Endothelial cell–cell contacts are essential for vascular integrity and physiology, protecting tissues and organs from edema and uncontrolled invasion of inflammatory cells. The vascular endothelial barrier is dynamic, but its integrity is preserved through a tight control at different levels. Inflammatory cytokines and G-protein-coupled receptor agonists, such as histamine, reduce endothelial integrity and increase vascular leakage. This is due to elevated myosin-based contractility, in conjunction with phosphorylation of proteins at cell–cell contacts. Conversely, reducing contractility stabilizes or even increases endothelial junctional integrity. Rho GTPases are key regulators of such cytoskeletal dynamics and endothelial cell–cell contacts. In addition to signaling-induced regulation, the expression of junctional proteins, such as occludin, claudins and vascular endothelial cadherin, also controls endothelial barrier function. There is increasing evidence that, in addition to protein phosphorylation, ubiquitylation (also known as ubiquitination) is an important and dynamic post-translational modification that regulates Rho GTPases, junctional proteins and, consequently, endothelial barrier function. In this Review, we discuss the emerging role of ubiquitylation and deubiquitylation events in endothelial integrity and inflammation. The picture that emerges is one of increasing complexity, which is both fascinating and promising given the clinical relevance of vascular integrity in the control of inflammation, and of tissue and organ damage.
The inner lining of all blood and lymphatic vessels is formed by a monolayer of vascular endothelial cells (ECs), which preserves integrity through dynamic but well-controlled cell–cell contacts. Loss of endothelial integrity, due to increased actomyosin-based contractility and reduced cell–cell contact, is among the first signs of inflammation and is associated with vascular pathology that accompanies chronic disorders, such as diabetes, atherosclerosis or rheumatoid arthritis (Bordy et al., 2018; Huveneers et al., 2015; Saharinen et al., 2017). Because of its clinical relevance, there is much interest in the mechanisms that govern endothelial integrity. This integrity is mainly determined by the adhesive function of vascular endothelial (VE)-cadherin, which acts in complex with F-actin-binding adapter proteins, such as β-catenin, α-catenin and vinculin. The adhesive function of this VE-cadherin (also known as CDH5) complex is mediated and controlled by actin dynamics and tyrosine (de)phosphorylation of VE-cadherin and β-catenin (Hordijk, 2016; Huveneers et al., 2012; Komarova et al., 2017; Orsenigo et al., 2012; Vestweber et al., 2014; Wessel et al., 2014). Recently, there has been growing interest in another post-translational modification that controls endothelial integrity, namely protein ubiquitylation (also known as ubiquitination).
Protein ubiquitylation is a three-step process, in which the 76-amino-acid peptide ubiquitin is transferred from an E1 to an E2 ligase, after which an associated E3 ligase catalyzes covalent linkage of the ubiquitin moiety to the substrate, in most cases on a lysine residue, or to lysine residues of a previously linked ubiquitin. This results in ubiquitin chain formation through, for example, K63 or K48 linkages (Fig. 1) (Rape, 2018). In addition to ubiquitin, cells also use ubiquitin-like proteins such as small ubiquitin-like modifier (SUMO) to modify target proteins (Akutsu et al., 2016; Kirkin and Dikic, 2007). It is estimated that there might be up to 600 ubiquitin E3 ligases, which can be subdivided in several families (Rape, 2018). Homologous to the E6AP C-terminus (HECT) ligases obtain the ubiquitin from the E2 ligase, prior to linkage to the substrate. In contrast, really interesting new gene (RING) ligases, multi-protein complexes comprised of scaffold, adapter and substrate recognition proteins, do not bind ubiquitin directly, but mediate its transfer from the E2 ligase to the substrate (Kirkin and Dikic, 2007; Schaefer et al., 2012). The ring-between-ring (RBR) E3 ligases (for example PARKIN) are a relatively small subgroup (14 members in humans) that combines features of HECT and RING ligases in their mode of ubiquitin binding and transfer to their substrates (Spratt et al., 2014; Walden and Rittinger, 2018). Finally, a new class of ubiquitin ligase was recently identified, designated RING-Cys-relay (RCR), which transfers ubiquitin to its substrate in a unique way, through esterification of a threonine, rather than a lysine, residue (Pao et al., 2018).
In addition to the variety in the types of ubiquitin ligase, there is also considerable complexity in the site-specific modifications with ubiquitin, including mono-, di- or poly-ubiquitylation, as well as linear or branched ubiquitin chains (Akutsu et al., 2016; Swatek and Komander, 2016) (Fig. 1). Moreover, as an example, the linear ubiquitin chain assembly complex (LUBAC) has a preference for sites that are already modified by K63-linked ubiquitin chains. Its addition of linear, methionine-linked ubiquitin results in a K63-polyUb/M1-polyUb hybrid or mixed chains (Cohen and Strickson, 2017; Emmerich et al., 2013).
There is increasing evidence that ubiquitylation not only serves to target the substrate for proteasomal degradation, but in fact controls cellular functions in many ways, including regulation of protein–protein interactions, vesicular trafficking, receptor internalization and subcellular localization of signaling proteins (Rape, 2018; Schaefer et al., 2012). Since ubiquitylation is reversible through the action of deubiquitylating enzymes (DUBs), it qualifies as a bona fide signal transduction event, similar to (de)phosphorylation or (de)acetylation (Kovačević et al., 2018; Li et al., 2018; Lv et al., 2018; Nethe et al., 2010, 2012; Nethe and Hordijk, 2010; Schaefer et al., 2012).
Increasing numbers of ubiquitin E3 ligases and DUBs have recently been linked to endothelial cell–cell contact and inflammation. There is evidence that ubiquitylation regulates proximal signaling induced by inflammatory cytokines, such as tumor necrosis factor (TNF), that increase endothelial permeability, leading to edema and, eventually, tissue damage (Heger et al., 2018; Kovačević et al., 2018; Kwon et al., 2016; Lv et al., 2018; Nanes et al., 2017). Here, we will first briefly address ubiquitin modifications that have been described in inflammatory cytokine signaling, before discussing the regulation of Rho GTPases by ubiquitylation. Rho GTPases act downstream of inflammatory and other activating agonists, and are considered master regulators of endothelial cell–cell contact and inflammation. Finally, we discuss ubiquitylation of junctional proteins and the relevance of ubiquitylation for vascular disease.
Ubiquitylation in cytokine-induced inflammation and vascular integrity
Inflammatory signaling in ECs serves to protect tissues from excessive damage by initiating, directly or indirectly, the removal of infectious or damaging agents. Uncontrolled or low-grade chronic inflammation, however, leads to pathologies, such as rheumatoid arthritis or atherosclerosis. ECs are among the first cells to participate in an inflammatory response. In ECs, this response comprises the production of reactive oxygen species (ROS), upregulation of adhesion molecules that recruit activated leukocytes and lymphocytes, a disruption of the endothelial barrier and increased leukocyte diapedesis (Nourshargh et al., 2010). One of the key pathways that drives these effects downstream from the pro-inflammatory cytokine TNFα, the interleukins IL-1β and IL-17, or Toll-like receptor (TLR) ligands such as lipopolysaccharide (LPS), is the nuclear factor κB (NFκB) pathway (Ebner et al., 2017; Kawasaki and Kawai, 2014). The NFκB pathway is complex, and its activation comprises a series of different components that are regulated by both phosphorylation and ubiquitylation (Grabbe et al., 2011). The first level of ubiquitin-mediated regulation concerns the transmembrane receptors and associated adaptor proteins. These are typified by the E3 ligases cellular inhibitor of apoptosis protein (cIAP)1 and cIAP2 (also known as BIRC2 and BIRC3, respectively), LUBAC and Itch, which have all been linked to the TNF-induced ubiquitylation of receptor interacting protein 1 (RIP1, also known as RIPK1) (Boisson et al., 2012; Bradley, 2008; Ikeda, 2015; Micheau and Tschopp, 2003). Ubiquitylation by these ligases and regulated deubiquitylation by the DUBs cylindromatosis (CYLD), A20 (also known as TNFAIP3) and Otulin are crucial for steering the TNF pathway towards pro-survival NFκB-dependent signaling (Heger et al., 2018; Keusekotten et al., 2013; Kovalenko et al., 2003; Wertz et al., 2016, 2004). Furthermore, polyubiquitin chains assembled by these E3 ligases are required for the activation of the IκB kinase (IKK)α–IKKβ complex and subsequent phosphorylation of IκBα, the inhibitory subunit of NFκB (Kanarek and Ben-Neriah, 2012). Phosphorylated IκBα is subsequently ubiquitylated by the E3 ligase SCF-βTRCP [the Skp, cullin, F-box-containing complex containing βTRCP (also known as BTRC) as the F-box protein] and degraded by the proteasome. This results in the release and nuclear translocation of NFκB, which then activates the transcription of genes required for cell survival and leukocyte-endothelial cell interactions (Boisson et al., 2012; Christian et al., 2016; Kanarek and Ben-Neriah, 2012).
In TNF-treated retinal ECs, NFκB-dependent loss of vascular integrity has been linked to reduced expression and altered subcellular localization of the tight junction proteins Zonula occludens 1 (ZO-1, also known as TJP1) and claudin-5 (Aveleira et al., 2010). In lymphatic ECs, IL-1β and TNF have been found to decrease the expression of VE-cadherin and activate actomyosin-based contraction, partially through an endothelial nitric oxide synthase (eNOS, also known as NOS3)-dependent mechanism whose molecular details remain to be established (Cromer et al., 2014). Finally, in murine brain ECs, IL-17 induces ROS production and actomyosin contraction, resulting in the downregulation of the tight junction protein occludin and disruption of the endothelial barrier (Huppert et al., 2010).
Together, while ubiquitylation is an abundant post-translational modification (PTM) in cytokine-induced proximal signaling, its regulation of the associated reduced expression of junctional proteins and the loss of endothelial barrier function remains to be further investigated. In this context, the role of actomyosin-based contraction, which may weaken junctions mechanically and thereby indirectly induce junctional protein internalization and possibly degradation, also warrants more detailed analysis.
Rho GTPase ubiquitylation and endothelial integrity
The integrity of the endothelial barrier is dependent on the dynamic stability of cell–cell contacts in the monolayer. Each EC exerts pushing and pulling forces on its neighboring cells, and the net result of these forces determines junctional integrity and permeability. Changes in these forces, for instance due to vascular contraction or relaxation, will determine the response of ECs to maintain and restore the barrier. ECs control the strength of their cell–cell contacts in part through the actin cytoskeleton, for instance by inducing actin (de)polymerization or the formation and bundling of F-actin stress fibers. The key molecular players that control these actin dynamics are members of the family of Rho GTPases.
Rho GTPases switch between GTP-bound ‘on’ and GDP-bound ‘off’ states through the action of guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs), respectively (Bos et al., 2007). When GTP bound, Rho GTPases are typically located at the plasma membrane where they interact with their effectors, including protein kinases and actin-binding proteins. This way, they affect the local assembly or disassembly of F-actin and allow polarized regulation of cell motility (Sit and Manser, 2011; Bravo-Cordero et al., 2013; Bustelo et al., 2012; Kraynov et al., 2000; Machacek et al., 2009). In their GDP-bound state, most Rho GTPases are bound to a member of the cytosolic Rho GDP dissociation inhibitor (GDI) family, which prevent nucleotide dissociation and thus the activation of Rho GTPases. The GDI also protects the Rho GTPase from degradation (Boulter et al., 2010; Dovas and Couchman, 2005).
More than 20 members of the family of Rho GTPases have been described thus far (Schaefer et al., 2014). The best characterized of these, and which are implicated in endothelial barrier function, are RhoA, RhoB, RhoC, Rac1 and Cdc42, each with distinct roles in cell adhesion (Pronk et al., 2017; Ridley, 2015). Typically, RhoA activity leads to stress fiber formation and actomyosin-based contraction, whereas activation of Rac1 and Cdc42 gives rise to actin polymerization and cell spreading due to the formation of lamellipodia and filopodia, respectively. The mechanical force exerted by contracting F-actin filaments on VE-cadherin complexes, and thereby on cell–cell contacts in an intact monolayer, can lead to barrier disruption. Consequently, tightly coordinated (in)activation of Rho GTPase signaling is essential for the stabilization, disruption and reformation of endothelial junctions (Huveneers et al., 2015). Localization of Rho GTPases is key to their activation and downstream signaling and is regulated by PTMs, including phosphorylation, isoprenylation, palmitoylation and, as identified more recently, sumoylation and ubiquitylation (Hodge and Ridley, 2016).
A clear role for ubiquitylation in the regulation of both Rho GTPase localization, as well as proteasomal degradation has been shown by us and others. Treatment of human umbilical vein endothelial cell (HUVEC) monolayers with the neddylation inhibitor MLN4924, an inhibitor of cullin RING ligase (CRL) activity, induces a rapid loss of endothelial barrier function (Sakaue et al., 2017; Kovačević et al., 2018). Neddylation is the covalent attachment of the ubiquitin-like Nedd8 protein to the cullin scaffold proteins, a modification that is required for their activity. MLN4942-induced loss of barrier function is accompanied by an increase in the levels of RhoB, which results from a decrease in RhoB degradation (Kovačević et al., 2018). This indicates that, in resting conditions, HUVECs constantly ubiquitylate and degrade RhoB through CRLs to preserve endothelial integrity. The de-neddylation inhibitor CSN5i-3 caused similar barrier-disruptive results, which were also accompanied by an increase in RhoB levels, although in this case, this was a result of increased IκB degradation, NFκB activation and a subsequent increase in RhoB transcription (Kovačević et al., 2018; Marcos-Ramiro et al., 2016; Schaefer et al., 2014; Wojciak-Stothard et al., 2001). Thus, CRLs play divergent roles in endothelial integrity, in part direct, by ubiquitylating key signaling molecules, such as RhoB, and indirect, through the activation of inflammatory signaling pathways. Since the role of ubiquitylation in the localization and activity of Rho GTPases has been previously reviewed (Cai et al., 2018; de la Vega et al., 2011; Ding et al., 2011; Hodge and Ridley, 2016), here, we limit the discussion to their regulation of endothelial barrier function (Fig. 2).
In human embryonic kidney 233 (HEK293T) and mouse embryonic fibroblast (MEF) cells, RhoA is targeted for ubiquitylation and degradation at K5 and K6 by the HECT ligase Smurf1 (Boyer et al., 2006), an event localized at cellular protrusions (Wang et al., 2003). Smurf1 acts in a complex formed by Cdc42, PAR6 and PKCζ, which induces the ubiquitylation and degradation of RhoA. This in turn stimulates the localized activation of Rac1 and Cdc42, thereby identifying ubiquitylation as part of the mechanism by which RhoA inhibits signaling by Rac1 and/or Cdc42 (Izzi and Attisano, 2006; Sander et al., 1999). In brain ECs, crebral cavernous malformation protein 2 (CCM2) localizes Smurf1 to the plasma membrane, thereby targeting RhoA for degradation (Crose et al., 2009). To counteract RhoA degradation, the actin-associated protein synaptopodin can directly bind to RhoA, which blocks Smurf1-mediated ubiquitylation and degradation in podocytes (Asanuma et al., 2006). Recently, it was reported that HUVECs express synaptopodin when they are under laminar shear stress, suggesting that similar signaling takes place in the vascular endothelium (Mun et al., 2014).
The regulation of RhoB is significantly different from that of RhoA and RhoC. This is mainly due to the difference in their hypervariable C-terminal regions. The hypervariable region of RhoB contains more polar amino acids than that of RhoA and RhoC, and additional isoprenylation modifications have been reported for RhoB (Allal et al., 2002; Baron et al., 2000; Pronk et al., 2019). Unlike RhoA and RhoC, RhoB does not bind the ubiquitously expressed RhoGDI1 (also known as ARHGDIA) and is therefore prone to ubiquitylation and degradation (Garcia-Mata et al., 2011). As a result, ECs typically express only low levels of RhoB in resting conditions (Kovačević et al., 2018; Kroon et al., 2013).
Mechanistically, we found that in HUVECs, knockdown of components of the RING E3 ligase, which comprises the scaffold protein Cullin3, the adapter protein Rbx1 and its RhoB-binding substrate receptor KCTD10, increases the protein level and activation of RhoB, resulting in cell contraction and barrier disruption (Kovačević et al., 2018). We further identified the relevant ubiquitylation acceptor sites in RhoB as K162 and K181, following ectopic expression of RhoB loss-of-ubiquitylation (K-R) mutants in HEK293T cells. In HUVECs, ubiquitylation of RhoB at these two lysine residues leads to its lysosomal targeting and degradation (Kovačević et al., 2018). A recent study showed that KCTD10-mediated RhoB degradation in epithelial cells serves to allow Rac1 activation (Murakami et al., 2018). In HUVECs, Rac1 is a barrier-stabilizing Rho GTPase (see below). This suggests that KCTD10-mediated downregulation of RhoB not only limits contractility, but also promotes cell spreading and endothelial barrier stability. Conversely, RhoB has been shown to drive internalization of Rac1 in TNFα-treated HUVECs, which limits the capacity of Rac1 to stabilize the endothelial barrier (Marcos-Ramiro et al., 2016).
Unlike what is seen for RhoB, the majority of Rac1 in resting ECs is localized in the cytosol, where it is bound to the chaperone RhoGDI and therefore inactive (Boulter et al., 2010; Ren et al., 1999). Upon cell stimulation, Rac1 is released from the GDI to be activated at cellular membranes by a local GEF, followed by its interaction with nearby effectors (Bustelo et al., 2012). These include p21-activated kinases (PAKs), which induce lamellipodia formation (Byrne et al., 2016), partitioning defective (PAR)6, which is important for cell polarity (Lin et al., 2000) and IQGAPs, which increase cell–cell adhesion, proliferation and angiogenesis (Meyer et al., 2008), as well as, specifically, Rac1-associated 1 (SRA1), WASP-family verprolin-homologous (WAVE) proteins and p67phox, part of the NADPH oxidase complex that leads to ROS production (Hordijk, 2006). To limit localized Rac1 signaling, active Rac1 is ubiquitylated, which is accompanied by its internalization and leads to its proteasomal degradation (Nethe and Hordijk, 2010; Pop et al., 2004).
Inhibition of Rac1 degradation increases ROS production and disrupts the endothelial barrier; this occurs through various mechanisms, including disruption of the plasma and mitochondrial membrane through membrane lipid peroxidation, which reduces ATP generation and decreases metabolism and cell survival (Daugaard et al., 2013; Farber, 1994; Kovacic et al., 2001; van Wetering et al., 2002). To inhibit ROS production, the HECT E3 ubiquitin ligase HACE1 targets active Rac1 for degradation by ubiquitylation at K147 (Daugaard et al., 2013; Mettouchi and Lemichez, 2012; Torrino et al., 2011; Visvikis et al., 2008). In contrast, Rac1-mediated ROS production is increased by the E3 ligase TRAF6. TRAF6-mediated ubiquitylation of Rac1 occurs in response to H2O2 and IL-1β stimulation, and after ischemia-reperfusion injury, all of which also lead to a loss of endothelial barrier function (Li et al., 2006, 2017).
Several other ubiquitin ligases also target Rac1. K147 polyubiquitylation and Rac1 degradation can be mediated by X-linked IAP (XIAP), cIAP1 and cIAP2 in HeLa and HEK293T cells (Oberoi et al., 2012). In addition, the SCF-FBXL19 E3 ligase targets Rac1 K166 for ubiquitylation and degradation, an event which requires AKT-mediated phosphorylation of Rac1 at S71 (Zhao et al., 2013). Through this pathway, FBXL19 negatively regulates Rac1 signaling; this impairs cell migration and reduces endothelial barrier integrity (Pronk et al., 2019).
Collectively, these studies illustrate that Rho GTPases are subject to ubiquitylation by HECT and RING E3 ligases, which, in most cases, alters their localization and limits their abundance and signaling capacities. This is not unique for Rho-like small GTPases, as the activation and output of Ras and Rab GTPases are also controlled through ubiquitylation (Shin et al., 2017; Thurman et al., 2017).
Ubiquitin modifications at endothelial cell–cell junctions
Cell–cell contacts between ECs comprise different types of junctions (adherens, tight and gap junctions) with different cell adhesion molecules and regulators. A large, and growing, number of E3 ligases and some DUBs have been implicated in the control of intercellular contacts, both in epithelial and ECs (Table 1; Fig. 3). A selection of these, i.e. those relevant for endothelial integrity, is discussed below.
Kowalczyk and colleagues provided the first indications that the expression levels of VE-cadherin and its localization at junctions are regulated by controlled, lysosomal degradation in primary microvascular dermal ECs (Xiao et al., 2003). Subsequently, it was shown that VE-cadherin is internalized in a clathrin-dependent fashion in ECs (Chiasson et al., 2009; Xiao et al., 2005) The notion that the clathrin-mediated internalization of VE-cadherin is the result of its ubiquitylation was proposed by Orsenigo et al. (2012). These authors demonstrated that bradykinin induces tyrosine phosphorylation of VE-cadherin at Y658 and Y685, which drives its consequent ubiquitylation, internalization and degradation. Although the responsible ligase was not identified in this study, VE-cadherin ubiquitylation was suggested to occur through K63-linkage, in line with its lysosomal degradation (Orsenigo et al., 2012). Interestingly, VE-cadherin-associated p120 catenin (also known as CTNND1) protects VE-cadherin from internalization (Chiasson et al., 2009; Xiao et al., 2005); this is caused by the binding of p120 catenin to an endocytic motif in the intracellular part of VE-cadherin, which limits both its internalization as well as ubiquitylation. In Kaposi sarcoma, which is caused by human herpesvirus 8, VE-cadherin is ubiquitylated by the virus-encoded, transmembrane MARCH-family ubiquitin ligase K5 (Mansouri et al., 2008; Nanes et al., 2017) (see Box 1). This E3 ligase thus contributes to the sarcoma-associated vascular leakage and tumorigenesis though its effect on VE-cadherin ubiquitylation and degradation.
Viral infections with the highly pathogenic strains of the influenza A virus, or even more severely with hemorrhagic viruses (e.g. Hanta, Dengue or West Nile Virus) can induce disruption of the endothelial barrier, acute lung injury and shock. Viruses can both directly and indirectly disrupt the endothelial barrier (Armstrong et al., 2012; Basu and Chaturvedi, 2008). Accordingly, strengthening of the endothelial barrier by signaling through Tie2, the receptor for angiopoietin, or by Robo-4, the receptor for the chemorepellent SLIT1, improves lung injury and survival of mice in an influenza model (London et al., 2010). Influenza A virus induces degradation of the tight junction component claudin-5 in primary microvascular ECs and of ZO-1 in primary HUVECs and in mice (Armstrong et al., 2012; Wang et al., 2010). Furthermore, the H1N1 influenza strain enhances hyper-phosphorylation of β-catenin and its proteasomal degradation in primary HUVECs (Hiyoshi et al., 2015). In addition, the E3 ubiquitin ligase Itch was found to be necessary for uncoating of influenza A virus and its transport from endosomes to the nucleus (Su et al., 2013). In epithelial cells, Itch mediates degradation of occludin (Traweger et al., 2002); however, direct interaction partners for Itch that mediate virus-induced loss of endothelial cell–cell contact remain to be identified. Finally, infection of HUVECs or HMECs with dengue virus also causes disruption of adherens and tight junctions, changes in the actin cytoskeleton and reduced expression of several junctional proteins including PECAM-1/CD31 and VE-cadherin (Dewi et al., 2008; Kanlaya et al., 2009; Talavera et al., 2004). Although the molecular mechanisms of this downregulation are currently incompletely understood, examples of virus-encoded ubiquitin ligases do exist; an example is the MARCH-family ubiquitin ligase K5 from herpesvirus (see main text) (Mansouri et al., 2008).
Ubiquitylation of the tight junction protein occludin in ECs has been implicated in junctional stability (see Box 1 and Table 1). PKCβ-mediated phosphorylation of occludin at S490 (Murakami et al., 2012) and its subsequent ubiquitylation by the E3 ligase Itch have been linked to vascular endothelial growth factor (VEGF)-induced permeability in primary bovine retinal ECs (Murakami et al., 2012). Interestingly, these authors showed that a C-terminal occluding–ubiquitin chimera was internalized, bypassing the requirement for phosphorylation (Murakami et al., 2009). Thus, VEGF, through PKCβ-mediated phosphorylation, promotes Itch-mediated ubiquitylation of occludin, which is required for its internalization and degradation, thereby enhancing retinal endothelial permeability.
Furthermore, another study has shown that ischemia in the rat brain, induced by permanent middle cerebral artery occlusion, is associated with increased Itch-mediated ubiquitylation of occludin and loss of vascular integrity (Zhang et al., 2013). The ischemia also induced a downregulation of Notch1 and, in line with Notch regulation by γ-secretase, intraventricular treatment with the γ-secretase inhibitor DAPT prevented the ubiquitylation and degradation of occludin. This also reduced Evans Blue leakage of the brain vasculature, indicative of restored barrier function. The notion that Notch may stabilize junctions by limiting occludin ubiquitylation is intriguing, and is in line with the finding that DAPT-mediated inhibition of γ-secretase induces vascular barrier protecting properties (Zhang et al., 2013). Finally, the tight junction protein ZO-1, is targeted for ubiquitylation by the E3 ligase N-recognin-1 (Ubr1) (Chen et al., 2014). This is initiated through the inflammatory cytokine IL-6, which, in turn, is produced following an infection of brain pericytes with Japanese encephalitis virus (Chen et al., 2014). Thus, a growing number of ubiquitin ligases appears to be involved in regulation of junctional integrity. Some of these target junctional proteins directly, or control barrier function indirectly, by ubiquitylating regulatory proteins that, for example, control cytoskeletal dynamics.
Ubiquitin ligases and junctional regulation
A number of different E3 ligases have been linked to the control of endothelial junctional integrity (Table 1). For instance, ectopic expression of MARCH family of E3 ligases, MARCH2 and MARCH4, whose mRNA is expressed in primary dermal microvascular ECs, impairs VE-cadherin localization to junctions (Nanes et al., 2017). However, this does not exclude a mechanism that involves ubiquitylation of VE-cadherin-associated proteins, which could regulate its internalization. The MARCH3 E3 ligase also negatively regulates endothelial integrity, but in an indirect manner. In human brain ECs in which MARCH3 was downregulated, histamine-induced permeability was reduced (Leclair et al., 2016). In brain ECs transfected with siRNA targeting MARCH3, both the mRNA and protein levels of claudin-5 and occludin were increased, whereas the mRNA and protein levels of VE-cadherin were only increased slightly. Here, MARCH3 was found to act through the transcription repressor FoxO1 to reduce the mRNA expression levels of the tight junction proteins occludin and claudin-5. Thus, MARCH3-mediated reduction of mRNAs that encode tight junction proteins impairs adherens junction stability and endothelial integrity (Leclair et al., 2016). This mechanism is thus clearly different from the ubiquitylation of claudin-5 in human brain microvascular ECs by the HECT domain E3 ubiquitin protein ligase (HECTD)1 (Rui et al., 2018). This pathway was identified following Streptococcus infection, which induced the ubiquitylation and degradation of claudin-5, which weakened the blood–brain barrier and promoted further infection with the pathogen (Rui et al., 2018).
Furthermore, it has recently been shown that activation of CRLs by neddylation is required for endothelial integrity (Sakaue et al., 2017). Here, pharmacological inhibition of CRL neddylation by MLN4924 induces a strong increase in the permeability of HUVECs that is associated with the loss of VE-cadherin protein, but not a decrease in its mRNA levels. The study further showed that the cullin 3 scaffold protein is required for the stabilization of VE-cadherin and endothelial integrity. However, it was not established whether the reduced VE-cadherin protein levels were an indirect result of the loss of integrity, or whether CRLs directly control VE-cadherin stability by ubiquitylation (Sakaue et al., 2017).
As discussed above, the cullin3 scaffold, in complex with the adapter protein Rbx1, and the substrate receptor KCTD10, stabilizes the integrity of endothelial monolayers by limiting RhoB levels and signaling output (Kovačević et al., 2018). Intriguingly, KCTD10 also stimulates developmental angiogenesis in mouse embryos, as determined by knockout studies (Ren et al., 2014). Here, KCTD10, in conjunction with cullin 3, was suggested to ubiquitylate Notch1, which reduces Notch signaling in the vasculature. Notch1 positively regulates endothelial integrity by activating a non-canonical VE-cadherin–Trio–Rac1 signaling pathway (Polacheck et al., 2017). Thus, the role of the Cul3–Rbx1–KCTD10 RING ligase in ECs is complex, as it not only negatively regulates Notch1, thus reducing barrier function, but also downregulates RhoB, which results in barrier stabilization.
Another E3 ligase that negatively regulates endothelial integrity is PDZ domain-containing ring finger 3 (PDZRN3) (Sewduth et al., 2017). PDZRN3 acts downstream of the PAR3 polarity complex and targets the protein discs lost-multi-PDZ domain protein 1 (MUPP1, also known as MPDZ) for poly-ubiquitylation and degradation. This pathway was identified in infarcted mouse brain, thereby implicating PDZRN3 as an important mediator of a compromised blood–brain barrier and tissue damage in acute ischemic stroke (Sewduth et al., 2017).
In contrast to MARCH3 and PDZRN3, the endothelial E3 ligase HECW2 (also known as NEDL2), a member of the NEDD4 family of ligases that also includes Itch, stabilizes endothelial junctions (Choi et al., 2016). Accordingly, siRNA-mediated loss of HECW2 reduces endothelial barrier function and promotes angiogenic sprouting. The authors suggest that this occurs through ubiquitylation-mediated stabilization, rather than destabilization, of the junctional protein AMOTL1 (Choi et al., 2016).
Although there is accumulating data on the role of ubiquitin ligases in endothelial integrity, only very little is known with regard to the role of DUBs in barrier regulation. For instance, the DUB Cezanne (also known as OTUD7B) has been shown to protect against hypoxia-induced NFκB-mediated inflammation in kidney ECs in vivo. This occurred through de-ubiquitylation of the E3 ligase TRAF6, which is part of the NFκB pathway (Luong et al., 2013). Another example is the DUB USP40, which is particularly highly expressed in glomerular ECs, as well as in podocytes in rats and mice (Takagi et al., 2017). In in vivo experiments in zebrafish, USP40 morpholinos induced cardiac edema and loss of glomerular permeability. Although the USP40 targets were not identified, its association with the intermediate filament protein nestin suggests that it is an endothelial integrity-stabilizing DUB that acts on cell–cell junctions through intermediate filaments (Takagi et al., 2017).
Ubiquitin-dependent modifications in vascular disease
It is not surprising, given the abundance and importance of protein ubiquitylation, that this process (and its deregulation) contributes to a range of disorders, including neurodegenerative and inflammatory diseases, as well as cancer (Lipkowitz and Weissman, 2011; Rape, 2018). Consequently, proteasome inhibition has already been in use as a therapeutic intervention for more than two decades; however, such an approach obviously shows limited specificity, and the use of proteasome inhibitors is accompanied by (cardiovascular) side effects (Cole and Frishman, 2018; Gavazzoni et al., 2018).
Most, if not all, vascular disorders are accompanied by a loss of endothelial integrity, which causes edema and tissue damage owing to elevated interstitial pressure and increased influx of activated leukocytes (Huveneers et al., 2015; Nourshargh et al., 2010). It is therefore imperative to understand in detail the different molecular mechanisms that control stable, as well as disrupted, endothelial barrier function. While the role of ubiquitylation in vascular pathologies is perhaps less well established compared to that in, for instance, cancer, some clear connections exist.
Probably the best-studied pathway involves the regulation of (tumor) angiogenesis through the degradation of hypoxia inducible (HIF) transcription factors by the von Hippel–Lindau (VHL) protein, a substrate receptor that is part of a Cul2–Rbx1-containing RING ubiquitin ligase (Kamura et al., 2004). Loss of VHL promotes VEGF expression and tumor vascularization as a result of HIF1 being stabilized (Robinson and Ohh, 2014). Several other studies have implicated (de-)ubiquitylation in hypoxia and angiogenesis, either through ubiquitin-mediated interactions between VEGFR2 and epsin1, which drives angiogenesis and wound healing (Rahman et al., 2016), or through sumoylation of Notch1, which controls VEGF receptor (VEGFR) signaling and angiogenesis (Zhu et al., 2017). Interestingly, Notch1 ubiquitylation by the FBXW7 RING ligase is required for angiogenesis in vitro and in vivo (Izumi et al., 2012). In good agreement with this, loss of Usp10, the DUB for Notch1, promotes in vivo vessel sprouting (Lim et al., 2019). Finally, VEGFR2 has been identified as a substrate for several ubiquitin ligases, including SCF-βTRCP, CHIP (also known as STUB1) and Cbl, which all regulate angiogenesis through directly targeting VEGFR2 (Duval et al., 2003; Shaik et al., 2012; Sun et al., 2015).
As discussed above, ubiquitin ligases and/or DUBs have been implicated in inflammatory vascular disorders, for example in the lung or the brain (Hartz et al., 2016; Li et al., 2018; Liu et al., 2017; Rape, 2018). Aberrations in the TNF signaling pathway underlie several human pathologies, as has been corroborated by functional studies in animal models. Mutations in TNFR1 or the LUBAC component HOIL-1 (also known as RBCK1) cause inflammation in affected individuals (Boisson et al., 2012; McDermott et al., 1999). Furthermore, genetic depletion of the DUBs A20 or CYLD in mice rendered them more susceptible to inflammatory bowel disease (Vereecke et al., 2014; Zhang et al., 2006). A20 was recently proposed to act as a DUB for VE-cadherin, thereby preserving endothelial barrier function, and its re-expression in A20-deficient mice was found to limit their lung permeability (Soni et al., 2018). Conversely, the RBR E3 ligase Parkin, originally linked to Parkinson disease, was recently shown to mediate vascular permeability both in vitro and in vivo in a study showing that Parkin-deficient mice are protected from LPS-induced acute inflammation and leakage in the lung (Letsiou et al., 2017).
A deletion in the cullin 3 scaffold (cullin3Δ9, a deletion of 57 amino acids encoding exon 9) drives vessel wall stiffness and hypertension due to impaired turnover of RhoA, resulting in increased smooth muscle cell contractility (Agbor et al., 2016). Conversely, the cullin 3 substrate adapter RhoBTB1 protects from arterial stiffness and hypertension through its ubiquitylation and the consequent degradation of the phosphodiesterase PDE5. Lower amounts of PDE5 lead to increased cGMP levels, which in turn promotes smooth muscle cell relaxation (Mukohda et al., 2019). This important role for cullin 3 in vascular smooth muscle cells is in good agreement with its degradation of RhoB in ECs, which also limits contraction and preserves endothelial integrity (Kovačević et al., 2018; Sakaue et al., 2017). Taken together, a growing number of ubiquitin ligases and DUBs has been implicated in vascular disorders. Targeting these individual ligases, based on detailed structural and mechanistic studies, is of key importance to selectively limit chronic inflammation, together with its associated loss of barrier function and tissue damage.
The ubiquitous nature of protein regulation through controlled degradation makes it obvious that this process is important, and it is also important for endothelial permeability and associated vascular pathologies, such as tumor angiogenesis and inflammation. The above overview aims to highlight our growing knowledge on the role of protein ubiquitylation in endothelial integrity. Although several of these ubiquitin ligases have already been implicated in vascular permeability, many questions remain unanswered. For instance, what is the ubiquitin ligase for VE-cadherin? Do endothelial-specific ubiquitin ligases or DUBs exist for the control cell–cell adhesion? If so, do these show vascular-bed-specific distribution and how might they be targeted to preserve vascular integrity and limit inflammation? Another interesting aspect is that autophagy, which comprises lysosomal degradation and recycling of proteins, is vascular barrier-protective in both the brain and the lung (Slavin et al., 2018; Yang et al., 2019). Since K48 ubiquitylation drives lysosomal degradation, this type of ubiquitin modification may directly link cellular homeostasis, controlled by autophagy, to endothelial integrity. This may also prove pivotal in the aging-related loss of autophagy, which correlates with an increase in cardiovascular disease (Leidal et al., 2018).
The above overview also underscores how many different ubiquitin ligases are mechanistically, both directly and indirectly, linked to the control of endothelial integrity. This apparent excess of regulators is not unique and is similar to, for example, the large number of RhoGEFs and GAPs (in total over 150) that regulate only ∼20 Rho GTPases (Bos et al., 2007). It is likely that the relatively crude way in which ubiquitin ligases have been studied so far obscures differences in their specific localization or in the conditions during which one or the other ligase or DUB is most relevant. On top of this, cell-type-specific differences in expression, even between EC subtypes, is likely to play a role, in addition to divergent culture conditions or cellular stimulation. The development of specific antibodies for detection of individual E3 ligases or DUBs, as well as of the ubiquitin chains on specific substrates, use of super-resolution imaging and careful definition of the cell type and conditions used, will increase our insight in this complex mode of cellular signaling. Clearly, protein ubiquitylation and its intersection with critical regulatory pathways predicts a busy, but also very interesting future for this field of research.
Our work in this area is supported by the Landsteiner Foundation for Blood Transfusion Research (to I.K., grant 1311).
The authors declare no competing or financial interests.