acute lung injuryis a common cause of respiratory failure and is characterized by physiological dysfunction attributed to increases in vascular endothelial cell permeability and pulmonary edema formation. Vascular barrier dysfunction in this setting results from endothelial cell activation, leading to cell shape changes and the formation of paracellular gaps (41, 42). The cytoskeletal microfilament proteins actin and myosin constitute 16% of the total endothelial cell protein and exert a critical role in determining cell shape and cellular migration (angiogenesis, wound repair) and in the regulation of endothelial monolayer permeability or barrier function in multiple models (15). The molecular motor that drives nonmuscle actin cytoskeletal rearrangement is the myosin II-associated ATPase, which generates mechanical force by promoting the translational cross-bridge movement of myosin heads across the actin fibers. This process is regulated by the phosphorylation of myosin light chain (MLC) on Ser¹⁹ and Thr¹⁸ residues catalyzed by the Ca²⁺/calmodulin-dependent MLC kinase (MLCK). MLC phosphorylation triggers myosin ATPase activity and actin polymerization and is essential to smooth muscle and nonmuscle tension development. Endothelial cell MLCK (214 kDa) is significantly larger than the smooth muscle MLCK isoform (130–160 kDa), although >95% homology exists between the COOH-terminal portions of the endothelial cell and smooth muscle MLCK isoforms. The 922-amino acid NH₂ terminus of the endothelial cell MLCK, however, constitutes a completely novel sequence not shared by the smooth muscle MLCK isoform (64). Several distinct endothelial cell MLCK variants, characterized by several specific deletions of nucleotides formed by alternative splicing of a unique transcribed mRNA precursor, have been described (38, 64) and are differentially distributed in various human tissues and cells (38). Endothelial cell MLCK triggers force development after stimulation with rapid activating agonists such as thrombin (13), and this activity is intimately involved in vascular inflammatory responses. For example, inhibition of endothelial cell MLCK activity reduces neutrophil influx into tissues (17) and exerts a marked barrier protective effect in models of ischemia-reperfusion injury (33) and after treatment with vasoactive agonists (13, 55).

Tumor necrosis factor (TNF)-α is a proinflammatory cytokine produced by activated leukocytes and endothelial cells, resulting in upregulation of endothelial adhesion molecules, alterations in endothelial cell permeability, and increased edema in isolated perfused lungs (21, 27). Increased TNF-α levels are found in bronchoalveolar lavage fluid from patients with acute respiratory distress syndrome (29, 46, 62) and contribute significantly to the organ dysfunction observed in acute inflammatory syndromes. Although the mechanisms by which TNF-α increases vascular barrier dysfunction are not well understood, there is evidence to support an active role for actin rearrangement (20), potentially evolving in a MLCK-dependent manner (43). TNF-α is also a well-recognized potent stimulus for programmed cell suicide or apoptosis, an important feature of inflammatory processes including acute lung injury (24, 26, 65, 66). TNF-α induces endothelial cell apoptosis in vivo and in vitro, and direct administration of TNF-α produces widespread endothelial cell apoptosis, predominantly in the pulmonary vascular bed, as well as pulmonary edema (24, 53, 54, 56,60).

Although information is limited, there is increasing appreciation that the microfilamentous cytoskeleton may be intrinsically involved in the apoptosis process by regulating intracellular signaling or by transmitting death messages to downstream effectors. Cytoskeletal components such as actin, the actin binding protein gelsolin, and focal adhesion kinase are all cleaved by caspases (1, 19, 32, 36, 39,40, 45), and actin polymerization is required for the initiation of membrane blebbing, a recognized feature of apoptotic cell death (37). It has been suggested that MLCK may be required for membrane blebbing in specific models of apoptosis induced by serum deprivation (12, 47) and, in a single report (69), TNF-α-mediated DNA fragmentation in the tumor cell line U937 was reduced by MLCK inhibition. In this study, we examined the hypothesis that the microfilament-based cytoskeleton is a critical participant in TNF-α-induced endothelial cell permeability and apoptosis. Our results indicate that MLCK and Rho GTPases, two effectors that work in tandem to increase MLC phosphorylation, are significantly involved in TNF-α-induced apoptosis and actomyosin rearrangement but do not appear to be critical participants in TNF-α-induced endothelial cell barrier dysfunction. Furthermore, inhibition of the execution phase of apoptosis did not alter the TNF-α-mediated increase in permeability. Elucidation of cytoskeletal participation in cell survival or apoptosis signaling may provide important clues regarding the role of endothelial cell apoptosis in the pathogenesis of acute lung injury.

METHODS

Cell culture conditions and reagents.

For most experiments, the bovine pulmonary artery endothelial cells (BPAECs) were obtained either at the 16th passage (CCL 209; American Type Culture Collection, Manassas, VA) or at the 2nd passage (caspase-8 activity assay, diphospho-MLC-specific immunoblotting experiments; Cell Systems, Kirkland, WA). The cells were maintained in complete culture medium consisting of 20% bovine serum, 17 μg/ml of endothelial cell growth supplement (Upstate Biotechnology, Lake Placid, NY), and 100 U/ml of penicillin-streptomycin (GIBCO BRL) at 37°C in an atmosphere of 5% CO₂ and 95% air. Caspase inhibitor I [Z-Val-Ala-Asp (ZVAD)-fluoromethylketone (fmk)], caspase-3 inhibitor I [N-acetyl-Asp-Glu-Val-Asp-aldehyde (Ac-DEVD-CHO)], caspase-8 inhibitor II [Z-Ile-Glu(OMe)-Thr-Asp(OMe) (Z-IETD)-fmk], 1-(5-iodonapthalene-1-sulfonyl)homopiperazine hydrochloride (ML-7), and cholera toxin (type Inaba 569B) were from Calbiochem-Novabiochem (La Jolla, CA). The Rho-associated protein kinase inhibitor Y-27632 was obtained from Upstate Biotechnology. Thrombin, TNF-α (with a biological activity of 2 × 10⁷ U/mg), and monoclonal anti-human TNF-α antibody (clone TA-31) were from Sigma-Aldrich (St. Louis, MO). Texas Red-X phalloidin was purchased from Molecular Probes (Eugene, OR).

Anti-diphosphorylated MLC antibody generation.

Antisera production and peptide syntheses were performed by Sigma-Genosys (The Woodlands, TX). Phospho-specific MLC antibodies were generated by injecting rabbits with the KLH-conjugated phosphorylated peptides KKRPQRATS[p]NVFA (monophospho-specific antibodies) or KKRPQAT[p]S[p]NVFA (diphospho-specific MLC antibodies), where [p] indicates that the residue to the left has been chemically phosphorylated. These peptides encompass MLC phosphorylation sites Thr¹⁸ and Ser¹⁹. Phospho-specific antibodies were obtained by a two-step affinity purification procedure. The first step involved removing antibodies generated against the nonphosphorylated region of the peptide by passing the crude antisera through a column containing nonphosphorylated peptide (KKRPQRATSNVFA) conjugated to Sepharose. The “flow-through” from this column showed a large enrichment for specific antibodies to monophospho-MLC and diphospho-MLC. This was then affinity purified in a second step by passing it over a column containing either the mono- or diphosphorylated peptides and collecting the eluant.

Apoptosis assays: endothelial cell annexin V staining.

BPAECs were stained with annexin V and propidium iodide with the ApoAlert annexin V apoptosis kit (Clontech Laboratories, Palo Alto, CA). Detached cells were recovered by centrifugation, resuspended in the binding buffer, and added back to the tissue culture plate before being stained. Apoptotic cells were detected by fluorescence microscopy with an Eclipse TE300 inverted microscope (Nikon, Melville, NY). Apoptotic cells stained green at the plasma membrane. Necrotic or late apoptotic cells took up propidium iodide and stained red. A quantitative estimation was made by counting apoptotic cells relative to the total number of cells seen within the counted fields (with the use of bright-field microscopy). The apoptosis index reported reflects the number of annexin-stained cells divided by the total number of cells on the monolayer within the same microscopic field and is expressed relative to the number obtained in unstimulated control conditions (multiple of increase).

DNA electrophoresis.

Both DNA laddering and nucleosome ELISA methods have high specificity and recognize a more downstream apoptotic event, internucleosomal DNA cleavage. Genomic DNA was obtained from cultured bovine endothelial cells after specific interventions with the Puregene DNA isolation kit (Gentra Systems, Minneapolis, MN), following the manufacturer's instructions. Briefly, cells were lysed with a solution containing SDS and Tris-EDTA. RNase was added, and the proteins were precipitated with ammonium acetate solution. DNA was precipitated with isopropanol, resuspended in Tris-EDTA, and electrophoresed on a 1.5% agarose-ethidium bromide gel at 6 V/cm. The resulting gel was then photographed under ultraviolet luminescence with a Polaroid camera. Apoptosis was recognized by the presence of discrete bands of DNA that migrated in the range of 100–1,000 bp (laddering), whereas necrosis was characterized by the appearance of a “smear” of DNA. Commercially obtained DNA fragments of 1-kb and 100-bp size were used as markers.

Nucleosome ELISA.

This method has the advantage of being easily quantifiable. Quantitation of apoptotic endothelial cells was obtained with a nucleosome ELISA kit (Oncogene Research Products, Cambridge, MA) following the manufacturer's protocol. In these experiments, endothelial cells were lysed, and the supernatant was loaded onto precoated DNA-binding protein wells. The nucleosomes were detected with anti-histone H3-biotinylated antibody followed by streptavidin-horseradish peroxidase, with absorbance (450 nm) compared with lyophilized standards with designated nucleosome unit values.

Caspase activity assay.

Caspase-8 activity was assayed with the ApoAlert caspase-8 colorimetric assay kit from Clontech used following the manufacturer's instructions. Cells were lysed and centrifuged, and the supernatant was assayed for protease activity with the specific chromogenic substrate Ac-IETD-p-nitroanilide (pNA), with optical density measurements at 400 nm taken every 30 min for up to 180 min in a Vmax microplate reader (Molecular Devices, Sunnyvale, CA). The slopes of the curves obtained were normalized to those of blank samples (buffer and lysis buffer only). Caspase activity units were calculated by dividing the values obtained by the slope of a standard curve of absorbance of the chromogen alone (pNA).

Endothelial monolayer resistance measurements.

The electrical resistance of BPAEC monolayers was measured with the electrical cell impedance sensor technique that our laboratory (16) has previously described. In this system (Applied Biophysics, Troy, NY), endothelial cells were cultured on a small gold electrode (10⁻⁴ cm²) in complete medium. The endothelial monolayers act as insulating particles, and the total resistance across the monolayers is composed of the resistance between the ventral cell surface and the electrode and the resistance between cells. A 4,000-Hz AC signal with 1-V amplitude through a 1-MΩ resistor created an approximate constant-current source. The lock-in amplifier attached to the electrodes detected changes in both magnitude and phase of the voltage that appeared across the endothelial cell and was controlled by an IBM-compatible personal computer that was used both for data accumulation and processing. Transcellular electrical resistance (TER) increased immediately after cell attachment and achieved a steady state when endothelial cells became confluent. Thus experiments were conducted after the electrical resistance achieved a steady state. Resistance data were normalized to the initial voltage and plotted as a normalized TER. Only wells in which the TER achieved >5,000 Ω were utilized.

MLC immunoprecipitation.

For immunoprecipitation under denaturing conditions, confluent endothelial cell monolayers in 60-mm tissue culture dishes were labeled with [³²P]orthophosphate (0.5 mCi/plate) for 2.5 h in phosphate-free DMEM (Sigma) with 1% serum, followed by stimulation with either vehicle alone, TNF-α, or thrombin. The stimuli were then removed, and the monolayers were rinsed twice with 2 ml of medium, further rinsed with 2 ml of PBS, and scraped into 200 μl of SDS-denaturing stop solution (PBS, pH 7.4, 1 mM EDTA, 1 mM EGTA, 50 mM NaF, 10 mM sodium pyrophosphate, 0.2 mM orthovanadate, 1% SDS, and 14 mM β-mercaptoethanol). The homogenate was prepared by passing the cell suspension through a 16-gauge needle several times. Homogenates were heat treated at 110°C for 5 min, diluted 1:10 with 900 μl of PBS, and incubated with 50 ml of 10% Pansorbin suspension (formalin-hardened and heat-killed Cowan 1 strainStaphylococcus aureus cells; Calbiochem) for 30 min at room temperature. Samples were clarified by microcentrifugation for 5 min, and supernatants were incubated first with 20 μl of anti-MLC antibody (Biodesign International, Kennebunk, ME) for 60 min at room temperature or overnight at 4°C and then with 50 μl of 10% Pansorbin suspension for 60 min at room temperature. Immunocomplexes were pelleted by microcentrifugation for 5 min, washed three times with 1 ml of PBS, solubilized in 100 μl of boiled 2× SDS-Laemmli sample buffer, and then were separated from the Pansorbin beads by microcentrifugation and subjected to SDS electrophoresis. After electrophoresis, the proteins were transferred to nitrocellulose membranes, and ³²P signals were detected by autoradiography at −70°C. The relative intensities of the ³²P-labeled MLC were quantified by scanning densitometry.

Actin and diphosphorylated MLC immunofluorescence.

BPAECs were cultured in 12-well dishes on coverslips coated with gelatin until a confluent monolayer was achieved. After exposure to experimental conditions, endothelial cell monolayers were fixed in 3.7% formaldehyde and were permeabilized with 0.25% Triton X-100. After being stained with Texas Red-X phalloidin (1:200), the coverslips were mounted on slides and examined under oil immersion (×60 magnification) with an Eclipse TE300 inverted microscope (Nikon). Actin was visualized by Texas Red-X phalloidin staining (Molecular Probes) for 1 h at room temperature and stained red. After exposure to experimental conditions, the coverslips were fixed in 3.7% formaldehyde and permeabilized with 0.25% Triton X-100. This method enabled the examination of endothelial cell morphology (cellular rounding, shrinkage), intercellular gap formation in confluent monolayers, and intracellular actin filament reorganization (stress fiber formation, cortical or perinuclear actin organization). Staining for MLC was performed in a similar manner with a primary antibody that was immunoreactive with diphosphorylated MLC. After three washes with PBS, the monolayers were incubated with an appropriate secondary antibody conjugated to immunofluorescent dyes (Alexa 488 for green fluorescence or Alexa 546 for red fluorescence) for 1 h at room temperature. After three washes in PBS, the coverslips were mounted and analyzed with a Nikon video imaging system consisting of a phase-contrast inverted microscope connected to a digital camera connected to an image processor, and the images were recorded and saved in an Adobe Photoshop 4 program on a Pentium II PC.

Western immunoblotting.

Endothelial cell proteins were separated by SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted for 2 h with a polyclonal antibody directed at diphosphorylated MLC. This was followed by the addition of the appropriate horseradish peroxidase-conjugated secondary antibody (1:10,000). The reaction was visualized by enhanced chemiluminescence and autoradiography (Amersham) according to the manufacturer's instructions. The quantification of the immunoreactive proteins present was made with a Bio-Rad GL-670 scanning densitometer.

RESULTS

Effects of TNF-α on TER.

Our initial studies examined TNF-α-induced changes in endothelial cell permeability utilizing the electrical cell impedance sensor technique in which endothelial cells are grown to confluence on gold microelectrodes and TER is measured and expressed as normalized resistance to the initial measured voltage. TNF-α caused significant time- and dose-dependent decreases in TER beginning at 4–5 h of exposure and reaching a maximum decline at 10 h, with changes persisting for 24–48 h (Fig. 1). The dose-dependent effects of TNF-α on endothelial cell barrier function plateaued at 100 ng/ml of TNF-α and were specific to TNF-α because an anti-TNF-α monoclonal antibody completely abolished the TNF-α-induced changes in TER (Fig. 1). In contrast to the delayed TNF-α-induced decreases in TER, thrombin, a proinflammatory mediator our laboratory (13) has previously shown to cause significant endothelial cell barrier dysfunction and vascular permeability, produced a prompt and immediate decline (onset in minutes) in TER, with subsequent recovery (Fig. 1).

Effect of TNF-α on endothelial cell MLC phosphorylation and actin rearrangement.

We next examined whether the myosin-driven endothelial cell contractile apparatus is involved in TNF-α-induced endothelial cell activation using several complementary strategies to test this hypothesis. We first immunoprecipitated MLCs from ³²P-labeled endothelial cells with anti-MLC antibody at specified periods (5 min, 30 min, and 1 h) after TNF-α challenge. These studies demonstrated an almost threefold increase in MLC phosphorylation at 1 h compared with baseline values (Fig.2A). The TNF-α-induced increase in MLC phosphorylation was similar to that in the positive control, diperoxovanadate, and was sustained up to 6 h as demonstrated by diphospho-MLC-specific immunoblotting (Fig.2B). TNF-α-induced increases in MLC phosphorylation were also evaluated and confirmed by immunofluorescence microscopy with anti-diphosphorylated specific MLC antibody, where a significant increase in fluorescence at 1 (data not shown) and 4 h (Fig.2C) was demonstrated after TNF-α. Pretreatment (30 min) with the selective MLCK inhibitor ML-7 (48) inhibited TNF-α-induced MLC phosphorylation in both an in vitro MLC phosphorylation assay with purified recombinant human endothelial cell MLCK and recombinant MLC (data not shown) and in vivo (Fig.2B), suggesting that TNF-α-induced MLC phosphorylation is catalyzed, at least in part, by MLCK. Rho kinase also participates in MLC phosphorylation regulation by phosphorylation of the regulatory subunits of the myosin-associated phosphatase (8). Endothelial cell pretreatment with Y-27632, a Rho kinase inhibitor (4), prevented TNF-α-induced MLC phosphorylation (Fig.2, B and C), implicating the involvement of Rho kinase activation in TNF-α-induced increases in MLC phosphorylation.

To better understand the effects of TNF-α on the actin cytoskeleton and the contribution of MLCK activation to these changes, we next investigated the effects of TNF-α on endothelial cell actin rearrangement by analyzing the spatial distribution of polymerized F-actin by immunofluorescence microscopy. Under control conditions, the BPAECs maintained tight intercellular contacts, with a prominent dense peripheral actin-containing cortical band (Fig.3A). TNF-α produced a significant decrease in circumferential actin staining in association with a dramatic increase in stress fiber formation that spanned the length of the cell, consistent with cellular contraction in concert with intercellular gap formation (Fig. 3B). These changes began 2 h after TNF-α exposure and were essentially abolished by ML-7, a direct MLCK inhibitor (Fig. 3C), cholera toxin, a protein kinase (PK) A activator that reduces MLC phosphorylation (data not shown) (17), and Y-27632, the Rho kinase inhibitor (Fig. 3D). These results reflect the central role of MLC phosphorylation in TNF-α-mediated microfilament dynamics and indicate the direct involvement of MLCK and Rho kinase in TNF-mediated MLC phosphorylation and actin-myosin rearrangement.

Effect of MLCK and Rho kinase inhibition on TNF-α-induced endothelial cell barrier dysfunction.

We next examined whether MLCK and Rho kinase were involved in TNF-α-induced endothelial cell barrier dysfunction. Endothelial cell monolayers grown on gold microelectrodes were treated with either the MLCK inhibitor ML-7 or the Rho kinase inhibitor Y-27632 before TNF-α challenge. Despite significant evidence of MLCK activation, neither inhibition of MLCK (2–100 μM; Fig.4A) nor Rho kinase (data not shown) altered TNF-α-mediated declines in TER. These data suggest that neither MLCK nor Rho kinase activation is a critical participant in the permeability response elicited by TNF-α. Identical results were obtained with the additional MLCK inhibitor ML-9 (data not shown) as well as with cholera toxin (Fig. 4B), which, via cAMP-dependent PKA activation, reduces MLCK activity and MLC phosphorylation (13) and is a remarkably effective barrier-protective agent in multiple models of vascular permeability (50, 59). These data suggest that despite significant MLCK- and Rho kinase-dependent MLC phosphorylation and active cytoskeletal rearrangement, these events are not central to the evolution of TNF-α-induced endothelial barrier dysfunction.

TNF-α-induced apoptosis in BPAECs.

We next defined the in vitro conditions required for the well-recognized ability of TNF-α to increase endothelial cell apoptosis. Complementary assays including DNA electrophoresis, annexin V immunofluorescence microscopy, and nucleosomal ELISA were used. TNF-α (at both 20 and 100 ng/ml) induced significant increases in the number of annexin-positive and propidium iodide-negative endothelial cells beginning as early as 4 h after exposure (Fig. 5A). The higher concentration of TNF-α (100 ng/ml) also increased the number of propidium iodide-positive cells (90-fold compared with control value), reflecting either necrotic or late apoptotic cells. A concentration of 20 ng/ml of TNF-α was therefore used in subsequent experiments involving apoptosis. To confirm the early membrane-related changes associated with apoptosis, we performed DNA electrophoresis and nucleosomal ELISA assays, which detect later stages of apoptosis (i.e., internucleosomal DNA cleavage). The results showed a strong time-dependent increase in the apoptotic index, up to 96% at 20 h by ELISA (Fig. 5B), with the detection of typical DNA laddering at 12 h by DNA electrophoresis (Fig.6B). As with changes in electrical resistance, the effects of TNF-α on apoptosis were specific because the addition of a monoclonal anti-TNF-α antibody completely abolished TNF-α-induced cell death (data not shown). Predictably, pretreatment with the irreversible general caspase inhibitor ZVAD-fmk (100 μM for 40 min) also effectively inhibited TNF-α-induced endothelial cell apoptosis (Fig. 6).

Effect of MLCK and Rho kinase inhibition on TNF-α-induced apoptosis.

We next tested the hypothesis that the apoptotic process requires active MLCK-driven microfilamentous cytoskeletal rearrangement. Inhibition of MLCK, either directly by ML-7 or indirectly by cholera toxin (2 μg/ml for 1 h) (51), significantly reduced endothelial cell apoptosis triggered by TNF-α (Fig. 7). The inhibitory action of the cell-permeable ML-7 was particularly dramatic when assessed with the sensitive annexin assay at 4 h (Fig. 7), whereas cholera toxin-mediated protection was more effective at later times, perhaps reflecting the inherently slower inhibition of MLCK by cholera toxin-elicited ADP ribosylation of G_s- and subsequent cAMP-mediated PKA activation (50) (Fig. 7). To further characterize the participation of the MLC phosphorylation-dependent cytoskeletal changes in the apoptotic process, we next investigated the effect these changes may have on the activation of the caspase cascade. Inhibition of MLCK by ML-7 (10 μM) resulted in a 54% attenuation of the TNF-α-induced increase (sixfold) of caspase-8 activity (Fig. 8), an upstream component of the caspase cascade, as demonstrated by cleavage of a specific chromogenic substrate, Ac-IETD-pNA. Given the temporal similarities in both the increases in permeability and the onset of apoptosis evoked by TNF-α, we next examined the potential role of apoptosis in endothelial cell permeability by inhibiting apoptosis with ZVAD-fmk and evaluating TNF-α-induced reductions in the TER. As seen in Fig. 9, ZVAD-fmk failed to significantly alter the onset of TNF-α-mediated permeability. Similar results were obtained with Ac-DEVD-CHO, a cell-permeable, reversible general caspase inhibitor, and with Z-IETD-fmk, a more specific caspase-8 inhibitor (data not shown).

DISCUSSION

The participation of TNF-α in pulmonary pathophysiology is well recognized, with specific activation of vascular cellular components resulting in endothelial cell barrier dysfunction (3, 10, 22,49), transendothelial leukocyte diapedesis into tissues (23, 63), and marked increases in endothelial cell apoptosis (30, 52-54, 56, 58). In this report, we have addressed potential endothelial cell cytoskeleton-dependent mechanisms that may underlie the effects of TNF-α on vascular barrier dysfunction and endothelial cell apoptosis. The endothelial cytoskeleton, a key determinant of cell shape and migration (angiogenesis, wound repair), is responsible for the dynamic nature of the paracellular junction regulation that is so vital to models of endothelial permeability (15). Our data provide several lines of evidence that significant activation of the endothelial cell contractile apparatus occurs in response to TNF-α. We noted that TNF-α-induced actin rearrangement and intercellular gap formation temporally coincided with the phosphorylation of MLCs (1–2 h) and preceded the onset of both permeability and apoptosis, suggesting a linkage between the actin cytoskeletal changes and both apoptosis and the permeability responses to the cytokine. Consistent with the role of MLCK in cytoskeletal rearrangement, MLCK inhibition with ML-7 markedly attenuated TNF-α-induced MLC phosphorylation, cell contraction, and stress fiber formation. Furthermore, MLCK inhibition significantly attenuated TNF-α-mediated endothelial apoptosis. Unlike models of vascular permeability after ischemia-reperfusion injury or treatment with thrombin, MLCK inhibition did not prevent the delayed permeability response that ensues after TNF-α, reflecting the mechanistic specificity of different proinflammatory and edemagenic agents.

Our data are consistent with the evolving concept that the actomyosin-based cytoskeleton is a critical participant in TNF-α-induced endothelial cell activation, particularly with respect to programmed cell death. TNF-α produces active actin cytoskeletal rearrangement in endothelium (Fig. 3) associated with an increase in the G-actin pool (20), stress fiber formation, and intercellular gap formation involving Rho-related GTPases (3,67). The cytoskeletal rearrangement in response to TNF-α is associated with MLC phosphorylation mediated by both MLCK and Rho kinase (Fig. 2). These results are in agreement with the work of Medina et al. (43), who previously observed a 23-kDa protein comigrating with MLC standards after TNF-α stimulation (100 ng/ml). It is well recognized that Rho kinases increase MLC phosphorylation via inhibition of the regulatory subunit of the myosin-specific phosphatase, whereas Rho inhibition decreases MLC phosphorylation (18). We have shown that biochemical inhibition of Rho kinase and MLCK reduced both MLC phosphorylation and actin filament rearrangement in response to TNF-α (Figs. 2 and 3), implicating Rho kinase and MLCK activation as essential steps in the mechanism of TNF-α-mediated cytoskeletal changes. Interestingly, Rho inactivation by Clostridium botulinum C3 toxin also decreased membrane blebbing, suggesting that Rho kinase may also participate in bleb formation (47). The activation of MLCK by TNF-α appears critical for the execution of the programmed cell death induced by the cytokine because inhibition of MLC phosphorylation significantly attenuated TNF-α-induced apoptosis (Fig. 7). This effect suggests an essential role of microfilament cytoskeleton rearrangement in the assembly of the intracellular death-activating pathways. Several authors (32, 40) have detected the cleavage of actin during programmed cell death, although this remains controversial (57). Fas-induced damage of actin was associated with caspase-3-induced cleavage of the actin-severing and -capping protein gelsolin. Fragmented gelsolin overexpression induces DNA fragmentation, suggesting an effector role of gelsolin cleavage in apoptotic morphological changes (36). Destabilization of the actin cytoskeleton may contribute to the activation of key apoptotic regulators because cytochalasin D, which potently induces actin depolymerization, also triggers significant apoptosis (6,61). Overexpression of the actin monomer binding protein thymosin-β10 accelerates apoptosis in a manner consistent with the notion that cytoskeletal rearrangement may be a critical event in endothelial cell apoptosis (6, 25). The existence of a direct link between actin depolymerization and DNA degradation is not known with certainty; however, it has been suggested that DNase I may reside in the ends of actin filaments, subsequently becoming liberated by G-actin destruction (32).

The direct involvement of MLC phosphorylation in apoptosis has also been suggested by studies of membrane blebbing and DNA fragmentation in cultured cells. Huot et al. (28) have shown that the membrane blebbing during H₂O₂-induced endothelial cell (human umbilical vein) apoptosis is regulated by F-actin reorganization, which, in turn, is dependent on stress-activated protein kinase/p38 activation via phosphorylation of heat shock protein-27, an actin polymerization modulator (28). The actin changes lead to disassembly of focal adhesions and membrane blebbing. Actin-myosin interaction, regulated by MLCK, proved to be a key factor in membrane blebbing of PC-12 cells induced by serum withdrawal (47). However, the membrane blebbing and the execution phase of apoptosis may be dissociated, as previously reported (44). Furthermore, TNF-α-mediated DNA fragmentation in the tumor cell line U937 was blocked by a potent inhibitor of MLCK but was unaffected by inhibitors of cAMP- or cGMP-dependent protein kinases (69).

TNF-α fragments form active trimers, which exert systemic effects via ligation of two classes of TNF-α receptors, 55-kDa receptor type 1 (TNFR1) and 75-kDa receptor type 2 (TNFR2), both of which are found on the membranes of virtually all nucleated cells. Apoptotic TNF-α signals to endothelial cells are transmitted through TNFR1, one of the five known apoptosis surface receptors that contains an intracellular protein motif known as the death domain. Once clustered at the activated receptor, these proteins relay as yet unknown signals downstream to a procaspase (procaspase-8), with subsequent activation of the caspase cascade (caspases 3, 6, and 7) and simultaneous inhibition of the nuclear factor κB-mediated antiapoptotic pathways, resulting in execution of the apoptotic program and the typical morphological changes of cellular shrinkage and apoptotic body formation. It was recently demonstrated that apoptotic regulation in myocytes involves a protein, ARC, an apoptosis repressor with a caspase recruitment domain activated by hypoxia or reactive oxygen species, which regulates apoptosis by acting upstream of caspase activation (35). ARC overexpression rescues HEK293 cells from apoptosis caused by cotransfection of the death receptor FAS and TNFR1 as well as the death domain adaptors TNF receptor-associated death domain and caspase and receptor-interacting protein adapter with death domain, which regulate TNFR1-evoked bifurcating pathways, resulting in caspase-8 activation (7). One potential mechanism for MLCK involvement in TNF-α-induced apoptosis in endothelium may involve caspase cascade activation facilitated by actomyosin contractile rearrangement that is critical to the assembly of the TNF-α death receptor complex. Garcia et al. (14) previously cloned endothelial cell MLCK and subsequently identified five splice variants whose exact role in cytoskeletal regulation has not yet been characterized (38). Endothelial cell MLCK is highly homologous in the COOH-terminal portion of the enzyme to smooth muscle MLCK, and recently, two protein kinases having significant homology to both smooth muscle and endothelial cell MLCK isoforms were identified and exhibited discrete kinase, calmodulin-binding, and ankyrin repeat regions and a death domain. One of these kinases is the DAP kinase, which has been suggested to promote apoptosis (31). Whether a specific MLCK splice variant participates in cytokine-stimulated programmed cell death is not yet known, but the idea provides a potentially intriguing mechanism for isoform-specific MLCK regulation of endothelial cell apoptosis.

Both apoptosis and increased vascular permeability have been observed in in vitro and in vivo models of TNF-α-induced injury, but their interdependency is not known (24). The onset of permeability after TNF-α exposure (4–5 h, maximum at 10 h) is in agreement with previous studies (10, 11, 22, 49) that relied primarily on the measurement of albumin flux across the endothelial cell monolayer measured over a 6-h exposure time. We confirmed that TNF-α induces apoptosis in BPAECs as early as 4 h, which coincides with the onset of permeability; hence, we evaluated the contribution of apoptotic cell death to TNF-α-induced endothelial cell permeability by inhibiting the execution phase of apoptosis. We have demonstrated that the changes in permeability secondary to TNF-α are not the direct result of caspase-mediated apoptotic death and, in fact, were quite impressed that after an initial decline in TER, barrier function remained stable, maximizing at 10 h, despite continued and substantial increases in endothelial cell apoptosis. These results suggest the presence of highly efficient cell-cell connections that contribute to the maintained stability of the endothelial monolayer and barrier by preventing paracellular gap formation. One potential explanation for the dichotomous TNF-α response is that TNF-α-induced permeability may be completely independent of apoptosis, involving discrete signaling cascades such as TNF-α-induced release of reactive oxygen and nitrogen species (11). Nevertheless, it would appear clear that although the precise mechanisms of TNF-α-induced endothelial cell barrier dysfunctions have not yet been elucidated, the rearrangement of the actin-based cytoskeleton is not the critical determinant of the TNF-α-mediated endothelial cell barrier dysfunction as is the case, for example, for specific bioactive agonists such as thrombin. MLC phosphorylation occurred in response to TNF-α in association with cytoskeletal changes and temporally before the onset of endothelial cell barrier dysfunction (Figs. 1 and 2). However, inhibition of MLC phosphorylation by either MLCK or Rho kinase did not attenuate TNF-α-induced endothelial cell permeability, indicating that MLCK activation alone is insufficient to induce barrier dysfunction. Conversely, the mechanisms by which TNF-α induces endothelial cell barrier dysfunction may not primarily involve MLCK-regulated actin-myosin interaction but involve other cytoskeletal components such as intermediate filaments, microtubules, and adherens junction proteins, which may also play a role in TNF-α-induced vascular barrier dysfunction (2, 5, 9, 68). It is thought that cyclic nucleotide-dependent protein kinases might be important mediators of the TNF-α-induced barrier dysfunction and that protein kinase C and calmodulin do not appear to be involved in this response (34, 70). More recently, however, Ferro and colleagues (10, 11) have implicated PKC and nitric oxide in at least the initial (first 4–6 h) phases of TNF-α-induced endothelial cell permeability.

In summary, we have examined the role of the actin-myosin microfilament system in TNF-α-induced endothelial cell apoptosis and barrier dysfunction. We have shown that MLCK is critical for the execution of TNF-α-induced apoptosis signal progression in endothelial cells, which is consistent with an active role of the actin-myosin interaction in the TNF-α-induced apoptotic signaling in endothelial cells. It is possible that the actomyosin contraction is needed for the spatial changes that favor molecular interactions of proapoptotic signals. The differential role of MLCK-dependent cytoskeletal changes in TNF-α-induced endothelial cell apoptosis and permeability, as well as their apparent dichotomy, underscores the vast complexity of the regulation of the biological response to this cytokine in inflammatory injury cascades.

FOOTNOTES

• This work was supported by National Heart, Lung, and Blood Institute Grants HL-50533, HL-58064, and HL-04396.

• Address for reprint requests and other correspondence: J. G. N. Garcia, Johns Hopkins Asthma and Allergy Center, 5501 Hopkins Bayview Cir., Rm. 4B.44, Baltimore, MD 21224-6801 (E-mail:[email protected]jhu.edu).

• 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. Section 1734 solely to indicate this fact.