PDZRN3 destabilizes endothelial cell-cell junctions through a PKCζ-containing polarity complex to increase vascular permeability

See allHide authors and affiliations

Science Signaling  31 Jan 2017:
Vol. 10, Issue 464, eaag3209
DOI: 10.1126/scisignal.aag3209

Destabilizing endothelial cell connections

Interconnected endothelial cells that line blood vessels form a barrier between the circulatory system and tissues. The integrity of the intercellular junctions between endothelial cells regulates vascular permeability, which is the leakiness of blood vessels. In stroke patients, too much fluid can accumulate in the brain and cause irreparable damage. Sewduth et al. elucidated a signaling pathway mediated by the E3 ubiquitin ligase PDZRN3 that regulated endothelial intercellular junctions and vascular permeability. Developing mice that overexpressed Pdzrn3 in endothelial cells died of hemorrhaging in multiple tissues, whereas genetic ablation of Pdzrn3 in adult mice decreased the brain edema that occurred after stroke, an effect mimicked by a pharmacological inhibitor of PKCζ, a kinase that destabilizes endothelial intercellular junctions. Manipulating Pdzrn3 abundance in endothelial cells affected the localization of PKCζ to intercellular junctions, activation of PKCζ, and permeability. These results suggest that inhibiting PDZRN3 or its downstream effector PKCζ may prevent the pathological edema that occurs in conditions such as stroke.


Endothelial cells serve as a barrier between blood and tissues. Maintenance of the endothelial cell barrier depends on the integrity of intercellular junctions, which is regulated by a polarity complex that includes the ζ isoform of atypical protein kinase C (PKCζ) and partitioning defective 3 (PAR3). We revealed that the E3 ubiquitin ligase PDZ domain–containing ring finger 3 (PDZRN3) regulated endothelial intercellular junction integrity. Endothelial cell–specific overexpression of Pdzrn3 led to early embryonic lethality with severe hemorrhaging and altered organization of endothelial intercellular junctions. Conversely, endothelial-specific loss of Pdzrn3 prevented vascular leakage in a mouse model of transient ischemic stroke, an effect that was mimicked by pharmacological inhibition of PKCζ. PDZRN3 regulated Wnt signaling and associated with a complex containing PAR3, PKCζ, and the multi-PDZ domain protein MUPP1 (Discs Lost–multi-PDZ domain protein 1) and targeted MUPP1 for proteasomal degradation in transfected cells. Transient ischemic stroke increased the ubiquitination of MUPP1, and deficiency of MUPP1 in endothelial cells was associated with decreased localization of PKCζ and PAR3 at intercellular junctions. In endothelial cells, Pdzrn3 overexpression increased permeability through a PKCζ-dependent pathway. In contrast, Pdzrn3 depletion enhanced PKCζ accumulation at cell-cell contacts and reinforced the cortical actin cytoskeleton under stress conditions. These findings reveal how PDZRN3 regulates vascular permeability through a PKCζ-containing complex.


Cell-cell junctions are important gatekeepers in the maintenance of adhesion and barrier properties of endothelial cell monolayers. Regulation of cell-cell contact is essential for physiological processes including vascular development, tissue regeneration, and organ morphogenesis (1). Disruption of cell-cell contacts is associated with human diseases including edema, ischemia, and inflammation.

The integrity of endothelium is mediated by adhesive interactions between endothelial cells, which involve adherens junctions and tight junctions (2). Tight junctions seal endothelial cells to create a primary barrier and maintain cell polarity. At the molecular level, tight junctions are primarily composed of claudin (3), occludin, and junctional adhesion molecule family members. They are linked to the cytoskeleton and adherens junction by adaptor molecules, including postsynaptic density-95/disc large/zona occludens 1 (ZO1) (PDZ) proteins. Tight junctions are regulated by two protein complexes: one containing Pals1 (protein associated with Lin seven), PATJ (PALS-1–associated tight junction protein) or its paralog MUPP1 (Discs Lost–multi PDZ domain protein 1) (4, 5), and PAR3 (partitioning defective 3), and the other containing PAR3, PAR6, and aPKC (6, 7). In epithelial cells, these complexes maintain apicobasal polarity, maintain the integrity of tight junction complexes, and act as anchor proteins, recruiting other proteins to cell junctions (810). However, the molecular mechanisms by which endothelial tight junctions are maintained in specialized areas remain largely unknown.

Wnt signaling is crucial to vascular formation and endothelial cell functional differentiation (11). At the cellular level, Wnt signaling is facilitated through activation of the canonical pathway, which leads to the stabilization of β-catenin, and the noncanonical Wnt-Ca2+ and Wnt–planar cell polarity (PCP) pathways (12). Initially identified as a signaling mechanism promoting polarization and organization of cells within the plane of the epithelium (12, 13), the Wnt-PCP pathway plays a role in endothelial cell reorganization and directional migration (14) and acts through adherens junction stabilization during lymphatic valve formation (15). Many conserved downstream effectors of PCP signaling, including the small guanosine triphosphatases Rho, Rac, and Cdc42 and c-Jun N-terminal kinase (JNK), participate in the regulation of adherens junction and tight junction formation and maturation (1618). However, the mechanistic link between activation of PCP signaling, downstream effector activation, and fine intercellular junction organization and stabilization is poorly understood. In cultured brain endothelial cells, the Wnt-PCP signaling pathway regulates the formation of tight junctions by recruitment of molecules involved in the formation of cell polarity, such as PAR3 and aPKC (19). These findings suggest that Wnt-PCP signaling may regulate junction protein clustering to maintain polarized organization of endothelial cells during vascular morphogenesis in vertebrates.

Our group has shown that the E3 ubiquitin ligase PDZ domain–containing ring finger 3 (PDZRN3) drives noncanonical Wnt-PCP signaling to regulate polarized migration of endothelial cells and vascular morphogenesis (20). Here, we investigated whether PDZRN3 contributed to the regulation of endothelial junction stability with in vivo and in vitro endothelial cell–specific Pdzrn3 gain- and loss-of-function approaches. Our data demonstrated that endothelial cell–specific Pdzrn3 overexpression induced severe hemorrhage in embryos, and conversely, endothelial cell–specific Pdzrn3 depletion reduced rupture of the blood-brain barrier (BBB) and, consequently, edema formation in an adult mouse model of stroke after middle cerebral artery occlusion (MCAO). We found that PDZRN3 regulated Wnt signaling and interacted with the polarity proteins PAR3, PKCζ, and MUPP1 at endothelial cell junctions to regulate tight junction stabilization. Elucidating a molecular link between Wnt-PCP signaling and polarity protein complexes may provide the basis for further understanding the process of turnover and signaling of cell adhesion molecules.


Overexpression of Pdzrn3 induced vascular leakage, leading to early embryonic lethality

To investigate the functional role of PDZRN3 signaling on vascular integrity, we selectively induced Pdzrn3 expression in endothelial cells by crossing transgenic mice, which express the transcriptional activator tTA under the control of the Tie2 promoter (Tie2-tTA), with reporter mice, which harbor a bidirectional Tet-promoter cassette with genes for PDZRN3-V5 and β-galactosidase (Tie2-tTA;Pdzrn3-V5). Efficient transactivation was verified by β-galactosidase staining and Western blotting for PDZRN3-V5 expression in heterozygous mouse embryos (Fig. 1A and fig. S1A). No Tie2-tTA;Pdzrn3-V5 pups were produced, suggesting that Pdzrn3 overexpression was embryonically lethal. Genotypic analysis demonstrated no overt defect until embryonic day 9.5 (E9.5), and by E15.5, all mutants were resorbed (Fig. 1B). Upon further dissection, abnormal Tie2-tTA;Pdzrn3-V5 embryos exhibited a hemorrhagic phenotype at E10.5, in which fetal blood had hemorrhaged into the lumen of cephalic ventricles (Fig. 1C).

Fig. 1 Overexpression of Pdzrn3-V5 in EC induces early embryo lethality and vascular dysfunction.

(A) Lysates of Tie2-tTA;Pdzrn3-V5 (V5) and control embryo heads at E10.5 were analyzed by immunoblotting using indicated antibodies and representative of three independent experiments. (B) To determine the stage of embryonic lethality, timed pregnancies were performed, and embryos were isolated and genotyped at various stages of gestation. Embryo numbers were n = 31 at E9.5, n = 34 at E10.5, n = 12 at E11.5, n = 9 at E14.5, and n = 27 at P0. (C) Representative images showing the gross phenotypes of Pdzrn3-V5–overexpressing (n = 3) and control (n = 3) embryos. Mutant embryos exhibited hemorrhage-filling ventricles at E10.5. Scale bars, 1 μm. (D) Representative electron microscopy images of E10.5 embryo cerebral microvessels from Tie2-tTA;Pdzrn3-V5 (n = 3) and control (n = 3) embryos. In Tie2-tTA;Pdzrn3-V5 embryos, rupture of endothelium was seen, and large cytosolic vacuoles were detected in the cytoplasm of the cells. E, endothelial cells; RBC, red blood cells; Pl, platelets; P, pericyte. Scale bars, 1 μm. Representative of three control and three Tie2-tTA;Pdzrn3-V5 embryos. (E) Representative images of leakage from brain cortical vessels Tie2-tTA;Pdzrn3-V5 and control embryos injected in utero with dextran tracer (red) at E14.5. Capillaries were labeled with CD31 antibody (green). Scale bars, 50 μm. Data are representative of two independent experiments.

We compared the whole-mount CD31 staining pattern of wild-type and Pdzrn3-overexpressing mutant embryos at E10.5. During this developmental period, the vascular plexus appeared to be developed in Tie2-tTA;Pdzrn3-V5 and wild-type embryos. However, upon closer examination, we observed enlargement of cranial vessels, indicating defects in vessel maturation (fig. S1B). Hemorrhages were not seen in the yolk sac, but embryonic vessels appeared dilated in mutant embryos with a hemorrhagic phenotype. In addition, some embryonic vessels in the labyrinth of the placenta exhibited a leaky phenotype (fig. S1C).

We examined brain embryos by electron microscopy and reported that, in control embryos, endothelial cells formed in linear junctions with red blood cells contained in blood vessels (Fig. 1D). The vessels had mature endothelial cell junctions, and endothelial cells are interconnected with overlapping flaps in their luminal surface. Moreover, pericytes were closely associated with the microvascular endothelium and had long, narrow cytoplasmic extensions surrounding the endothelial cells. In the Tie2-tTA;Pdzrn3-V5 embryos, some cytoplasmic regions of endothelial cells appeared to be thinner, and there were ruptures visible in the junctions between endothelial cells, along with evidence of red blood cell extravasation. In contrast to control embryos, endothelial cells in these mutant embryos had large vacuoles in the cytoplasm, had visible ruptures in the luminal portion of the cytoplasm, and were poorly covered by pericyte processes (Fig. 1D). These data confirm that endothelial cells were dysfunctional in Tie2-tTA;Pdzrn3-V5 embryos and that the barrier function of the neurovascular unit was not well maintained. Tracer experiments performed in E14.5 embryos that received intracardiac injection of 10-kDa rhodamine dextran showed that, although the tracer was contained to brain vessels of control mice, it was extravasated in the parenchymal tissues of Tie2-tTA;Pdzrn3-V5 mutant mice (Fig. 1E). These results revealed that the endothelial cell barrier function was perturbed in mutant mice and confirmed the hypothesis that PDZRN3 signaling alters organization of junction complexes during vessel formation between E10.5 and E14.5.

Loss of Pdzrn3 reduced BBB disruption in a stroke model of MCAO

Disruption of tight junctions leads to impaired BBB function. We surmised that the loss of Pdzrn3 could strengthen tight junctions and reduce vessel permeability in a mouse model of stroke because PDZRN3 is detected in brain endothelial cells by immunofluorescence analysis (fig. S2). To confirm this hypothesis, we used mice with an endothelial cell–specific deletion of Pdzrn3 (Pdzrn3iECKO), which was achieved by crossing Pdzrn3 conditional knockout (Pdzrn3fl/fl) mice (20) with Pdgf tamoxifen-inducible Cre+/− mice (Pdgfb-ICreER) (21). Because Pdzrn3 deletion impaired precocious events during vessel morphogenesis (20), we depleted Pdzrn3 in adult Pdzrn3iECKO mice and verified depletion by immunofluorescence analysis of isolated kidney endothelial cells (fig. S3A) and by Western blot analysis of brain lysates from Pdzrn3iECKO and Pdzrn3iECWT mice (fig. S3B). One month after tamoxifen treatment, micro–computed tomography (micro-CT) was performed to image the whole-brain vascular network and verify that Pdzrn3 depletion did not induce any vascular modification under basal conditions (fig. S4).

Tamoxifen-treated mice were subjected to occlusion of the middle cerebral artery followed by a reperfusion period. Vascular damage to the BBB was analyzed by the injection of 70-kDa fluorescein isothiocyanate–conjugated dextran (FITC dextran) 2 hours after reperfusion. This tracer leaked into the brain interstitial space about 10 min after injection in the injured hemisphere, whereas none leaked into the contralateral hemisphere (Fig. 2A). We noted that significantly less tracer leaked in Pdzrn3iECKO mice than in Pdzrn3iECWT mice (Fig. 2A). To measure BBB integrity, the amount of Evans Blue dye extravasation was analyzed 6 and 24 hours after reperfusion. Consistently, the quantity of Evans Blue dye content in the injured hemisphere was significantly reduced in Pdzrn3iECKO mice compared to control mice at both time points (Fig. 2B). To confirm these observations, we performed mouse immunoglobulin G (IgG) staining of brains of mice subjected to MCAO. The area of extravasation in the infarcted hemisphere was significantly decreased in Pdzrn3iECKO mice compared to that in control Pdzrn3iECWT mice (Fig. 2C). These results demonstrated that loss of Pdzrn3 limits BBB breakdown and leakage after transient MCAO.

Fig. 2 Loss of Pdzrn3 reduces vascular edema after MCAO in adult mice.

(A) MCAO leading to infarct was induced in one infarcted hemisphere but not in the contralateral hemisphere. To evaluate vascular leakage, extravasation of 70-kDa FITC dextran (green) was assessed 2 hours after MCAO on sagittal sections of the brain. Quantification of leakage is presented in the graph. iEC wild-type (WT) (Pdzrn3iECWT) mice, n = 6; iEC knockout (KO) (Pdzrn3iECKO) mice, n = 5. Error bars indicate SEM. P values were determined using Mann-Whitney test. Scale bars, 100 μm. (B) To evaluate vascular leakage, Evans Blue extravasation was assessed 6 and 24 hours after MCAO by spectrophotometry after extraction from brain lysates. OD, optical density. iEC WT mice, n = 10; iEC KO mice, n = 9. Error bars indicate SEM. P values were determined using Mann-Whitney test. (C) Representative images of BBB rupture as assessed by diffusion of IgG autoantibody out of the vessels in the parenchyma. Quantification of the area of leakage is presented in the graph. Scale bars, 1 mm. iEC WT mice, n = 6; iEC KO mice, n = 6. Error bars indicate SEM. P values were determined using Mann-Whitney test.

Pdzrn3 deletion impaired disassembly of ZO1 and c-Jun from tight junction complexes

To study the molecular mechanisms involved in mediating the protective effect of Pdzrn3 deletion on BBB disruption, we used a biochemical fractionation protocol (22) that enriches microvessels on infarcted and contralateral brain hemispheres. Because different Wnt genes have been reported to be involved during brain vascular formation and maintenance, we analyzed whether they were induced 6 hours after MCAO injury. Quantitative reverse transcription polymerase chain reaction (RT-PCR) analysis revealed a significant increase in the abundance of Wnt3a, Wnt5a, Wnt7a, and Wnt7b in both Pdzrn3iECKO and Pdzrn3iECWT microvessel-enriched fractions (fig. S5), suggesting that Wnt signaling could be activated during stroke. We also found that phosphorylation of c-Jun and PKCζ was increased after 6 hours of reperfusion in Pdzrn3iECWT vessel lysates from the infracted hemisphere, as compared to the control hemisphere (Fig. 3A). In Pdzrn3iECKO mutants, endothelial depletion of Pdzrn3 impaired the increase in phosphorylation of both c-Jun and PKCζ, providing further evidence that PDZRN3 signaling is required for c-Jun and PKCζ activation in endothelial cells after MCAO injury.

Fig. 3 Loss of Pdzrn3 protects from tight junction disruption induced by MCAO surgery.

(A) Enriched microvessel fractions from contralateral (Ctl) and infarcted (Inf) hemispheres from Pdzrn3iECWT (iEC WT) and Pdzrn3iECKO (iEC KO) mice were analyzed by Western blot with indicated antibodies. The band intensities were quantified. Data are means ± SEM of three separate experiments. P values were determined using one-way analysis of variance (ANOVA), followed by post hoc Bonferroni test. p, phosphorylated. (B) Representative images of cortical microvessels of contralateral and infarcted areas from iEC WT and iEC KO mice were analyzed by immunostaining with CD31 (green) and ZO1 (red) antibodies (top) or CD31 (green) and aquaporin 4 (red) antibodies (bottom). Scale bars, 20 μm. Data are representative of two independent experiments.

Canonical Wnt signaling controls BBB integrity and organization of endothelial intercellular junctions through β-catenin stabilization and induction of claudin protein (23, 24). After injury, a decrease in Ve-cadherin at cell-cell junctions correlates with an accumulation of phosphorylated β-catenin and forkhead box protein O1 (FOXO1) in the nucleus where they repress the expression of the claudin 5 gene in tight junctions. Conversely, under resting conditions, Ve-cadherin limits endothelial cell permeability by maintaining a pool of β-catenin at the cell-cell junctions and localization of the transcriptional factor FOXO1 in the cytosol, in a phosphorylated inactive form (25). Thus, we measured the fraction of phosphorylated β-catenin, phosphorylated FOXO1, and claudin 5 after MCAO in the enriched microvessel fractions (Fig. 3A). Phosphorylation of FOXO1 was decreased in both Pdzrn3iECKO and Pdzrn3iECWT mice after 6 hours of reperfusion, suggesting that FOXO1 was activated under conditions of cerebral ischemia in a Pdzrn3-independent manner. We detected a strong increase in the phosphorylation of β-catenin in Pdzrn3iECKO vessel-enriched extracts after MCAO (Fig. 3A). Claudin 5 and Ve-cadherin amounts were not substantially modified under either condition. Together, these data suggest that impairing endothelial PCP signaling by deleting Pdzrn3 may favor endothelial Wnt canonical signaling, which may participate in the stabilization of the BBB during ischemic injury in adult mice.

Immunofluorescence and confocal analysis revealed that endothelial cell junctions in vessels from the contralateral noninfarcted areas of the cortex had clearly delineated circular lumens, as indicated by ZO1 and CD31 staining and aquaporin 4 staining, which identifies astrocytes around the vessels (Fig. 3B). In infarcted areas of the cortex of Pdzrn3iECWT mice, vessels appeared to be irregularly shaped in some regions, had discontinuous ZO1 labeling, and exhibited low amounts of aquaporin 4 labeling. In Pdzrn3iECKO mice, cortical vessels were less affected in the infarcted areas, as demonstrated by strong and abundant ZO1 labeling of CD31-positive vessels and maintenance of aquaporin 4 labeling (Fig. 3B). Together, these results suggest that endothelial cell–specific deletion of Pdzrn3 was associated with reduced activation of c-Jun and PKCζ in a model of ischemic stroke and increased canonical Wnt signaling through enhanced accumulation of phosphorylated β-catenin. We propose that decreased expression of Pdzrn3 reduces injury-induced breakdown of the BBB and may also limit vascular edema after injury.

PDZRN3 signaling destabilized endothelial cell-cell junctions through PKCζ

PKCζ is localized in junction complexes in epithelial (26) and endothelial cells (27) and is involved in regulating endothelial cell permeability (28). We therefore tested whether PKCζ was required for the ability of PDZRN3 to regulate endothelial cell permeability and whether PDZRN3 signaling regulated PKCζ localization in endothelial cells. In vitro permeability tests indicated that endothelial cells transduced with Pdzrn3 lentivirus were more permeable to FITC dextran, an effect attenuated by blocking PKCζ with pseudosubstrate (Fig. 4A) or by depleting PKCζ (Fig. 4B and fig. S6B). These results were further supported by Western blot analysis of PKCζ activation, which showed that the phosphorylation of c-Jun and PKCζ was increased by Pdzrn3 overexpression, an effect blocked by PKCζ pseudosubstrate treatment (Fig. 4C). Note that ectopic expression of Pdzrn3 led to an increased amount of total c-Jun, which may result in an enhancement of total c-Jun activity.

Fig. 4 Pdzrn3 overexpression induces endothelial cell permeability through PKCζ.

(A and B) Control or Pdzrn3-V5–transduced human umbilical vein endothelial cells (HUVECs) were grown on Transwell inserts and either treated with PKCζ pseudosubstrate (PS) (A) or transfected with small interfering RNA (siRNA) directed against PKCζ (si-PKCζ) or a control siRNA (si-Control) (B). Leakage of FITC dextran added to the upper wells into the lower wells over the indicated time intervals was measured. Data were from three independent experiments carried out in triplicate and are presented as means ± SD. P values were determined using two-way ANOVA, followed by post hoc Bonferroni test. In (A), **P < 0.01 for Pdzrn3-V5 compared to control. In (B), *P < 0.05 and **P < 0.01 for Pdzrn3-V5 + si-Control compared to control + si-Control; $P < 0.05 and $$P < 0.05 for control + si-PKCζ compared to control + si-Control. (C) Lysates of control or Pdzrn3-V5–transduced HUVECs after PKCζ PS treatment were analyzed by immunoblotting (IB) using the indicated antibodies. (D) Vascular leakage was evaluated in mice with Evans Blue injection after PKCζ PS or no treatment by spectrophotometry after extraction from brain lysates (in micrograms per gram of tissue) 24 hours after MCAO. Control mice, n = 9; PKCζ PS–treated mice, n = 10. Error bars indicate SEM. P values were determined using Mann-Whitney test. Scale bars, 1 mm. (E) Enriched microvessel fractions from infarcted and contralateral hemispheres from control and PKCζ PS–treated mice were analyzed by Western blot with the indicated antibodies. Blots in (C) and (E) are representative of two independent experiments.

Finally, to examine the physiological importance of PKCζ activation, we analyzed the effects of PKCζ activity blockade on brain edema and BBB breakdown after MCAO and reperfusion. At 24 hours after reperfusion, the quantity of extravasated Evans Blue was significantly reduced in mice treated with PKCζ pseudosubstrate (Fig. 4D), suggesting that the induction of the PKCζ pathway could contribute to stroke-induced barrier damage. PKCζ pseudosubstrate treatment also resulted in the impaired activation of PKCζ and c-Jun in microvessel-enriched fractions from the injured hemisphere compared to those from the noninjured contralateral hemisphere (Fig. 4E). These results mimicked the molecular effects induced by endothelial cell–specific Pdzrn3 deletion. Note that the total abundance of PKCζ increased after ischemia, which may contribute to the extent of ischemic injury. Together, these data suggest that PDZRN3 signaling reduces the accumulation of the polarity effector PKCζ at cell-cell contacts and is associated with an alteration of intercellular junctions that are sufficient for increased endothelial cell monolayer permeability.

PDZRN3 targeted MUPP1 for polyubiquitination to destabilize tight junctions

Endothelial cell junction integrity is regulated by the recruitment of polarity complexes to cell-cell junctions; these complexes can include the adaptor PAR3, the effector PKCζ (29), and large scaffolding PDZ-containing proteins including MUPP1 (5, 30). Because molecules that regulate signaling upstream of these polarity complexes remain unknown, and because PDZRN3 is a PDZ domain–containing protein, we speculated that it could be recruited to these polarity complexes. Coimmunoprecipitation experiments using HeLa cells expressing PAR3, PKCζ, and PDZRN3 demonstrated that PKCζ, MUPP1, and a small amount of endogenous PDZRN3 were detected in PAR3 complexes, whereas a large amount was associated with PKCζ in a PAR3-independent manner (Fig. 5A).

Fig. 5 PDZRN3 interacts with a polarity complex containing PAR3, PKCζ, and MUPP1 to regulate endothelial junction stability.

(A) Extracts from HeLa cells transfected with Pdzrn3-V5 vector were immunoprecipitated with antibodies against PAR3 and PKCζ or nonspecific immunoglobulin (IgG). Immunoprecipitates (IPs) and lysates were then immunoblotted with the indicated antibodies. (B) Coprecipitation of MUPP1 and PDZRN3 on enriched microvessel fractions from the infarcted and contralateral hemisphere from Pdzrn3iECWT (iEC WT) and Pdzrn3iECKO (iEC KO) mice subjected to MCAO. IPs and lysates were then immunoblotted with the indicated antibodies. (C) PDZRN3 interacts with MUPP1. Extracts from HeLa cells transfected with Pdzrn3-V5 (+) and MUPP1-Flag (+) vectors were immunoprecipitated with antibody against V5. IPs and lysates were then immunoblotted with the indicated antibodies. Blots in (A) to (C) are representative of two independent experiments. (D) PDZRN3 induced ubiquitination of MUPP1 and its degradation. HeLa cells were transfected with MUPP1-Flag and Pdzrn3-V5 and then treated with cycloheximide (cyclohex.) for 3 or 6 hours. Lysates were then immunoblotted with indicated antibodies. Results are representative of three independent experiments. (E) PKCζ, PAR3, and ZO1 immunostaining (white) were performed on si-Control, si-Pdzrn3, or si-MUPP1 HMVECs. Cell nuclei were labeled with DAPI (4′,6-diamidino-2-phenylindole). Scale bars, 20 μm. (F) PDZRN3 controls actin cytoskeleton organization: Staining for F-actin showed that Pdzrn3 depletion impaired stress fiber formation induced by serum deprivation during 4 hours (stress condition). Scale bars, 20 μm. Data in (E) and (F) are representative of three independent experiments.

We hypothesized that PDZRN3 might be an effector of the PAR3/aPKC polarity complex that mediates MUPP1 degradation to destabilize endothelial cell junctions. PDZRN3 was immunoprecipitated with MUPP1 from enriched microvessel fractions from contralateral and infarcted brain hemispheres in Pdzrn3iECWT mice (Fig. 5B). The increase in MUPP1 ubiquitination in vessel-enriched lysates from the infarcted brain hemisphere suggested that MUPP1 could be a substrate for PDZRN3 (Fig. 5B). Furthermore, the amount of ubiquitinated MUPP1 was greater in lysates from Pdzrn3iECWT mice than in those from Pdzrn3iECKO mice, suggesting that MUPP1 may be a specific substrate of PDZRN3-mediated ubiquitination in endothelial cells. Coimmunoprecipitation analysis indicated that PDZRN3-V5 coimmunoprecipitated MUPP1-Flag and vice versa from transiently transfected HeLa cells (Fig. 5C). Cycloheximide chase experiments in cells transiently expressing PDZRN3-V5 and MUPP1-Flag revealed that the abundance of MUPP1-Flag was maximally reduced at 6 hours, which correlated with a decrease in the membrane fractions, suggesting that PDZRN3 affected MUPP1 turnover. Addition of the proteasome inhibitor MG132 prevented this decrease in the amount of MUPP1, indicating that the reduction in MUPP1 abundance was mediated by the proteasome (Fig. 5D). These results suggest that Pdzrn3 overexpression enhances proteasomal degradation of MUPP1, thereby enhancing dissociation of tight junction complexes.

To determine the importance of endogenous PDZRN3 on polarity complex stabilization, we used human microvascular endothelial cells (HMVECs), which express PDZRN3 and display well-developed cell-cell junctions. Pdzrn3-depleted HMVECs (fig. S6A) showed increased PKCζ recruitment at cell-cell membrane junctions (Fig. 5E). Conversely, MUPP1 knockdown (fig. S6C) resulted in partial disruption of the continuous ZO1 staining pattern in HMVECs, which was changed to a zigzag staining pattern (Fig. 5E), and PKCζ and PAR3 were no longer localized at cell-cell junctions. Examination of the actin cytoskeletal organization revealed that Pdzrn3-depleted cells had only a thin rim of cortical actin, whereas control siRNA-treated cells showed an increase in actin fiber density. Serum depletion enhanced actin fiber density in control siRNA-treated HMVECs but not in Pdzrn3-depleted cells (Fig. 5F), suggesting that decreasing PDZRN3 abundance acts to prevent stress fiber formation and to favor actin cytoskeleton stabilization for stable cell-cell junctions. Collectively, these data suggest that junction stability is maintained by a PDZRN3-dependent fine-tuning mechanism that balances polarity protein complex stabilization with intercellular tensional cytoskeletal forces.


We used conditional mouse mutants to study the in vivo Pdzrn3 function and signaling pathway in the vascular system. We report that overexpression of endothelial Pdzrn3 led to embryonic lethality with hemorrhages and endothelial cell junction alteration. Conversely, endothelial loss of Pdzrn3 in adult mice led to impaired endothelial cell junction rupture events during ischemic stroke and limited consequent permeability.

Vascular growth is a complex cellular process, which requires endothelial cells to proliferate, change their morphology, and move directionally. Endothelial cells do not keep their position but continuously shuffle from the back to the front position (31), which requires dynamic remodeling of their junctions with neighboring cells to follow the shape of the vascular tube and coordination between adjacent cells. We found that mice lacking global expression of Pdzrn3 from E7.5 displayed vascular defects, such as impaired vascular network remodeling in the extraembryonic vasculature, but no obvious defects were observed in the embryo proper during vascular formation. Mutant embryos die in utero between E11.5 and E12.5 (20). We found that endothelial overexpression of Pdzrn3 induced vascular defects with localized vascular hemorrhages in both the placenta and the head of embryos without altering developmental vasculogenesis and angiogenesis. These results suggest that PDZRN3 signaling did not perturb early vessel formation steps.

Electron microscopy revealed that Pdzrn3-overexpressing endothelial cells appeared more elongated and had more vacuoles, which could be linked to interendothelial junction weakness, causing vascular breaks and leakages. β-Catenin null mutants have focal hemorrhages within E11.5 and E13.5, which may be due to molecular remodeling of endothelial cell junctions with an increase of fenestrae (32). The vascular phenotypes in both loss- and gain-of-function Pdzrn3 endothelial mutants underscore the fact that Pdzrn3 gene dosage is critical for vascular network remodeling and stability.

Although various studies have linked the angiogenic process with the PCP pathway, the underlying molecular mechanism has yet to be clearly delineated (3335). Endothelial cell–specific conditional deletion of Evi (evenness interrupted), which blocks Wnt release from the Golgi (36), or treatment with the Wnt inhibitor TNP470 led to reduced microvessel density and an increase in vessel regression in postnatal retina (37), revealing an important role of the Wnt noncanonical pathway in vessel stability and remodeling. Accordingly, we have shown that the intracellular effector PDZRN3 regulates PCP signaling in endothelial cells and that deletion of Pdzrn3 in endothelial cells impairs the polarized organization of the intermediate vascular plexus in postnatal retina with an increase in vascular regression (20). Wnt-PCP signaling has been proposed to control cell polarity and cell movement through the activation of Rac1, Cdc42, RhoA/B, and downstream JNK signaling cascades (14, 38).

We speculated that the vascular phenotype caused by Pdzrn3 overexpression may be linked to its stimulatory role in the Wnt-PCP pathway. Several pieces of evidence have implicated Wnt canonical signaling in regulating the formation of cerebral vessels. Wnt7a and Wnt7b have been implicated in brain angiogenesis (39, 40) by promoting canonical Wnt β-catenin signaling. Moreover, canonical Wnt β-catenin signaling induces BBB maturation and tight junction stabilization [reviewed in (11, 41, 42)]. Conversely, endothelial depletion of frizzled4 leads to vascular leakage in the cerebellum (24). In addition, the transcriptional factor ETS-related gene promotes endothelial stability through the Wnt β-catenin pathway (43). However, the role of Wnt-PCP signaling in brain vessels has not yet been fully elucidated. We propose that blocking the Wnt-PCP signaling pathway may protect the brain from a disruption of the BBB under pathological ischemic conditions. We demonstrated that endothelial depletion of Pdzrn3 reduced tight junction disruption and vascular leakage in the brain after acute ischemic stroke, which correlated with an impairment of c-Jun activation. We found that Wnt ligand expression increased after MCAO in both Pdzrn3-deleted and control enriched microvessel fractions, which correlated with an accumulation of phosphorylated forms of c-Jun, a terminal effector that stimulates vascular permeability (44). We reported here an increased amount of active phosphorylated β-catenin in enriched microvessel fractions from Pdzrn3-deleted mutants after MCAO injury. We have proposed that PDZRN3 plays a role in switching from canonical to Wnt-PCP signaling (20) and suggest that, when endothelial Pdzrn3 is depleted, canonical Wnt signaling is activated, which may help maintain and/or reinduce barrier function in brain microvessels in the infarcted area. However, because we used Pdzrn3 gain- and loss-of-function mouse mutants as molecular tools to decrypt the role of PDZRN3 signaling on maintaining vessel barrier functions in vivo, future experiments are required to further clarify the precise role of PDZRN3 signaling under physiological conditions at the tissue or organ level. Furthermore, our study was focused on the role of PDZRN3 signaling in endothelial cells. Brain microvasculature is an interdependent, multicellular unit with endothelial cells, astrocytes, and pericytes, all of which interact through numerous autocrine and paracrine signaling pathways (45). Deletion of PDZRN3 signaling in endothelial cells may modify cell interdependence, which may account for the fragility of blood vessels. Future studies investigating the role of PDZRN3 signaling in BBB unit regulation at the cellular level are warranted.

We speculated that PDZRN3 could interact with these polarity complexes because maintenance of tight junction stability relies on the recruitment of large scaffolding proteins to polarity complexes and because PDZRN3 contains PDZ domains, which are often found in scaffold proteins (46). MUPP1 assembles protein complexes through their cytoplasmic domains to regulate biological processes, such as cell polarity in epithelial cells (7, 47) and directional cell migration (48). Translocation of MUPP1 to the cytoplasm in cells infected with tumorigenic human viruses induces disruption of tight junctions and endothelial cell polarity (49). The Rho-specific synectin-binding exchange factor (Syx) is involved in endothelial junction stabilization and endothelial barrier function during angiogenesis (50) and has been proposed to interact with MUPP1 and Crumbs polarity complexes. The MUPP1/Syx complex is recruited to endothelial cell-cell junctions to regulate cell junctions and to maintain tight junction integrity (30). Here, we propose a model in which PDZRN3 signaling induces the ubiquitination of MUPP1 at tight junctions, leading to its degradation, which, in turn, loosens tight junctions. This body of work supports a model in which PDZRN3 may occupy a key position at the crossroads of Wnt signaling and the PAR3/PKCζ/MUPP1 polarity protein complex to control vascular permeability and, notably, BBB integrity (fig. S7). The potential interactions between these two pathways remain to be found. It will be of interest to gain more insight into the biological importance of regulating such PDZRN3-dependent pathways to determine its impact during vascular morphogenesis.


Experimental animals

For endothelial cell–specific deletion, PDGF-icre (21) mice were crossbred to Pdzrn3fl/fl mice (20). For Pdzrn3 gene deletion in adults, 1 mg of tamoxifen was injected intraperitoneally for three successive days 2 weeks before surgery. For endothelial Pdzrn3 overexpression, we generated TRE-Pdzrn3-V5 mutant mice by microinjection of a construct containing the bidirectional tet-responsive promoter, which allows simultaneous expression of PDZRN3 and LacZ (51). The Pdzrn3 transgene fused in the C terminus with a V5 epitope in pcDNA3 expression vector (Invitrogen) was excised from the plasmid backbone with Eco RI and Pme I restriction enzymes and cloned in the Pst I sites. These mice were subsequently backcrossed onto a C57BL6/J background. Interbred, homozygous Pdzrn3-V5 mice exhibited apparently normal development and were viable and fertile. For endothelial cell–specific overexpression, Tie2-tTA males were mated with Pdzrn3-V5 females. Tails of pups were genotyped by PCR using the P1 and P2 primer set to detect the Pdzrn3-V5 coding gene (P1, 5′-CAGCTTGAGGATAAGGCGCT-3′; P2, 5′-CTTCGAGCTGGACCGCTTC-3′) and using the P3 and P4 primer set to detect the tTA coding gene (P3, 5′-GCTGCTTAATGAGGTCGG-3′; P4, 5′-CTCTGCACCTTGGTGATC-3′.

Middle cerebral artery occlusion

Focal ischemia was induced by transient (90 min) MCAO, as described previously (52). Briefly, core body temperature was regulated at 37°C. Under isoflurane anesthesia (30% O2, 70% N2O), a silicon-coated nylon monofilament (Dermalon) was introduced into the external carotid artery and advanced along the internal carotid artery until the origin of the middle cerebral artery was occluded. After 90 min, the filament was withdrawn to establish reperfusion. Evans Blue [100 μl at 2% (w/v)] was injected just after reperfusion. PKCζ pseudosubstrate (3 mg/kg; Santa Cruz Biotechnology) was injected after reperfusion. Neurological status was assessed according to a neurological grading score of increasing severity of deficit: 0, no observable deficit; 1, torso flexion to right; 2, spontaneous circling to right; 3, leaning, falling, or barreling to right; 4, no spontaneous movement or unconscious; 5, death.

For assessment of brain edema, mice were perfused transcardially with phosphate-buffered saline (PBS) and then with 4% formalin. For measurement of the area of brain edema, brains were removed and embedded in agarose. Slices of 100 μm were then observed under a Zeiss Axio Observer microscope, taking advantage of the fact that Evans Blue fluorescence was at 540 nm. For quantification of Evans Blue using spectrophotometry, brain hemispheres were separated and frozen. After sonication in PBS, the lysates were centrifuged and resuspended in trichloroacetic acid (volume/volume). After overnight incubation at 4°C, the lysates were centrifuged, and the quantity of Evans Blue in the supernatant was estimated by measuring the optical density at 610 nm.

Vessel extraction

Brain tissues were frozen at −20°C and, after thawing, potterized 20 times on ice in PBS. After a 10-min centrifugation at 1000g, the pellet was resuspended in PBS and placed on top of a PBS/17.5% dextran gradient. After a 10-min centrifugation at 4400g, the pellet was resuspended and filtered through a 40-μm filter (BD Biosciences). The fraction that was retained on the filter was then eluted with PBS by inversion of the filter. For optimal extraction of vessels, the vessel fraction can be loaded on a column containing glass beads. After washing with PBS, the purified vessel fraction was harvested by gentle agitation.

Isolation of primary endothelial cells from mouse kidney

Kidneys of sacrificed mice were removed in PBS containing CaCl2 and MgCl2 (Gibco), minced in slices, and incubated for 45 min at 37°C in a solution of PBS Ca/Mg containing collagenase type I (Sigma-Aldrich). The suspensions were triturated with a cannula and filtered first with a 70-μm filter (BD Biosciences) and then with a 30-μm filter (Miltenyi). After centrifugation, cells were seeded on petri dishes in minimum essential medium supplemented with d-valine. Five days later, cells were trypsinized and incubated with antibodies against CD31 (Pharmingen) and endoglin (Santa Cruz Biotechnology) for 30 min at room temperature. The cell suspension was then placed in the presence of microbeads coupled with rat IgG (Miltenyi) for 15 min at 4°C. After several washes with Ca/Mg PBS and passaging through a magnetic column (MACS MS columns, Miltenyi), the endothelial cell preparation was harvested.

FITC dextran or rhodamine dextran injection

FITC dextran (70 kDa) was injected in the ventricle and left for circulation for 5 min in adult mice (500 μl at 10 mg/ml). Rhodamine dextran (10 kDa) was injected in the liver of E14.5 embryos and left for circulation for 5 min (5 μl at 100 mg/ml) using a microinjector. The head was then decapitated, and the brain was removed and placed in a tube containing paraformaldehyde for 24 hours; the brain was then incubated in 30% sucrose for 48 hours and embedded in OCT (optimal cutting temperature). Sections of 7 μm were then cut for visualization.

Micro-CT analysis

The brain vasculature was imaged with a high-resolution micro-CT imaging system (Bruker MicroCT), set to an effective detector pixel size of 0.08 mm. The apparatus used for imaging the vessels was the Bruker SkyScan microCT scanner from Bruker with spatial resolutions of 7 to 36 μm and used with the Bruker Acquisition and DigiXCT reconstruction utility programs from Digisens. Data were acquired in axial mode, as described previously (53). The volume was visualized with the Avizo program.

Expression constructs

From human Pdzrn3 complementary DNA (cDNA) provided by M. Inui (54), the sequence of Pdzrn3 was subcloned in frame into the pcDNA3.1-V5-His plasmid (Invitrogen) to generate Pdzrn3-V5. The Pdzrn3-V5 sequence was subcloned into the lentiviral vector pRRLsin-MND-MCS-WPRE. Lentivirus preparations were produced at the Bordeaux 2 lentivirus platform. The MUPP1-Flag cDNA was provided by T. Miller (55).

Immunofluorescence staining

For immunostaining, P7 and adult brains were fixed in 4% paraformaldehyde/PBS, embedded in OCT/paraffin or agarose, and sectioned in 10-μm/7-μm or 100-μm sections, respectively. Endothelial cells were fixed in 4% formalin. Immunofluorescence staining was performed as previously described (20, 35), with antibodies specific for CD31 (Pharmingen), ZO1 (Invitrogen), aquaporin 4 (Millipore), PKCζ (Santa Cruz Biotechnology), V5 (Invitrogen), PAR3 (Millipore), and F-actin (Molecular Probes). Images were taken with a confocal microscope (Olympus FSV 1000). Visualization was obtained after digitized image multichannel deconvolution by AutoDeblur (Media Cybernetics Inc.), and three-dimensional projections were digitally reconstituted from stacks of confocal optical slices by Imaris software (Bitplane).

Cell culture/DNA transfection and drug treatments

HUVECs and HMVECs were cultured in EGM-2 or in EGM-2MV medium, respectively. HeLa cells were cultured in RPMI supplemented with 10% fetal bovine serum and penicillin-streptomycin. Primary endothelial cells from control (iEC WT) and Pdzrn3-deleted (iEC KO) mice were isolated from kidneys after magnetic separation with MACS MS columns (Miltenyi), as described previously (20). Plasmids were transfected using jetPRIME (Polyplus) following the manufacturer’s protocol [final concentration of 1.5 μg/ml for human embryonic kidney (HEK) 293 cells and 0.5 μg/ml for HeLa cells]. SiRNAs were transfected using interferin (Polyplus) at a final concentration of 30 nM. The oligonucleotides used were designed by Origene for h-Pdzrn3 (SR307944A, 5′-CGGCACTAGAGTATACAATTCCTTC-3′; SR307944B, 5′-CGATCCAAATGACTACATTGGAGAC-3′; SR307944C, 5′-GCATGTACAGGAAAGTTGTGGATG-3′), for hMUPP1 (04, 5′-GGAAUAUUUGUGUAGGUGAtt-3′; 08, 5′-AUUGCCUAAGGACUUGUGCUUtt-3′) as previously reported (47), and for hPKCζ Stealth RNAi (#HSS18348) as previously reported (19). Lentiviruses were tranduced at a multiplicity of infection of 20. MG132 (Santa Cruz Biotechnology) at a concentration of 0.25 M was used. The efficacy of Pdzrn3, MUPP1, and PKCζ knockdown was assessed by immunoblotting (fig. S6, A to C).

Western blot, immunoprecipitation, ubiquitination analysis, and in vitro pulldown assay

Immunoprecipitation and immunoblotting were performed as described previously (20, 35). Antibodies against PKCζ (Santa Cruz), Par3 (Millipore), V5 (Invitrogen), and MUPP1 (Abcam) were used for immunoprecipitation assays. Proteins were then resolved by SDS–polyacrylamide gel electrophoresis and blotted with antibodies specific for V5 (Invitrogen), PDZRN3 (Santa Cruz), phospho-c-Jun (Cell Signaling), c-Jun (Upstate), phosphorylated β-catenin (ABC, clone 8E7; Millipore), β-catenin (Sigma-Aldrich), α-tubulin (Sigma-Aldrich), Ve-cadherin (Santa Cruz Biotechnology), claudin 5 (Millipore), ZO1 (Invitrogen), PKCζ (Santa Cruz Biotechnology), and phospho-PKCζ (Ozyme Cell Signaling). Binding of antibodies to the blots was detected using the Odyssey Infrared Imaging System (LI-COR Biosciences).

Cell fractionation

For cell fractionation, cell or tissue lysates were resuspended in a buffer [containing Pipes (pH 6.8), 10 mM KCl, 1.5 mM MgCl2, 250 mM sucrose, 1 mM EGTA, and 1 mM EDTA] and potterized 20 times on ice. After a centrifugation at 800g for 5 min, the pellet was reserved for nuclear extract preparation, and the supernatant was collected and spun at 100,000g for 1 hour. The pellet and supernatant were classified as the “cell membrane fraction” and “cytoplasmic fraction,” respectively. The pellet obtained after the centrifugation at 800g was washed and placed as a cushion on top of a 1 M sucrose buffer. After a centrifugation at 2700g for 5 min, the pellet was resuspended in a buffer containing tris (pH 7.4), 10 mM NaCl, 3 mM MgCl2, 1 mM EGTA, and 0.05% NP-40 and spun again at 2700g for 20 min. The pellet was then resuspended in a buffer containing Hepes (pH 7.9), 300 mM NaCl, 1.5 mM MgCl2, and 0.2 mM EDTA. After 20 min of incubation, the lysate was spun at 10,000g for 20 min. The pellet was classified as the “nuclear membrane fraction,” and the supernatant was classified as the “soluble nuclear fraction.”

RNA preparation and quantitative PCR

Mouse tissues were homogenized in TRI Reagent (Euromedex), and RNA was extracted according to the manufacturer’s instructions. Quantitative PCR was performed as described previously (20).

Permeability assay

HUVECs were seeded on inserts with a pore size of 0.4 μm (BD Biosciences), cultured in medium, and assayed for permeability to FITC dextran (40 kDa), as previously described (56).

Statistical analysis

The experimental results represent means ± SEM. Each experiment was conducted at least three times. When multiple experiments using different numbers of animals were pooled for the statistical analysis, the range of number of animals was indicated in the figure legend. Comparison of continuous variables between two groups was performed using Mann-Whitney test. Comparison of triplicate data between four groups was performed by one-way ANOVA with post hoc Bonferroni test. Comparison of multiple measures was performed by two-way ANOVA with post hoc Bonferroni test. P < 0.05 was considered statistically significant. All analyses were performed with appropriate software (GraphPad Prism 6, GraphPad Software). The statistical test is indicated for each data analysis.


Fig. S1. Ectopic expression of Pdzrn3 in embryos.

Fig. S2. Immunostaining of PDZRN3 in vessels in brain sections.

Fig. S3. Analysis of Pdzrn3 depletion in isolated endothelial cells.

Fig. S4. Analysis of mutant and control mouse brain vasculature by micro-CT.

Fig. S5. Quantitative RT-PCR analysis of Wnt3a, Wnt5a, Wnt7a, and Wnt7b expression after MCAO.

Fig. S6. Knockdown efficiency of siRNAs directed against Pdzrn3, Pkcζ, and MUPP1.

Fig. S7. Schematic figure.


Acknowledgments: We thank P. Couraud (Paris, France) for discussions, D. Massey-Harroche (Institut de Biologie du Développement de Marseille, Marseille, France) for her support with MUPP1 reagents, D. Vivien (GIP Cyceron, Caen, France) for discussions and his help in AVC training, and D. Daret (Inserm U1034, Bordeaux, France) for electron transmission technical expertise. We thank the SEAT-TAAM CNRS Phenomin mouse facility (France) for the generation of transgenic mice by DNA construct microinjection, M. Inui (Yamaguchi University, Yamaguchi, Japan) for providing the human Pdzrn3 cDNA, and T. Miller (Case Western Reserve University, Cleveland, USA) for providing the FLAG-tagged MUPP1 plasmid. Funding: This work was supported by grants from Fondation pour la Recherche Médicale (DVS20131228623) and from Region Aquitaine (20101301039MP). R.N.S. was supported by a grant from Region Aquitaine (20101301039MP) and from Fondation pour la Recherche Médicale (FDT20130928208). V.J. was supported by a grant from Fondation pour la Recherche Médicale (DEA20130726937). Author contributions: R.N.S. conducted and designed the experiments, analyzed and interpreted the data, and wrote the manuscript; B.J.-V. conducted the initial experiments and analyzed and interpreted the data; B.J.-V., H.K., and S.J. conducted the biochemical experiments; V.J., T.W., and N.F. helped technically with the experiments; M.P., E.R., and P.D. were involved in study design; T.C. and C.D. designed the experiments, analyzed and interpreted the data, and edited the manuscript. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Signaling

Navigate This Article