TNIK: A redox sensor in endothelial cell permeability

Abstract

Dysregulation of endothelial barrier integrity can lead to vascular leak and potentially fatal oedema. TNF-α controls endothelial permeability during inflammation and requires the actin organizing Ezrin-Radixin-Moesin (ERM) proteins. We identified TRAF2 and NCK-interacting kinase (TNIK) as a kinase directly phosphorylating and activating ERM, specifically at the plasma membrane of primary human endothelial cells. TNIK mediates TNF-α-dependent cellular stiffness and paracellular gap formation in vitro and is essential in driving inflammatory oedema formation in vivo. Unlike its homologs, TNIK activity is negatively and reversibly regulated by H₂O₂-mediated oxidation of C202 within the kinase domain. TNIK oxidation results in intermolecular disulfide bond formation and loss of kinase activity. Pharmacologic inhibition of endogenous reactive oxygen species production in endothelial cells elevated TNIK-dependent ERM phosphorylation, endothelial cell contraction, and cell rounding. Together, we highlight an interplay between TNIK, ERM phosphorylation, and redox signalling in regulating TNF-induced endothelial cell permeability.

Fig. 1.. MSN and TNIK are ERM kinases in Drosophila and HUVEC, respectively.

(A) Kinome-wide RNAi screen in D. melanogaster S2R⁺ cells. A score of p-Dmoesin staining intensity (arbitrary scale) was assigned to each kinase knockdown, with a cutoff score of ≤50 regarded as a kinase positively regulating Dmoesin phosphorylation. (B) Confocal images from high-content screen in (A), cells were fixed and stained with anti–p-Dmoesin (green) or 4′,6-diamidino-2-phenylindole (DAPI) (red) to stain nuclei. Scale bars, 50 μm. Ctrl, control S2R⁺ cells, not treated with dsRNA. (C) Alignment of Drosophila (D) moesin with human (hs) moesin showing the highly conserved amino acid region around Thr⁵⁵⁸ (human)/Thr⁵⁵⁹ (Drosophila), highlighted in green, which is detected by the anti–p-ERM antibody. (D) GST-TNIK KD mixed with GST-Moesin C-terminal domain and subjected to in vitro kinase assays before Western blotting. (E) In vitro kinase assays using full-length WT YFP-TNIK or kinase-dead K54A YFP-TNIK, followed by Western blotting as in (D). Arrowhead denotes GST-Moesin C-terminal domain. Autophos., autophosphorylation. (F) HUVEC transfected with control siRNA (Ctrl, RISC-free) or siRNA duplexes targeted to TNIK and stimulated with TNF-α for 15 min before Western blotting. p-Ezr, p-Ezrin. p-Moe, p-Moesin. (G and H) Quantification of (F), two-way analysis of variance (ANOVA), means ± SEM. ***P ≤ 0.001, n = 3 independent experiments. RF, RISC-free control siRNA. (I) HUVEC were treated with the indicated concentration of TNIK inhibitor KY-05009 for 2 hours before Western blotting. (J) HUVEC transfected with YFP-TNIK reveal that only membrane-localized TNIK is coincident with p-ERM. Scale bar, (C) 10 μm. KD, knock down; ATP, adenosine 5′-triphosphate.

Fig. 2.. TNIK positively regulates TNF-α–induced permeability and KY-05009 reverses permeability after TNF-α stimulation.

(A) Schematic illustrating transwell in vitro endothelial permeability assay (see also Materials and Methods). (B) HUVEC monolayers were treated with RISC-free (RF) control siRNA or TNIK siRNA before transwell permeability assays. Monolayers were stimulated for 4 hours with TNF-α, and fluorescent tracers (67 kDa tetramethyl rhodamine isothiocyanate (TRITC)–bovine serum albumin (BSA) and 4 kDa fluorescein isothiocyanate (FITC)–dextran) were subsequently added for a further hour before detection of fluorescence emerging in the lower chamber (see Materials and Methods). Statistical analysis using two-way ANOVA, means ± SEM, ***P ≤ 0.001, four independent experiments. (C) HUVEC monolayers were stimulated with TNF-α for 4 hours, followed by the addition of fluorescein isothiocyanate (FITC)–dextran or tetramethyl rhodamine isothiocyanate (TRITC)–bovine serum albumin (BSA), alongside dimethyl sulfoxide (DMSO) or 10 μM KY-05009 for a further 1 hour to monitor resealing or reversal of permeability (see Materials and Methods). Statistical analysis using two-way ANOVA, means ± SEM, **P ≤ 0.01 and ***P ≤ 0.001, four independent experiments. (D) Evaluation of the effect of KY-05009 on endothelial barrier integrity using endothelial cell-substrate impedance sensing. Cells were pretreated with 10 μM KY-05009 (2 hours) followed by addition of TNF-α (10 ng/ml, arrow). Resistance values were normalized to baseline (see fig. S2A for non-normalized graph). Green arrowheads indicate where resistance was decreasing mid-point at 4 hours and where resistance was plateauing at 8 hours. (E and F) Graphs of the normalized resistance trace represent TNF-α induced drop in endothelial resistance as compared to baseline at t = 4 hours and t = 8 hours. Mean and 95% confidence interval of 12 biological repeats from n = 4 independent experiments. Unpaired Student’s t test, **P ≤ 0.0013 and ****P ≤ 0.0001.

Fig. 3.. TNIK positively regulates TNF-α–induced paracellular gap formation in vitro, and TNF-α–induced oedema formation in vivo.

(A) Time course of HUVEC gap formation in response to TNF-α stimulation. Four fields of view (FoV) were acquired per experiment (see fig. S2B for representative FoV). Data represent n = 4 and are expressed as a percentage of gaps per FoV. Exp, experiment. Data analyzed using a repeated measures one-way ANOVA with Dunnett’s test for differences relative to unstimulated cells: *P ≤ 0.05. (B) Two images showing approach for quantifying endothelial gaps (see Materials and Methods for more details). Scale bar, 50 μm. (C) Quantification of n = 4 p-ERM Western blots of HUVEC monolayers treated with TNF-α for 7.5 hours, followed by 30 min with KY-05009 or DMSO carrier. Unpaired Student’s t test, ****P ≤ 0.0001. (D) Paracellular (PC) gaps were quantified as indicated in left image of (B). Paracellular gaps are resealed after addition of KY-05009. (E) Representative images of data acquired in (D), showing resealing of paracellular gaps (see fig. S2C for larger FoV). Scale bar, 50 μm. (F) Schematic of in vivo skin plasma extravasation assay (see also Materials and Methods). IP, intraperitoneal injection. IV, intravenous injection. ID, intradermal injection. (G) Tnik endothelial-specific conditional knockout mice were subjected to skin permeability assays. n = 6 to 7 mice per group. Statistical analysis using two-way ANOVA, means ± SEM, *P ≤ 0.05, **P ≤ 0.01, and ****P ≤ 0.0001.

Fig. 4.. TNF-α–induced activation of HUVEC leads to bimodal ERM phosphorylation, contributing to cellular stiffness at 5 min and stable paracellular gaps at 4 to 8 hours.

(A) HUVECs were stimulated with TNF-α (10 ng/ml) at indicated time points, and whole-cell lysates were harvested for Western blotting for p-ERM. Data are from n = 4 independent experiments. Western blot below bar graph is representative of the data. One-way ANOVA with Tukey’s post hoc test *P ≤ 0.05 and **P ≤ 0.01. (B) Confocal images of HUVEC before and after TNF-α stimulation. Fluorescence signals corresponding to VE-cadherin (red), p-ERM (green), and DNA (DAPI, blue). Scale bar, 10 μm. (C) Still taken from movie S1: HUVEC transduced with lentivirus to express T558D (TD) moesin-GFP, a phospho-mimic mutant of moesin (98). Yellow arrows indicate the direction of migration. Scale bar, 66 μm. (D and E) Quantification of Young’s Modulus (YM) of control (i.e., before TNF-α treatment). −/+ TNF-α treatment in the presence of DMSO (carrier) or KY-05009. Mean whole-cell YM values were plotted over time, approximately 21 min (i.e., 3.5 min per scan number). The first scan was acquired before adding TNF-α and DMSO/KY-05009. Values are represented as a change from control. N/n = 3/38 to 40. Data are presented as means ± SEM, Kruskal-Wallis test, *P < 0.05, **P < 0.01, and ***P < 0.001. (F and G) Representative 100 μm by 100 μm SICM images of topography and YM maps using the loop mode. (H) Western blot of p-ERM, performed similarly to (A), but separate series of experiments altogether, showing acute rises in p-ERM. p-MLC2 Ser¹⁹ = phospho myosin light chain kinase at position serine-19. Actin loading control and total TNIK protein are both represented. (I) Quantification of (H), n = 3. Two-way ANOVA with Tukey’s post hoc test. *P ≤ 0.05 and ****P ≤ 0.0001. h, hours.

Fig. 5.. Membrane-localized TNIK undergoes reversible oxidation through C202.

(A) Schematic of PEG-switch assay (51). (B) HUVEC monolayers were treated with hydrogen peroxide (H₂O₂) for 15 min and washed into EGM-2 medium without H₂O₂ for 10 min before PEG-switch assays and Western blotting. Bands corresponding to oxidized (Ox) and reduced (Red) forms of endogenous TNIK are indicated. (C) HUVEC monolayers were treated with 50 μM H₂O₂ for 15 min and washed into EGM-2 medium without H₂O₂ for the indicated time before PEG-switch assays and Western blotting. Bands corresponding to oxidized and reduced forms of endogenous TNIKs are indicated with arrowheads. (D) HUVEC monolayers were treated with H₂O₂ for 15 min before PEG-switch assays and Western blotting. Bands corresponding to oxidized and reduced forms of endogenous TNIK are indicated. (E) HUVEC monolayers were treated with 200 μM H₂O₂ for 15 min, before subcellular fractionation, PEG-switch assays and Western blotting. Nuc, nuclear pellet. PNS, post-nuclear supernatant. Cyt, cytosol. NaCl, 500 mM NaCl washed membranes. DRM, 1% Triton X-100 4°C detergent resistant membranes. Bands corresponding to oxidized and reduced forms of endogenous TNIK are indicated with arrowheads. (F) Recombinant GST-tagged TNIK KD was oxidized in vitro with 50 μM H₂O₂ at 30°C for 15 min followed by reducing (+DTT) or nonreducing (−DTT) SDS-PAGE and Western blotting. (G) YFP-TNIK or YFP-TNIK mutants were expressed in HEK293T cells by plasmid transfection and immunoprecipitated. Immunoprecipitates were oxidized with 200 μM H₂O₂ at 30°C for 15 min followed by PEG-switch assays and Western blotting. Oxidised/reduced forms of YFP-TNIK are indicated. (H) YFP-TNIK or YFP-TNIK C202S were expressed in HEK293T cells and immunoprecipitated. Immunoprecipitates were oxidized with increasing amounts of H₂O₂: 0, 0.2, 0.5, and 1 mM at 30°C for 15 min followed by nonreducing SDS-PAGE and Western blotting. WB, Western blot;

Fig. 6.. MD simulations predict apposed TNIK C202 residues in the TNIK dimer are able to form reversible disulfide bonds.

(A) Distance between apposed C202 residue sulphur atoms in WT-TNIK docked dimer during 50-ns MD simulation. See also movie S4. Relates to fig. S4C, replica 1. (B) 3D structure of the minimum distance between C202-C202, retrieved from MD simulation in (A). (C) Frequency (number of frames) from MD simulations of the TNIK dimer where apposed C202-C202 distances are within the reversible disulfide bond range (>3.0 Å and ≤6.2 Å) (54) (red). The distances between apposed C202 residues in the WT-TNIK crystal structure dimer (PDB: 5CWZ) were calculated from an extended simulation trajectory of 124 ns (gray). The C202-C202 distances for the wt-TNIK docked model (blue) were calculated using a concatenated trajectory containing all four simulations from fig. S4C and 20,000 frames in total. (D) Distance between apposed in silico mutated S202 residues (side-chain hydroxyl hydrogen atoms) in mt-TNIK–docked dimer during 50-ns MD simulation. See also movie S5. Relates to fig. S4F, replica 4. (E) 3D structure of the maximum distance between S202-S202, retrieved from MD simulation in (D). (F) Frequency (number of frames) from MD simulations of the mt-TNIK–docked dimer where apposed in silico mutated S202-S202 distances are within the reversible disulfide bond range (>3.0 Å and ≤6.2 Å) (54) (red). The S202-S202 distances for the mt-TNIK–docked model (blue) were calculated using a concatenated trajectory containing all four simulations from fig. S4F and 20,000 frames in total.

Fig. 7.. TNIK kinase activity is reversibly inhibited through cysteine oxidation by hydrogen peroxide.

(A) GST-TNIK KD was mixed with MBP and subjected to in vitro kinase assays before Western blotting. Reaction mixtures were oxidized with the indicated concentration of hydrogen peroxide before the addition of ATP. (B) Quantification of (A), statistics was performed using a one-way ANOVA, means ± SEM, *P ≤ 0.05, ***P ≤ 0.001, and ****P ≤ 0.0001, n = 3 independent experiments. (C) GST-TNIK KD was oxidized with the indicated concentration of hydrogen peroxide, followed by reduction with DTT where indicated. In vitro kinase assays were then performed, followed by Western blotting.

Fig. 8.. Blocking endogenous ROS in HUVEC induces cell rounding and large paracellular gaps in a KY-05009-sensitive manner.

(A) HUVEC monolayer before and after treatment with DPI for 30 min (left and middle), followed by TNIK inhibition with KY-05009 for 2 hours (right). Stills from live-cell imaging time-lapse (see also movies S7 and S8). Scale bars 60 micrometers. (B) Quantification of cell rounding in HUVEC treated with DMSO (carrier), DPI, KY-05009 (KY), and DPI for 30 min and then KY for 2 hours or KY for 30 min and then DPI for 2 hours (see movie S9 for KY and then DPI). Statistical analysis using one-way ANOVA, means ± SEM, ****P ≤ 0.0001, four independent experiments. (C) Western blot of HUVEC whole cell lysates, representative of two independent experiments (one experiment resolved per lane). HUVEC treated with DMSO vehicle control, DPI or simultaneously with both DPI and KY-05009 (TNIKi, TNIK inhibitor) for 15 min followed by Western blotting. p-Ezr, p-Ezrin. p-Moe, p-Moesin. p-MLC2, phospho-myosin light chain at position serine 19. Arrowhead indicates the position of p-MLC2. Actin loading controls are for TNIK and p-ezrin, total ezrin, and p-MLC, respectively.