INTRODUCTION

Top Abstract Introduction Materials and methods Results Discussion References

DURING AN IMMUNE inflammatory response, endothelium at postcapillary venules participates in the recruitment of circulatory leukocytes through expression of adhesion molecules, chemokines, and cytokines (41). The proinflammatory cytokines tumor necrosis factor- (TNF-) and interferon- (IFN-) act synergistically in vitro and in vivo to activate endothelium, resulting in cellular responses such as altered morphology, loss of barrier function, and adhesion molecule upregulation and/or redistribution (12, 44, 46, 51). Rearrangement of the actin cytoskeleton and tight junction components to TNF- and IFN- has also been well documented in cultured endothelial cells (8, 18, 38). Cytokine-induced failure in microvascular barrier function resulting in tissue edema is evident in a variety of pathological conditions such as septic shock, adult respiratory distress syndrome, reperfusion injury, and postperfusion syndrome in cardiopulmonary bypass surgery.

Endothelial intercellular junctions have been shown to constitute the paracellular pathway of diffusive transendothelial transport (43). Morphological analysis of intercellular junctions in microvascular endothelium by freeze-fracture electron microscopy shows a segmental distribution of junctional structures (48). In mesenteric venules, endothelial intercellular junction structure shows that they are composed of loosely organized and discontinuous tight junctions. Arterioles, in contrast, show a complex combination of tight junctions, suggesting a less permeable barrier. Tight junctions, or zonula occludens, consist of a complex of proteins, including ZO-1, ZO-2, cingulin, and occludin, and are responsible for the macromolecular permeability barrier in endothelial and epithelial monolayers (37). The organization and hierarchy of assembly of the various components of the tight junction have yet to be completely elucidated.

Continuity of the endothelial barrier is also maintained by adhesive proteins of the adherens junction complex (zonula adherens). Disruption of cell-cell contact between endothelial cells by reducing the level of extracellular calcium leads to opening of tight junctions (47). Adherens junctions are closely apposed regions of adjacent membranes associated with peripheral actin filaments and, in addition to cytoskeletal proteins, are composed of a complex of cadherins, tyrosine kinases, and phosphatases (20). Cadherins are one of the major families of transmembrane cell-cell adhesion molecules and function in Ca²⁺-dependent homotypic adhesion to maintain the structure of normal tissue (53). Cadherin-5 (also called VE-cadherin) is localized to the intercellular junctions of endothelium in arterioles, capillaries, and venules (32, 52). Cadherin-5 is associated with the cytoplasmic proteins -catenin or plakoglobin, and this complex binds -catenin, which mediates binding to the actin cytoskeleton (31). The catenins function in the dynamic regulation of the cadherin complex with the actin cytoskeleton (1). Formation of the endothelial permeability barrier requires prior self-association of the extracellular domains of cadherin-5 at the anchoring junctions, in addition to binding to the cytoskeleton (22, 32). Localized in a discrete separate subdomain of adherens junctions (2) is another transmembrane adhesion molecule, PECAM-1 (CD31), which is a member of the Ig superfamily. Regulation of the balance of multiple cell-cell adhesion molecules is likely to play a role in the control of endothelial cell integrity and tight junction assembly.

Inflammatory mediators transiently increase the permeability of blood components by focal disruption of the peripheral actin microfilaments and induce gap formation between endothelium (3, 5, 36). In vitro models show a synergistic role in the regulation of PECAM-1 in endothelium by the proinflammatory cytokines TNF- and IFN- (44, 46). In the present study, we examined the immediate effects of TNF-- and IFN--mediated endothelial activation in vivo on the organization of cadherin-5 and the associated peripheral actin microfilaments. We used an in situ system, the microvascular endothelium of the rat mesentery, to make our observations. A major advantage of the mesenteric window preparation is the capability of examining changes occurring at sites precisely where the interendothelial junctional barrier has failed (3, 57).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and methods Results Discussion References

Materials. Human recombinant TNF- (Boehringer Mannheim, Indianapolis, IN), specific activity  1 × 10⁸ U/mg; rat IFN- (GIBCO BRL, Gaithersburg, MD), specific activity = 4 × 10⁶ U/mg; and endotoxin (LAL test), <10 EU/mg (TNF- and IFN-) were used.

Antibodies. The murine monoclonal antibody (MAb) to cadherin-5, 9H7, was developed as described previously (22). The cadherin-5 MAb, 9H7, is immunoreactive with cadherin-5 in human, bovine, and rat endothelium. Murine monoclonal antibody to rat CD11b/c, OX-42, was purchased from Pharmingen (San Diego, CA). Indocarbocyanine (Cy3)-conjugated rat anti-mouse IgG was purchased from Jackson ImmunoResearch Laboratories (West Grove, PA).

Surgical procedure for the isolation of the mesenteric window. The surgical procedures for the rat mesenteric preparation have been described previously (57) and are summarized here. The mesentery of male Sprague-Dawley (Harlan Sprague Dawley, Indianapolis, IN) rats (350-450 g) was used for all experiments. The rats were preanesthetized with an intramuscular injection of a cocktail containing ketamine-HCl (20 mg/100 g body wt) and acepromazine maleate (1.25 mg/100 g body wt). After the animal was sufficiently sedated, the primary anesthetic, which consisted of pentobarbital sodium, was injected into the intraperitoneal cavity (6 mg/100 g body wt). The animal was tracheotomized to facilitate artificial ventilation. The mast cell stabilizer cromolyn (10 mg/kg; Sigma, St. Louis, MO) was injected into the internal jugular vein at this point and again after an interval of 30 min (17).

An abdominal incision along the linea alba was performed using a cautery to seal cut vessels in both the skin and the fascia. Care was taken to keep the underlying intestine from coming into contact with blood, thus reducing the chance of exposure to inflammatory mediators generated during the incision. A segment of intestine was then externalized on moist gauze pads and continuously superfused with HEPES-buffered saline solution (HBS; pH 7.4, Sigma) at 37°C. A loop of suture thread was placed around the hepatic portal vein in preparation for its later incision to create an outflow vent for experimental perfusates. A well-vascularized series of mesenteric windows was selected, with care taken to examine the microvascular networks under a dissection microscope for the lack of microhemorrhages and blood stagnation.

The superior mesenteric artery (SMA) was cannulated distal to the selected series of mesenteric windows, and the appropriate bordering arteries and veins were ligated to allow perfusion only to these chosen windows. Before the windows were completely isolated from systemic blood flow and the subsequent commencement of the experiment, the windows were examined again under the dissection microscope to confirm that surgical manipulations did not initiate new microhemorrhages or the clotting off of microvessels. The SMA proximal to the windows was then clamped, and the window was flushed clear of blood with HBS containing 1 U/ml sodium heparin (Pharmacia and Upjohn, Kalamazoo, MI) and 0.5% (wt/vol) BSA (Sigma). The animal was euthanized via removal of the tracheal tube and intracardial injection (0.5 ml) of Eutha-6 CII, a pentobarbital sodium-based euthanizing solution (Western Medical Supply, Arcadia, CA). The portal vein was cut, and the isolated windows were perfused with either the heparinized HBS-BSA in control experiments or cytokine (TNF-/IFN-)-treated heparinized HBS-BSA in treated experiments (cytokine dosages reported in legends to Figs. 2-7). These solutions were allowed to incubate for 7 min with the hepatic portal outlet clamped. The incubating solution was then flushed (hepatic portal outlet clamp removed) with an identical solution, this time containing 0.05% FITC-BSA (Sigma), and allowed to incubate for three additional minutes, thus bringing the total treatment time to 10 min. After treatment, 3 ml of fixative, 4% paraformaldehyde (Electron Microscopy Sciences, Ft. Washington, PA) diluted in calcium-magnesium free (CMF)-PBS was infused and, with the hepatic portal outlet clamp replaced, allowed to incubate for 30 min. The fixative was also superfused onto the windows during this incubation period.

Pressure monitoring was performed using a pressure transducer (Gould Instrumentation Systems, Cleveland, OH) connected by a three-way stopcock to the SMA cannula. Pressure traces were generated on a Gould RS3400 chart recorder (Gould Instrumentation Systems, Cleveland, OH). Systemic arterial blood pressure measured at the superior mesenteric artery was typically 110/75. All perfusates were introduced (3 ml over ~5 s) manually using syringes. Buffer solutions were infused at pressures fluctuating around 50 mmHg (range: 0-85 mmHg). Fixatives were infused at pressures fluctuating around 75 mmHg (range: 0-130 mmHg). Lumen pressures during incubation periods were essentially 0 mmHg. Using this method of introducing the perfusates, it was not possible to accurately document any changes in vascular tone between cytokine-treated and control mesenteric windows.

For experiments in which immunohistochemistry was performed, the windows were excised after the initial 30-min fixation period, rinsed with CMF-PBS, and processed as described. To examine the distribution of the actin microfilaments, the windows were subsequently perfused with 2.5 ml of a cocktail of 4% paraformaldehyde, 0.1% Triton X-100 (Boehringer Mannheim), and rhodamine-phalloidin (10 U/ml) (Sigma) in CMF-PBS. This cocktail was incubated for 30 min before being flushed with more fixative. The mesenteric window was then carefully excised and mounted in the aqueous mounting medium Vectashield (Vecta Labs, Burlingame, CA) on a glass slide.

Immunohistochemistry. The excised mesenteric tissue was permeabilized in cytoskeletal stabilization buffer (0.5% Triton X-100, 10 mM PIPES, pH 6.8, 50 mM NaCl, 300 mM sucrose, 3 mM MgCl₂) for 30 min at room temperature. The tissue was washed once with CMF-PBS and incubated at room temperature in blocking solution (1% BSA in CMF-PBS) for 30 min. Next, the tissue was incubated with a primary antibody diluted in blocking solution for 1 h, washed 3 times for 5 min each in CMF-PBS containing 0.1% BSA, and then incubated with the Cy3-conjugated secondary antibody for 30 min. After another set of three 5-min washes, the tissue was mounted on a glass slide in Vectashield.

Laser confocal microscopy. The microvascular network of the specimens was examined with a laser scanning confocal microscope (Zeiss LSM 410; Carl Zeiss, Thornwood, NY) equipped with a krypton-argon laser and a ×40 (numerical aperture 1.3) planneofluar oil immersion objective lens. Because the FITC-BSA incubation step was followed immediately by the infusion of the fixative, the lumen of the entire microvasculature was lightly stained and thus appeared outlined. Areas of FITC-BSA leakage could be identified by the observation of intense extravascular FITC-BSA staining. In the z-axis, extravascular staining was confirmed by confocal microscopy, i.e., observing continuity of intense FITC-BSA staining when scanning planes several micrometers above and/or below the outlined lumen of the blood vessel. Having identified a leakage site on a vessel of interest, we collected a Z-series (step size 1 µm) at excitation wavelengths of 488 nm (for FITC-BSA) and 568 nm (for Cy3 secondary antibody or rhodamine phalloidin). The collection of image sets for each fluorophore was conducted separately (i.e., two automated runs over precisely the same set of planes in the z-axis, each collecting information for one fluorophore) to eliminate "bleed through," which would be apparent due to the intensity of the extravascular FITC staining. The two sets of images were then stacked in the z-axis and then merged. Metamorph Imaging System 3.0 (Universal Imaging, West Chester, PA) and Adobe PhotoShop 4.0 (Adobe Systems, Mountain View, CA) were used to manipulate and merge the stacked images.

Assessment of microvascular albumin permeability. The sites of FITC-BSA leakage were assessed on venules ranging in size from 15 to 75 µm in diameter. Venules were distinguished by their larger diameters and their positions relative to their corresponding smaller diameter arteries. With the use of a Zeiss Axiopan microscope (×20 objective, numerical aperture 0.6, fitted for epifluorescence), the number of sites and the size of extravascular FITC-BSA-stained areas were observed. Using a charge-coupled device video camera (VI-470; Optronics Engineering, Goleta, CA) mounted at the camera port of the microscope, we recorded scans of the microvascular network on a video recorder. The video recording was later converted into digital images and analyzed using National Institutes of Health (NIH) Image 1.61 (public domain; on the Internet at http://rsb.info.nih.gov/nih-image/). From several (6-13) random frames obtained of each window scanned, the length and diameter of vessels (both leaky and nonleaky) were measured as well as the number and size (using density-slicing macros) of extravascular FITC-BSA sites in that field. The data were then pooled for each experimental condition, and the average number of leaks per micrometer of venule was determined. For each experimental condition data were collected from at least three animals. The resulting data were compared using the two-tailed Student's t-test with a P value <0.05 to indicate statistical significance.

Mast cell staining. Mast cells were visualized by labeling with ruthenium red (Sigma) (21). Mesenteric windows that had been scanned for albumin permeability were lifted from their glass slides and incubated for 1 min in 0.05% (wt/vol) ruthenium red and were then rinsed in HBS and remounted. Ruthenium red stained both degranulated and nondegranulated mast cells but did not wash off the extravascular FITC-BSA staining, which was fixed by the paraformaldehyde. Correlations between degranulated mast cells and areas of vascular barrier failure could thus be investigated.

	RESULTS

Top Abstract Introduction Materials and methods Results Discussion References

Immunolocalization of cadherin-5 in mesenteric microvascular endothelium. The rat mesenteric window preparation is a well-characterized in situ system for examining the response of microvascular beds to edemagenic agents (3, 33). Individual arterioles and venules within the thin transparent film of connective tissue of the window can be readily visualized at the light-microscopic level (Fig. 1A). These microvessels were characterized in categories according to their sizes: large arterioles and venules (>50 µm ID), midsized to small arterioles and venules (10-50 µm ID), and capillaries. Small venules were further categorized as postcapillary venules if their internal diameters ranged from 10 to 20 µm.

View larger version (72K): [in this window] [in a new window]   Fig. 1.   Photomicrograph of cadherin-5 immunolocalization in endothelium of rat mesenteric microvasculature. A: microvascular network within a rat mesenteric window is shown. White arrows on branches of the mesenteric artery and vein indicate the direction of blood flow to and from the network. Black arrow points toward the location of the intestines. B: adjacent large arteriolar and venular vessels immunolabeled with anti-cadherin-5 (9H7) primary antibody followed by indocarbocyanine (Cy3)-conjugated anti-mouse IgG. C: typical staining pattern of adjacent midsized to small arteriole and venule immunolabeled for cadherin-5. A, arteriole; V, venule. Scale bars: 400 µm (A) and 8 µm (B and C).

Immunohistochemistry of a control mesenteric window using a murine MAb specific for cadherin-5 (9H7) revealed a pattern delineating endothelial cell borders of the microvasculature. Immunoreactivity of the cadherin-5 antibody was selected for reactivity with the extracellular domain of cadherin-5. Characteristic cadherin-5 immunolabeling is shown in Fig. 1. The immunolocalization of cadherin-5 was distinctive for arteriolar and venular endothelial cells (Fig. 1, B and C, large and small vessels, respectively). Endothelial cells of both the large arteriole in Fig. 1B and the small arteriole in Fig. 1C were narrow (5-7 µm at widest points) and elongated. In comparison, venular endothelial cells were much wider (~15 µm at widest point) and less elongated. In capillaries, cadherin-5 immunoreactivity was limited to just one or two "strands," as it should be when the circumference of the entire vessel is composed of only one or two endothelial cells (data not shown). In all vessels, cadherin-5 localization was well defined and continuous along the entire border of each endothelial cell with almost no presence in the cytoplasmic regions. The immunolabeling pattern for cadherin-5 was very similar to that shown for the peripheral F-actin (compare Figs. 1, B and C, and 6, A and B).

Quantification of microvascular barrier function in response to cytokines. The microvasculature in control experiments (10-min incubation with buffer solution) showed few spontaneously occurring focal albumin leaks (Fig. 2A). Epifluorescence microscopy, performed after fixation, revealed outlines of entire microvascular networks due to residual FITC-BSA remaining on the luminal surfaces. At high magnification, endothelial cell outlines could be observed, presumably as a result of additional FITC-BSA accumulated in interendothelial junctional grooves (data not shown). In addition, extravascular areas that were marked by the extravasated FITC-BSA indicated precise locations where the vessel's barrier function had been compromised (Fig. 2, A and C). These sites, occurring randomly in both arterioles and venules, have been noted by other investigators (33, 40).

View larger version (109K): [in this window] [in a new window]   Fig. 2.   Digitized images of the extravascular FITC-albumin labeling pattern used to evaluate the microvascular paracellular permeability in response to cytokines tumor necrosis factor- (TNF-) and interferon- (IFN-). A and C: representative digitized images of microvascular networks from control experiment (A; n = 72) and from a window treated with TNF- (12.5 U/ml) and IFN- (20 U/ml) for 10 min (C; n = 68). B and D: areas of intense extravascular FITC-albumin staining were then delineated and quantitated by NIH Image. Scale bar: 50 µm.

Mesenteric windows perfused under physiological conditions with buffered saline to remove blood and then treated for 10 min by infusion with the cytokines TNF- and IFN- (12.5 U/ml and 20 U/ml, respectively) had more sites of FITC-BSA extravascular leakage (Fig. 2C) compared with control (Fig. 2A) mesentery. Moreover, the extravascular FITC-BSA staining marking these leakage sites had diffused further from the vessel into the interstitium of the surrounding tissue. This suggests that changes had occurred in the permeability barrier function of the vascular endothelial lining in response to the combined cytokine treatment. To quantify these observations, video scans of random fields of venules within the microvasculature of the mesentery were digitized and analyzed by NIH Image. As shown in Fig. 2, B and D, the sites that stained brightly due to extravasated FITC-BSA were delineated by the software. The area and numbers of these delineated locations were tabulated and analyzed for statistical significance between sets of experiments. For the cohort of animals used in this study, the average number of albumin leakage sites per micrometer of venule length was 2.3 ± 1.7 in control experiments. As shown in Fig. 3, there was a significant increase in the number of albumin leakage sites when either of the TNF-- and IFN--cotreated groups of experiments were compared with control [P < 0.005, n = 7 (125 U/ml TNF- + 200 U/ml IFN-); P < 0.05, n = 6 (12.5 U/ml TNF- + 20 U/ml IFN-)]. There was no significant difference when the control group (n = 6) was compared with the group treated with TNF- alone (125 U/ml) (n = 3). When the two groups of windows treated with the 12.5 U/ml TNF- + 20 U/ml IFN- and 10-fold higher dosages of combined cytokines were compared with each other, there were no significant differences as determined by Student's t-test. However, the data suggest that experiments with the 10-fold higher treatment concentration of TNF- and IFN- induced an even greater increase in permeability compared with the low-dose regimen.

View larger version (22K): [in this window] [in a new window]   Fig. 3.   Quantitation of the average number of albumin leakage sites per micrometer of venule (diameters of 10-75 µm) for each experimental group as indicated. Multiple cytokine-treated windows have significantly more leaks than control (* P < 0.05,  P < 0.005). Error bars are SE.

Correlation of endothelial barrier failure with cadherin-5 redistribution. Using immunolabeling and confocal microscopy, we examined changes in the endothelial intercellular junctions occurring specifically at areas where the vascular barrier had failed as evidenced by macromolecular extravasation. In particular, we focused on the changes occurring to one constituent of the adherens junction, cadherin-5, which has been shown to be important in the barrier formation of cultured endothelium (22, 32). Experiments were conducted whereby isolated rat mesenteric windows were perfused and incubated for 10 min with a control buffer solution (Fig. 4, A, C, and E), 12.5 U/ml TNF- and 20 U/ml IFN- of cytokines in combination (Fig. 4, B, D, and F), a 10-fold higher dose of cytokines (125 U/ml TNF- and 200 U/ml IFN-) (Fig. 5, A, C, and E), and a 100-fold higher dose of cytokines (Fig. 5, B, D, and F). The tissues were then perfusion fixed, subjected to immunohistochemistry, and analyzed.

View larger version (60K): [in this window] [in a new window]   Fig. 4.   Confocal image pairs and overlays showing colocalization of cadherin-5 immunoreactivity and sites of increased FITC-albumin permeability in TNF-- and IFN--treated venules. Comparison between control (A, C, and E) and low-dose TNF- (12.5 U/ml) and IFN- (20 U/ml) cytokine treatments (B, D, and F). A and B: FITC-albumin staining pattern. C and D: immunolabeling was with anti-cadherin-5 (9H7) primary antibody followed by Cy3-conjugated anti-mouse IgG. E and F: overlays of image pairs. Representative of 3 separate experiments per condition. Scale bar: 40 µm.

View larger version (66K): [in this window] [in a new window]   Fig. 5.   Confocal image pairs and overlays showing colocalization of cadherin-5 immunoreactivity and sites of increased permeability in TNF-- and IFN--treated venules. Comparison between 10-fold higher dose [TNF- (125 U/ml) + IFN- (200 U/ml)] cytokine treatments (A, C, and E) and 100-fold higher dose [TNF- (1,250 U/ml) + IFN- (2,000 U/ml)] cytokine treatments (B, D, and F). A and B: FITC-albumin staining pattern. C and D: immunolabeling with anti-cadherin-5 (9H7) primary antibody followed by Cy3-conjugated anti-mouse IgG. L, lymph vessel. E and F: overlays of image pairs. Representative of 3 separate experiments per condition. Scale bar: 80 µm.

As shown in Fig. 4C, confocal microscopy analysis on mesenteric venules from control experiments immunostained with anti-cadherin-5 demonstrated a normal interendothelial junctional staining pattern even at areas where small, spontaneous, focal FITC-BSA extravasations were detected (Fig. 4E). Reorganization of cadherin-5 immunolabeling was seen both in individual optical sections (not shown) and in the three-dimensional reconstructions shown in Figs. 4 and 5. Analysis of areas where there was extravascular FITC-BSA leakage in mesenteric windows treated with multiple cytokines consistently revealed a striking degree of disruption in the distinctive intercellular borders of the endothelium (merged images in Figs. 4F and 5, E and F). Cadherin-5 immunolocalization was no longer confined to the intercellular junction. Instead, a rather diffuse labeling pattern indicated that the remaining cadherin-5 molecules were free floating within the plasma membrane and were no longer anchored at the intercellular junction (Figs. 4D and 5, C and D). With the experimental protocol used in these studies, however, we were not able to follow the development of FITC-BSA leakage sites and cadherin-5 reorganization at different time points using the same preparation.

When FITC-BSA leakage sites in the experiments with varying TNF- and IFN- dosages were compared, increasing the cytokine concentrations resulted in observations of lengthier stretches of endothelium exhibiting FITC-BSA extravascular leakage and disruption of cadherin-5 immunolocalization. These results are illustrated in Figs. 4 and 5 with a comparison of representative image sets with tenfold increments of the cytokine concentration. Cadherin-5 disruption involved ~140 µm of venule (Fig. 4, B, D, and F) when caused by treatment with 12.5 U/ml TNF- + 20 U/ml IFN-, whereas the 10-fold higher cytokine dosage (125 U/ml TNF- + 200 U/ml IFN-) affected ~240 µm of venule (Fig. 5, A, C, and E). The 100-fold cytokine dosage (1,250 U/ml TNF- + 2,000 U/ml IFN-) tested was able to affect several hundred micrometers of venule, a segment of which is illustrated on Fig. 5, B, D, and F. Although we do not regard the highest cytokine dosage used in this study to be physiologically relevant (as described in DISCUSSION), the differences in the extent of leakage sites represented in Figs. 4 and 5 provide evidence that there is a dose-dependent response, i.e., the number of endothelial cells affected by the cytokines increases in a dose-responsive manner.

To quantitate the relative frequency with which leakage sites correlate with disorganized cadherin-5 immunolabeling, the albumin leakage sites on a representative mesenteric window treated with 125 U/ml TNF- + 200 U/ml IFN- were counted, and each leak was evaluated under high magnification (×100) for changes in the cadherin-5 staining pattern. Only extended leaks that involved more than two endothelial cells were evaluated in the collection of this data. Cadherin-5 immunolabeling at leakage sites was categorized as either "normal," "reorganized," or "not determined" and tabulated under the vessel type it was noted on (Table 1). The sites were categorized as not determined when a comprehensive examination of the cadherin-5 organization could not be made due either to peculiarities in the morphology of the microvascular network or to high background.

Of 39 leakage sites observed, 22 correlated with altered cadherin-5 immunolabeling. A minority of the leaks involved arterioles and capillaries (3 leaks per vessel type), some of which (100% and 33%, respectively) correlated with cadherin-5 organizational changes in the treated window. Together, these observations indicate that although venular endothelial cells were more likely to be affected by the cytokines (as indicated by alterations in cadherin-5 immunolabeling), some endothelial cells from arterioles and capillaries were also susceptible. Failure in the permeability function of arteriolar vessels has rarely been reported in acute inflammatory studies using histamine-type mediators but has been reported to occur as a result of chronic nonspecific irritation (15).

Cytokine-induced changes in the actin cytoskeleton. The endothelium of mesenteric venules shows a prominent peripheral rim of actin microfilaments in a pattern similar to cadherin-5. We examined whether combined cytokine-induced changes occur in our in situ experimental model using rhodamine-phalloidin to visualize the actin microfilaments. As shown in the epifluorescence micrographs of Fig. 6, both postcapillary venules (Fig. 6A) and large venules (Fig. 6B) from control experiments had fairly continuous peripheral actin rims as previously reported (55). Central stress fibers were not apparent.

View larger version (98K): [in this window] [in a new window]   Fig. 6.   Rhodamine-phalloidin staining patterns in mesenteric vessels within untreated and cytokine-treated microvascular networks. A and B: untreated postcapillary venules showing continuous, well-defined peripheral actin rims. C: treated [TNF- (12.5 U/ml) + IFN- (20 U/ml)] postcapillary venule showing disorganized cytoskeleton and undefined peripheral actin rims. D: treated [TNF- (12.5 U/ml) + IFN- (20 U/ml)] postcapillary venule showing punctate staining patterns and gaps within the barely defined peripheral actin rim in a larger postcapillary venule. Representative of 3 separate experiments per condition. Scale bar: 8 µm.

Treatment of the mesenteric window with 12.5 U/ml TNF- and 20 U/ml IFN- showed that there were subtle changes in the actin microfilament pattern in venular endothelium. However, due to the high background caused by the leakage of perfused rhodamine-phalloidin, we could not characterize the peripheral F-actin microfilament changes at major sites of compromised endothelial permeability. Even so, a range of different labeling patterns for F-actin filaments was observed elsewhere in the microvasculature, i.e., not correlated to leakage (FITC-BSA stained) sites. In some venules, the organization of peripheral actin cytoskeleton was no longer discernible, while disorganized cytoplasmic staining increased drastically (Fig. 6C). In other venules where the peripheral actin rims could still be visualized, more subtle changes were apparent, such as stretches of punctate interendothelial junctional staining and the presence of small gaps in the peripheral F-actin microfilaments (Fig. 6D). We conclude that the cytokines TNF- and IFN- do indeed cause global changes in the peripheral F-actin cytoskeleton as reported by others (51). We postulate that cytokine-stimulated changes in the peripheral actin cytoskeleton may precede the disorganization of the adherence junctions and the increased macromolecular permeability.

Leukocytes and mast cells at sites of endothelial barrier failure. To isolate the effects of the cytokines TNF- and IFN- on the endothelial lining of the mesenteric microvasculature, we attempted to eliminate the effects of the blood-borne cellular and humoral inflammatory mediators by flushing the isolated microvascular network with heparinized, buffered saline before beginning each experiment. We verified whether potential endothelial barrier damaging residual leukocytes remained within the intravascular space of the microvascular networks by immunodetection using the anti-CD11b/c MAb OX-42 (45).

As shown by immunofluorescence microscopy (Fig. 7, A and B), the venular intravascular space from TNF-- and IFN--treated (12.5 U/ml TNF- and 20 U/ml IFN-) experiments was free of OX-42 positively immunolabeled cells. There were no extravasating OX-42 positively labeled cells observed. Extravascular OX-42 positively immunolabeled cells were common in interstitial areas surrounding the microvascular networks (Fig. 7, A and B). Since cells other than leukocytes express CD11b/c, these other cells probably accounted for some of the positively labeled cells detected. These cells include other granulocytes, B cell subsets, T cell subsets, natural killer cells, and macrophages. Nevertheless, no correlation could be made regarding the positions of positively labeled cells and the presence of extravascular FITC-BSA leakage. For example, examination of 47 extended FITC-BSA leakage sites from two cytokine-treated windows revealed only five sites with immediately adjacent CD11b/c positively labeled cells present. In comparison, three of nine leakage sites in a window from a control experiment had CD11b/c positively labeled cells immediately adjacent to the abluminal surface. In addition, the microvascular barrier was often intact in specific areas where CD11b/c positively immunolabeled cells were immediately adjacent to a venule (Fig. 7A). Conversely, areas of FITC-BSA leakage occurred at areas where no CD11b/c positively labeled cells were immediately present (Fig. 7B).

View larger version (92K): [in this window] [in a new window]   Fig. 7.   Photomicrographs showing that endothelial albumin permeability due to multiple cytokine treatment (TNF-, 12.5 U/ml, and IFN-, 20 U/ml) is not mediated via leukocyte extravasation or mast cell degranulation. A and B: double-exposure epifluorescence micrographs reveal that CD11b/c-positive cells (including leukocytes) identified by monoclonal antibody OX-42 are exclusively extravascular in this model. CD11b/c-positive staining cells immediately adjacent to venule (arrows, A) do not correlate with FITC-albumin leakage. Also, FITC-albumin leakage can occur at areas where no CD11b/c-positive staining cells were immediately present (arrows, B). Scale bars: 13 µm. C and D: normal light and epifluorescence micrographs taken of the same field show nondegranulated mast cells stained with ruthenium red in the vicinity of a venule with one FITC-albumin leakage site (white arrows). Black outline in C delineates the position of the venule in the field. Scale bars: 20 µm. E and F: pair of degranulated mast cells (black arrows, E) present in the same field, but at a distance (>100 µm) from a venule with a FITC-albumin leakage site (white arrows). Representative of 2 separate experiments per condition. Scale bars: 20 µm.

The increase in both leukocyte adhesion and albumin flux across the endothelial barrier in the rat mesenteric preparations has previously been attributed, in part, to the activation of the endothelium by proinflammatory mediators released from degranulating mast cells. Kubes and Gaboury (28) demonstrated that when connective tissue mast cells within the rat mesentery were activated (using compound 48/80), the resulting degranulation caused the release of mediators that increased both leukocyte recruitment and FITC-albumin extravasation.

To minimize mast cell degranulation in our experiment, the mesenteric microvascular network was pretreated with cromolyn, which is a well-established stabilizer of mast cells and granulocytes (17). To explore the possibility of whether the observed TNF-- and IFN--induced increase in permeability was mediated by mast cells, combined cytokine-treated mesenteric windows were stained for mast cells with ruthenium red (21) and examined with light microscopy. Because OX-42 was characterized by Robinson et al. (45) as not recognizing rat mast cells, we predicted that ruthenium red would stain a separate population of cells. This was confirmed by restaining OX-42-treated windows with ruthenium red (data not shown). Because the FITC-BSA staining could still be detected in these tissue samples under epifluorescence, a correlation between areas of microvascular barrier failure and of degranulated mast cells could be investigated.

FITC-BSA leakage sites were found regardless of whether mast cells were present adjacent to the microvessels. Images taken under both normal light and epifluorescence (Fig. 7, C and D) reveal relatively intact mast cells in the vicinity of a venule with just one focal FITC-BSA leakage site. In Fig. 7, E and F, two well-degranulated mast cells (Fig. 7E), as evidenced by the reduced density of the ruthenium red stain and the surrounding released granules, were captured in the same field but distant from a small venule with a focal albumin leak (Fig. 7F). These observations provide evidence that the perfusion of these mesenteric windows with the cytokines TNF- and IFN- did not cause widespread mast cell activation and degranulation

	DISCUSSION

Top Abstract Introduction Materials and methods Results Discussion References

In this study, we provide evidence that the cytokines TNF- and IFN-, in combination, affect the barrier function of the vascular endothelial lining by direct stimulation of endothelium that ultimately results in the disruption of cadherin-5-mediated cell-cell adhesion. The microvasculature of the mesentery was perfused with cytokines and subsequently fixed under physiological conditions. Disruption of the endothelial barrier occurred in the absence of leukocytes in the lumen of microvessels and did not correlate with the presence of mast cells or leukocytes in the interstitium. Our data suggest the following conclusions. 1) The combined infusion of TNF- and IFN- was effective in triggering rapid vascular barrier failure at physiological doses, i.e., at least one order of magnitude lower than that which has often been used in in vitro studies (8, 38, 46, 51). 2) The rapid opening of the interendothelial junctions required the synergy generated by the presence of both cytokines. TNF-, even at 125 U/ml, was not able to elicit a significant, consistent response in the time period examined. However, longer incubation times in vitro have shown that TNF- alters endothelial permeability (12, 18, 38, 56). We did not present data addressing the effects of IFN- treatment because of the lack of support in the literature of IFN-- induced increase in endothelial permeability. 3) Although not statistically significant, there was a trend toward more albumin leakage sites and larger leakage areas (comparison of image sets in Fig. 4 and 5) when higher cytokine dosages were used. This lack of significance, however, may be due to limitations in the analysis of kinetics in the mesenteric window preparation. For example, the merging of closely situated large extravascular leakage sites (observe leaks 4 and 7 on Fig. 2D) tended to cause an underestimation in leakage site counts. For this reason quantification of leakage sites in mesenteric windows treated with the highest cytokine dosage was not plausible and therefore not conducted.

Serum levels for cytokines reported from clinical studies indicate that the 12.5 U/ml TNF- + 20 U/ml IFN- and the 10-fold higher cytokine dosage (125 U/ml TNF- + 200 U/ml IFN-) concentrations used in this study are likely to be physiologically relevant. Measurement of plasma TNF- in patients with acute myocardial infarction and septic shock ranged from 10 to 1,510 pg/ml, whereas the concentration used in this study ranged from 12.5 to 125 pg/ml (6, 13). Clinical measurements of cytokine plasma levels can, however, be inaccurate due to their short half-lives in plasma, as well as the cyclical nature of cytokine release by immune cells (7, 30). In addition, plasma levels of cytokines reported in most studies may not reflect the local tissue environment where immune cells releasing cytokines are active and may produce environments with even higher cytokine concentrations. Together, these complications in plasma measurements may account for the fact that some studies have not been able to report a correlation of IFN- serum levels with acute inflammatory responses such as those during transplant rejection or septic shock (13, 19).

Agonist-induced macromolecular extravasation via the paracellular pathway in the microvascular endothelium has previously been shown to be linked to changes in the endothelial cytoskeleton and junctional proteins (35). The presence of interendothelial gaps, which mediated macromolecular leakage into the interstitium, was observed in rat cremaster microvessels treated with histamine (36) and substance P (5). In the rat mesenteric microvasculature, sites of increased venular permeability in response to histamine were correlated with disruption of the peripheral F-actin cytoskeleton (3). Formation of focal transient intercellular spaces or gaps between endothelium were found to correspond to sites of plasma protein leakage and cytoskeletal changes. Leach et al. (34) demonstrated a generalized loss of cadherin-5 detection in frozen sections of human placental microvessels after perfusion with histamine. Our studies further demonstrated that, in response to TNF- and IFN-, changes in the organization of cadherin-5 occurred precisely at areas where the endothelial barrier had failed, as did changes in the associated actin cytoskeleton.

The intercellular junction complex in microvascular endothelium consists of the tight junction barrier and the actin-associated adherens junction (48). Evidence suggests that together these junctional structures form a functional unit. The distinctions between the tight junction and the adherens junction are based on morphological and biochemical criteria (8, 26, 32, 37, 48). These criteria, however, do not describe the functional interrelationships between proteins of the junctional complex. In cultured endothelium, cadherin-5-mediated cell-cell contacts are critical in the formation of tight junctions, which confer permeability barrier properties (22, 32). Moreover, the protein complexes, which form the specialized intercellular structures of tight and adherens junctions, have been shown to share common proteins. Components of cadherin-based adherens junctions function in the assembly of ZO-1 into tight junctions (42). The catenins facilitate mobilization of ZO-1 from the cytosol to the plasma membrane during junction assembly in epithelium, which is initiated by E-cadherin adhesion.

The increase in endothelial permeability in response to inflammatory mediators has previously been attributed to disruption in cellular localization of components of both tight and adherens junctions. Morphological and immunohistochemical evidence of tight junction and adherens junction changes induced by histamine perfusion has been reported in intact human placental model (34). Changes in the localization of both cadherin-5 and occludin, an integral membrane component of the tight junction, have recently been reported to occur in cultured endothelial cells treated with vascular permeability factor (26). In addition, TNF- and IFN- induced changes in the cellular localization of the tight junction-associated protein ZO-1, and these changes closely correlated with alterations in the F-actin microfilament system along the cell junctions (8). The changes in ZO-1 localization were dependent on peripheral actin organization and were blocked by cytochalasin D. Together, these results suggest that TNF- and IFN- induce the disassembly of the cadherin-5/catenin complex, which occurs in conjunction with a series of other changes in components of the tight junction and the actin cytoskeleton and contributes to increasing permeability.

The rapid kinetics of cytokine-induced reorganization of cadherin-5 in venular endothelium suggests modification of the cadherin-5/catenin complex at the distal end of the signaling cascade. Adhesion of cadherin-5 requires the interaction of the cytoplasmic domains with the F-actin cytoskeleton through the catenins, which have been shown to be a site for regulation of the complex (31). At the proximal end of the signaling cascade, signal transduction by TNF- and IFN- is initiated through cell surface receptors. Binding of IFN- to its receptors initiates intracellular signal transduction primarily through activation of a cascade of subunits in the JAK/STAT pathway such as Jak-1, Jak-2, and Stat1 (16). Two cellular receptors for TNF-, p55 and p75, have been identified. Although both receptors are present in endothelial cells, p55 has been found to be largely responsible for the activation of the diverse intracellular signaling pathways. (49). On TNF- binding, adapter proteins associate with the cytoplasmic domain of members of the TNF receptor superfamily and mediate downstream signaling (e.g., TRADD, FADD, RIP, TRAF-1, and TRAF-2) (25). These proteins associate with the receptors via "death domain" sequences, termed as such because their activities have been linked to apoptotic responses. It has been postulated that TRADD association with the p55 receptor in endothelial cells is followed by the recruitment of multiple adapter proteins (e.g., FADD or TRAF-2), thereby generating multiple signaling pathways (25, 50).

The dynamic disorganization of the cadherin-5/catenin complex from intercellular junctions in response to TNF- and IFN- is likely to involve the regulation of interaction of the cadherin/catenin complex with the actin cytoskeleton. One mechanism could be via the alteration of the affinity or conformation of the cadherin-5/catenin complex, thus modifying its association with actin filaments or the cadherin-mediated intercellular adhesion of the extracellular domains. A second potential mechanism is the disruption of the peripheral F-actin filaments, which could break up any linkage of the cadherin-5/catenin complex to the cellular cytoskeleton. Either mechanism could result in the redistribution of cadherin-5 molecules within the plasma membrane or, if internalized, the cytoplasm. We propose four distinct signaling cascades that may operate via one or both of our proposed mechanisms to redistribute cadherin-5 in response to TNF- and IFN- stimulation. These signaling cascades could involve either 1) tyrosine (31) or serine/threonine phosphorylation, 2) small GTP-binding proteins (10, 24), 3) apoptosis/caspases activity, or 4) nitric oxide activity (4, 29).

Rho family members are key players in the signal transduction mechanisms that regulate a diverse range of intracellular activities such as the formation of actin stress fibers, lamellipodia and filopodia, as well as a number of intracellular tyrosine and serine/threonine kinases (54). Hippenstiel et al. (24) demonstrated that the inactivation of Rho members of the Ras superfamily of small GTP-binding proteins disrupted the barrier function of porcine pulmonary artery endothelial monolayers. Braga et al. (10) recently reported that the small GTPases Rho and Rac are required for the establishment of cadherin-dependent cell-cell contacts in keratinocytes. Inhibition of Rho and Rac resulted in loss of E-cadherin localization at cell-cell junctions. Thus, for TNF- and IFN- to disrupt cadherin-5 localization, either Rho subfamily proteins must activate kinases that phosphorylate members of the cadherin-catenin complex or certain Rho subfamily members must be inactivated.

Numerous studies show that TNF- and IFN- are able to activate signaling cascades implicated in the apoptotic response (14, 50). For example, the TRADD/FADD pathway leads to the activation of a cascade of cysteine proteases known as caspases (39). Caspase-8 (MACH) has been found to be recruited to FADD on TNF-p55 binding and is thus likely to be the first member of the caspase cascade to be activated (9). Cytoskeletal substrates for activated caspases include proteins associated with the actin cytoskeleton, such as gelsolin or -catenin (11, 27). Notably, a recent report demonstrated the caspase-mediated cleavage of -catenin in cultured human endothelial cells undergoing apoptosis triggered by growth factor deprivation (23). Cleavage of -catenin leaves it incapable of maintaining an association with the cadherin-catenin junctional complex, thereby leading to the loss of cadherin-mediated cell-cell adhesion and finally cadherin relocalization.

The cadherin-5/catenin complex and the associated actin cytoskeleton appear to play a role in control of vascular permeability by stabilization of the interendothelial junction. In a microvascular network, postcapillary venules are principally involved in the pathophysiological response to proinflammatory cytokines. We have shown that treatment of microvascular endothelium in situ with TNF- and IFN- results in loss of the venular endothelial barrier function with the disruption of cadherin-5 adhesion in intercellular junctions. Self-association of the extracellular domains of cadherin-5 on adjacent cells as well as the stable association of cadherin-5 with the actin cytoskeleton are important in the formation of the endothelial barrier formed by tight junctions (22, 32). Future investigations are directed toward determining the mechanism of regulation of the cadherin-5/catenin complex in modulating the permeability barrier in endothelium exposed to proinflammatory cytokines.