grouped into the categories according to the β subunit they contain. The β₂ integrins include leukocyte function-associated antigen 1 (LFA-1 or α_Lβ₂) and Mac-1 (αMβ₂) and β₁ integrins such as very late antigen 4 (VLA4 or α₄β₁), α₉β₁, and α₅β₁ are common antigens that are expressed in leukocytes (114, 181). LFA-1 is the principal adhesion molecule in neutrophils that binds to ICAM-1 in endothelial cells and mediates transition from rolling to adhesion on the endothelial cell surface (350). Mac-1 also binds to ICAM-1 and mediates crawling of neutrophils along endothelial cells to the entry of endothelial junctions for paracellular migration (350). There is a 10-fold decrease in crawling displacement in Mac-1 null neutrophils, which forces them to transmigrate close to the initial site of adhesion and leads to transmigration (350). The activation of LFA-1 is induced by the activation of chemokine receptors on neutrophils by ligands that are secreted by endothelial cells such as CXCL1 (465). Fully activated LFA-1 mediates the stop of rolling and makes firm adhesion along with Mac-1 (175). Deficiency of β₂ integrin chain causes Leukocyte Adhesion Deficiency Type 1, an inherited disorder that is characterized by inability of leukocytes to undergo adhesion-dependent migration (49). These patients suffer from recurrent bacterial infections. β₂ integrin Mac-1 favors paracellular migration of leucocytes and in Mac-1^−/− (CD11b) mice, neutrophils predominantly emigrate transcellularly (through endothelial cells) and were delayed by 20–30 min compared with the more rapid paracellular emigration route (351). In contrast to β₂ integrins, the β₁ integrin family mediates neutrophil adhesion to extracellular matrix proteins such as fibronectin and VLA4 or α₄β₁ integrin. The coating of fibronectin in an in vitro migration assay can greatly enhance the migration of human PMNs (115). VLA4 or α₄β₁ integrin in murine neutrophils mediates leukocyte adhesion to endothelial cells by interacting with VCAM-1 (9). However, VLA4 is not expressed in human neutrophils, but its function instead is served by α₉β₁ integrin. α₉β₁ integrin is a major β₁ integrin that is induced on neutrophil activation which mediates binding to VCAM-1 (265). However, β₁-integrin-mediated adhesion has a limited role in adhesion and transendothelial migration of neutrophils compared with β₂ integrins (265). After β₂-integrin-mediated neutrophil arrest, there is a polarization of neutrophils at the leading edge referred to as lamellipodium that orchestrates receptors for chemokines and phagocytosis. Polarization of neutrophils, which is important for efficient transendothelial migration, is mediated by the organization of filamentous actin (F-actin) cytoskeleton at the leading edge (375, 461).

It consists of CAMs that have an Ig-like domain and are expressed in endothelial cells, platelets, and leukocytes. Included in this family are members of inter-cellular adhesion molecules (ICAM-1 and -2), vascular cell adhesion molecule (VCAM-1), platelet endothelial cell adhesion molecule (PECAM-1 or CD31), junctional adhesion molecules (JAM-A, -B and -C), CD99 antigen-like 2 (CD99L2), and endothelial cell-selective adhesion molecule (ESAM) (45). The expression of ICAM-1 and VCAM-1 is increased on activated endothelium (166). At the time of neutrophil transmigration, ICAM-1 and VCAM-1 are recruited in specialized preformed tetraspanin-enriched microdomains (TEMs) and promote nano-clustering of adhesion receptors to increase leukocyte binding to endothelium (25). These specialized TEM domains have been termed endothelial adhesive platforms and are formed independent of ligand binding and actin cytoskeleton anchorage (25). The binding of LFA-1 integrin on neutrophils with ICAM-1 on endothelial cells generate signals that activate Rho A, leading to actin polymerization and formation of transmigratory cups (423). Transmigratory cups are microvilli-like projections that surround leukocytes and help in efficient transendothelial migration of leukocytes (24, 52). The adaptor proteins of ERM family (ezrin, radixin, and moesin) link ICAM-1 and VCAM-1 with actin cytoskeleton and help in reorganization of actin cytoskeleton and formation of transmigratory cups (25). The deep penetration of neutrophils between endothelial cells is mediated by JAM-A (Fig. 7b), as neutrophils accumulate deeper down the junctions in JAM-A^−/− mice treated with IL-1β (458). In the resting endothelium, PECAM-1 is stored along the cell borders in the intracellular compartment known as lateral border recycling compartment (LBRC) and translocates to the site of diapedesis when leukocyte migration occurs (Fig. 7b). The pool of PECAM-1 in LBRC is approximately 30% of the total (266). Neutrophils in PECAM-1^−/− mice become impacted between endothelial cells and basement membrane in IL-1β-stimulated mice, indicating that PECAM-1 is essential for the neutrophil to traverse the junctions (458). Interestingly, the role of JAM-A, PECAM1, and ICAM-2 in neutrophil migration appears to be stimulus dependent. Neutrophil migration was suppressed in PECAM1, ICAM-2, or JAM-1 knockout mice on treatment of IL-1β and ischemia reperfusion injury; whereas TNF-α-induced neutrophil migration was unaffected in these knockout models (458). The mechanism behind stimulus-dependent specificity of these junctional proteins is not clear. CD99L2, also known as leukocyte antigen, is a small membrane protein of 100 amino acids that is expressed in endothelial cells. It is essential for neutrophil transmigration but not for lymphocytes (40). CD99 L2 is not relevant for the adhesion between leukocytes and endothelium but rather for the transmigration step. The blocking of antibodies against CD99L2 arrested the neutrophils between endothelial cells and basement membrane similar to the phenotype observed in PECAM1^−/− neutrophils (39). The expression of ESAM, another endothelial adhesion molecule, is restricted to endothelial tight junctions (TJs) and plays a role in Rho A activation and junctional opening. The migration of leukocytes was decreased to almost 50% in mice deficient in ESAM on treatment with TNF-α and IL-1β (444). Thus, there appears to be a great deal of redundancy in the role of these endothelial adhesion molecules in mediating leukocyte transmigration. One of the key questions is whether they are activated sequentially and whether they have distinct roles in the transmigration response. Only with dissection of the specific elements of the transmigration response into discrete steps will it become clear whether the function of these adhesion molecules is, in fact, distinct.

The postcapillary vascular endothelium of all the secondary lymphoid organs except spleen is constitutively enriched in CAMs. These regions were classified as HEVs, enabling circulation of naive lymphocytes between the blood and lymph node. Approximately 25% of all the lymphocytes circulating in HEVs bind and emigrate into a single lymph node (via HEVs) every second (134). In addition to commonly expressed CAMs in normal venules, HEVs abundantly express highly glycosylated and sulfated forms of adhesion molecules, including glycosylation-dependent cell-adhesion molecule 1 (GLYCAM1), CD34, MAdCAM-1, and peripheral LN addressin (PNAd), which are commonly referred to as vascular addressins, enabling rapid homing of lymphocytes in lymphoid organs (296). Certain chronic inflammatory conditions such as rheumatoid arthritis, psoriasis, Hashimoto's thyroditis, Crohn's disease, ulcerative colitis, and multiple sclerosis are accompanied by development of HEVs in normal organs, facilitating a large influx of lymphocytes that contribute to chronic inflammation (8, 231). This de novo formation of organized lymphoid tissue in chronic inflammation diseases has been termed lymphoid neogenesis. Lymphotoxin-α, a member of TNF family, has been implicated in the development of secondary lymphoid organs in these diseases (179). Psoriasis is a common cutaneous inflammatory disorder in which the dermal microvasculature undergoes distinctive changes, including HEV formation, which facilitates trafficking of lymphocytes to the skin (253). There is an interesting correlation between Ang II-mediated activation of the immune system and hypertension (158, 450) with the increased risk of hypertension and cardiovascular mortality in psoriasis patients that is associated with elevated levels of Ang II (253, 285).

There is considerable evidence indicating that extravasation of leukocytes in response to inflammatory stimuli is regulated by oxidative stress that is produced by leukocytes. The adhesion of neutrophils to the surface of endothelial cells has been demonstrated to elicit a biphasic response that is related to endogenous ROS production in endothelial cells (221). Oxidative stress can regulate the expression of endothelial CAMs by a direct activation of CAMs that presents on the surface and also by a transcription-dependent mechanism involving redox-sensitive transcription factors (i.e., NF-κB and AP-1). In this regard, Patel et al. (337) demonstrated that external application of H₂O₂ or t-butylhydroperoxide as early as 1 h in HUVECs was accompanied by transcriptional-independent surface translocation of P-selectin, leading to increased PMN adherence. Treatment of cells with an antibody against P-selectin or antioxidant treatment abolished neutrophil adhesion to endothelial cells. Antibodies directed against P-selectin abolished an O₂ ^•−mediated increase in rolling leukocytes and decreased the number of adherent leukocytes, suggesting an increase in surface expression of P-selectin by O₂ ^•−. Similar to this study, Gaboury et al. (124) demonstrated using the intra vital microscopy that the application of hypoxanthine/XO (O₂ ^•− generating system) in rat mesenteric artery caused a decrease in leukocyte-rolling velocity and increased the number of rolling and adherent leukocytes which could be blocked by antibodies against P-selectin. In addition to oxidative stress, inflammatory mediators such as thrombin, VEGF, IL-17, and TNF-α are known to enhance the surface expression of P-selectin (154, 254). Unlike P-selectin, expression of ICAM-1, VCAM-1, and E-selectin is regulated at transcription level in endothelial cells by oxidative stress, although there is also considerable evidence that a significant component of ICAM-1 up-regulation on the endothelial cell surface is due to phosphorylation of cell-surface-bound ICAM-1 at tyrosine residues (249).

The promoter region of ICAM-1 contains binding sites for inducible redox sensitive transcription factors such as NF-κB and AP-1 (239, 361). H₂O₂ increases the transcripts of ICAM-1 in HUVECs within 30 min, the response was sustained for approximately 2 h, and antioxidant treatment (catalase) or application of ICAM-1 antibody abrogated the H₂O₂-induced PMN adhesion (251). TNFα-induced NF-κB activation and ICAM-1 expression in endothelial cells is dependent on oxidants that are generated by the PMN NADPH oxidase complex (117). In p47^phox−/− mice, TNFα-induced NF-κB activation and ICAM-1 expression were significantly attenuated as compared with the wild-type mice (117, 432). Moreover, PMNs from p47^phox−/− mice showed markedly reduced adhesion to mouse lung vascular endothelial cells, indicating the essential role of oxidant signaling in NF-κB activation and ICAM-1 expression in endothelial cells (117). Oscillatory shear stress (OS) induces endothelial expression of ICAM-1, and monocyte adhesion is also dependent on ROS that is produced from NADPH oxidases (393). ECs obtained from p47phox^−/− mice showed impaired adhesion to monocytes in response to OS and were restored when the cells were transfected with p47^phox plasmid. Furthermore, OS increased mRNA levels of NOX2 and NOX1, and monocyte adhesion was shown to be blocked by siRNA against NOX1. Interestingly, Matheny et al. have shown that VCAM-1 ligand binding induced ROS generation via NADPH oxidases in endothelial cells and promoted lymphocyte migration by actin restructuring in endothelial cells (278). Inhibiting NADPH oxidase activity with DPI or Apocynin blocked lymphocytic migration across mouse endothelial cells to greater than 65% (278). Expression of VCAM-1 in HUVECs was up-regulated by cytokine IL-1β and enhanced ROS generation. Treatment with antioxidant pyrrolidinedithiocarbamate (PDTC) and N-acetyl cysteine (NAC) repressed greater than 90% of the IL-1β-induced increase in VCAM-1 expression (276). Increased expression of ICAM-1 and P-selectin was also observed in the mouse model of acute pancreatitis that was dependent on enhanced ROS production (412). In the mouse model of atherosclerosis, Ang II-induced expression of VCAM-1 was dependent on mitochondria-derived ROS (354). In addition, in patients suffering from obstructive sleep apnea, an increased expression of adhesion molecules CD15 (a counter-receptor for selectins on ECs), CD11c (counter-receptor for ICAM-1 on ECs) on monocytes was correlated to increased ROS generation (105, 381). These monocytes showed increased adherence in culture to human endothelial cells and generated more intracellular ROS compared with control subjects (105). Continuous positive airway pressure treatment of these patients decreased basal ROS production in monocytes and concomitantly decreased the expression of CD15 and CD11c on monocytes (105). The circulating levels of ICAM-1, VCAM-1, and E-Selectin are elevated in hyperlipidemic and hyperglycemic and noninsulin-dependent diabetic-patients in direct correlation with increased oxidative stress (58, 69, 86, 313, 352). Reducing oxidative stress by GSH administration or NAC or low-fat diet improved the circulating levels of adhesion molecules (5, 57, 85, 86). All of these reports indicate that increased expression of CAMs induced by oxidative stress can contribute to the pathogenesis of the inflammatory disorders.

TJs are composed of transmembrane proteins occludins, claudins, and JAMs. The transmembrane proteins of TJs are connected to actin cytoskeleton by cytoplasmic adaptor proteins (zonula occludens [ZO]-1, -2, and -3, AF6, and cingulin). TJs in endothelial cells represent ∼20% of all the junctional complexes (454). The tightness of TJs depends on the protein composition and varies in different tissues ranging from almost complete tightening of the paracellular cleft for solutes (e.g., as in the blood brain barrier) to the formation of paracellular pores for specific ions. All transmembrane proteins of TJs, including occludins, claudins, and JAMs, possess a PDZ (postsynaptic protein disc large ZO-1)-binding motif in the C-terminal that binds to cytoplasmic adaptor proteins containing PDZ domain. Occludin consists of four transmembrane domains and two extracellular loops that enable a homotypic interaction with other occludin molecules and contributes to the TJ assembly. Arterial endothelial cells that are less permeable compared with vein endothelial cells have almost 18-fold greater expression of occludin (211). The C-terminal of occludin is linked to actin cytoskelton by cytoplasmic adaptor protein ZO-1. The role of occludin in maintaining TJ stability is not clear. In vitro studies showed that disruption of occludin function using antisense oligonucleotide or peptide antagonists against occludin molecules decreased transendothelial electrical resistance, whereas occludin^−/− mice were still viable and showed no alternations in the structural and functional barrier properties of TJs (41, 211, 368). A compensatory up-regulation of other junctional proteins may account for the observed normal phenotype in these mice.

Claudins are transmembrane proteins of TJs that are similar to occludins and possess four transmembrane domains, two extracellular loops, and an N-terminal and C-terminal cytosolic domains. At least 24 claudins are identified in humans. Similar to occludin, the PDZ domain in the c-terminus of claudin interacts with ZO-1 and connects claudins with actin cytoskeleton. The composition of claudins varies in different endothelial barriers and, hence, determines whether the barrier is selectively permeable (e.g., the ion selective pores in the kidney) or impermeable (e.g., blood–brain barrier) (328). Pore-forming claudins such as claudin 2 (Cldn-2) form pores that are specific to monovalent cations and water and are found in Madin–Darby Canine Kidney (MDCK) cells. The claudins involved in the sealing of TJs include claudin 1, 3, 5, 11, and 19 (328). Of these claudins, claudin 5 is highly expressed in the blood–brain barrier, and claludin 5^−/− null mice die within 10 h after birth (320). Although in these mice the development and morphology of blood vessels were not altered and showed no brain edema, however, it selectively increased the permeability to small molecules with MW <800 D (320).

JAMs belong to Ig superfamily of proteins consisting of a single-pass membrane protein with a long extracellular domain. At least three different isoforms of JAMs are known, including JAM-A, JAM-B, and JAM-C (18). The expression pattern of the JAMs varies significantly in different cell types. JAM-A is more widely distributed in epithelial cells, endothelial cells, monocytes, and neutrophils. The expression of JAM-C is exclusively restricted to endothelial cells, whereas JAM-B is more abundant in HEVs of lymphatic organs (16 –18). The homophilic interaction between different JAMs contributes to the tightness of the TJs. Studies show that expression of JAM-A or JAM-C in CHO cells reduced the paracellular permeability to FITC dextran (17, 273). Moreover, antibodies directed against JAM-A significantly delayed the recovery of transepithelial electrical resistance during TJ reformation in epithelial cell monolayers (246). Orlova et al. (325) showed that JAM-C was mainly localized intracellularly in quiescent microvascular endothelial cells, and was recruited to junctions on stimulation with VEGF or histamine. Notably, disruption of JAM-C function decreased basal permeability and prevented the further increase in permeability induced by VEGF and histamine in human dermal microvascular endothelial cells in vitro and skin permeability in mice. JAMs are also known to interact with different integrins on adjacent cells. This heterophilic interaction regulated leukocyte–endothelial cell and leukocyte–platelet interactions (326, 329, 373).

ZOs belong to a membrane-associated guanylate kinase (GUK) homologue protein family. There are three ZO subtypes (ZO-1, -2, -3), and all of them possess three domains: PDZ, an Src homology 3 (SH3), and GUK (140, 305). ZO is an acronym for ZO. ZO-1 was the first TJ-specific protein identified (398). ZO1 acts as a scaffolding protein, as it interacts directly with claudins and also binds to the other Adherens junction (AJ) proteins such as α-catenin and to connexins in gap junctions (228). Through their PDZ, SH3, and GUK domains, ZOs are critical in recruiting signaling molecules to TJs, and thereby linking TJ proteins to the actin cytoskeleton (140).

AJs are composed of the cadherin family of transmembrane adhesion proteins that are anchored to the cytoskeletal network by intermediate cytoplasmic proteins from the catenin family, including α-, β-, γ-, and p120 catenins (31). There are two types of cadherins present in endothelial cells: VE-cadherin (vascular endothelial cadherin) and N-cadherin (N for neuronal) (315). Although both are abundantly expressed in endothelial cells, VE-cadherin mainly promotes the homotypic interaction between endothelial cells, and when present, it excludes N-cadherin from those sites (315). VE-cadherin is a homophilic binding protein that mediates the interaction of endothelial cells in a calcium-dependent manner. It has five cadherin-like repeats in its extracellular domain that oligomerize in a cis- and trans-manner with other VE-cadherin molecules between the same and adjacent cells (Fig. 8). The cytoplasmic tail of VE-cadherin contains two functional domains: juxtamembrane domain (JMD) and C-terminal domain (CTD). JMD of VE-cadherin binds p120-catenin, whereas CTD binds β-catenin and γ- catenin (also known as plakoglobin) (Fig. 8) (31). α-catenin by binding to β- and γ- catenin links the catenin-cadherin complex to the actin cytoskeleton. Both the N- and C-terminus of VE-cadherin are essential in the regulation of endothelial barrier function. Monoclonal antibodies targeted against N-terminus of VE-cadherin prevented the formation of IEJs and increased the microvascular permeability (70, 71). Interestingly, the C-terminal-truncated mutant of VE-cadherin in CHO cells did not impair the ability of cells to form aggregates; however, it greatly enhanced the intercellular junction permeability to high-molecular-weight molecules (56, 314). This finding suggests that the N-terminal region of VE-cadherin is required for cell–cell adhesion, whereas the C-terminal provides a tight restrictive barrier regulating paracellular permeability. VE-cadherin is essential for the development of endothelial barrier in embryos, and VE-cadherin knockout mice die in the embryonic stage due to abnormal vascular development (53). Moreover, expression of dominant negative VE-cadherin mutants also resulted in leaky junctions and an increase in microvascular permeability (430). Homotypic binding of VE-cadherin molecule between endothelial cells is stabilized by calcium, and chelation of calcium using EDTA destabilized the junctions and increased transendothelial permeability (125).

Catenins are equally important in regulating the stability of AJs by connecting them with actin cytoskelton. Deletion of β-catenin binding site on VE-cadherin caused lethality at 9.5 days of gestation (53). Similarly, conditional deletion of β-catenin in cultured endothelial cells reduced the ability of endothelial cells to maintain intercellular contacts and increased paracellular permeability (55). p120 catenin regulated the surface expression of VE-cadherin, and the depletion of p120 by siRNA decreased the VE-cadherin expression and increased the junctional permeability. In addition, p120 catenin acted as a scaffold protein by binding with two different tyrosine phosphatases, including Src homology 2 domain-containing tyrosine phosphatase-1 (SHP-1) and protein tyrosine phosphatases (PTPμ), which regulated VE-cadherin phosphorylation and stability of the junctions (210, 271, 477) (see next section).

Oxidative stress produced by leukocytes at the site of inflammation plays a crucial role in initiating junctional disassembly. At the site of inflammation, AJs are disrupted by several mediators that are released by inflammatory cells, including ROS, cytokines, chemokines, thrombin, histamine, PAF, VEGF, and bradykinins. Most of these molecular mechanisms converge on mediating disruption of AJs and TJs, leading to gap formation between cells. These pathways in relation to oxidative stress are described next.

Endothelial barrier integrity is formed by tight cell–cell and cell–matrix adhesions, and it is coupled to cytosolic Ca²⁺ levels (284). Increases in [Ca²⁺]_i induced formation of inter-endothelial cell gaps and vascular hyperpermeability through different signaling pathways (Fig. 10) (256, 264, 371). Several reports suggest that an increase in [Ca²⁺]_i leads to activation of Ca²⁺/calmodulin-dependent myosin light chain kinase (MLCK), which facilitates reorganization of actin cytoskeleton to induce changes in endothelial cell shape. Increasing cytosolic Ca²⁺ by thapsigargin (which depletes ER Ca²⁺ store) and calcium ionophore increased permeability in an HUVEC-derived cell line ECV304 (103, 127). Interestingly, thapsigargin induced activation of PKC-α, and phosphorylation of VE-cadherin-associated proteins was prevented by PKC inhibitor, calphostin C. These findings suggest that activation of PKC-α is critical in Ca²⁺-mediated increase in endothelial permeability (370, 371). An increase in [Ca²⁺]_i has also been associated with activation of small GTPase RhoA. The inflammatory mediator such as thrombin is known to induce receptor-mediated Ca²⁺ entry followed by RhoA activation, thereby resulting in increased endothelial permeability (390).

In addition, evidence indicates that ROS activates Ca²⁺ signaling and influences different cellular events which trigger inflammation. Chelation of extracellular Ca²⁺ by BAPTA suppressed H₂O₂ formation (312). An increase in intracellular Ca²⁺ was associated with enhanced H₂O₂ generation (101). The converse is also true. H₂O₂ caused a dose-dependent rise in [Ca²⁺]_i of endothelial cells (172). ROS-mediated regulation of intracellular free Ca²⁺ concentration is a major mechanism of increased vascular permeability, the hallmark of inflammation. However, to date, little is known about the mechanism of ROS-induced Ca²⁺ entry in endothelial cells and how this relationship promotes the loss of endothelial barrier integrity.

The principal Ca²⁺ entry pathway in endothelial cells is via store-operated calcium (SOC) channels and receptor-operated calcium (ROC) channels. Since transient receptor potential (TRP) channels are believed to underlie the molecular makeup of many ROC channels (TRPC3, 6 and 7) and at least some SOC channels (TRPC1 and 4) (333), it is logical to presume that TRPC proteins are sensitive to regulation by ROS. TRPC7 (also known as transient receptor potential melastatin 2 (TRPM2), TRPC6, and TRPC3/TRPC4 have been shown to be regulated by ROS in the endothelium (156, 408). In one study, Balzer et al. (22) provided strong evidence for a central role of TRP channels in oxidant-induced vascular pathophysiology. They showed that expression of a dominant negative fragment of TRPC3 suppressed oxidant-induced membrane currents in endothelial cells. To further analyze the role of TRPC3 as a redox sensor, Poteser et al. (353) stably transfected HEK293 cells with TRPC3, and these cell types displayed a large increase in membrane conductance in response to cholesterol oxidase. Generation of redox cation conductance by over-expression of TRPC3 suggests that TRPC3 is an important redox-sensitive cation channel. Using immunoprecipitation and Forster Resonance Energy Transfer (FRET), it was shown that TRPC4 interacts with TRPC3 to form heteromers, representing a potential redox-sensitive channel complex. The requirement of TRPC3/C4 heteromerization to serve as a sensor for cellular redox state was further evident by the finding that redox-activated TRPC3 cation conductance was inhibited by expression of dominant negative mutants of TRPC4 (353).

Recent studies have demonstrated that TRPC6 also represents a target of ROS in different disease models (98, 443, 446). Using lung ischemia–reperfusion edema (LIRE) model, Weissmann et al. (446) elegantly showed that NOX2-mediated O₂ ^•− activates TRPC6 and thereby induces Ca²⁺ influx in endothelial cells. They also found that NOX2^−/− and TRPC6^−/− mice were equally protected against ischemia-induced lung injury. Thus, TRPC6 also represents a target for ROS in different pathophysiological conditions.

TRPM2, of the melastatin subfamily, originally named TRPC7 (307), is no longer considered a member of the TRPC subfamily. The TRPM subfamily only exhibits limited homology to the other TRP family ion channels, and functional channels comprising members of this subfamily are believed to be homotetramers (277). CTD of TRPM2 share 39% sequence identity to the human nucleoside diphosphate-linked moiety X-type motif 9 (NUDT9) ADPR pyrophosphatase (173, 343, 372, 386) Despite significant homology with NUDT9 ADP pyrophosphate, TRPM2 has low levels of ADP-ribose (ADPR) pyrophosphatase activity (343, 372). Intracellular application of ADPR, the substrate of the NUDT9 enzyme, specifically induced TRPM2 gating (343). A further mode of TRPM2 activation is through oxidative stress (164, 392, 445). Different mechanisms have been proposed for H₂O₂-mediated TRPM2 gating. Perradud et al. (344) showed that oxidative stress activates production of ADP-ribose in mitochondria, which are then released to the cytosol and function as a second messenger to stimulate TRPM2 gating. In contrast, Hara et al. (164) demonstrated that TRPM2 is a Ca²⁺-permeable cation channel and can be activated by H₂O₂ via action of NAD⁺. Direct oxidation of TRPM2 by H₂O₂ has also been reported as a mechanism of activation of the channel (445).

Hecquet et al. (171) showed that H₂O₂ induced TRPM2-like cation current that was increased by over-expression of TRPM2 and by 3-deaza-cADP ribose but inhibited by a specific TRPM2 small interfering RNA and by PARP inhibitors in human pulmonary artery endothelial cells (Fig. 11). These data clearly favor an indirect mechanism of activation of TRPM2 channels by H₂O₂ via the formation of ADPR. Using Ca²⁺ add-back protocol designed to rule out indirect effects of H₂O₂ on Ca²⁺ entry via Ca²⁺ store depletion, Hecquet et al. also showed that H₂O₂ increased intracellular Ca²⁺ by stimulating Ca²⁺ entry without provoking Ca²⁺ store depletion. More importantly, this study showed the importance of TRPM2 in mediating the H₂O₂-induced increase in endothelial permeability, which could be prevented partly by application of TRPM2 small interfering RNA and TRPM2-specific antibody (171). The short variant of TRPM2, which lacks the pore domain and acts as a dominant-negative form by inhibiting the formation of functional homotetrameric channels, also significantly reduced H₂O₂-mediated microvascular hyper-permeability (171). In another article, Di et al. (95) addressed the role of TRPM2 in phagocytic cells and showed a protective role for TRPM2 in LPS-induced lung inflammation. This involved the influx of cations via TRPM2, which prevented membrane depolarization induced by transfer of electrons from NADPH to NADPH oxidase. Therefore, the net effect of TRPM2 activation was to reduce the production of ROS, which, in turn, modulated the lung inflammatory response to E. coli.

Endothelial cells contain contractile machinery, which includes actin and myosin. Phosphorylation of regulatory myosin light chain (MLC) leads to activation of the endothelial contractile elements and disrupts endothelial barrier function. (127). Permeability increasing mediators such as thrombin, histamine, and H₂O₂ enhance MLC phosphorylation, which precedes the onset of endothelial contraction (Fig. 10) (303, 387, 470). The myosin phosphorylation in the endothelial cells is regulated by two enzymes: nonmuscle MLCK and myosin-associated protein phosphatase. Activated MLCK directly phosphorylates MLC at Thr-18 and Ser-19 site, which induced isometric tension development and actin polymerization in endothelial cells (136). Zhao et al. (470) reported that H₂O₂ also induced MLC phosphorylation in endothelial cells. Though the inhibition of MLCK blocked H₂O₂-induced MLC phosphorylation, yet it did not affect H₂O₂-induced actin and myosin assembly. Therefore, MLC phosphorylation does not appear to regulate H₂O₂-induced actin and myosin assembly but may play a role in induced actin rearrangement. The role of MLCK was also studied in hyperoxia-induced lung injury model (422). This study demonstrated that NADPH oxidase-mediated ROS generation during hyperoxia was inhibited by silencing nmMLCK in lung endothelial cells. In addition, nmMLCK^−/− mice show decreased ROS production, phosphorylation of MLC, and pulmonary vascular leak. These results provide strong evidence for the involvement of nonmuscle MLCK in the assembly and activation of NADPH oxidase and ROS formation during hyperoxia.

Depletion of MLCK by using siRNA was also shown to reduce LPS-induced phosphorylation of mono and diphospho-MLC and restored barrier function (43). Inflammatory mediators are known to inactivate myosin-associated phosphatases, which further promoted MLC phosphorylation (111, 112). Phosphorylation of MLC phosphatase subunit (MYPT1) was increased by ROS followed by cell contraction (199). Taken together, ROS-mediated activation of MLC kinase contributes to actin reorganization in endothelial cells.

The Ras family of small GTP-binding proteins are activated by a variety of agonists, growth factors, TNF-α, G-proteins, and ROS and regulate a wide range of biological processes, including reorganization of the actin cytoskeleton, co-ordinated migration, proliferation, tumorigenesis, and cell barrier functions (2, 32, 147). This small GTPases superfamily is subdivided into five subfamilies: Rho, Ras, Rab, Ran, and Arf (394). These proteins act as switches by cycling between active (GTP-bound) and inactive (GDP-bound) conformations. The activation state of these small molecules is regulated by: GEFs, which promote exchange of GTP for GDP; GTPase-activating proteins (GAPs), which inactivate G proteins by enhancing the GTP-hydrolysis; and guanine-nucleotide-dissociation inhibitors, which bind to GDP-bound small proteins and regulate translocation between membranes and the cytosol (394). The active form of these small molecules interacts with their downstream targets and mediates their biological functions.

The mammalian Rho GTPase family consists of more than 20 distinct members, while the Ras family has 36 members. In this section, we focus on the regulation of endothelial barrier functions by the best-characterized members of RhoGTPase family: RhoA, Rac1, and CDC42 and Ras family GTPase Rap1 (449). RhoA activation downstream of several endothelial permeability-increasing mediators causes loss of endothelial junctions by enhancing actomyosin contractility and increasing isometric tension at the cell margins (426). However, Rac1 and Cdc42 are required for the assembly and maturation of endothelial junctions, and their activity increases during junction formation (144, 226, 234, 449). In contrast, other studies showed that activation of Rac1 downstream of VEGF and other growth factors increased endothelial permeability (226, 234). This points to the notion that an optimal level of Rac activity is required to maintain the stability of endothelial junctions and either inhibition or hyper-activation can disrupt the AJs.

Sundaresan et al. (400) showed that activated Rac1 increased the ROS generation in fibroblasts. This was further confirmed by multiple reports in response to different stimuli in endothelial cells (92, 463). Over-expression of active Rac1 induced tyrosine phosphorylation of VE-cadherin and increased endothelial permeability in an ROS-dependent manner (427). Contrary to this, active RhoA induced pronounced stress fiber formation but did not induce loss of cell–cell adhesion, indicating that Rac-mediated ROS generation was specifically involved in loss of cell–cell adhesion (427).

The role of RhoA family GTPases has also been investigated in endothelial cells under hypoxic condition. Rac1 and RhoA oppose each other during changes in oxygen tension. In addition, expression of dominant negative Rac1 (N17Rac1) strongly inhibited RhoA and NADPH oxidase activity, suggesting that Rac1 acted upstream of RhoA during hypoxia/reoxygenation stress (427, 455). Interestingly, there are also other mechanisms through which ROS can directly affect RhoA activity and RhoA-mediated cytoskeletal rearrangement. Campbell et al. (176) found that the treatment of purified recombinant GTPase with superoxide in vitro resulted in guanine nucleotide dissociation. Furthermore, using cysteine to alanine mutants, Aghajanian et al. (2) demonstrated that ROS-mediated activation of RhoA was dependent on cysteines 16 and 20.

It has been known for several years that Cdc42 activation is required to maintain endothelial barrier function. Barrier-stabilizing effect of oxidized phospholipids was blocked by siRNA targeting Cdc42 (38), and expression of constitutively active Cdc42 abolished LPS-induced lung vascular permeability in vivo (357). These studies suggest an obligatory role of Cdc42 for barrier maintenance. Although there is little information regarding the direct effects of ROS on Cdc42 function, studies showed the involvement of Cdc42 in ethanol-induced NOX activation and ROS generation and subsequently, reorganization of actin filaments in endothelial cells (437). Further studies are needed to fully understand the mechanisms by which this important modulator of the cell cytoskeleton regulates endothelial barrier function under oxidant stress conditions.

The PKC isoforms are divided into three subclasses according to their structure, activation, and substrate requirements. The calcium-dependent conventional or classical PKC isoforms (α, β1/2, and γ) are regulated by diacylglycerol (DAG) or phosphatidylserine (PS), whereas the novel PKCs are Ca²⁺ independent (δ, ε, η, θ, and μ) and activated by DAG or PS. The atypical PKCs (ζ, λ) are regulated by PS and are independent of DAG and Ca²⁺ (250, 334). The PKC activity is regulated by its translocation to the membrane and phosphorylation and generally denotes enzyme activation (334). PKC is the target for multiple tyrosine phosphorylations by various tyrosine kinases, including Src family kinases (304, 396). These modifications further increase kinase activity of PKC (243). Several PKC inhibitors such as staurosporine, H7, calphostin C, bisindolylmaleimide, and chelerytherine inhibited endothelial hyper-permeability in response to different pro-inflammatory agents (133, 200, 341, 358, 460). A number of studies have demonstrated that membrane translocation and activation of the atypical isoforms of PKCζ is required in thrombin-mediated increase in endothelial permeability (244, 294). Depletion of PKCζ by siRNA or overexpression of dominant negative mutants of PKCζ abolished thrombin-induced hyper-permeability by inhibiting phosphorylation of MLC and activation of RhoA (294). In agreement with these findings, PKC-α was shown to induce phosphorylation of p120-catenin in response to thrombin or LPS, which reduced p120-catenin binding affinity for VE-cadherin and subsequently disrupted endothelial junction integrity (429). TNF-α-mediated endothelial barrier dysfunction was shown to be dependent on an increase in PKC-α activity. Furthermore, the antisense of PKC-α reduced the TNF-α-induced increase in endothelial permeability (120). Other studies have reported that VEGF can increase vascular permeability through activation of PKC-β in vivo and, importantly, oral administration of PKC β-isoform-selective inhibitors partially inhibited VEGF-induced vascular permeability (4). Moreover, Titchenell et al. (414) have shown that pseudo-substrate peptide inhibitors of atypical PKC isoform such as aPKC-I-PD and aPKC-I-diCl can inhibit VEGF-induced endothelial permeability. These studies support a prominent role of different PKC isoforms in microvascular leakage associated with different pro-inflammatory mediators. Increased PKC activity has also been associated with various vascular disorders such as atherosclerosis, hypoxia, and ischemia reperfusion (138, 174, 286, 447).

In the endothelium, H₂O₂ is also known to activate multiple PKC isoforms (224, 225). There are different mechanisms through which ROS can activate PKC. All PKC isoforms contain redox-sensitive cysteine residues in both regulatory and catalytic domains, which are required for autoinhibition and catalytic activity, respectively. Direct oxidative modification of the regulatory domain results in increased PKC activity (143). ROS can also regulate PKC activation by generating different mediators and cofactors, including PLC and PLD (267, 438). Oxidant-mediated activation of PKC also occurred through phosphorylation of distinct residues of different PKC isoforms (224). Several lines of evidence suggest that PKC is likely to play a permissive role in H₂O₂-mediated endothelial permeability and lung edema (200, 389, 470). Johnson et al. (200) showed that H₂O₂-induced lung edema and stress fiber formation in ECs were inhibited by protein kinase blocker H-7. Another study demonstrated that the selective inhibition of PKC by calphostin C abolished H₂O₂-mediated myosin light-chain phosphorylation in ECs (470). Taher et al. (406) demonstrated H₂O₂-inducing translocation of PKC from cytosol to membrane and increased membrane-associated PKC activity. Thus, PKCs act as a key mediator of inflammation and microvascular hyper-permeability under stimulated conditions.

TLRs are a subset of pathogen-associated pattern recognition receptors family that are classified on the basis of homology to the cytoplasmic domain of the interleukin-1 receptor (IL-1R) family, also known as the Toll/IL-1R (TIR) domain (391). A total of 10 and 14 TLRs have been identified in human and mouse, respectively (63, 405). TLRs are expressed in a wide range of immune cells, including DCs, macrophages, PMNs, and certain nonimmune cells such as endothelial and epithelial cells (28, 220). Among all members, TLR4 is the most widely characterized and well-established receptor for LPS signaling. LPS-binding protein recognizes the lipid-A moiety of LPS and brings LPS to the cell surface to form a ternary complex with CD14, which facilitates the transfer of LPS to TLR4 and MD-2 (306). This complex then initiates intracellular signals through two adaptor proteins, TRIF and myD88, which recruit multiple cytoplasmic signaling molecules TRAF6, TIRAP and the kinases RICK and IRAK, and activates downstream signaling components involving JNK, p38, and NF-κB (84). A number of studies showed that LPS-TLR4 axis is involved in endothelial barrier disruption by different pathways. LPS-mediated TLR4 activation induced tyrosine phosphorylation of junctional proteins, including VE-cadherin, p120-catenin, and γ-catenin through activation of Src family kinases to open the endothelial paracellular pathway (139). Liu et al. (248) showed that LPS/TLR4 signaling activated Src family kinases by enhancing the interaction of TRAF6 with cSrc and Fyn and promoting transendothelial albumin flux. In another study, Wang et al. (442) reported a direct role of p120-catenin in the regulation of LPS-mediated increase in endothelial permeability. They showed that the inhibition of p120-catenin expression in endothelial cells increased MyD88 and TLR4 association, ICAM-1 expression and enhanced the severity of LPS-induced lung inflammation.

Increasing evidence suggests that enhanced ROS generation is central in the activation of TLR-mediated signaling pathways. Ryan et al. (365) showed that ROS are involved in LPS/TLR4-mediated IL-8 signaling. Antioxidant treatment decreased LPS-induced nuclear translocation of NF-κB as well as IL-8 production in human monocytes. Park et al. (336) have proposed a direct mechanism through which TLR4 regulates ROS formation. They demonstrated that TLR4 interacted with the COOH-terminal region of NOX4 in HEK293 cells, and knockdown of NOX4 inhibited LPS-mediated NF-κB activation (335). NOX4 was also shown to be required for LPS-induced ROS generation and NF-κB activation in endothelial cells (335). Knockdown of NOX4 significantly inhibited LPS-induced MCP1 production, ICAM-1 expression, and adhesion of monocytes to endothelial cells (335). TLR4 also induced TLR2 expression in human endothelial cells in an NF-κB-dependent manner (119). A unique mechanism of cross-talk between TLR4/TLR2 has been proposed; that is, the role of PMN NADPH oxidase complex in mediating the cross-talk between TLR4 and TLR2 (116). Impairment of PMN NADPH oxidase markedly decreased the LPS-induced NF-κB activation and TLR2 up-regulation in endothelial cells (116). In addition, LPS did not up-regulate the TLR2 expression in TLR4 knockout mice, indicating that PMN NADPH oxidase mediates TLR2 up-regulation in endothelial cells and is an important mechanism that is responsible for amplifying PMN transmigration to the site of infection (118).

The extrinsic pathway of cell death is mediated by cell death receptors, which after binding with their ligands initiate protein–protein interactions, resulting in activation of the initiator caspases. There are four major cell death receptors, including TNF receptor 1 (TNFR1), TNF-related apoptosis-inducing ligand receptor 1 (TRAIL-R1), TRAIL receptor 2 (TRAIL-R2), and Fas receptor. The ligands that bind to these receptors belong to TNF superfamily of cytokines TNFα, Fas ligand, and TRAIL (64). TNF-α-induced cell death has been attributed to excessive ROS generation-sustained JNK activation and caspase activation (Fig. 12) (90, 369, 457). TNF-α is a principal plieotropic cytokine that is secreted by activated macrophges which possesses a variety of biological properties such as production of inflammatory cytokines, cell proliferation, and cell death. It was first defined by William Coley in his findings that cancer patients have necrotic tumors when they developed bacterial infections and was formally referred to as Coley's toxins (452). The active component inducing cell death in tumors was later identified as a cytokine secreted by macrophages and hence, termed tumor necrosis factor (68). TNFR1 signaling under resting conditions is inhibited by association with a 60-kD cytoplasmic protein known as silencer of death domains (SODD). Within minutes of TNFR1 activation by TNF-α, this auto-inhibition is released, the adaptor proteins TNFR1-associated death domain (TRADD) and Fas-associated death domain (FADD) are recruited, and the collective death-inducing signaling complex (DISC) is internalized. Release of endosomal DISC to the cytosol further recruits initiator procaspase-8, resulting in its activation (64). Similar to TNFR1, the apoptosis induced by FasR has been attributed to enhanced ROS generation via NADPH oxidases and caspase activation. The apoptosis induced by FasR likewise depends on the recruitment of FADD and pro-caspase-8 activation. A key feature of FasR-mediated apoptosis is regulation by FLICE inhibitory protein (FLIP), which inhibits apoptosis by preventing the binding of procaspase-8 to FADD (189). The enhanced ROS generation down-regulates FLIP protein and executes FasR-dependent apoptosis (436). The extent of caspase-8 activation is an essential determinant of cells undergoing mitochondrial independent or dependent apoptosis (377). Extensive caspase-8 activation directly activated effector caspases such as caspase-3 and induced cell death, whereas low caspase-8 activation activated an amplification loop that was dependent on the mitochondria (described in the next section).

However, TNF-α does not always induce apoptosis in all cells, because it can alternatively activate at least two cell survival pathways (Fig. 12). TNF-α activates NF-κB by recruiting TNFα-receptor-associated factor 2 (TRAF2) and receptor-interacting kinase 1 (RIP1), which has pro-survival effects by induction of anti-apoptotic proteins (e.g., anti-apoptotic proteins of Bcl2 family and caspase inhibitors) and suppressed TNF-α induced ROS generation by enhancing expression of MnSOD (205, 369). Further, TNF-α mediated an early and transient activation of JNK/Jun, which contributed to cell survival in conjunction with NF-κB by activation of the AP-1 transcription factor (232). In the absence of NF-κB, a massive accumulation of H₂O₂ leads to prolonged activation of JNK kinase, which induced cell death (90, 205, 369).

Although caspase activation and enhanced ROS generation are central to the mechanism of apoptosis, there are reports suggesting no significant role of ROS in the progression of apoptosis and showing that reduced ROS generation is, in fact, the cause of apoptosis (15, 122). This is surprising, as the efficacy of pharmacological antioxidants such as GSH, NAC, deferoxamine (283, 420, 425), and antioxidant enzymes such as catalase or GPx (149, 283) in preventing apoptosis have been shown by several groups. Other studies, however, were unable to demonstrate any role of antioxidants in preventing apoptosis (15, 66, 122). These differences may be ascribed to the use of different cell types, different methods of measuring ROS, and the particular antioxidants tested. In addition, plasma membrane permeability of different cell types to antioxidant enzymes can greatly influence the analysis.

The intrinsic pathways of cell death are dependent on increased mitochondrial outer membrane permeability (MOMP), which causes mitochondria-to-cytosol release of apoptogenic proteins such as Cytochrome-c (Cyt-c), Smac/Diablo, apoptosis-inducing factor (AIF), and endonuclease G, which trigger cell death in either a caspase-dependent or -independent manner (Fig. 12) (151, 230). This phenomenon of increased MOMP can be induced by excessive calcium entry, high oxidative stress, and compounds, which result in mitochondrial membrane depolarization (417). The released Cyt-c then binds to apoptosis activation factor-1 (Apaf-1) and recruits initiator pro-caspase 9, which undergoes auto-activation. This whole complex is collectively known as “apoptosome” (151) (Fig. 12). Caspase-9 then activated the effector caspases such as caspase-3. Smac/Diablo also facilitates the activation of effector caspases by removing the blockage by inhibitor of apoptosis proteins (IAPs) (417). AIF, on activation, induced DNA condensation and cleavage (50). EndoG-mediated DNA fragmentation is the final event in the apoptosis. The mitochondrial release of Cyt-c is considered the major determinant of cell death.

The pro-apoptotic member of Bcl-2 family such as Bid (BH3 interacting-domain death agonist) can enhance MOMP by cleavage with caspase-8. The truncated bid after caspases cleavage then induces oligomerization of pro-apoptotic proteins Bax/Bak, which then integrated in the outer mitochondrial membrane and triggered Cyt-c release (342). However, recombinant oligomeric Bax protein alone contributed only 18% of Cyt-c release from brain mitochondria (342). When combined together, recombinant Bax and complex I inhibitors (which is known to induce oxidative stress) induced approximately 65% of the mitochondrial Cyt-c release (342), indicating that oxidative stress is the major trigger in Cyt-c release from mitochondria. The release of Cyt-c further enhanced ROS generation from mitochondria because of uncoupling of ETC. An anionic phospholipid specific to mitochondria known as cardiolipin (CL) helps in anchoring of Cyt-c to the inner mitochondrial membrane. Enhanced oxidative stress leads to oxidation of CL, which decreases the affinity binding with Cyt-c (218, 388). A growing number of reports have strongly implicated the role of oxidized CL in Cyt-c release (64, 203, 309, 327). The content of oxidized CL was increased in mitochondria isolated from ischemic myocardium, which potentially contributed to the ischemia/reperfusion injury (345). However, the mechanism of oxidized CL-mediated Cyt-c release is not yet clear. According to Gonzalvez et al. (141, 142), oxidized CL is translocated to the outer mitochondrial membrane provided a docking site for tBid, enabling mitochondrial outer membrane permeabilization and Cyt-c movement. However, contradictory reports indicate that oxidized CL does not play a role in Bax-mediated pore formation (194, 195). Interestingly, Kagan et al. showed that a pool of CL-bound mitochondrial Cyt-c functions as a peroxidase that catalyzes peroxidation of CL (204). The peroxidase activity of Cyt-c can be stimulated at a lower H₂O₂ concentration when bound to CL-containing membranes compared with unbound (35), suggesting that increased oxidative stress can enhance the peroxidase activity of Cyt-c, enabling CL peroxidation and execution of cell death.

Independent of apoptosis by Bcl-2 family members, the mitochondrial permeability transition (MPT) induced by the opening of mitochondrial permeability transition pore (mPTP) can also lead to cell death. Unlike MOMP, mPTP enhances permeability of inner mitochondrial membrane and is independent of Cyt-c release. The opening of the mPTP causes mitochondrial swelling due to massive influx of solute and water, resulting in mitochondrial depolarization and cessation of ATP synthesis (215). mPTP is the underlying event in primary necrosis. However, apoptotic cells can undergo secondary necrosis if their removal is delayed. In contrast to primary necrosis, mitochondrial depolarization in secondary necrosis can be coincident or followed by Cyt-c release. mPTP is a megapore spanning the inner and outer mitochondrial membranes, which is hypothesized to consist of at least three components (72): outer membrane channel known as VDAC, inner membrane channel known as adenine-nucleotide translocase (ANT), and cyclophilin-D (CypD) present in mitochondrial matrix. CypD, although present in the mitochondrial matrix, becomes associated with the inner mitochondrial membrane during the MPT (72, 160). Although MOMP and mPTP appear to be distinct pathways, a cross-talk has been reported between them. Mice deficient in Bax/Bak render cells resistant to mPTP opening and necrosis (451). Superoxide-induced mitochondrial Cyt-c release in HepG2 cells depends on VDAC, and blockers of VDAC inhibited the release (259). ANT and cyclophilin D are also targets of oxidative stress, and their oxidative modification has been implicated in an increase in mitochondrial permeability and Cyt-c release (20, 135). Mice deficient in CypD were protected from ischemia/reperfusion-induced cell death in vivo and primary hepatocytes and fibroblasts isolated from CypD null mice were protected from Ca²⁺ overload and oxidative stress-induced cell death (20). Therefore, it seems that there is an intricate relationship between these two pathways which is likely modulated by oxidative stress during inflammation.