Rearrangement of adherens junctions by transforming growth factor-β1: role of contraction

Victor Hurst IV, Peter L. Goldberg, Fred L. Minnear, Ronald L. Heimark, Peter A. Vincent

Abstract

The signal transduction pathways that lead to disruption of pulmonary endothelial monolayer integrity by transforming growth factor-β1 (TGF-β1) have not been elucidated. The purpose of this investigation was to determine whether disassembly of the adherens junction is temporally associated with the TGF-β1-induced decrease in pulmonary endothelial monolayer integrity. Measurement of albumin clearance and electrical resistance showed that monolayer integrity started to decrease between 1 and 2 h post-TGF-β1 treatment and continued to slowly decrease over the next 6 h. Immunofluorescence microscopy of monolayers between 2 and 3 h post-TGF-β1 showed that β-catenin, plakoglobin, α-catenin, and cadherin-5 were colocalized both at the cell periphery and in newly formed bands that are perpendicular to the cell-cell border. At 4 h post-TGF-β1, cells began separating; however, β- and α-catenin, plakoglobin, and cadherin-5 could still be found at the cell periphery at areas of cell separation and in strands between separated cells. By 8 h, these junctional proteins were no longer present at the cell periphery at areas of cell separation. The myosin light chain kinase inhibitor KT-5926 prevented the TGF-β1-induced change in integrity but did not inhibit the formation of actin stress fibers or the formation of bands containing adherens junction proteins that were perpendicular to the cell-cell junction. Overall, these results suggest that adherens junction disassembly occurs after cell separation during TGF-β1-induced decreases in pulmonary endothelial monolayer integrity and that the loss of integrity may be due to the activation of a myosin light chain kinase-dependent signaling cascade.

  • cadherin
  • catenin
  • myosin
  • myosin light chain
  • myosin light chain kinase
  • actin
  • vascular endothelial cells
  • immunofluorescence microscopy
  • KT-5926

increases in the paracellular flux of macromolecules can result from changes in endothelial cell shape and the formation of intercellular gaps. Endothelial cell shape and gap formation are believed to be controlled by signal transduction pathways that alter the balance of competing adhesive and contractile forces (reviewed in Refs. 16, 27). Endothelial cell-cell and cell-matrix contacts tether endothelial cells to each other and to the extracellular matrix, respectively, and act against centripetal tension generated by actomyosin motors. Thus the loss of cell-cell or cell-matrix attachment or the activation of contractility will alter the balance of forces within the cell, resulting in cell retraction and a decrease in endothelial monolayer integrity.

A number of investigators (15, 20, 29, 30, 39) have demonstrated that thrombin and histamine stimulate an actin-myosin motor system within endothelial cells that causes cells to contract. The signaling cascade is similar to what has been observed in smooth muscle cells where activation of myosin light chain (MLC) kinase (MLCK) leads to the phosphorylation of MLC and, subsequently, cell contraction. Treatment of endothelial monolayers with thrombin has been shown to activate MLCK, causing decreases in barrier integrity as demonstrated by observed increases in paracellular protein permeability (15, 20). Taken together, these studies suggested that physiological agents can alter the endothelial monolayer integrity through the activation of MLCK, which subsequently results in cell contraction.

The adherens junction, which is composed of cadherins and catenins, is one type of cell junction that maintains endothelial cell-cell adhesion. The cadherins are a family of structurally related proteins that bind in a Ca2+-dependent, homophilic fashion and have been implicated in modulating the integrity of endothelial cell monolayers (2, 26). Cadherin-5, also called VE-cadherin, is the cadherin found specifically on endothelial cells (25, 26). The cytoplasmic tail of cadherins is very well conserved between the different types of cadherins as well as across species (17). This cytoplasmic tail has been shown to bind to the actin cytoskeleton via cytoplasmic proteins called catenins. The catenins have been classified on the basis of molecular mass as α-catenin (102 kDa), β-catenin (92 kDa), and γ-catenin (82 kDa), with γ-catenin being homologous to plakoglobin (reviewed in Refs. 21,44). The cytoplasmic tail of cadherin binds to either β-catenin or plakoglobin but not both as demonstrated by the inability to coprecipitate β-catenin and plakoglobin in the same cadherin-catenin complex (1, 23). α-Catenin, which has been shown to associate with the actin cytoskeleton, binds with β-catenin and plakoglobin, thereby linking the cadherin-catenin complex to the actin cytoskeleton (1, 32,34).

Recently, Lampugnani et al. (25) demonstrated that the association of cadherin-5 with β-catenin or plakoglobin in endothelial cells was dependent on the duration of cell-cell interaction. These investigators showed that complexes consisting of cadherin-β-catenin-α-catenin were localized to the cell junction before those complexes consisting of cadherin-plakoglobin-α-catenin, with the latter adherens junction complex being formed after the cell monolayers were confluent for 72 h. This order was reversed because cell-cell adhesion decreases when endothelial cells prepare to migrate; that is, plakoglobin was found to leave the junction before the disappearance of β-catenin. These findings led the investigators to postulate that the localization of β-catenin and plakoglobin to the junction may be related to the strength and/or integrity of the endothelial cell junction. Similar findings have also been reported by Schnittler et al. (38).

Transforming growth factor (TGF)-β1 is a 25-kDa disulfide-linked homodimer that has been implicated as a mediator in altering the structure and function of the vascular wall. TGF-β1 has been shown to alter endothelial cell phenotype as demonstrated by a decrease in cell-cell contact, inducing a reorganization of the actin cytoskeleton of the cell, and by a decrease in the localization of the tight junction protein ZO-1 at the cell periphery (11, 28). The ability of TGF-β1 to alter endothelial cell phenotype suggests that this cytokine-like growth factor may play a role in the regulation of vascular endothelial integrity. This may occur with pulmonary hypertension where there is an elevated level of TGF-β1 in the vessel wall (8) and an increased transvascular flux of water and macromolecules across the endothelial cell monolayer (36, 40). Botney et al. (8) found that TGF-β1-expression in vessels of hypoxia-induced pulmonary hypertension was decreased in the medial smooth muscle layer. However, immunohistochemical staining of TGF-β1 was more intense in endothelial cells of hypertensive vessels compared with that in endothelial cells in normotensive vessels. As reviewed by Stenmark et al. (41), a loss of endothelial integrity and the increased edema and transudation of mitogenic plasma factors into the subendothelial space have been implicated in initiating proliferation and protein synthesis in both smooth muscle cells and fibroblasts, both important factors in vascular remodeling found in pulmonary hypertension. Thus TGF-β1-induced changes in endothelial morphology may play a role in mediating decreased vascular integrity that leads to vascular remodeling.

The signal transduction pathways that lead to the TGF-β1-induced decrease in endothelial cell monolayer integrity are not known. We hypothesized that disassembly of the cell-cell adherens junction was, in part, responsible for the TGF-β1-induced decrease in endothelial monolayer integrity. Studies were undertaken to determine the time course and dose response of the TGF-β1-induced increase in permeability immediately after its addition to endothelial monolayers. Endothelial monolayer permeability was assessed by measuring changes in albumin clearance and electrical resistance. Immunofluorescence microscopy was then used to determine the kinetics of disassembly of cell-cell adherens junction after treatment of endothelial cell monolayers with TGF-β1. Colocalization of β-catenin and plakoglobin was performed to determine whether cadherin-catenin complexes consisting of plakoglobin would disassemble before those complexes consisting of β-catenin as described by Lampugnani et al. (25). Colocalization of β-catenin and actin was also performed to determine whether the changes in these two structures are temporally associated. We also used the MLCK inhibitor KT-5926 to determine whether TGF-β1-induced changes in endothelial integrity and rearrangement of the adherens junction were dependent on MLC phosphorylation, which would suggest a role for cell contraction.

METHODS

Endothelial cell culture. Calf pulmonary arterial endothelial cells (CPAECs; American Type Culture Collection) were grown in modified Eagle’s medium (MEM; GIBCO BRL) supplemented with 20% fetal bovine serum (Sterile Systems), nonessential amino acids (10 mM), penicillin (100 U/ml), and streptomycin (100 μg/ml; all from GIBCO BRL). The cells were split 1:4 every fifth day. All experiments were performed betweenpassages 18 and24.

Experimental protocol. The endothelial cells were seeded (8 × 104cells/cm2) and grown to confluence (3–5 days) on tissue culture-treated polycarbonate micropore membranes (13-mm diameter, 0.4-μm pore size; Transwell, Costar) for albumin clearance, on gelatinized wells for electrical resistance measurement, on glass coverslips for immunofluorescence microscopy, and on tissue culture-treated 35- and 60-mm-diameter plates (Sarstedt) for assessment of the soluble and insoluble protein pools and MLC phosphorylation, respectively. The day before the experiment, the medium was changed to MEM containing penicillin-streptomycin and 5% serum. TGF-β1 (porcine platelet) was purchased from R&D Systems. Antibodies to TGF-β1 and preimmune serum were also purchased from R&D Systems. KT-5926, an inhibitor of MLCK, was from Kamiya Biomedical. All incubations with TGF-β1 and KT-5926 were performed in MEM containing penicillin-streptomycin and 5% serum.

Assessment of 125I-labeled albumin clearance.

Measurement of 125I-labeled albumin clearance (in μl/min) as described by Cooper et al. (12) was used to assess the changes in the diffusive permeability of albumin across the endothelial monolayers. This technique allowed the measurement of transendothelial albumin transport in the absence of an oncotic pressure gradient and a changing hydrostatic pressure gradient (12). The dual-chamber monolayer system consisted of a 0.2-ml luminal chamber (containing a filter lined with a confluent endothelial monolayer) that floated in a larger 25-ml abluminal chamber. To ensure complete mixing, the abluminal chamber was stirred constantly, and both chambers were kept at a constant 37°C in a thermostatically controlled water bath. A concentration gradient of purified125I-labeled albumin was established across the endothelial monolayer. Present in the abluminal chamber was 25 ml of MEM containing 0.5% bovine serum albumin (BSA; Sigma). To the luminal chamber, we added 200 μl of the same BSA-containing MEM solution as well as the125I-labeled albumin. The protocol for the measurement of albumin clearance consisted of obtaining 400-μl samples from the abluminal chamber every 5 min for 60 min.

As discussed by Cooper et al. (12), the volume of the luminal chamber, which is cleared of the albumin tracer into the abluminal chamber, represents the total activity of the abluminal chamber. The change in clearance volume during the interval between sampling points was calculated by dividing the amount of albumin flux during the interval by the luminal tracer concentration. The clearance volume of albumin at each time point (Valb, t) was calculated by summing the incremental clearance volumes up to that time pointValb,t=i=1tVAt×Δ[A]t[L]t wherei is the numbered time point in the equation, VAt is the volume in the luminal chamber at each time point, Δ[A]t is the increase in tracer concentration between time points, and [L]t is the tracer concentration in the abluminal chamber at each time point. This approach accounts for changes in the chamber volume due to sampling because the chamber volume is constant between sampling points. The change in albumin volume over time (dValb/dt), which is equal to the clearance expressed in microliters per minute, was determined by weighted least-squares nonlinear regression for the experimental 5-min time intervals from 0 to 60 min (12).

BSA was iodinated with Na125I with the chloramine T method as previously described (7). Five millicuries of 125I were combined with 100 mg of albumin. The 125I-labeled albumin was maintained in dialysis against phosphate-buffered saline (PBS; pH 7.4) until used. Free-to-bound125I was determined by trichloroacetic acid (TCA) precipitation before and after each study.

Assessment of electrical resistance.Changes in electrical impedance were assessed with the electric cell-substrate impedance sensor (ECIS) from Applied Biophysics. This system was designed to study the dynamic behavior of cells in culture and is explained in detail by Giaever and Keese (18, 19). The measurement of changes in endothelial cell shape with ECIS has been previously described in detail (43). Briefly, endothelial cells were plated (8 × 104cells/cm2) in a well containing a small gold electrode (10−3cm2) and a larger gold counter electrode. Culture medium was used as the electrolyte. The small size of the active electrode is the critical feature of the system. When electrodes of 10−3cm2 or smaller are used, the major impedance at 4,000 Hz is found at the electrode-electrolyte interface, allowing the morphology of cells seeded at this interface to be assessed. Larger electrodes do not detect changes in cell shape because the resistances in the solution predominate over the impedance of the electrode and mask changes produced by the cells. The small electrode and the larger counter electrode were connected to a phase-sensitive lock-in amplifier. A 1-V, 4,000-Hz AC signal was supplied through a 1-MΩ resistor to approximate a constant current of 1 μA. This has the advantage of having the measured in-phase voltage proportional to the resistance and the out-of-phase voltage proportional to the reactance. The resistance is used to assess endothelial cell shape.

Dual-label immunofluorescence microscopy. CPAECs were seeded and grown to confluence on 22-mm2 glass coverslips (Corning). The cells were incubated in the presence and absence of TGF-β1 for the designated times as stated inresults. After incubation, the medium was removed, and the coverslips were washed twice with PBS containing Ca2+ and Mg2+, pH 7.4 (GIBCO BRL), on ice. The cells were fixed for 10 min on ice with cold 3% formaldehyde (Sigma), rinsed with cold PBS, and permeabilized for 10 min on ice with cold 1% Nonidet P-40 (Sigma). The coverslips were washed three times with cold PBS and subsequently blocked for 1 h at 37°C with 1% BSA (in PBS). Primary antibodies (dilutions of 1:50 to 1:250, 200 μl/coverslip) were applied and allowed to incubate overnight at 4°C. The coverslips were washed three times with PBS containing 0.2% BSA and incubated at 37°C for 1.5 h with both fluorescein isothiocyanate (FITC)- and tetramethylrhodamine isothiocyanate-conjugated antibodies (dilutions of 1:40 to 1:100). The coverslips were washed three times with PBS and mounted onto glass slides (Fisher) with 2% n-propyl gallate. The slides were viewed on a Nikon Microphot SA microscope, and pictures were taken with a Nikon microflex UFX-DX camera on Kodak T-Max 400 film.

Different combinations of primary and secondary antibodies were used to determine the level of colocalization between β-catenin, plakoglobin, α-catenin, cadherin-5, and actin after treatment of the endothelial cells with TGF-β1. Primary antibodies produced in different species were used, allowing the conjugated secondary antibodies to be specific for each ligand. The following antibodies or agents were used: β-catenin mouse monoclonal antibody (clone 14; Transduction Laboratories), β-catenin polyclonal antibody (Sigma Immunochemicals), γ-catenin (plakoglobin) monoclonal antibody (Transduction Laboratories), α-catenin polyclonal antibody (Sigma Immunochemicals), cadherin-5 monoclonal antibody [clone 9H7 (22)], tetramethylrhodamine isothiocyanate-conjugated goat anti-mouse IgG (Jackson Immunoresearch Laboratories), FITC-conjugated goat anti-rabbit IgG (Sigma Immunochemicals), and FITC-conjugated phalloidin (Sigma).

Assessment of the actin-associated (insoluble) and cytoplasmic (soluble) pools of myosin. CPAECs were seeded and grown to confluence on 35-mm-diameter tissue culture dishes. The cells were incubated in the presence and absence of TGF-β1 (1 ng/ml) for 3 h. After incubation, the cells were washed once with PBS (with Ca2+ and Mg2+). Soluble buffer (60 mM PIPES, 25 mM HEPES, 10 mM MgCl2, 5 mM EGTA, 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and 10 μg/ml of aprotinin, pepstatin, and leupeptin) was then added to the cells and allowed to incubate on ice for 5 min. The soluble buffer was aspirated from the dish and spun at 10,000g for 10 min at 4°C. The supernatant was collected, diluted 1:2 with lysis buffer [0.002% bromphenol blue, 0.5 M Tris ⋅ HCl, 4% sodium dodecyl sulfate (SDS), 20% glycerol, and 0.1 mM sodium orthovanadate] and boiled for 2–5 min. After removal of the soluble buffer, the insoluble buffer (60 mM PIPES, 25 mM HEPES, 10 mM MgCl2, 5 mM EGTA, 1% SDS, 0.5% deoxycholate, 1 mM phenylmethylsulfonyl fluoride, and 10 μg/ml of aprotinin, pepstatin, and leupeptin) was added to the cells and allowed to incubate on ice for 5 min. The cells were scraped with a rubber policeman, diluted 1:2 with lysis buffer, and boiled for 2–5 min. SDS-PAGE was performed with a 7.5% acrylamide running gel and 3% acrylamide stacking gel (Mini-PROTEAN II, Bio-Rad). Proteins from the gel were transferred to nitrocellulose paper (Bio-Rad) for 14 h at 35 V. The transfer buffer was composed of 25 mM Tris base, 192 mM glycine, and 20% methanol. Nitrocellulose was blocked with 3% BSA and 1% Tween 20 and subsequently immunoblotted with nonmuscle myosin IgG (dilution 1:2,000; Biomedical Technologies). After incubation with horseradish peroxidase-conjugated anti-rabbit IgG (dilution 1:20,000; Jackson Immunoresearch Laboratories), enhanced chemiluminescence (LumiGLO, Kirkegaard & Perry Laboratories) was used for protein detection.

MLC phosphorylation assay. Endothelial monolayers were analyzed for MLC phosphorylation by urea PAGE with a modification of the method by Garcia et al. (15). CPAECs were seeded and grown to confluence on 60-mm-diameter tissue culture dishes. The cells were incubated with either control medium for 6 h or TGF-β1 (1.0 ng/ml) for 0.5, 1, 2, 4, or 6 h. After incubation, 10% TCA-0.01 M dithiothreitol was added to the dish. The cells were scraped with a rubber policeman, transferred to a centrifuge tube, and placed on ice for 30 min. The tubes were spun at 10,000g for 7 s at 4°C, the supernatant was removed, and ether was added to remove any remaining TCA from the protein pellet. This process was repeated four times. Samples were then dried, combined with sample buffer (6.7 M urea, 20 mM Tris base, 22 mM glycine, 9 mM dithiothreitol, 0.004% bromphenol blue, and 1% Triton X-100), and sonicated for 30 min. Proteins were separated by charge on a native gel (10% acrylamide, 0.5% bis-acrylamide, 40% glycerol, 20 mM Tris base, 22 mM glycine, and 0.22 μM ammonium persulfate) for 1 h at 400 V and transferred to nitrocellulose for 14 h at 40 V. MLC was detected by immunoblotting with MLC monoclonal antibody (dilution of 1:2,000; clone 21; Sigma). After incubation with horseradish peroxidase-conjugated anti-mouse IgM (dilution of 1:50,000; Jackson Immunoresearch Laboratories), enhanced chemiluminescence as described in Assessment of the actin-associated (insoluble) and cytoplasmic (soluble) pools of myosin was used for protein detection.

Statistical analysis. One-way analysis of variance was used to determine significant changes in albumin clearance, MLC phosphorylation, and shifts in myosin between soluble and insoluble protein pools after treatment with TGF-β1. Newman-Keuls post hoc analysis was used to determine significant differences between groups, with the level of significance set atP < 0.05.

RESULTS

TGF-β1-induced increase in endothelial monolayer permeability. Experiments were first performed to characterize the change in permeability after the addition of TGF-β1 to CPAECs. Figure 1 shows the dose response of the TGF-β1-induced increase in albumin clearance. Active TGF-β1 was added at concentrations of 0.001, 0.01, 0.1, 1.0, and 10.0 ng/ml to the luminal chamber of the monolayer system and incubated for 18 h. At the end of 18 h, albumin clearance was determined as outlined inmethods. The addition of 1.0 and 10.0 ng active TGF-β1/ml to the endothelial monolayer increased permeability threefold. Doses of 0.1 ng active TGF-β1/ml and lower showed no change in monolayer permeability at 18 h. Because 1.0 ng/ml of TGF-β1 produced a maximal response, this dose was used to determine the time course of the change in albumin clearance (Fig.2). Time-course experiments showed that TGF-β1 produces a significant increase in permeability within 3 h after its addition to endothelial monolayers (P < 0.05). The maximal threefold increase in albumin clearance was achieved at 6 h because there was no significant difference in albumin clearance determined at 6 or 18 h after the addition of TGF-β1. To demonstrate the specificity of this response, we coincubated monolayers with TGF-β1 and antisera to TGF-β1. As seen in Fig. 3, the addition of TGF-β1 antisera or preimmune chicken serum alone did not alter endothelial monolayer permeability. In contrast, the addition of TGF-β1 antisera in conjunction with TGF-β1 inhibited the typical threefold increase in albumin clearance.

Fig. 1.

Dose response of transforming growth factor (TGF)-β1-induced increase in endothelial monolayer albumin clearance. Endothelial monolayers were incubated for 18 h with 0.001, 0.01, 0.1, 1.0, and 10.0 ng TGF-β1/ml before determination of125I-labeled albumin clearance. Values are means ± SE from 3 experiments; totaln = 12 monolayers/group. Treatment with 1.0 and 10.0 ng TGF-β1/ml resulted in a significant increase in albumin clearance compared with control group (medium alone; * P < 0.05).

Fig. 2.

Time course of TGF-β1-induced increase in endothelial monolayer albumin clearance. Endothelial monolayers were incubated for 3, 6, and 18 h with 1.0 ng TGF-β1/ml before determination of125I-labeled albumin clearance. Albumin clearance was determined over a 1-h period after preincubation time. Values are means ± SE from 3 experiments; totaln = 6 monolayers/group. Albumin clearance in TGF-β1-treated monolayers was significantly greater than that in control groups at all time points (* P < 0.05). Maximum increase in permeability was generated after a 6-h incubation period.

Fig. 3.

Effects of antiserum to TGF-β1 (anti-TGF-β1) on TGF-β1-induced increase in endothelial monolayer albumin clearance. Endothelial monolayers were preincubated for 6 h with either medium, 1.0 ng/ml of anti-TGF-β1, preimmune (normal) serum (NS), or 1.0 ng/ml of TGF-β1 or were coincubated with anti-TGF-β1+TGF-β1 or NS+TGF-β1. Values are means ± SE from 3 experiments;n = 12 monolayers. Treatments with TGF-β1 alone or NS+TGF-β1 resulted in a significant increase in permeability (* P < 0.05). Anti-TGF-β1 prevented TGF-β1-induced increase in endothelial monolayer albumin clearance.

ECIS was used next to determine the changes in endothelial cell permeability by measuring changes in electrical resistance. This technique has the benefit of assessing changes in endothelial shape at 1-min intervals, giving near to real time assessment for the changes in permeability. Figure4 B shows tracings of the first 6 h from two typical experiments where monolayers were treated with TGF-β1 (1.0 ng/ml). The dotted and solid tracings show that electrical resistance started to drop 1 and 2 h, respectively, after the addition of TGF-β1, demonstrating the initial variability in endothelial monolayer integrity. Dose-response experiments were performed, and the data were plotted over 20 h at 30-min intervals (Fig. 4 A). Dose-response experiments (Fig. 4 A) revealed that concentrations of TGF-β1 ranging from 0.05 to 1.0 ng/ml produce similar decreases in electrical resistance over the first 4–5 h, beginning between 1 and 2 h after the addition of TGF-β1. Higher doses (0.5 and 1.0 ng/ml) continued to decrease electrical resistance, with maximum levels being reached at 8–9 h, whereas the electrical resistance of the lower doses (0.05 and 0.01 ng/ml) began to return to basal levels during the next 10 h of the experimental period. Comparison of ECIS and albumin clearance data revealed that the time course of changes in protein permeability and electrical resistance is consistent. The addition of TGF-β1 did not affect cell viability because the cells were able to exclude trypan blue and tested negative for apoptosis after 24 h of treatment with 1.0 ng TGF-β1/ml (data not shown).

Fig. 4.

Measurement of electrical resistance across confluent endothelial monolayers after treatment with TGF-β1.A: confluent calf pulmonary arterial endothelial cells (CPAECs) were treated with either control medium or 0.05, 0.1, 0.5, or 1.0 ng TGF-β1/ml, and electrical resistance was measured over a 20-h period. All doses induced similar decreases in electrical resistance over the 1st several hours. Lower doses began to return to control level 6–7 h posttreatment, whereas resistance measurements for monolayers exposed to higher doses of TGF-β1 (0.5 and 1.0 ng/ml) continued to decrease until 8–10 h. Values are means ± SE at 30-min intervals (minimumn = 10 monolayers/treatment). B: tracings of 2 representative electric cell-substrate impedance sensor experiments with confluent CPAECs being exposed to TGF-β1 (1.0 ng/ml). Electrical resistance was assessed at 1-min intervals for 6 h. Note how resistance begins to decrease between 1 and 2 h posttreatment with TGF-β1.

TGF-β1-induced rearrangement of adherens junctions. The changes in both the actin cytoskeleton and β-catenin were examined by immunofluorescence microscopy during treatment with 1 ng/ml of TGF-β1. Shown in Fig.5 is staining for F-actin (A,C, E, and G) and β-catenin (B,D, F, and H) for control cells (A andB) and after 2 (C andD), 4 (E andF), and 8 (G andH) h of incubation with TGF-β1. Figure 5 C shows that there was a loss of the actin peripheral band and an increase in actin stress fibers 2 h after the addition of TGF-β1. Although some cell-cell separation started at 2 h, the majority of cell-cell contacts was still intact as shown by the β-catenin staining at this time (Fig.5 D). At 4 h, the cells were separated, and strands were formed that linked the endothelial cells to one another. Interestingly, β-catenin was still located at the cell periphery between strands when the cells were separated at 4 h (Fig.5 F, arrow); however, by 8 h post-TGF-β1, β-catenin was no longer present at the cell periphery (Fig. 5 H). These results suggest that the actin cytoskeleton reorganizes before disruption of the cell-cell junction and that disassembly of the adherens junction as demonstrated by the loss of β-catenin occurs after the cells have separated.

Fig. 5.

Time course of actin and β-catenin rearrangement in endothelium treated with TGF-β1. CPAECs were treated with vehicle alone for 8 h (A andB) or 1.0 ng TGF-β1/ml for 2 (C andD), 4 (E andF) or 8 (G andH) h, and cells were dual labeled for actin (A,C, E, and G) and β-catenin (B,D, F, and H). At 2 h, there was a decrease in peripheral bands of actin and an increase in actin stress fibers. A 4-h incubation of TGF-β1 revealed cell separation, formation of strands between cells, and an increase in actin stress fibers. β-Catenin can still be found at cell periphery in areas of cell separation (F, arrow). Loss of peripheral β-catenin appears to coincide with complete cell separation in cells treated with TGF-β1 for 8 h. Bar, 10 μm.

Previous studies (25, 38) have suggested that the loss of cell-cell junctions containing plakoglobin results in a decrease in junction strength. We used dual-label immunofluorescence microscopy to determine whether plakoglobin was lost from the cell-cell junction before or during TGF-β1-induced cell separation. We also colocalized cadherin-5 and α-catenin with β-catenin because these are also important in the formation of adherens junctions. CPAECs were stained for β-catenin in conjunction with staining for actin, plakoglobin, α-catenin, and cadherin-5 after treatment with TGF-β1 (1.0 ng/ml) for 2–3 h (“initial” stage of cell separation). Figure6, A,C, E, and G, shows that β-catenin remained at the cell border before cell separation, but β-catenin was also found in projections that were perpendicular to the cell periphery. Whether these structures are filipodia, as suggested by Esser et al. (14) for vascular endothelial growth factor (VEGF)-treated endothelial cells, or are fingerlike projections, as shown by Baluk et al. (6), could not be determined from these micrographs. The colocalization of β-catenin and actin during the initial stage of cell separation (Fig.6, A andB) suggests that these fingerlike projections may be precursors to the perpendicular bands containing β-catenin at the cell border (Fig.7). Plakoglobin, α-catenin, and cadherin-5 are also present at the cell border before cell separation and are colocalized with β-catenin (Fig. 6,D, F, and H, respectively). Plakoglobin, α-catenin, β-catenin, and cadherin-5 were all found to colocalize at the cell-cell junction in control cells (data not shown). Single staining of each of these proteins also produced staining patterns similar to what is shown in Figs. 6 and 7, indicating that colocalization is not due to background fluorescence from the different probes (data not shown). Together, the data indicate that β-catenin, α-catenin, plakoglobin, and cadherin-5 appear to realign with reorganizing actin polymers into fingerlike projections that are perpendicular to the cell border.

Fig. 6.

Colocalization of actin, plakoglobin, α-catenin, and cadherin-5 with β-catenin during initial stages of TGF-β1-induced endothelial cell separation. CPAECs were treated with 1.0 ng TGF-β1/ml for 2–3 h and then dual labeled for β-catenin (A,C, E, and G) and actin (B), plakoglobin (D), α-catenin (F), or cadherin-5 (H). β-Catenin was present in bands that are perpendicular to cell periphery and at points of cell-cell adhesion. Plakoglobin, α-catenin, and cadherin-5 colocalized with β-catenin in a similar perpendicular band “pattern” at cell border during initial stages of TGF-β1-induced cell separation. Bar, 10 μm.

Fig. 7.

Colocalization of actin, plakoglobin, α-catenin, and cadherin-5 with β-catenin during transitional stages of TGF-β1-induced endothelial cell separation. CPAECs were treated with 1.0 ng TGF-β1/ml for 3–4 h and then dual labeled for β-catenin (A,C, E, and G) and actin (B), plakoglobin (D), α-catenin (F), or cadherin-5 (H). β-Catenin, plakoglobin, α-catenin, and cadherin-5 remained at cell border after cells underwent separation (C,D, F, and H, arrows, respectively). Bar, 10 μm.

We next assessed the colocalization of α-catenin, β-catenin, plakoglobin, cadherin-5, and actin immediately after the cells began to separate. CPAECs were stained for β-catenin in conjunction with staining for actin, plakoglobin, α-catenin, and cadherin-5 after treatment with TGF-β1 (1.0 ng/ml) for 3–4 h (“transitional” stage of cell separation). Figure 7 shows that all four adherens junction proteins, as well as actin, were colocalized in the strands that connected the cells to one another. Interestingly, β-catenin remained colocalized with plakoglobin, α-catenin, and cadherin-5 at the cell periphery between strands in cells that were separated (Fig. 7,CH), suggesting that the adherens junction had yet to disassemble. Colocalization of actin with β-catenin was not consistent because areas showing β-catenin staining did not always show actin stress fibers. The realignment of adherens junction proteins into perpendicular bands at the cell border may be a precursor to the strand formation observed later in the TGF-β1 time course (Fig. 7).

TGF-β1-induced increase in MLC phosphorylation. Previous studies (15, 20) have demonstrated that activation of an MLCK-dependent signaling pathway contributes to thrombin-induced endothelial cell separation, suggesting that cell separation is caused by an increase in cell contraction. In addition, Goeckeler and Wysolmerski (20) have shown that after MLC phosphorylation, myosin interacts with reorganizing actin as demonstrated by myosin shifting from a detergent-soluble protein pool to an insoluble (actin-associated) protein pool during cell separation. We were interested in whether this observed MLC phosphorylation and shift in myosin to the actin-associated pool occurs in endothelial cells after treatment with TGF-β1. Figure8 shows that TGF-β1 induces MLC phosphorylation beginning at 2 h (23.0 ± 1.5% monophosphate and 8.2 ± 3.6% diphosphate) compared with that in control treated cells (15.0 ± 3.0% monophosphate and 0.0 ± 0% diphosphate). Maximum levels of MLC phosphorylation were observed at 6 h posttreatment with TGF-β1 (21.9 ± 2.7% monophosphate and 32.8 ± 9.0% diphosphate). The elevation in MLC phosphorylation led to an increased interaction of myosin with actin as demonstrated by the significant shift of myosin to the actin-associated pool (vehicle: 40 ± 5% soluble and 60 ± 5% insoluble; TGF-β1: 14 ± 3% soluble and 86 ± 3% insoluble; P< 0.05; Fig. 9). Together, these results indicate that TGF-β1 increases the phosphorylation of MLC and causes an increased interaction between myosin and actin. Thus TGF-β1-induced endothelial cell separation may be caused by an increase in cell contraction.

Fig. 8.

TGF-β1 induces an increase in myosin light chain phosphorylation in endothelial monolayers. Confluent CPAECs were treated with 1.0 ng TGF-β1/ml for 0.5, 1, 2, 4, and 6 h. After incubation, proteins were precipitated and collected with 10% TCA. Proteins were processed as described in methods, separated by charge with native gels, transblotted to nitrocellulose, and subsequently detected via Western blotting with myosin light chain antibody. A: representative immunoblot of myosin light chain with unphosphorylated (UN), monophosphorylated (MONO), and diphosphorylated (DI) forms.B: quantitation of 3 separate experiments. TGF-β1 induced a significant increase in myosin light chain phoshorylation (P < 0.05).

Fig. 9.

Myosin shifts from soluble (non-actin-associated) protein pool to insoluble (actin-associated) protein pool during TGF-β1-induced endothelial cell separation (3 h post-TGF-β1). Confluent CPAECs were treated with 1.0 ng TGF-β1/ml for 3 h. After incubation, proteins were extracted with soluble (SOL) or insoluble (INS) buffers (seemethods).A: representative immunoblot of SOL and INS pools of myosin from control (CTRL) and TGF-β1-treated monolayers. B: quantitation of 3 separate experiments (n = 3 monolayers). TGF-β1 induced a redistribution of myosin toward INS protein pool by 3 h, suggesting that there is an increased interaction between myosin and actin. * Significant difference from CTRL value, P < 0.05.

To determine whether activation of the MLCK-dependent pathway plays a role in TGF-β1-induced decreases in endothelial integrity, we added the MLCK inhibitor KT-5926 to endothelial monolayers and assessed the changes in integrity with ECIS (30, 33). Figure10 shows that coincubation of KT-5926 (1.0 μM) with TGF-β1 (1.0 ng/ml) prevents the decrease in barrier function caused by TGF-β1 alone. Dual-label immunofluorescence analysis revealed that endothelial cell monolayers treated with both KT-5926 and TGF-β1 (for either 3 or 8 h) do not form the intracellular gaps that are observed in monolayers treated with TGF-β1 alone (compare Fig. 11 with Fig.5). Interestingly, the actin stress-fiber formation and the fingerlike projections that are observed in TGF-β1-treated cells were also seen in cells that were cotreated with the MLCK inhibitor KT-5926 (Fig.11 J, arrows). Together, the results indicate that KT-5926 inhibits the formation of intracellular gaps by TGF-β1, thus preventing the loss of endothelial barrier integrity that is induced by TGF-β1 alone.

Fig. 10.

Measurement of electrical resistance across confluent endothelial monolayers after treatment with TGF-β1 and KT-5926. Confluent CPAECs were treated with either 1.0 ng TGF-β1/ml, 1 μM KT-5926, or both, and electrical resistance was measured over an 18-h period. TGF-β1 induced a decrease in electrical resistance similar to that observed in Fig. 4. KT-5926 initially increased electrical resistance across endothelial monolayers, but resistance values returned to original levels within 7–8 h. Coincubation of KT-5926 with TGF-β1 prevented the change in electrical resistance observed with TGF-β1 alone. DMSO at a dilution of 1:1,000 served as a vehicle control for KT-5926 and did not induce any change in electrical resistance (data not shown). Values are means ± SE at 30-min intervals; minimumn = 3 monolayers/treatment.

Fig. 11.

KT-5926 prevents TGF-β1-induced intercelluar gap formation. Confluent CPAECs were treated with either 1 μM KT-5926 alone (C,D, G, and H) or 1 μM KT-5926 plus 1.0 ng TGF-β1/ml for either 3 (CF) or 8 h (GJ) and then dual labeled for actin (A,C, E,G, andI) or β-catenin (B,D, F,H, andJ). Cells were exposed to DMSO (dilution of 1:1,000) for 8 h as a control (A andB). KT-5926 prevented formation of intercellular gaps that were induced with TGF-β1 alone (compare with Figs. 5 and 7). β-Catenin remained at cell border either as a continuous band or in bands that are perpendicular to cell border (J, arrows). KT-5926 did not prevent formation of actin stress fibers induced by TGF-β1 alone (E andI). Bar, 10 μm.

DISCUSSION

Previous studies (11, 28, 42) have demonstrated that TGF-β1 causes changes in the endothelial cell phenotype that would result in a decrease in endothelial monolayer integrity. We have extended these findings by showing that TGF-β1 causes a dose-dependent decrease in monolayer integrity that begins 1–2 h after treatment. We have further demonstrated that the loss of endothelial integrity induced by TGF-β1 is not due to a disassembly of the adherens junction but is dependent on MLC phosphorylation, suggesting that contraction plays an important role in this response. Interestingly, we also found that inhibition of MLC phosphorylation did not prevent the morphological changes in adherens junction structure or actin stress-fiber formation induced by TGF-β1 alone, even though gap formation and increased permeability were inhibited.

Our results agree with those of Coomber (11), who, using monolayers of CPAECs, found that endothelial cells lose their cobblestone morphology and assume a pleiomorphic shape after an 18-h incubation with TGF-β1. The data presented here are also similar to those of Sutton et al. (42), who showed that when TGF-β1 was added to newly confluent cultures of bovine aortic endothelial cells for 5 days, a number of cells in the monolayer detached, whereas the remaining adherent cells pulled away from one another and became “enlarged and ragged.” These investigators found that the response of these cells to TGF-β1 was dependent on the proliferative state and the degree of confluence, with quiescent endothelial cells not showing any response to TGF-β1.

Our data extend these findings by measuring the changes in albumin clearance and electrical resistance to assess decreases and/or increases in monolayer integrity. The use of ECIS allowed the changes in electrical resistance to be measured in small intervals (1 min) over a 20-h period. This method allows permeability to be assessed for long experimental periods and in real time; thus the start of a response and the transient nature of the response can be accurately determined. As shown in Fig. 4, electrical resistance began to change between 1 and 2 h after the addition of TGF-β1. Resistance continued to decrease over the next 6 h, reaching a maximum decrease at 8–9 h post-TGF-β1 treatment. The duration of the decrease was dependent on the dose of TGF-β1 applied to the monolayer, with doses > 0.5 ng/ml maintaining the decrease in electrical resistance, whereas resistance returned toward baseline in monolayers treated with doses < 0.5 ng/ml. The reversibility of the TGF-β1-induced increase in permeability at the lower doses, as demonstrated by ECIS, may be explained by the internalization of the TGF-β1 receptor. Anders and colleagues (3, 4) have demonstrated that on activation of the TGF-β1 receptor, the receptor-ligand complex is internalized, with downregulation of heteromeric TGF-β1-receptor activity. These investigators also found that surface binding recovered from downregulation in 6–8 h. This would suggest that the lower doses (0.1 ng/ml and below) of TGF-β1 are metabolized during the experimental period shown in Fig. 4, allowing resistance to return to control levels. However, the higher doses are not depleted during this time frame; thus the increase in permeability is maintained for the entire experiment.

Previous investigators (24, 25, 38) have suggested that disassembly of the adherens junction, i.e., a loss of binding between cadherins and catenins, with redistribution of the catenins into the cytoplasmic pool, contributes to the formation of gaps between cells after the addition of cytokines or other inflammatory mediators. This can be demonstrated morphologically by a disappearance or loss of immunofluorescence localization of junctional components at the cell periphery. An example of this would be the recent work by Kevil et al. (24) that showed that cadherin-5 and occludin were lost from the cell border at areas of gap formation after the addition of VEGF to endothelial monolayers. Although our data show that adherens junction proteins are lost from the cell periphery after the cells have completely separated (Fig. 5 H), these junctional proteins were still present at the cell periphery when the cells initially separated. This is demonstrated in Fig. 7 where β-catenin, plakoglobin, α-catenin, and cadherin-5 are all found at the cell periphery (arrows) as well as in strands connecting separated cells. Indeed, the presence of adherens junction proteins at the cell periphery immediately after cell separation suggests that the loss of cell-cell contact is not the result of a disappearance of junctional proteins from the cell-cell contact area. This does not rule out the possibility that homotypic binding of the cadherins was decreased even though the junctional components are still located at the cell periphery. Recent studies have suggested that the adhesion strength of cadherin-mediated junctions may be altered by changes in the phosphorylation state of the junctional proteins (14), a change in the lateral clustering of cadherins (45), or a loss of certain adherens junction proteins from the cadherin-catenin complex (25, 38), all of which may occur independently of a disappearance of all junctional components from the cell-cell contact area. Esser et al. (14) recently showed that VEGF causes an increase in the tyrosine phosphorylation of cadherin, β-catenin, plakoglobin, and p120 that is associated with a decrease in endothelial barrier function. Similar to our findings, Esser et al. also found that the cadherin-catenin complex remained intact after VEGF treatment and that the adherens junction components were still localized at the cell periphery when monolayer integrity was decreased. These findings suggest that phosphorylation of adherens junction proteins plays a role in governing the integrity of endothelial monolayers without causing disassembly of the adherens junction. Whether TGF-β1 alters phosphorylation levels of adherens junction proteins, thereby decreasing cadherin affinity and strength of the cell-cell junction, remains to be determined.

β-Catenin and plakoglobin have been shown to form distinct complexes with the cytoplasmic domain of E-cadherin in epithelial cells (1, 10,23). Distinct cadherin-catenin complexes are also believed to occur in endothelial cells. A recent study by Schnittler et al. (38) as well as the study by Lampugnani et al. (25) suggested that a loss of plakoglobin-containing adherens junctions will weaken endothelial cell-cell contacts. We used immunofluorescence microscopy to determine whether plakoglobin was lost from cell-cell junctions before other adherens junction proteins after TGF-β1 treatment. As seen in Fig. 6, both plakoglobin and β-catenin remain colocalized at the cell border before cell separation in TGF-β1-treated monolayers. Immediately after cell-cell separation (Fig. 7), plakoglobin and β-catenin were also found at the cell periphery between strands as well as in strands connecting separating cells. This would suggest that plakoglobin does not preferentially leave the cell junction before β-catenin during TGF-β1-induced endothelial cell separation.

The colocalization of β-catenin, plakoglobin, α-catenin, and cadherin-5 at the cell periphery immediately after cell-cell separation indicates that disassembly of the adherens junction does not occur before cell separation. Alternatively, TGF-β1-induced cell separation could be caused by an increase in tension due to cell contraction. The appearance of strands containing actin and adherens junction proteins between separated cells (Fig. 7) has also been observed during thrombin-induced decreases in endothelial monolayer integrity (37). Thrombin has been shown to decrease monolayer integrity by increasing endothelial cell contraction via a signaling cascade similar to that observed in smooth muscle cells (15, 20). Thrombin stimulation increases the activity of MLCK, which subsequently phosphorylates MLC, thus allowing myosin to interact with actin stress fibers within the cell. Using a native gel system to separate MLC based on the level of phosphorylation, we found that TGF-β1 treatment increased the phosphorylation of MLC. As seen in Fig. 8, this increase was temporally associated with the decrease in monolayer integrity shown in Figs. 2and 4. In association with the increase in MLC phosphorylation, there was also an increase in the amount of myosin associated with actin (Fig. 9), suggesting that TGF-β1 causes endothelial cell separation by increasing MLC phosphorylation, which leads to cell contraction.

To verify that MLCK plays a role in TGF-β1-induced endothelial cell separation, we applied KT-5926, a specific inhibitor of MLCK (33), to both control and TGF-β1-treated endothelial monolayers. Dual-label immunofluorescence microscopy revealed that KT-5926, when coincubated with TGF-β1, prevented the formation of intercellular gaps (Fig. 11). Assessment of the changes in monolayer integrity with ECIS showed that KT-5926 also prevented the decrease in electrical resistance (Fig. 10) typically observed in monolayers treated with TGF-β1 alone. The dose of KT-5926 used in these studies (1 μM) is lower than that used in a previous study (15) that used pretreatment with KT-5926 to prevent the change in barrier integrity and the increase in MLC phosphorylation observed after treating bovine pulmonary arterial endothelial cells with thrombin. Although KT-5926 will also inhibit protein kinase (PK) C, the inhibition constant value of KT-5926 for PKC is 40-fold higher than that for MLCK (33). Also, PKC stimulation via the addition of phorbol esters does not result in the phosphorylation of MLC in bovine pulmonary arterial endothelial cells, suggesting that PKC is not involved in this response (15). Thus the ability of KT-5926 to inhibit these changes suggests that TGF-β1 alters barrier integrity through an MLCK-dependent pathway that decreases the contractile activity of the endothelial cell. Alternatively, KT-5926 may preserve monolayer integrity by preventing a loss of homotypic cadherin-mediated adhesion induced by TGF-β1. As previously stated, there is a growing body of evidence indicating that phosphorylation of components of the cadherin-catenin complex may lead to a loss of adhesive function and the destabilization of the adherens junction (reviewed in Ref. 13). Although MLCK or a serine/threonine kinase with similar characteristics has not been identified as altering cell-cell adhesion or phosphorylating components of the cadherin-catenin complex, the data shown in Fig. 11 would allow one to hypothesize that such a mechanism may exist.

Baluk et al. (6) demonstrated that substance P caused the formation of fingerlike projections before gap formation in tracheal microvascular endothelial cells and that after gap formation strands formed between cells. We found similar results after TGF-β1 treatment. Before the formation of large gaps, actin, β-catenin, plakoglobin, α-catenin, and cadherin-5 are all colocalized in fingerlike projections that are perpendicular to the cell border (Fig. 6). These structures are also similar to the filipodia found in endothelial cells treated with VEGF (14). Whether these are precursors to the strands seen during cell separation (Fig. 7) remains to be determined. Interestingly, the formation of these projections as well as increased actin stress-fiber formation occurs even in the presence of the MLCK inhibitor KT-5926 (Fig. 11, I andJ). Two interpretations may explain this finding. The first is that the dose of KT-5926 did not completely inhibit all of the TGF-β1-induced MLC phosphorylation. Thus, in these experiments, the amount of MLC phosphorylation inhibited is enough to prevent the severe changes in morphology that result in the disruption of endothelial monolayer integrity, but the amount of MLC that is phosphorylated still permits the formation of stress fibers.

A second interpretation is that a signal transduction pathway independent of both MLC phosphorylation and increased cell tension is responsible for the generation of actin stress fibers and the formation of perpendicular strands containing proteins of the adherens junction. Rho and Rac, members of the Rho family of small GTPases, have been shown to be important in the formation of actin stress fibers (35). Recently, Rho and Rac have been implicated in the formation of cadherin-mediated junctions as well as in the stabilization of these junctions in epithelial cells (9), suggesting that these small GTPases may also be involved in the reorganization of adherens junction proteins. Although Rac and Rho have been shown to be activated after TGF-β1 exposure in other cell types (5, 31), activation of these small GTPases by TGF-β1 in endothelial cells has yet to be reported. Further experiments need to be performed to determine the mechanism of stress-fiber formation in the presence of KT-5926.

In summary, the addition of TGF-β1 to confluent endothelial monolayers induces a significant decrease in the integrity of these monolayers. The decrease in endothelial monolayer integrity occurs after the reorganization of the actin cytoskeleton but before the disassembly of the adherens junction complex as demonstrated by the presence of α-catenin, β-catenin, plakoglobin, and cadherin-5 at the cell periphery after cell separation. It is also apparent that complexes consisting of plakoglobin and cadherin-5 do not leave the cell-cell junction before those complexes consisting of β-catenin and cadherin-5 during TGF-β1-induced cell separation. TGF-β1 elevates the level of MLC phosphorylation, causing an increased interaction between myosin and reorganizing actin, both of which are associated with an increase in cell contraction. Taken together, the data indicate that TGF-β1 induces endothelial cell separation by initiating cell contraction and not through the disassembly of adherens junctions.

Acknowledgments

We thank Wendy Ward, Debbie Moran, and Maureen Davis for assistance in manuscript preparation.

Footnotes

  • Address for reprint requests and other correspondence: P. A. Vincent, Dept. of Physiology and Cell Biology (MC-134), Albany Medical College, 47 New Scotland Ave., Albany, NY 12208-3479 (E-mail:peter_vincent{at}ccgateway.amc.edu).

  • This work was supported by American Heart Association (New York Affiliate) Grant RG-148-N and National Heart, Lung, and Blood Institute Grant HL-54206.

  • V. Hurst IV is a predoctoral trainee supported by National Heart, Lung, and Blood Institute Grant T32-HL-07194.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

REFERENCES

View Abstract