Interferon-gamma (IFN-γ) is well known for its immunomodulatory activity; however, it is worth to mention that IFN-γ is also a proinflammatory cytokine that disrupts the cell barrier integrity. In the past 30 years, numerous works have been performed to investigate the effect of IFN-γ on the regulation of barrier function both in vivo and in vitro, but most of these previous studies focus on its impact in modulation of the barrier function of epithelial cells rather than endothelial cells. The impact of IFN-γ in the regulation of epithelial permeability has been reviewed in [1].

In in vivo model, IFN-γ was reported to increase vascular leakage in dextran sulfate sodium-induced colitis mice, concomitant with disorganization of vascular networks in colonic mucosa [2]. However, till now, the knowledge on mechanism of actions of IFN-γ as a regulator of the endothelial barrier function remains largely unclear. Given its critical impact in pathological vascular inflammation and diseases, in the recent decade, much more attention has been given on exploring IFN-γ mechanisms on the disruption of endothelial barrier function. To date, however, there is an absence of articles reviewing the roles and mechanisms of IFN-γ in the regulation of endothelial barrier function. This article emphasizes on the effects of IFN-γ on the regulation of endothelial barrier function and signaling pathways involved in IFN-γ-mediated endothelial barrier dysfunction.

IFN-γ was first identified as a soluble factor secreted by human leukocytes which interferes with viral replication upon stimulation with phytohemagglutinin [3]. This pleiotropic cytokine is therefore, well known for its anti-viral and immunomodulatory activities. IFN-γ exerts direct effect on halting various stages of viral life cycle including entry into host cells, replication of viral genome and viral gene transcription, thereby possesses anti-viral effects in many cell types [4]. In the context of immunomodulation, IFN-γ is the key cytokine that drives Th1 response during viral infection. IFN-γ has been reported to enhance antigen presentation and increase the expression of major histocompatibility complex (MHC) I and II in antigen presenting cells [[5], [6]]. IFN-γ also activates macrophages by stimulating the release of nitric oxide [7], upregulating production of reactive oxygen species [8] and increasing phagocytic activity [9]. Furthermore, IFN-γ mediates leukocyte trafficking into blood vessels, thereby allowing recruitment of immune cells to the site of infection. Using a mouse model of herpes simplex virus 1 corneal infection, IFN-γ was shown to promote neutrophil infiltration to the infected areas via an increase in platelet endothelial cell adhesion molecules (PECAM)-1 expression on the vascular endothelium [10]. In IFN-γ knockout (IFN-γ-/-) mice with parasitic infection, the number of cerebral vessels expressing vascular cell adhesion molecule (VCAM)-1 and number of CD8+ T cells recruited into the brain were significantly reduced compared to infected wild-type mice. This showed a significant role of IFN-γ in promoting an upregulation of VCAM-1 expression and leukocytes trafficking during infection [11].

IFN-γ exerts its effector actions by binding to a specific cell surface receptor which in turn activates multitude signaling cascades in various cell types. The cell surface receptor is composed of two subunits, i.e. IFN-γ receptor (IFNGR)-1 and IFNGR2. IFNGR1, but not IFNGR2, binds with IFN-γ via its extracellular domain whereas both IFNGR1 and IFNGR2 interact with signaling molecules through their intracellular domains. Among all the explored IFN-γ-activated signaling cascades, the most well studied is the Janus-activated kinase (JAK)-signal transducer and activator of transcription (STAT) pathway. In the cell, JAK1 and JAK2 are constitutively bound to IFNGR1 and IFNGR2, respectively. Upon ligand binding, JAKs interact with the intracellular domains of the IFN-γ receptor and become phosphorylated. This enables STAT-1 to dock at the IFN-γ receptor through its Src homology 2 (SH2) domains. STAT-1 then phosphorylates on Tyr701 and homodimerizes before it moves into the nucleus and binds with IFN-γ-activated sequences. This eventually results in upregulation of transcription of IFN-γ-regulated genes. Even though most of the transcriptional activities caused by STAT-1 are dependent on phosphorylation of Tyr701, IFN-γ also causes phosphorylation of STAT-1 on Ser727, which is important for recruitment of histone acetylases and association of STAT-1 with the coactivator CREB-binding protein [12].

IFN-γ is also known to stimulate non-canonical STAT signaling pathways through formation of STAT1-interferon regulatory factor (IRF)-9 complexes [13]. STAT-1-IRF 9 complexes were found to bind with interferon-stimulated response elements (ISRE) promoter sequences in IFN-γ-induced Vero cells [14]. The expression of CxCL10, an IFN-γ-inducible gene, was found to be important for the formation of STAT1-IRF9 complexes at the CxCL10 promoter [15]. These findings suggest that IFN-γ-induced transcriptional activity also depends on association of STAT1 with other proteins, apart from the STAT1-phosphorylation observed in the canonical JAK-STAT pathway. In addition, other studies working on the non-cannonical STAT pathway proposed a role of STAT-2 in IFN-γ response. In human lung epithelial cells, an IFN-stimulated gene factor 3 (ISGF3II) complex containing phosphorylated STAT-1, unphosphorylated STAT-2 and IRF-9 was recruited to the ISRE sequences at late phase of IFN-γ response [16]. Following IFN-γ stimulation, transcription of CxCL10 was shown to be dependent on the ISGF3II, albeit transcriptional activities of other IFN-γ-inducible genes were ISGF3II-independent [15]. As the activity of the ISGF3II complex is influenced by the IFN-γ concentration and it appears to be associated with only certain genes and stages of IFN-γ response, more evidence is needed to describe the exact role of STAT2 in IFN-γ-induced signaling pathways.

Interferon-gamma (IFN-γ) is primarily released by activated T cells. It can activate macrophages to release other types of proinflammatory cytokines such as TNF-α and IL-6. This pleiotropic cytokine mediates antiviral and antimicrobial immunity and regulates both innate and adaptive immunity [4]. Several studies have reported that IFN-γ receptors are highly expressed in atherosclerotic lesions, and IFN-γ deficiency delays neointimal formation, indicating the critical role of IFN-γ in the development of atherosclerosis [[17], [18], [19]]. The central role of IFN-γ in atherosclerosis has been reviewed in detail elsewhere [[20], [21], [22], [23], [24]].

IFN-γ has been reported to possess a synergistic effect with TNF-α in promoting inflammatory responses in in vitro and ex vivo experimental models [[17], [25]]. IFN-γ and TNF-α were found to synergistically increase monocytes and T-cell chemoattractant in atherosclerotic tissue ex vivo. In addition, combined treatment of TNF-α and IFN-γ not only synergistically increases chemokines and cell adhesion molecule expressions, but also disrupts inter-endothelial junctions in HAECs in vitro [17]. The combined effects of IFN-γ and TNF-α in epithelial and endothelial permeability have been reviewed previously [1].

IFN-γ exerts a potent antiviral activity that protects the host from harmful infections, however, it is not a toxic inducer to vascular cells as several reports have shown that prolonged IFN-γ exposure did not alter the total cell viability, at least in in vitro human umbilical vein endothelial cells (HUVECs) [[26], [27]]. IFN-γ has been shown to attenuate the proliferation of macro- and microvascular endothelial cells [27], but the anti-proliferative effect is reversible upon the removal of IFN-γ [28]. IFN-γ also causes reversible changes in human endothelial cell morphology and phenotype [[28], [29]]. Furthermore, IFN-γ exerts angiostatic effects by decreasing angiogenic sprout length and thickness in vitro, vessel outgrowth from mouse embryo metatarsals ex vivo and colon vessel proliferation in vivo [30]. Interferon induced protein 35, IFI35, an IFN-γ-induced protein important for anti-viral immunity, has been found to inhibit endothelial cell proliferation and migration in vitro and delay re-endothelialization of wire-injured mouse carotid arteries through suppression of NF-κB pathway [31].

One of the major effects of IFN-γ in endothelial cells is the disruption of barrier integrity [[26], [32], [33], [34], [35]], which eventually increases the permeability of endothelial cells to plasma proteins and immune cells. Importantly, the increase in endothelial permeability induced by IFN-γ occurs in a biphasic manner, in which the early phase involves a gradual increase of permeability followed by a delayed phase where the increased permeability returns back to a basal level, and a sustained phase after prolonged stimulation. The biphasic nature of endothelial hyperpermeability caused by IFN-γ implies that distinct mechanisms are involved in the regulation of endothelial barrier integrity. In addition to changes on the monolayer permeability, IFN-γ also promotes trans endothelial migration of T cells in ICAM-1 and VCAM-1 dependent manner [36].

The permeability changes in response to IFN-γ can be divided into an early phase (0–8 h), a delayed phase (12 h) and a sustained phase (16–24 h) [26]. The cellular mechanisms involve in these phases are primarily associated with actin cytoskeleton remodeling and junctional protein reorganization Fig. 1, Fig. 2 Table 1.

In the early phase of IFN-γ stimulation, endothelial cells change from a uniform cobblestone appearance into a rounded yet spiky-like morphology [26]. As a consequence, the stimulated cells lose their close contact with adjacent cells, thereby forming noticeable intercellular gaps. Nevertheless, the cells are still connected to their adjacent cells through fine retraction fibers [26]. The presence of contractile responses suggests that the increased cell permeability may be modulated by contractile mechanisms. Gap formation and discontinuous cell–cell contact were also reported in lung microvasculature following IFN-γ stimulation [37].

Caldesmon, a cytoskeleton-binding protein, inhibits cell contraction by preventing actomyosin ATPase activity, where these cellular events are reversible when caldesmon phosphorylation occurs. A study has shown that IFN-γ stimulates caldesmon phosphorylation at Ser789 and this event occurs in parallel with actin dissociation from caldesmon [38]. These data suggest that there are structural conformation modifications as a result of caldesmon serine phosphorylation, which alters the binding affinity of caldesmon to actin; however, in depth studies are required to verify this possibility. Several studies have shown that disassembly of actin from caldesmon is a prerequisite for cytoskeleton remodeling and serves as a fundamental for actomyosin-based contractility in endothelial cells [[39], [40], [41]].

In addition to alteration of cytoskeletal proteins, IFN-γ also stimulates reorganization of reticular adherens junctions in human endothelial cells into discontinuous adherens junctions, which arrange orthogonally to the cell edges. The reorganized adherens junctions are corresponded with endothelial cell contractions [[26], [38]]. This suggests that actomyosin contractility and its derived tension force are considered as a requisite for the formation of discontinuous adherens junctions. It has also been reported that IFN-γ reduced VE-cadherin staining and caused fragmentation of cellular junctions in the early phase, a similar response which was also observed in the sustained phase [37].

Lymphatic endothelial cells have been shown to respond differently from other type of endothelial cells in the presence of IFN-γ [42]. Cromer et al. showed that IFN-γ disrupts lymphatic endothelial barrier by reducing VE-cadherin expression while elevating β-catenin and phosphorylated myosin light chain expressions. These findings are in contrast to our previous observations in HUVECs where IFN-γ affects only the organization of adherens juntions but not affecting the total junctional protein expression. The study of barrier function of endothelial cells from different origins, therefore, needs to be interpreted carefully as each vasculature is unique although it might share some similar properties with others.

An increase in endothelial permeability caused by IFN-γ occurs in phases rather than continuously. In the delayed phase, endothelial cells undergo a transition period, which is accompanied by several impressive cellular events. Noteworthy, there is a dramatic change in the morphology of the endothelial cells from rounded and retracted into elongated, and this is accompanied by redistribution of peripheral condensed actin bands into cytoplasmic stress fibers [26]. Both of the changes occur concurrently with an increase in F-/G-actin ratio, implying a critical role of F-/G-actin ratio in the regulation of endothelial cell shape and actin cytoskeleton arrangement [26]. Furthermore, caldesmon rearranges from the cell edges, which is seen in the early phase, into the cytoplasm. Caldesmon also displays similar structural patterns as F-actin, indicating that both proteins are colocalized in the endothelial cells [[26], [38]]. Unexpectedly, there is an enhancement of myosin attachment to caldesmon [38]. The observations suggest a role of caldesmon as a regulator of cytoskeletal remodeling upon IFN-γ stimulation, but the significance of the myosin-caldesmon interactions remains unclear and therefore still need further investigations. Moreover, in this phase, restoration of basal endothelial permeability occurs concomitantly with reformation of reticulum structures [[26], [38]]. This suggests that adherens junction organization play a critical role in the regulation of endothelial permeability in response to IFN-γ. The mechanisms underlying this transition state remain unclear but it might involve a sequence of molecular switches, new gene transcriptional or de novo protein synthesis as these cellular processes could affect the structures and functions of various intracellular proteins, which subsequently modulates the endothelial barrier function.

In the sustained phase of IFN-γ stimulation, HUVECs remain elongated and the cytoplasm is enriched with stress fibers [[26], [28], [43]]. A similar change of morphological phenotypes has also been reported in cultures of HBMEC [29]. A previous research has demonstrated that IFN-γ does not alter the pinocytic activity of endothelial cells [29]. This indicates that an increase in junctional permeability is the primary mechanism responsible for the permeability changes of the cell monolayers. IFN-γ has been reported to disrupt adherens junctions, one of the main contributors that lead to an impairment of the endothelial barrier [[30], [38]]. In this phase, adherens junctions remodel from reticulum structures into linearized structures, and this parallels with the onset of reduction of junctional areas and adherens junctions expression [38]. The downregulation of VE-cadherin and β-catenin expressions is mainly due to a reduction of VE-cadherin and β-catenin expressions in both membrane and cytoskeletal fractions [38]. Furthermore, a previous study demonstrated that blockage of IFN-γ in reovirus-infected mice caused localization of β-catenin, claudin-5 and ZO-1 in junctions of brain vasculature but their total protein expressions remain unchanged. Using the mouse brain cell line bEnd.3 cells, the authors found that IFN-γ induced junction disorganization where membrane VE-cadherin, β-catenin and claudin-5 were relocalized to the cytoplasm [44].

Recent studies have shown that IFN-γ disrupts small intestine mouse organoids, a newly developed three-dimensional cell culture model, by causing disruption of the tight junction proteins [45]. IFN-γ treatment was also reported to induce increased barrier permeability in bovine intestinal organoids [46]. Although tight junctions are also one of the key regulators for the maintenance of barrier function, the roles of tight junctions in IFN-γ-induced endothelial barrier dysfunction are less extensively investigated than that of adherens junctions. The expressions of occludin and VE-cadherin in HUVECs were decreased after prolonged treatment of IFN-γ [[33], [34]]. On the contrary, Cong et al. showed that IFN-γ increased claudin-5 expression in human microvessel endothelial cell line [47]. IFN-γ also caused ZO-1 redistribution from cell-cell junctions [36] and the authors suggested an indispensable role of ZO-1 in IFN-γ-mediated barrier disruption through knockdown of ZO-1 in bEnd.3 cells [36]. These data imply that IFN-γ targets adherens and tight junctions whereby both junctional proteins regulate the disruption of endothelial barrier integrity. While structural reorganization and relocalization of junction proteins are prominent features stimulated by IFN-γ in the sustained phase, the differential effect of IFN-γ on the junctional protein expressions could be explained by variations in the vascular bed, dose and duration of treatment used.

Collectively, the phasic changes in endothelial permeability observed in response to IFN-γ denote that this cytokine differentially regulates the endothelial cell morphology, actin cytoskeletal arrangement and adherens junction’s organization throughout the three phases Table 1.

P38 MAP kinase has been implicated to mediate rearrangement of cytoskeletons and regulate the epithelial barrier integrity [48]. Prolonged treatment (hours) of IFN-γ activates p38 MAP kinase in HUVECs, the most commonly used endothelial cells in in vitro experiments, and the activation precedes remodelling of actin and cell morphology and endothelial hyperpermeability caused by IFN-γ [26]. Blocking of p38 MAP kinase with a pharmacological inhibitor SB203580 attenuates IFN-γ-stimulated actin cytoskeleton and cell morphological changes, which subsequently diminishes the endothelial hyperpermeability [26]. Moreover, p38 MAP kinase also regulates IFN-γ-mediated F-actin hyperpolymerization [26]. These data show the role of p38 MAPK in the modulation of cytoskeletal rearrangement and disruption of endothelial barrier integrity caused by IFN-γ. Surprisingly, the biphasic changes of caldesmon and adherens junction reorganization upon IFN-γ stimulation are independent of p38 MAP kinase, denoting the involvement of other signaling pathways such as ROCK. Further investigations are needed to elucidate the involvement of other signaling molecules activated by IFN-γ.

The small GTPases comprises several families including Rho, Ras, Arf and Rab. Rho family is amongst the most extensively studied of the small GTPases for its role in modulation of the endothelial barrier. RhoA, the best characterized Rho GTPase, activates its downstream target Rho-associated kinase (ROCK), which in turn inhibits myosin light chain phosphatase, causing myosin light chain phosphorylation, thereby stimulating actomyosin contractility and actin polymerization [49]. In bEnd.3 cells, IFN-γ was demonstrated to increase RhoC, RhoG, ROCK1 and ROCK2 expressions at 24 h [44]. Using pharmacological inhibitors for Rho-GTPases and ROCK, the authors showed that IFN-γ alters the permeability of the blood–brain barrier through an activation of ROCK activity but not Rho-GTPase activity [44]. ROCK is also required for actin remodeling, phosphorylation of myosin light chain and relocalization of junction proteins including VE-cadherin, β-catenin and claudin-5 triggered by IFN-γ [44]. Since small GTPases, RhoA and ROCK, play predominant roles in mediating the endothelial barrier function, the roles of other small GTPases such as Rap1 and Rac1 and their downstream effectors warrant further investigation.

The best studied signaling pathway activated by IFN-γ is the JAK-STAT pathway. IFN-γ acts on its receptor to activate JAK1 and JAK2 protein-tyrosine kinases which subsequently phosphorylates STAT1. Phosphorylated STAT1 translocates to the nucleus and binds to conserved IFN-γ activation site (GAS) DNA elements, which in turn initiates transcription of interferon-stimulated genes (ISGs). The IFN-γ-signaling cascade has been previously reviewed in detail [[13], [50], [51]].

IFN-γ promotes STAT1 and NF-κB activation in human aortic valve endothelial cells, which subsequently upregulates the expression of downstream molecules such as cell adhesion molecules [52]. In Caco-2 epithelial cell, IFN-γ was reported to impair tight junction by upregulating the expression of TLR2 and its downstream molecules MyD88, IRAK1, p-65, and MLCK [53]. Another study has reported that IFN-γ upregulates extra-hepatic cytochrome CYP1B1 in human T84 colon carcinoma cells. This could reduce the activity of the aryl hydrocarbon receptor (AhR) pathway and thereby favouring HIF1α/ARNT-mediated cell permeability [54]. A previous study also reported that IFN-γ induces the impairment of epithelial barrier function and disruption of tight junction proteins, by upregulation of HIF-1α expression through NF-κB pathway [55]. Smyth et al. has reported that IFN-γ activates Src kinase Fyn which subsequently causes the formation of a complex STAT5b/Gab2/p85α, and the complex causes PI3K activation and a subsequent increase in macromolecular permeability in human colon-derived T84 epithelial cell lines [56].

IFN-γ has been shown to impair blood–brain barrier in brain endothelial cells, suggesting its significant impact during encephalomyelitis [[36], [57]]. Additionally, IFN-γ promotes chemokine-mediated transendothelial migration of CD4+ T cells and elevates permeability of brain endothelial cells [36]. Interestingly, T cells were found to transmigrate through the transcellular route in response to IFN-γ. Transfection of the brain endothelial cells with STAT1 specific siRNA abrogated IFN-γ-mediated redistribution of VE-cadherin, ZO-1 and claudin-5 molecules and transendothelial migration of CD4+ T cells, suggesting that IFN-γ activates STAT1 which in turns modulate all these cellular events [36]. These findings indicate that IFN-γ disrupts barrier integrity and promotes leukocytes transmigration via STAT1 dependent pathway.

Nitric oxide (NO) is produced by three subtypes of nitric oxide synthase (NOS), including neuronal NOS (nNOS), inducible NOS (iNOS) and endothelial NOS (eNOS), of which only eNOS is constitutively expressed in the endothelium [58]. The effect of IFN-γ on endothelial barrier function is dependent on the cell types and experimental settings. IFN-γ has been shown to diminish endogenous NO in HUVECs and exogenous NO donated by sodium nitroprusside [59]. Although IFN-γ alters NO bioavailability in the cells, however, NO only partially participates in the regulation of IFN-γ-stimulated endothelial hyperpermeability [59]. Weng et al. demonstrated that IFN-γ downregulates eNOS mRNA and protein levels in immortalized endothelial cells (EAhy926) and HAECs [60]. IFN-γ was also shown to inhibit eNOS transcription by promoting an interaction between class II trans-activator and H3K9 trimethyltransferase (SUV39H1) on the eNOS promoter [60]. In contrast, NO production was enhanced in response to a prolonged treatment of IFN-γ through augmentation of iNOS expressions in a murine aortic endothelial cell line [[61], [62]].

Cyclic guanosine monophosphate (cGMP), produced by soluble and particulate guanylyl cyclases, is one of the downstream molecules in NO signaling pathway. Differential effects of cGMP on the endothelial permeability have been observed and the effect seems to be largely dependent on the expression of cyclic nucleotide phosphodiesterases in endothelial cells [63]. Using human brain microvessel endothelial cells, Wong et al. demonstrated the roles of NO and cGMP donors in reversing IFN-γ-stimulated endothelial hyperpermeability, indicating that both NO and cGMP regulate the increased brain blood barrier permeability caused by IFN-γ [64]. Furthermore, IFN-γ suppressed basal cGMP levels and exogenous cGMP donated by 8-bromo-cGMP in a confluent HUVEC model, however cGMP does not play a role in the regulation of IFN-γ-mediated endothelial hyperpermeability [59].