Differential role for rapid proteostasis in Rho GTPase-mediated control of quiescent endothelial integrity

Abstract

Endothelial monolayer permeability is regulated by actin dynamics and vesicular traffic. Recently, ubiquitination was also implicated in the integrity of quiescent endothelium, as it differentially controls the localization and stability of adhesion and signaling proteins. However, the more general effect of fast protein turnover on endothelial integrity is not clear. Here, we found that inhibition of E1 ubiquitin ligases induces a rapid, reversible loss of integrity in quiescent, primary human endothelial monolayers, accompanied by increased F-actin stress fibers and the formation of intercellular gaps. Concomitantly, total protein and activity of the actin-regulating GTPase RhoB, but not its close homolog RhoA, increase ∼10-fold in 5 to 8 h. We determined that the depletion of RhoB, but not of RhoA, the inhibition of actin contractility, and the inhibition of protein synthesis all significantly rescue the loss of cell–cell contact induced by E1 ligase inhibition. Collectively, our data suggest that in quiescent human endothelial cells, the continuous and fast turnover of short-lived proteins that negatively regulate cell–cell contact is essential to preserve monolayer integrity.

The most inner lining of blood vessels is formed by the endothelium which acts as a dynamic barrier controlling the extravasation of leukocytes and plasma from the circulation into the tissue. Dysfunction of the endothelial barrier leads to vascular leak and edema and is a hallmark of chronic inflammatory diseases. Thus, control of endothelial integrity is crucial for tissue and organ health. This integrity is critically dependent on cell–cell adhesion between endothelial cells (ECs), which is mediated by the homotypic adhesion protein vascular endothelial (VE)-cadherin. Intracellular adaptor proteins link VE-cadherin to the cortical actin cytoskeleton, stabilizing its adhesive function (1). Importantly, both this connection and the organization of the actin cytoskeleton are dynamic and subject to complex regulation. For example, agonist-induced changes in cytoskeletal organization in combination with the induction of myosin-based contraction can generate contractile forces that disrupt VE-cadherin–based complexes and impair vascular integrity (2, 3, 4).

The key regulators of cytoskeletal dynamics in ECs are members of the family of Rho GTPases. RhoA/B activity promotes F-actin stress fiber assembly and myosin activation leading to cell contraction, loss of cell–cell contact, and an increase in vascular permeability. In contrast, Rac1 promotes actin polymerization and the formation of a cortical F-actin network improving endothelial integrity (5, 6). Signaling by Rho GTPases is primarily controlled by their cycling between an inactive, GDP-bound state and an active, GTP-bound state. This cycle is tightly regulated both in time and space by activating guanine nucleotide exchange factors (GEFs) and inactivating GTPase-activating proteins (GAPs) (7, 8).

In addition to their GTP/GDP cycling, Rho GTPases can be posttranslationally modified by lipidation, phosphorylation, sumoylation, or ubiquitination to further regulate their localized signaling capacity. Ubiquitination entails the covalent attachment of the highly conserved, 76 amino acid ubiquitin peptide to lysine residues of a substrate and subsequently to lysine residues of ubiquitin itself (known as polyubiquitination). The formation of such chains through ubiquitin K48 or K63 residues targets the substrate for proteasomal or lysosomal degradation, respectively (9, 10). Mechanistically, E1, E2, and E3 ubiquitin ligases act in a cascade, with the E3 ligase transferring the activated ubiquitin to the substrate (11). Rho GTPases are targeted by several ubiquitin E3 ligases. For example, HACE1 and inhibitors of apoptosis proteins (IAPs) target Rac1 for degradation, thereby controlling the generation of reactive oxygen species in human umbilical vein endothelial cells (HUVECs) as well as cell spreading and migration in HEK293 and HeLa cells (12, 13, 14). RhoA stability is regulated by the Smurf1 and Cullin3-BACURD E3 ligases to control cell polarity and cell migration in tumor and HeLa cells, respectively (15, 16). In addition, Rho targeting by Cullin3 limits smooth muscle cell contractility and hypertension (17, 18).

The ubiquitously expressed E1 ligases ubiquitin-activating enzyme 1 (UBA1) and UBA6 initiate the ubiquitination cascade by activating ubiquitin and charging the E2 ligases (19). Recently, Hyer et al. identified a small molecule inhibitor named MLN7243 (a.k.a. TAK-243) targeting both UBA1 and UBA6, thus inhibiting cellular ubiquitination (20). Prolonged inhibition of the E1 enzymes leads to impairment of DNA damage repair, cell cycle progression, and proliferation and exhibits antitumor activity in various tumor cell lines and in mouse xenograft models for solid and hematological cancers (20, 21, 22, 23, 24). The therapeutic potential of MLN7243 is currently being assessed in patients with advanced solid tumors and leukemia (ClinicalTrials.gov Identifier: NCT02045095 and NCT03816319).

We previously identified RhoB, rather than RhoA, as a major negative regulator of monolayer integrity in quiescent ECs (5, 25). RhoB has a very short half-life (1–2 h), as compared to RhoA (8–12 h), and we showed that ubiquitination of RhoB by the Cul3-Rbx1-KCTD10 E3 ligase is a key regulatory event controlling endothelial cell–cell contact (26). However, whether there exists a more generic role for fast protein turnover in regulating endothelial integrity is unknown. To address this, we used short-term inhibition of the E1 ubiquitin ligases in quiescent, primary human ECs and analyzed the consequences for barrier integrity as well as for the turnover of RhoA, RhoB, and Rac1. Our results show that E1 inhibition rapidly (within 5–8 h) and reversibly induces marked cytoskeletal changes and a pronounced loss of endothelial integrity. This is accompanied by a fast accumulation of total and activated RhoB, but not RhoA or Rac1. Finally, this effect is completely dependent on protein synthesis. Together, these results support the concept that continuous and rapid protein turnover is required to preserve the integrity of quiescent endothelium.

Results

Ubiquitination preserves endothelial barrier function

To establish the importance of ubiquitination for endothelial barrier integrity, the ubiquitination pathway in primary HUVEC monolayers was inhibited by the E1 ligase inhibitor MLN7243 (TAK-243; Fig. 1A). MLN7243 caused a rapid, dose-dependent reduction of endothelial integrity within 4 to 8 h, as analyzed in real time using electric cell-substrate impedance sensing (ECIS) (Fig. 1, B and C). Additionally, increased permeability of endothelial monolayers was observed after treatment with MLN7243, as quantified by horseradish peroxidase (HRP) leakage in a transwell assay (Fig. 1D). The total pool of ubiquitinated proteins decreased accordingly in a dose- and time-dependent manner (Fig. S1A). Since ubiquitination plays a crucial, regulatory role in many signaling events, inhibition of the E1 ligase may eventually compromise cell viability. To address this, we determined the amount of cell death induced by treatment with MLN7243. Only 16 h and 24 h of treatment resulted in 24% and 46% of PI⁺ cells, respectively, whereas no increase in cell death was observed during the first 8 h of treatment (Fig. S1B). Additionally, we studied time-dependent changes in caspases, marking the induction of apoptosis. MLN7243 treatment of HUVECs resulted in a reduction of caspase 9 and an increase in cleaved caspase 3 as detected after 16 h, but not during the first 8 h of treatment (Fig. S1C), indicating that the MLN7243-induced loss of monolayer integrity was not due to apoptosis or cell death.

Figure 1.

MLN7243 disrupts endothelial barrier integrity. A, MLN7243 inhibits the E1 ligases UBA1 and UBA6. B and C, normalized resistance of HUVECs treated with different concentrations of MLN7243. C, bar graph represents normalized endothelial resistance after 2 h, 5 h and 8 h of MLN7243. Data are mean + SD of three individual experiments. D, macromolecule passage of HRP across HUVECs monolayer treated with 500 nM MLN7243, normalized to their respective DMSO control. Data are presented as mean + SD, n = 3. E, normalized resistance of HUVECs treated with 500 nM MLN7243 and refreshed with medium without MLN7243 after 7 h of treatment. F, monolayer of HUVECs treated with 500 nM MLN7243 for the indicated times were stained for F-actin (white) and VE-cadherin (green) and counterstained with DAPI (blue). Scale bars represent 50 μm in overview images and 15 μm in zoomed images. G, Western blot analysis of VE-cadherin expression after 500 nM MLN7243 treatment for the indicated times. β-tubulin was used as loading control. Bar graph shows quantification of VE-cadherin expression normalized to β-tubulin. Data are presented as mean + SD, n = 3. ns, nonsignificant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. HRP, horseradish peroxidase; HUVEC, human umbilical vein endothelial cell; UBA1, ubiquitin-activating enzyme 1; VE, vascular endothelial.

To further exclude that the MLN7243-induced loss of integrity was the result of irreparable cell damage, we tested the reversibility of its effect. We found that the reduction of transendothelial resistance, induced by MLN7243, was fully restored when the compound was removed after 7 h by washout. This induced a complete recovery of endothelial electrical resistance within 10 h (Fig. 1E), indicating that reformation of cell–cell contact, rather than increased cell proliferation, underlies the restoration of barrier function. MLN7243 did not induce an upregulation of ICAM1, suggesting that E1 inhibition does not trigger an inflammatory response, which is generally accompanied by a loss of barrier function (Fig. S1D). Similar as for HUVECs, inhibiting ubiquitination with MLN7243 in human microvascular endothelial cells (hMVECs) resulted in a dose-dependent impairment of endothelial integrity (Fig. S1E). Based on these initial findings, we decided for subsequent experiments to limit both the dose (500 nM) and duration (max. 8 h) of treatment with MLN7243.

To further understand the barrier-disruptive effect of E1 inhibition, distribution of VE-cadherin and F-actin was analyzed by immunofluorescence microscopy. At early time points of incubation with MLN7243 (1 h and 2 h), cell–cell junctions appeared stable as indicated by honeycomb-like structures marked by VE-cadherin and cortical F-actin. However, after 5 h and 8 h of exposure to MLN7243, F-actin stress fibers were strongly increased, and VE-cadherin localization at junctions was markedly reduced (Figs. 1F and S1F). The levels of VE-cadherin protein remained unaltered (Fig. 1G), suggesting an indirect effect of MLN7243 on VE-cadherin distribution. Thus, short term inhibition of ubiquitination in EC has inhibitory, but reversible effects on VE-cadherin–mediated cell–cell contact. These data show that in quiescent, primary human EC, there is continuous ubiquitin-mediated degradation of barrier-disrupting protein(s), as part of a dynamic proteostatic mechanism that acts at short time scales to preserve monolayer integrity.

MLN7243 causes accumulation of active RhoB, but not RhoA

Rho GTPases are key regulators of actin filament organization. RhoA/B/C, which are identical in their effector domain (27), promote formation of contractile F-actin fibers via their shared effector Rho-associated kinase (ROCK). ROCK induces the inhibitory phosphorylation of myosin phosphatase targeting subunit 1 (MYPT1), which, in turn, regulates phosphorylation of myosin light chain (MLC) and the induction of actomyosin contraction (28, 29). The increased F-actin stress fiber formation which we observed following E1 inhibition in HUVECs with MLN7243 (Fig. 1F) was accompanied by elevated levels of phosphorylated MLC (Fig. 2A) and phosphorylated MYPT1 (Fig. S2A). To test if cell contraction upon E1 inhibition is mediated by ROCK, HUVECs were pretreated with the ROCK inhibitor Y27632. This significantly reduced MLN7243-induced disruption of endothelial integrity, indicative for a role of ROCK-mediated contraction (Fig. 2, B and C).

Figure 2.

Inhibition of ubiquitination by MLN7243 leads to increased RhoB expression and activity.A, Western blot analysis for pMLC of HUVECs treated with 500 nM MLN7243 for the indicated times. GAPDH was used as loading control. Bar graph shows quantification of pMLC expression normalized to GAPDH. Data are presented as mean + SD, n = 3. B and C, normalized resistance of HUVECs treated with 500 nM MLN7243, 10 μM Y27632, or both, and (C) relative drop in resistance after 4 h. Data are presented as mean + SD, n = 3. D, Western blot analysis and quantification of RhoA, RhoB, and Rac1 protein expression in HUVECs treated with 500 nM MLN7243. Bar graph shows quantification of RhoA/B, Rac1 expression, relative to GAPDH and DMSO. Data are presented as mean + SD, n = 3 to 4. E, immunofluorescent staining for RhoB (red) and counterstained with DAPI (blue) of HUVECs after treatment with 500 nM MLN7243. Scale bars represent 50 μm in overview images and 15 μm in zoomed images. Bar graph shows fluorescent intensity of RhoB staining. Data are represented as mean + SD of 10 regions of interest normalized to DMSO. F, Rhotekin pulldown of HUVECs treated with 500 nM MLN7243 followed by Western blot analysis for the GTP-bound, active forms of RhoA and RhoB. GAPDH was used as loading control. Bar graphs show quantification of RhoA/B.GTP normalized to GAPDH. Data are presented as mean + SD, n = 3. G, endogenous RhoB was immunoprecipitated from HUVECs treated with 500 nM MLN7243 followed by Western blot analysis of RhoB and ubiquitinated proteins (FK2). Input equals 8%. H, Western blot analysis of RhoB expression after treatment with 500 nM MLN7243 for the indicated times and with a medium refresh without MLN7243 after t = 7 h (right panel only). ns, nonsignificant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. HUVEC, human umbilical vein endothelial cell; MLC, myosin light chain.

As ROCK acts downstream of Rho GTPases, we investigated if the short-term inhibition of ubiquitination alters the levels and activity of the main barrier regulating Rho GTPases RhoA, RhoB, and Rac1. Strikingly, MLN7243 induced a 5- to 10-fold, significant increase in RhoB protein level within 5 to 8 h, whereas the levels of RhoA and Rac1 showed only little (∼1.4 fold for RhoA) to no increase (Rac1) (Fig. 2D). Similar to the MLN7243-induced loss of endothelial integrity, the MLN7243-induced increase in RhoB level was dose-dependent (Fig. S2B). Similar as in HUVECs, MLN7243 caused a strong accumulation of RhoB in hMVECs, whereas RhoA was only slightly increased (Fig. S2C). In line with these findings, immunofluorescence microscopy revealed a significant increase in RhoB levels after 5 h and 8 h of exposure of HUVECs to MLN7243 (Fig. 2E). Under these conditions, RhoB localizes to intracellular vesicles throughout the cytosol and close to or at cell–cell contacts. In parallel, RhoB signaling activity, analyzed by Rhotekin pulldown, was significantly and markedly increased after 5 h and 8 h of MLN7243 treatment (Fig. 2F). This was in marked contrast to the effect on RhoA which showed only a limited, nonsignificant increase in activity (Fig. 2F). The loss of RhoB ubiquitination induced by MLN7243 was further confirmed by an immunoprecipitation (IP) of endogenous RhoB in HUVECs. After 5 h and 8 h of MLN7243 treatment, the amount of poly-ubiquitinated RhoB was clearly diminished (Fig. 2G). Finally, similar to the loss of integrity, the MLN7243-induced increase in RhoB was reversible since washout of MLN7243 after 7 h caused a strong reduction of RhoB protein level, comparable to control levels (Fig. 2H). These data show that continuous ubiquitination plays a crucial role in limiting the levels of endothelial RhoB protein and activity, with only minor effects on RhoA.

In various cancer cell lines, the inhibition of the E1 ligases results in the accumulation of myriad proteins and is associated with the induction of ER stress (20, 24, 30). In line with this, MLN7243 induced ER stress in HUVECs after 8 h as indicated by elevated phosphorylation of eIF2α (Fig. S2D). Induction of ER stress by Tunicamycin (TM) resulted in a rapid disruption of the endothelial integrity, comparable to the effects of MLN7243 (Fig. S2, E and F). While this is in line with the notion that ER-stress impairs endothelial integrity (31, 32), TM treatment did not increase RhoB levels (Fig. S2E). This indicates that the observed increase in RhoB protein, induced by MLN7243, is consequent to the inhibition of ubiquitination rather than to ER-stress.

Ubiquitination of RhoB, but not RhoA, is crucial for endothelial barrier function

Based on the strong upregulation of RhoB following E1 inhibition, we next analyzed the requirement for RhoB expression in the inhibitory effects of MLN7243 on endothelial integrity. Following the siRNA-induced loss of RhoB and RhoA expression, HUVECs were treated with MLN7243, and barrier function was measured by ECIS. Interestingly, silencing of RhoB partly rescued the loss of barrier function induced by MLN7243, whereas depletion of RhoA did not (Fig. 3, A and B). Importantly, siRNA-mediated loss of RhoB induces an increase of RhoA protein level and vice versa (Fig. 3C) (5). Since this might partially compensate for the loss of RhoB or RhoA function, confounding the interpretation of this experiment, we also tested the effects of simultaneous depletion of RhoA and RhoB. Combined knockdown of these two Rho GTPases resulted in a pronounced and significant protection of endothelial barrier integrity (Fig. 3, D and E), largely preventing the effect of MLN7243. This indicates that lack of, primarily, RhoB makes the endothelial barrier refractory to the short-term effects caused by inhibition of cellular ubiquitination. These data support the notion that continuous ubiquitination and degradation of RhoB is crucial to preserve barrier function in quiescent ECs.

Figure 3.

Depletion of RhoB partially rescues the MLN7243-induced loss of endothelial integrity. A–E, HUVECs were transfected with nontargeting control siRNA (N.T.), siRhoA, and/or siRhoB and cultured for 72 h. A and B, normalized resistance of HUVECs with single knockdowns treated with 500 nM MLN7243. B, bar graphs represent mean + SD of normalized endothelial resistance after 5 h of four individual experiments relative to DMSO. C, Western blot analysis of RhoB and RhoA expression. Bar graphs show quantified RhoB and RhoA expression normalized to GAPDH and N.T. or N.T./N.T., respectively. Data are presented as mean + SD, n = 3. D and E, normalized endothelial resistance of HUVECs with double knockdowns of RhoA/B treated with 500 nM MLN7243. E, bar graphs represent mean + SD of normalized endothelial resistance after 5 h, relative to the DMSO control, of four individual experiments. ns, nonsignificant, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. HUVEC, human umbilical vein endothelial cell.

Inhibition of protein translation prevents the loss of endothelial barrier function induced by MLN7243 and MLN4924

The rapid effects of MLN7243 on barrier integrity suggest that dynamic proteostasis controls endothelial integrity in quiescent monolayers. However, protein turnover is a function of both ubiquitin-mediated degradation and protein synthesis. To address the contribution of protein synthesis to this effect, HUVECs were briefly (2 h) pretreated with the protein translation inhibitor cycloheximide (CHX), and barrier function was measured with ECIS. Intriguingly, CHX by itself caused a marked improvement of endothelial integrity within 6 to 8 h (Fig. 4, A and B), indicating that in quiescent ECs, barrier-disrupting proteins with a very short half-life are continuously synthesized. In line with this notion, CHX completely prevented the barrier-disruptive effect of MLN7243 (Fig. 4, A and B). Since RhoB is known to have a short half-life (1–2 h) (Fig. S3A) (33), we tested if CHX affects the MLN7243-induced accumulation of RhoB. Indeed, inhibition of translation completely blocked the increase in RhoB, whereas the levels of RhoA and Rac1 (both with longer half-lives; Fig. S3A) remained unaffected (Fig. 4, C–F). Concomitantly, CHX prevented both the loss of junctional localization of VE-cadherin and formation of F-actin stress fibers, induced by E1 inhibition (Fig. 4G). Importantly, CHX did not alter VE-cadherin levels or junctional distribution (Figs. S3B and 4G) or Rac1 levels or activity within the time frame of the experiment (Fig. S2C). These data suggest that the ongoing synthesis of signaling-competent RhoB plays a prominent role in actomyosin contractility and (local) barrier instability, which is kept in check by its fast and continuous ubiquitination and degradation.

Figure 4.

Regulation of RhoB turnover is crucial for endothelial integrity. A–F, HUVECs were treated with 500 nM MLN7243, following pretreatment as indicated for 2 h with 0.5 μg/ml CHX to block protein synthesis. A and B, normalized resistance. B, bar graphs represent mean + SD of normalized endothelial resistance after 8 h, relative to DMSO, n = 3. C, Western blot analysis for RhoB, RhoA and Rac1 expression. GAPDH is used as loading control. Quantification of (D) RhoB, (E) RhoA, and (F) Rac1 expression relative to GAPDH and DMSO. Data are represented as mean + SD, n = 3. G, immunofluorescent staining for RhoB (red), F-actin (white), and VE-cadherin (green) and counterstained with DAPI (blue) of HUVECs after pretreatment of 0.5 μg/ml CHX for 2 h followed by 5 h MLN7243 or MLN4924. Scale bar represents 50 μm; ns, nonsignificant, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. CHX, cycloheximide; HUVEC, human umbilical vein endothelial cell; VE, vascular endothelial.

Previously, we showed that inhibition of Cullin-based E3 ligases by the neddylation inhibitor MLN4924 leads to an accumulation of RhoB and a concomitant loss of endothelial barrier function (Fig. S2D) (26). Thus, we investigated if these MLN4924-induced effects are also dependent on protein synthesis. In accordance with the above, CHX completely blocked barrier breakdown by Cullin inhibition and prevented the concomitant increase in RhoB protein level (Fig. S3, E and F). Consistently, inhibition of protein synthesis also prevented the stress fiber formation and cortical actin bundles induced by MLN4924 (Fig. 4G). This further confirms the notion that RhoB synthesis and its Cullin3-mediated degradation is tightly regulated to preserve endothelial integrity.

Regulation of RhoB synthesis in inflamed endothelium

While the above experiments are all performed in quiescent endothelial monolayers, various receptor agonists are known to activate ECs and disrupt monolayer integrity within a ∼4 to 8 h time frame. The pro-inflammatory cytokine tumor necrosis factor (TNF) α is known to enhance vascular permeability and to promote transcription of RhoB mRNA leading to elevated RhoB protein level and activity, all within 4 to 8 h (34). In accordance with these findings, both the TNFα-induced increase in RhoB protein and the concomitant loss of endothelial integrity were prevented by CHX (Fig. 5, A–C). Similarly, the pro-inflammatory cytokine interleukin-1β (IL-1β) induced significant accumulation of RhoB and disruption of barrier function, which were both prevented by CHX (Fig. 5, E–G). In contrast, RhoA protein levels were not affected by TNFα, IL-1β, or CHX treatment (Fig. 5, B, D, F and H). This suggests that tight regulation of RhoB turnover is crucial not only in quiescent but also in inflamed endothelium.

Figure 5.

CHXP prevents loss of endothelial integrity induced by pro-inflammatory cytokines.A–D, (A) normalized resistance and (B–D) Western blot analysis for expression of RhoB and RhoA expression after HUVECs were pretreated with 0.5 μg/ml CHX for 2 h and treated with 10 ng/ml TNFα. GAPDH and β-tubulin were used as loading controls. Bar graphs show quantification of (C) RhoB and (D) RhoA expression normalized to GAPDH or β-tubulin and ctrl. Data are presented as mean + SD, n = 3. E–H, (E) normalized resistance and (F–H) Western blot analysis for RhoA and RhoB expression after HUVECs were pretreated with 0.5 μg/ml CHX for 2 h and treated with 10 ng/ml IL-1β for indicated times. GAPDH was used as loading control. G and H, bar graphs show quantification of (G) RhoB and (H) RhoA expression normalized to GAPDH and PBS. Data are presented as mean + SD, n = 3. I, normalized resistance of HUVECs after pretreatment with 0.5 μg/ml CHX for 2 h and treatment with 100 μM TFLLR. ns, nonsignificant, ∗∗p < 0.01. CHX, cycloheximide; HUVEC, human umbilical vein endothelial cell; IL-1β, interleukin-1β.

In addition to TNFα, activation of the G-protein coupled protease-activated receptor 1 (PAR1) by thrombin or the thrombin receptor peptide (35) is a widely used approach to investigate the mechanisms underlying endothelial integrity. Activation of PAR1 induces a very fast and transient loss of endothelial barrier function, which is accompanied by, and dependent on, a pronounced activation of RhoA (36). Interestingly, inhibition of protein synthesis did not affect the thrombin receptor peptide–induced transient loss of barrier function (Fig. 5I). Together, these experiments show that the relatively slow and persistent induction of barrier loss induced by TNFα and IL-1β is mediated by (upregulation of) RhoB, while the fast and transient effect induced by PAR1 is mediated by the activation of RhoA and independent of protein translation.

Discussion

In this study, we report that inhibiting cellular ubiquitination by the E1 ligase blocker MLN7243 (a.k.a. TAK-243) in primary human venous and microvascular EC causes a marked increase in densely packed F-actin stress fibers and an increase in phosphorylation of MLC. The resulting contraction is accompanied by a redistribution of VE-cadherin away from cell–cell junctions and a loss of cell-cell contact. These effects occur already within 5 to 8 h, are fully reversible and cannot be explained by ER stress or apoptosis. In contrast, they are completely dependent on protein synthesis. Together, these data show that the continuous synthesis and ubiquitin-mediated degradation of proteins with a short half-life plays a central role in the integrity of quiescent endothelial monolayers.

We recently identified Cullin3-Rbx1-KCTD10 as the E3 ligase targeting RhoB for K63 poly-ubiquitination and lysosomal degradation in quiescent human ECs (26). Similarly, the Cullin3-KCTD10 E3 ligase has been recently shown to target RhoB for degradation in HER2-positive breast cancer cells (37). In marked contrast to RhoA and Rac1, RhoB has a very short half-life of 1 to 2 h (33). This is likely due to the fact that RhoB does not bind the ubiquitously expressed chaperone RhoGDI1, which is known to protect GTPases from activation and degradation (38, 39, 40, 41). Consequently and because of the abundance of GTP in the cytosol, newly synthesized RhoB rapidly becomes GTP-bound, signaling competent, and sensitive to degradation. Consistent with this idea, our data suggest that the regulation of RhoB output occurs predominantly at the level of its expression and degradation. We show that inhibiting cellular ubiquitination, i.e., blocking the E1 ubiquitin ligase by MLN7243, leads to a rapid ≥10-fold increase in total RhoB protein level and activity, coinciding with the induction of contraction and disruption of the endothelial barrier. Notably, silencing of RhoB, but not of RhoA, significantly rescues the MLN7243-induced loss of barrier function. These data support the idea that continuous ubiquitin-mediated degradation of RhoB is essential to preserve quiescent endothelial integrity.

RhoB and RhoA share 88% sequence homology (27) and are both well-known negative regulators of endothelial barrier function by inducing ROCK-mediated F-actin contraction (6). Ubiquitination of active and inactive RhoA by various E3 ligases and in various different cell types has been described previously. Smurf1, SCF(FBXL19), and Cullin3-BACURD were identified as E3 ligases that mediate ubiquitination of (active and inactive) RhoA, leading to its degradation. This mechanism contributes to the regulation of tumor cell motility, neurite outgrowth, proliferation, and blood pressure (15, 16, 17, 42, 43). Interestingly, in ECs, IAPs were suggested to modulate RhoA activity and, thereby, endothelial barrier function in response to thrombin (44). In contrast, our data show that in quiescent EC, RhoA ubiquitination does not significantly control barrier function as E1 ligase inhibition induces only a small increase in RhoA abundance and activity. Additionally, depletion of RhoA only marginally improves endothelial resistance when the ubiquitination cascade is inhibited. This indicates that ubiquitin-mediated degradation of RhoA plays only a minor role in the integrity of quiescent endothelial monolayers.

Protein turnover is not only governed by the rate of ubiquitin-dependent degradation but also by protein synthesis. Intriguingly, inhibition of protein synthesis is sufficient to enhance quiescent endothelial integrity within 4 h and to completely prevent MLN7243-induced cytoskeletal changes, loss of junctional VE-cadherin localization, and monolayer integrity. These data reveal that in quiescent EC, barrier function is controlled by negative regulators with a short half-life. In addition, MLN7243-induced accumulation of RhoB is abolished by CHX. In good agreement with these results, barrier disruption and increase in RhoB protein level, induced by the Cullin E3 ligase inhibitor MLN4924 (26), also requires protein synthesis. These data strongly suggest that the continuous synthesis of (active) RhoB contributes to the loss of endothelial integrity which is consequent to the inhibition of the ubiquitination cascade. In contrast, RhoA levels and activity are not altered by inhibiting protein synthesis during E1 ligase inhibition, which is in line with the much longer half-life of RhoA (Fig. S3A). Thus, these data suggest that regulation of RhoA protein synthesis does not play a major role in control of quiescent endothelial integrity.

Although RhoA and RhoB share high sequence homology and downstream signaling pathways, our data suggest that protein synthesis and ubiquitination play a strikingly different role in regulating their output. It is well established that RhoA activation mediates transient responses such as the rapid (at a time scale of minutes) contraction and relaxation induced by activation of G-protein-coupled receptors (GPCRs) such as PAR1, S1PR, and LPA receptor (Fig. 5I) (5, 45). The cycling between the active and inactive form, regulated by RhoGDI, GEFs, and GAPs, allows acute, fast (<1 h time scale) and transient activation of RhoA (Fig. 6A). The available pool of cytosolic, RhoGDI-bound, inactive RhoA allows these immediate responses to receptor agonists, which regulate endothelial integrity. In line with this, our data indicate that ubiquitin-mediated degradation and protein synthesis do not significantly contribute to regulate RhoA activity in quiescent endothelium.

Figure 6.

Model of differential regulation of endothelial integrity by RhoB and RhoA. A, GDI, GEF, and GAP proteins are primarily responsible to control RhoA activity conveying rapid responses to, mostly GPCR-binding, agonists. B, in contrast, the activity of RhoB is mainly regulated at the levels of its mRNA expression, protein synthesis, and ubiquitination-mediated degradation, which allows for a relatively stable level of constitutive signaling by RhoB. C, despite being differentially regulated, active RhoB and RhoA both signal toward their common effectors, including ROCK, driving the phosphorylation of MLC, which results in actomyosin contraction and an increase in endothelial permeability. GAP, GTPase-activating protein; GEF, guanine nucleotide exchange factor; MLC, myosin light chain; ROCK, Rho-associated kinase.

In contrast, the lack of RhoGDI binding and abundance of GTP leads to rapid activation of newly synthesized RhoB. RhoB signaling activity is thus more clearly regulated by a synthesis-degradation cycle (>2–4 h time scale; Fig. 6B). By this mode of regulation, RhoB expression is largely concomitant with its activation and downstream signaling. In contrast to RhoA, this allows RhoB to mediate long-term (>hours) and chronic effects. For example, RhoB levels and activity are increased for prolonged periods of time (time scale of hours) due to hypoxia and stimulation by inflammatory mediators such as TNF, IL-1β, and lipopolysaccharide (Fig. 5, B, C, F and G) (34, 46, 47, 48). Besides, high RhoB expression was detected in intestinal, inflamed blood vessels of patients with Crohn’s disease (47). Elevated RhoB levels have also been found in urine of patients with chronic kidney disease, and RhoB has been identified as an upregulated gene 2 to 24 h after renal ischemia-reperfusion injury following kidney transplantation (49, 50). However, once activated, RhoA and RhoB share overlapping downstream signaling, comprising activation of ROCK and the induction of actomyosin-based contraction (Fig. 6C).

These findings propose only a limited role for regulation of RhoB signaling activity by GEF and GAP activity in quiescent endothelium, deviating from the prevailing model of Rho GTPase regulation. Recently, Pierce et al. proposed p190BRhoGAP to selectively inactivate RhoB in the context of capillary leak syndrome (51). This could suggest that under certain conditions or in specific organs, constitutive regulation by GAP proteins may also contribute to RhoB output. Conversely, several studies have identified various E3 ligases targeting RhoA for degradation. For example, a role for Cullin3-based ubiquitin ligases targeting RhoA, both in smooth muscle cells as well as in neurons, has been described (52, 53). Apparently, under these conditions, GAP-dependent inactivation is insufficient to limit RhoA signaling, and ubiquitin-mediated degradation is prevalent. This may also relate to the fact that in these studies, the biological effects (such as blood pressure regulation or synaptic transmission) are of a more chronic rather than an acute nature.

Intriguingly, recent studies described transient RhoB activation upon thrombin or sphingosine-1-phosphate stimulation (47, 54). However, total RhoB protein levels remained unaffected, suggesting that under certain conditions, RhoB acts along with RhoA to mediate rapid responses to GPCRs. Whether there exists a small pool of inactive RhoB available for rapid and transient activation by a GEF in order to boost RhoA-induced signaling remains to be determined. Single RhoA/B/C depletion induces upregulation of some of the other Rho GTPases (5), indicating possible functional compensation among Rho GTPases.

While we here identify RhoB as part of a proteostatic mechanism that preserves endothelial integrity, there are likely several other proteins, with a similar short half-life that are also involved in controlling endothelial integrity (e.g., c-Jun (20, 55)). Future proteomic analyses will be required to identify such proteins and their associated regulatory mechanisms. This is important as knowledge on the regulation of endothelial integrity will contribute to our options to target dysregulation of vascular permeability in (inflammatory) disease.

Experimental procedures

Antibodies and reagents

The following primary antibodies were used in this study: anti-RhoB (#sc-180, Santa Cruz Biotechnology and #63876, Cell Signaling Technology), anti-VE-cadherin (#2500, Cell Signaling Technology), anti-RhoA (#2117, Cell Signaling Technology), anti-Rac1 (#610650, BD Transduction Laboratories), anit-peIF2α (#9721 Cell Signaling Technology), anti-caspase 9 (#9502, Cell Signaling Technology), anti-cleaved caspase 3 (#9661, Cell Signaling Technology), anti-GAPDH (#2118, Cell Signaling Technology), anti-β-tubulin (#2128, Cell Signaling Technology), anti-Vinculin (#V9131, Sigma-Aldrich), anti-pMLC (#3671, Cell Signaling Technology), anti-ubiquitin (FK2; #A-106, Boston Biochem and #BML-PW8810, Enzo Life Sciences), anti-ICAM1 (#sc-8439, Santa Cruz Biotechnology), anti-pMYPT1 (#4563, Cell Signaling Technology). As secondary antibodies for Western blotting, HRP-conjugated goat anti-rabbit and anti-mouse antibodies (Dako) were used.

For immunofluorescent staining the following primary antibodies were used: anti-RhoB (#sc-8048, Santa Cruz Biotechnology) and anti-VE-cadherin (#2500, Cell Signaling Technology). Alexa-488 donkey anti-rabbit and Alexa-555 donkey anti-mouse (Invitrogen) were used as secondary antibodies. The nucleus was stained with DAPI (#62248, Thermo Fisher Scientific) and F-actin with Acti-stain 670 phalloidin (#PHDN1-A, Cytoskeleton). In addition, propidium iodide (#P3566, Molecular Probes) and Hoechst Stain 33,342 (#sc-200908, Santa Cruz Biotechnology) were used.

In this study, the inhibitors MLN7243 (#8341, Selleck Chemicals), MLN4924 (#S7109, Selleck Chemicals), MG132 (#S2619, Selleck Chemicals), Y27632 (#Y0503, Sigma), TM (#5.04570, Sigma), Cycloheximide (#239764, Merck Millipore Calbiochem), the cytokine TNFα (#300–01A, Peprotech), and the thrombin receptor activator TFLLR-NH2 (trifluoroacetate salt, #T7830, Sigma) were used.

Cell culture

HUVECs were purchased from Lonza (#CC-2519) and cultured on fibronectin-coated plates with Endothelial Cell Medium (ScienCell Research Laboratories). The medium was refreshed every second day. The cells were cultured at 37 °C in 5% CO₂ and used for experiments until passage 5.

Primary hMVECs were isolated from the foreskin of healthy donors. Informed consents were obtained from all donors in accordance with the institutional guidelines and the Declaration of Helsinki. The cells were isolated as described before (56). The primary hMVECs were cultured in M199 medium supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin, 2 mM L-glutamine (all Bio Whittaker/Lonza), 10% heat-inactivated human serum (Invitrogen), 10% heat-inactivated new-born calf serum (Gibco), 150 μg/ml crude endothelial cell growth factor (prepared from bovine brains), and 5 U/ml heparin (Leo pharmaceutical products). The cells were cultured at 37 °C in 5% CO₂, and medium was refreshed every second day. For experiments, cells from single donors or pools of three donors in passages 5 to 6 were used.

siRNA transfection

HUVECs were seeded on fibronectin-coated ECIS or culture plates. When cells reached 70 to 80% confluency, siRNA transfection using Dharmafect reagent 1 (#T-2001, Dharmacon) in OptiMEM (Gibco) was performed. For gene silencing, a final concentration of 25 nM of ON-TARGETplus Human RhoA siRNA SMART pool (siRhoA) or ON-TARGETplus Human RhoB siRNA SMART pool (siRhoB) or combination of both (siRhoA/siRhoB, all Dharmacon) was used. ON-TARGET plus nontargeting control pool (N.T.) was used as negative control. After 7 h, the transfection medium was replaced by normal culture medium. Seventy-two hours post-transfected cells were used for experiments.

Endothelial barrier function assays

To measure endothelial barrier function, ECIS was performed. HUVECs were seeded on fibronectin-coated ECIS plates containing gold intercalated electrodes (Applied Biophysics). hMVECs were seeded on gelatin-coated ECIS plates. The resistance was monitored at 4000 Hz. When cells formed a stable monolayer, treatment with the compounds was performed as indicated.

Permeability of the endothelial monolayer was measured by passage of HRP. HUVECs were seeded on gelatin-fibronectin coated Thin-Certs (Greiner Bio-One), which have a pore size of 3 μm. After forming a stable barrier, HUVECs were treated as indicated. For the last hour of treatment 5 μg/ml HRP was added to the upper compartment. After the treatment, a sample was taken from the lower compartment and upper compartment (total HRP). The HRP concentration from the lower compartment was calculated as % of the total HRP after adding the substrate tetramethylbenzidine (TMB; Organon Teknika), stopping the reaction with 2M sulfuric acid and by measuring the absorbance at 450 nm with a microplate spectrophotometer (Epoch BioTek).

Western blot

After treatment with compounds as indicated, cells were washed with PBS supplemented with 1 mM CaCl₂ and 0.5 mM MgCl₂ and lysed in 2× SDS sample buffer (125 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 100 mM DTT, 0.02% Bromophenol Blue in MilliQ). Proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane. Membranes were blocked in 5% BSA in TBS-T for 1 h and incubated with designated primary antibodies in 5% BSA in TBS-T overnight at 4 °C. After incubation with secondary antibodies, proteins were visualized using enhanced chemiluminescence (Amersham/GE Healthcare) and an AI600 imager (Amersham/GE Healthcare). Densitometric analysis of the band intensities was perfomed using ImageQuantTL.

Immunofluorescence analysis

HUVECs were seeded on fibronectin-coated 13 mm coverslips (#0117530, Marienfeld superior), and when confluency was reached treated as indicated. After washing twice with PBS supplemented with 1 mM CaCl₂ and 0.5 mM MgCl₂, cells were fixed with warm 4% paraformaldehyde in PBS at room temperature for 15 min. After three washes with PBS, cells were permeabilized with 0.2% Triton X-100 in PBS for 3 min and blocked for 1 h with 1% human serum albumin in PBS. Hereafter, coverslips were incubated with primary antibodies in 1% human serum albumin in PBS for 1 h at room temperature or overnight at 4 °C. Coverslips were washed three times with PBS and incubated with a secondary antibody, Acti-stain 670 phalloidin (#PHDN1-A, Cytoskeleton) or DAPI (#62248, Thermo Fisher Scientific) for 1 h at room temperature. Subsequently, coverslips were washed with PBS and mounted on Mowiol4-88/DABCO solution (Calbiochem, Sigma Aldrich). Confocal scanning laser microscopy was performed on a Nikon A1R confocal microscope (Nikon). Images were analyzed and equally adjusted with ImageJ.

Cell death analysis

After treatment with compounds as indicated, HUVECs were double stained with 1 μg/ml propidium iodide and 10 μg/ml Hoechst 33,342, both prediluted in PBS, for 15 min at 37 °C. Hereafter, six images per condition were acquired with the ZOE Fluorescent Cell Imager. Images were equally adjusted, and Hoechst⁺ and PI⁺ nuclei were counted with Image J.

Rac1 and Rho GTPase activation assays

To analyze Rho activity, confluent HUVECs seeded on fibronectin-coated 55 cm² dishes were treated as indicated. Subsequently, cells were washed with ice-cold PBS supplemented with 1 mM CaCl₂ and 0.5 mM MgCl₂ and lysed on ice. The levels of RhoA and RhoB signaling activity were measured using the Rho Activation Assay Biochem Kit (Cytoskeleton) with Rhotekin pulldown according to the manufacturer’s protocol. The input and pulldown samples were analyzed by Western blot.

Rac1 activity was analyzed by performing a CRIB pulldown (57). HUVECs were seeded on fibronectin-coated 21 m² dishes and treated when confluent. Cells were washed with PBS supplemented with 1 mM CaCl₂ and 0.5 mM MgCl₂ and lysed in 500 μl ice-cold lysis buffer (150 mM NaCl, 50 mM Tris pH 7.6, 1% Triton-X 100, 20 mM MgCl₂) containing 30 μg biotinylated CRIB peptide. After centrifugation of lysates at 14000 rpm and 4 °C for 5 min, 10% of supernatant was taken as input and mixed with 3× SDS sample buffer (125 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 100 mM DTT, 0.02% Bromophenol Blue). The remaining lysate was incubated with streptavidin beads (Sigma) for 30 min rotating at 4 °C. Subsequently, the beads were washed five times with lysis buffer with freshly added 10 mM MgCl₂. The buffer was aspirated, and the beads were lysed in 30 μl 2× SDS sample buffer. The input and pulldown samples were analyzed by Western blot.

Endogenous RhoB IP

Confluent monolayers of HUVECs grown on fibronectin-coated 55 cm² dishes were treated as indicated. Additionally, all samples were treated with 5 μM MG132 (#S2619, Selleckchem) for 2 h. Cells were washed with PBS supplemented with 1 mM CaCl₂ and 0.5 mM MgCl₂ and lysed in 400 μl ice-cold lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP-40, and 1× protease inhibitor cocktail). Lysates were centrifuged at 14000 rpm and 4 °C for 10 min, and 8% of supernatant was taken as input and mixed with 3× SDS sample buffer (125 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 100 mM DTT, 0.02% Bromophenol Blue). The remaining supernatant was incubated with 1 μg anti-RhoB (#sc-8048, Santa Cruz Biotechnology) and rotated overnight at 4 °C. Twenty-five microliter Dynabeads Protein G (#10003D, Thermo Fisher Scientific) were added to the lysate and incubated for 1 h rotating at 4 °C. The beads were washed four times with lysis buffer and ultimately lysed in 30 μl 2× SDS samples buffer. The input and IP samples were analyzed by Western blot.

Statistical analysis

Data are presented as mean ± SD or mean + SD. Statistical analysis was performed using GraphPad Prism. One way ANOVA with Dunnett’s post-hoc test was applied when groups were compared to one (control) group. For comparison of several groups, one-way ANOVA with Tukey’s post-hoc test was used. When only two groups were compared, two-tailed Student’s t test was used. p-values <0.05 were considered as statistically significant.

Data availability

All data are in the manuscript.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Donita Brady