Disheveled-1 Interacts with Claudin-5 and Contributes to Norrin-Induced Endothelial Barrier Restoration

Abstract

Previous studies have revealed that norrin can reverse vascular endothelial-growth-factor (VEGF)-induced permeability in a β-catenin-dependent pathway. Here, we have explored the contribution of disheveled-1 (DVL1) in norrin-induced blood-retinal barrier (BRB) restoration. We provide evidence that in addition to canonical signaling, DVL1 promotes tight junction (TJ) stabilization through a novel, non-canonical signaling pathway involving direct claudin-5 (CLDN5) binding. Immunofluorescence staining of rat retinal cross-sections showed enriched expression of DVL1 and 3 at endothelial capillaries and co-localization with CLDN5 and ZO-1 at the TJ complex in primary bovine retinal endothelial cells (BRECs). Barrier properties of BRECs were determined via measurements of trans-endothelial electrical resistance (TEER) or permeability to 70 kDa RITC-dextran. These studies demonstrated that norrin restoration of barrier properties after VEGF treatment required DVL1 as an siRNA knockdown of Dvl1 but not Dvl2 or Dvl3, reduced basal barrier properties and ablated norrin-induced barrier restoration. However, loss of Dvl1 did not decrease β-catenin signaling activity as measured by Axin2 mRNA expression, suggesting the contribution of a non-canonical pathway. DVL and TJ protein interactions were analyzed via co-immunoprecipitation of endogenous protein in BRECs, which demonstrated that DVL1 interacts with both CLDN5 and ZO-1, while DVL3 interacts only with ZO-1. These interactions were most abundant after inducing BRB restoration by treating BRECs with VEGF and norrin. DVL has previously been shown to form intramolecular bindings between the C-terminal PDZ-binding motif (PDZ-BM) with an internal PDZ domain. Co-transfection of HEK293 cells with DVL1 and CLDN5 or relevant mutants revealed that DVL1 interacts with CLDN5 through the DVL PDZ domain binding, CLDN5 PDZ-BM, in competition with DVL1 PDZ-BM, since DVL/CLDN5 interaction increases with deletion of the DVL1 PDZ-BM and decreases by co-expressing the C-terminal fragment of DVL1 containing the PDZ-BM or through deletion of CLDN5 PDZ-BM. In BREC cells, transfection of the C-terminal fragment of DVL1 downregulates the expression of CLDN5 but does not affect the expression of other proteins of the TJs, including ZO-1, occludin, CLDN1 or VE-cadherin. Blocking DVL1/CLDN5 interaction increased basal permeability and prevented norrin induction of barrier properties after VEGF. Combined with previous data, these results demonstrate that norrin signals through both a canonical β-catenin pathway and a non-canonical signaling pathway by which DVL1 directly binds to CLDN5 to promote barrier properties.

Keywords: Wnt signaling; barriergenesis; blood–retinal barrier; claudin-5; disheveled; endothelium; norrin; permeability; retina; tight junction.

Figure 1

DVL1 and 3 are highly expressed in retinal capillaries, co-localizing with the TJ proteins CLDN5 and ZO-1. (a) Cross-sections of rat retinas immunostained with specific antibodies against DVL2 and DVL3 or with Pan-DVL antibody detecting both, DVL1 and DVL3 (green). Isolectin B4 (IB4) was used as a marker of blood-vessels (red); scale bar = 50 μM. (b) Co-localization of DVL (green) with the TJ proteins CLDN5 or ZO-1 (red) in rat retina whole mounts. Arrows point to the sites of co-localization in the TJs. (c) Co-immunofluorescence staining of DVL proteins (green), ZO-1 (red) and CLDN5 (magenta) in BREC monolayers; scale bar = 10 μM. (d) BRECs transfected with siDvl3 or scramble (Scr) before the IF staining with Pan-DVL (green), CLDN5 (red) and ZO-1 (white) antibodies; scale bar = 10 μM. (e) Protein co-localization, analyzed in xz stacks using Mander’s correlation coefficient, shows a high overlap of DVL1 with the TJ proteins in siDvl3 monolayers. p values were calculated via t-test analysis. Error bars, S.D.

Figure 2

The knockdown of Dvl1 specifically, reduced basal barrier properties despite increased β-catenin signaling. (a) The specificity of siRNA sequences was analyzed via the qRT-PCR of Dvl1 or Dvl2 or Dvl3 PCR in BRECs transfected with siDvl1, siDvl2, siDvl3, the combined siDvl2/3 or scramble (Scr) control. Barrier properties were analyzed through the permeability to 70 kDa RITC-dextran molecule (b) and Axin2 mRNA expression was measured, being accounted as β-catenin signaling activation (c). (d) Immunofluorescence staining of TJ proteins CLDN5 and ZO-1 in BREC monolayers after Dvl1 knockdown; scale bar = 10 μM. (e–h) Knockdown of Dvl1 with a second siRNA sequence. (e) qRT-PCR of Dvl1 or Dvl2, or Dvl3 PCR. (f) Solute flux. (g) qRT-PCR of Axin2. (h) Immunofluorescence staining of the TJ proteins; scale bar = 10 μM. p values were calculated via one-way ANOVA followed by Sidak post hoc test (a–c), or via t-test analysis (d–h). Error bars, S.D.

Figure 3

The knockdown of Dvl1 but not Dvl3-ablated norrin-induced blood–retinal barrier restoration after VEGF. (a,d) TEER or (b,c,e) permeability to a 70 kDa RITC-dextran molecule was measured on BRECs transfected with siDvl1a (a,b), a second siDvl1b (c), siDvl3 (d,e) or a scramble (Scr) control sequence. Monolayers were stimulated with vehicle (control), VEGF 50 ng/mL or VEGF and norrin 40 ng/mL (VEGF/Norrin; V/N) 72 h after transfection. Solute flux was measured after 72 h of stimulation. p values were calculated via two-way ANOVA, followed by Tukey´s post hoc test. Error bars, S.D.

Figure 4

DVL1 interacts with CLDN5 and ZO-1, while DVL3 interacts with ZO-1, especially with the addition of VEGF and norrin. BREC monolayers were stimulated with vehicle (control, C), norrin 40 ng/mL (N), VEGF 50 ng/mL (V) or both (V/N). After 72 h of stimulation, lysates were collected for total protein or the immunoprecipitation (IP) of DVL using Pan-DVL (a,b) or DVL3 (c,d) specific antibodies, followed by the immunoblot (IB) with Pan-DVL, DVL3, CLDN5 or ZO-1 antibodies. Actin was used as a loading control of total protein. (a,c) Representative Western blots of total and IP protein. (b,d) Densitometry of CLDN5 or ZO-1 proteins relative to IP protein. p values were calculated by one-way ANOVA, followed by Sidak post hoc test. Error bars, S.D.

Figure 5

DVL1 interacts with CLDN5 through its PDZ domain. HEK-293 cells co-transfected with full-length DVL1 (HA-DVL1-WT), HA-DVL1-WT and DVL1 C-terminus (CT) 169aa fragment (HA-DVL1-CT) or Dvl1 deleted in the last 7aa which includes the PDZ-BM (HA-DVL1-ΔC7), together with FLAG-CLDN5 or FLAG-CLDN5Δ-PDZBM. Cell lysates were collected for total protein or the immunoprecipitation (IP) of DVL mutants or CLDN5, using specific antibodies against HA or FLAG tags, respectively. (a,c,e) Representative immunoblot (IB) of total and IP protein, using HA or FLAG antibodies. Actin was used as a loading control of total protein. (b,d,f) Densitometry of IP proteins. p values were calculated via one-way ANOVA, followed by Sidak post hoc test. Error bars, S.D.

Figure 6

DVL1-CT fragment reduces CLDN5 content and DVL1/CLDN5 interaction in endothelial cells. BRECs were transfected with empty vector (EV), DVL1 WT (HA-DVL1) or DVL1 C-terminus (CT) 169aa fragment (HA-DVL1-CT). Some monolayers were stimulated with vehicle (control, C) or VEGF 50 ng/mL for 30 min, followed by Norrin 40 ng/mL (V/N) for additional 30 min. Cell lysates were collected for total protein or the immunoprecipitation (IP) of DVL using Pan-DVL antibody, followed by the immunoblot (IB) with Pan-DVL, CLDN5, ZO-1, VE-cadherin, occludin and CLDN1 antibodies. Actin was used as a loading control of total protein. (a,c) Representative Western blots. (b,d) Densitometry analysis. In IP experiments, CLDN5 protein was calculated relative to DVL IP protein. p values were calculated via one-way ANOVA, followed by Sidak post hoc test. Error bars, S.D.

Figure 7

Effect of DVL1-CT in barrier properties. BRECs were transfected with empty vector (EV) or DVL1 C-terminus (CT) 169aa fragment (HA-DVL1-CT) for measurements of (a) basal permeability to a 70 kDa RITC-dextran molecule, (b) Axin2 mRNA expression (β-catenin signaling activity) or (c) the IF staining of DVL1 mutants (HA-tag; green) and CLDN5 or ZO-1 TJ proteins (red); scale bar = 10 μM. Arrows point to sites of TJ protein disruption in transfected cells. (d) Quantification of four images per experiment by masked scoring combined from three individuals. Data show the frequency of CLDN5 or ZO-1 loss at the cell contacts ranked in five categories, where 100% corresponds to complete loss. (e) BREC monolayers transfected with EV or DVL1-CT fragment and stimulated with vehicle (control, C), VEGF (V) or VEGF/Norrin (V/N) for 24 h before measurements of their permeability to a 70 kDa RITC-dextran molecule. p values were calculated via t-test analysis (a,b,d) or via two-way ANOVA (e), followed by Tukey’s post hoc test. Error bars, S.D.

Figure 8

Norrin signals through DVL1 to stimulate barrier properties. Previous studies revealed that VEGF-induces endothelial permeability and simultaneously promotes norrin signaling by inducing FZD4 co-receptor TSPAN12 migration to the plasma membrane. Norrin signals through DVL1 promoting the canonical pathway with β-catenin stabilization, likely through the closed confirmation of DVL1 involving intramolecular binding of DVL1 PDZ domain and DVL1 PDZ-BM. Data here reveal an additional non-canonical signaling pathway necessary for tight junction organization and induction of barrier properties involving the open conformation of DVL1 with its PDZ domain binding to claudin 5 PDZ-BM.