angiopoietin-1 (Ang-1) is a ligand for Tie-2 tyrosine kinase receptors expressed on endothelial cells (9). Mutation of Ang-1 or its receptors is lethal during embryonic development because primitive vascular tubes do not evolve into mature functional blood vessels (11, 32). Ang-1/Tie-2 signaling is also thought to stabilize blood vessels through the recruitment of mural cells (32). Expression of Ang-1 ligand in mural cells and Tie-2 receptors in endothelial cells of quiescent vessels of the adult vasculature suggests a role for Ang-1 in maintenance of vascular stability (22, 30, 31, 44).

Plasma leakage, a hallmark of inflammation, has been attributed to the formation of intercellular gaps in the endothelium of postcapillary venules and collecting venules (24–26). Staining with silver nitrate or Ricinus communis agglutinin-I lectin, which mark sites of leakage, and ultrastructural studies have shown that endothelial gaps are focal openings resulting from transient separation of intercellular junctions (3, 16, 25, 33).

Ang-1 has a distinctive effect on plasma leakage. Dermal blood vessels in transgenic mice that overexpress Ang-1 in the skin are resistant to leakage induced by a variety of inflammatory mediators (35). Similar results have been obtained with COMP-Ang-1, a variant of Ang-1 that stimulates the formation of new blood vessels that are resistant to leakage (6). Moreover, systemic delivery of Ang-1 by adenoviral vector inhibits leakage induced by inflammatory stimuli (34). Ang-1 also has beneficial effects in models of endotoxic shock, retinopathy, and ischemia (28, 43, 45). The cellular mechanism of these effects of Ang-1 in vivo is unclear.

Studies of endothelial cells in vitro suggest that Ang-1 can decrease monolayer permeability, change the distribution and activation of junctional adhesion molecules, and stabilize cell-cell junctions. Ang-1 promotes recruitment of CD31 (platelet-endothelial cell adhesion molecule-1, PECAM-1) to junctions and decreases phosphorylation of CD31 and VE-cadherin (14, 18, 38). Ang-1 also reduces cytokine-induced expression of ICAM-1, VCAM-1, E-selectin, and tissue factor, all of which are implicated in the activation of endothelial cells and leukocyte adhesion and migration in inflammation (20, 21).

The present study sought to determine the magnitude and cellular mechanism of the antileak effect of Ang-1 in a mouse model of airway inflammation. Adult mice were injected intravenously with recombinant adenovirus encoding a genetically engineered version of angiopoietin-1 called Ang-1* (34). Adenovirally transduced hepatocytes generated systemic levels of Ang-1* in the bloodstream. Three days later the tracheal vasculature was challenged by intravenous injection of bradykinin (26). The dose response and time course of bradykinin-induced leakage were assessed by measuring extravasation of Evans blue dye. Leaky vessels were identified, and pore cutoff size was determined by quantifying extravasated fluorescent microspheres. The integrity of endothelial cell junctions and formation of endothelial gaps were evaluated by examining immunohistochemically stained tracheal whole mounts by confocal microscopy.

MATERIALS AND METHODS

Animals and Treatment

The effect of Ang-1* on vascular leakage was examined in pathogen-free FVB/n 8-wk-old male mice (Charles River, Hollister, CA). Mice were anesthetized by intramuscular injection of ketamine (87 mg/kg) and xylazine (13 mg/kg), and 10⁹ plaque-forming units of Ang-1*-expressing adenovirus were injected into a jugular vein (Ad-Ang-1* group) (34). At 3 days after the injection of Ad-Ang-1*, serum concentrations of Ang-1* were 1–15 μg/ml in all mice, consistent with published observations (34). Age-matched control FVB/n mice received an injection of adenovirus expressing green fluorescent protein (GFP; Ad-GFP group) or nothing. GFP gene expression in hepatocytes, which were the main target of adenoviral transduction, was confirmed by fluorescence microscopy (34). Ang-1* was not detected in blood of the control mice.

Three days after the injection of adenovirus, mice were challenged by intravenous injection of bradykinin (Sigma, St. Louis, MO) at a dose of 0.01, 0.1, 1, 10, 20, or 50 mg/kg diluted in 100 μl of 0.9% NaCl; control mice received 100 μl of 0.9% NaCl. All experiments were approved by the University of California, San Francisco Institutional Animal Care and Use Committee.

Measurement of Evans Blue Leakage

Mice used to measure the dose response and time course of bradykinin-induced plasma leakage of Evans blue dye were anesthetized with ketamine and xylazine. Evans blue (30 mg/kg in 100 μl of PBS; Electron Microscopy Sciences, Fort Washington, PA) was injected into a femoral vein, followed 2 min later by injection of bradykinin (4, 35). Five to sixty minutes after the Evans blue injection, the chest was opened, the atria were removed to create an exit route, the left ventricle was incised, an 18-gauge cannula was inserted through the ventricle into the ascending aorta, and then intravascular dye was flushed from the bloodstream by perfusion of 1% paraformaldehyde (PFA) in citrate buffer (0.05 M, pH 3.5) at a pressure of 120 mmHg for 3 min. Completeness of the perfusion was confirmed by uniform clearing of blood from the trachea and other organs. The effectiveness of this approach for rinsing blood from the vasculature was previously verified by infusion of a labeled lectin that documented the absence of blood by binding uniformly to tracheal blood vessels (33). Tracheas were removed, weighed, and incubated in formamide overnight at 67°C (Sigma). Extracted Evans blue was measured with a spectrophotometer at 620 nm and expressed as nanograms of dye per milligram of tissue wet weight.

Localization and Measurement of Microsphere Extravasation

Sites of leakage in tracheal blood vessels were marked by extravasated fluorescent polymer microspheres ranging in diameter from 25 to 1,000 nm (20 μl iv; Duke Scientific, Palo Alto, CA) (25). Because the microsphere preparations had a uniform weight per volume relationship, the number of microspheres in 20 μl varied inversely with diameter: 2 × 10¹³ (25 nm), 4 × 10¹¹ (100 nm), 3 × 10⁹ (500 nm), and 3 × 10⁸ (1,000 nm). Mice anesthetized with Nembutal (40 mg/kg ip) received an intravenous injection of bradykinin (1 or 10 mg/kg) or saline and 0–9 min later an injection of microspheres. One minute thereafter, intravascular microspheres were removed from the bloodstream by vascular perfusion of 1% PFA in PBS, following the same approach as for Evans blue.

Tracheas were removed and fixed for an additional 2 h, and then the vasculature was stained for CD31 immunoreactivity (2). Tracheal whole mounts were incubated for 1 h in 5% normal goat serum (Jackson ImmunoResearch, West Grove, PA) in PBS containing 0.3% Triton X-100 (PBS-plus) and then incubated overnight at 4°C with hamster anti-mouse CD31 (clone 2H8; Chemicon, Temecula, CA; 1:1,000 in PBS-plus) and/or rat anti-mouse CD31 (MEC13.3; Pharmingen, San Diego, CA; 1:500 in PBS-plus). The next day, tracheas were washed several times with PBS-plus and incubated for 4 h at room temperature with FITC- or Cy3-conjugated goat anti-hamster and/or anti-rat IgG (Jackson ImmunoResearch; 1:200 in PBS-plus). Tracheas were rinsed several times in PBS-plus, flattened, and mounted with the luminal surface facing up in Vectashield (Vector Laboratories, Burlingame, CA).

CD31-stained tracheal whole mounts were examined with a Zeiss Axiophot fluorescence microscope (×10 objective, ×1 Optovar, tissue region 960 × 1,280 μm) equipped with fluorescence filters and a low-light, externally cooled, three-chip charge-coupled device camera (480 × 640 pixel RGB color images; CoolCam, SciMeasure Analytical Systems, Atlanta, GA). Area density of extravasated microspheres (proportion of total pixels) was measured on digital images of regions of tracheal mucosa between five to eight cartilage rings where venules and postcapillary venules were most abundant. Camera settings were uniformly applied to images from all groups in each experiment (19). RGB color images were converted into 8-bit gray scale, and then ImageJ software ( http://rsb.info.nih.gov/ij) was used to determine the proportion of pixels having a fluorescence intensity equal to or greater than specific threshold values that maximized the microsphere signal and minimized tissue autofluorescence. Thresholds were found empirically to be ≥40 for green or red 100-nm microspheres or ≥20 for blue 500-nm microspheres (fluorescence intensity range: 0–255 in 8-bit images). Area densities of microspheres in a 205,920-pixel region of interest in each image were expressed as the percentage of pixels with a brightness equal to or greater than the threshold. An average area density value was calculated for all images from each mouse (3–4 mice/group).

Confocal Microscopic Imaging of Endothelial Gaps

For confocal microscopic studies of endothelial gaps, mice were anesthetized with Nembutal and then received an intravenous injection of bradykinin (1 mg/kg) or saline. Fluorescent microspheres (100 or 500 nm) were injected at the time of the bradykinin injection or from 1 to 9 min later. Fixation by vascular perfusion of 1% PFA began 1 min after injection of microspheres, thereby standardizing the microsphere circulation time in all experiments. Tracheas were stained for CD31 immunoreactivity with both hamster and rat anti-mouse CD31 antibodies, rinsed in PBS-plus, mounted in Vectashield, and examined with a Zeiss LSM-510 confocal microscope with argon, helium-neon, and UV lasers.

The two anti-CD31 antibodies labeled endothelial cell junctions in arteries and veins but played separate roles in the identification of endothelial gaps because they labeled different populations of CD31 molecules in endothelial cells of tracheal capillaries and venules. The hamster antibody labeled CD31 at intercellular junctions, and the rat antibody labeled a population of CD31 molecules more uniformly distributed in the plasma membrane. We took advantage of these properties to distinguish unstained intercellular gaps from regions of nonjunctional plasma membrane, which stained with the rat antibody but not the hamster antibody. When the hamster antibody was used alone, cell junctions were labeled but intercellular gaps could not be distinguished from nonjunctional regions of plasma membrane because neither was stained. Addition of the rat antibody solved that problem. By uniformly labeling endothelial cell plasma membranes, the rat antibody also made measurements of vessel diameter more accurate.

Measurement of Endothelial Gap Number and Size

Unlike measurements of microsphere area density, the number and size of endothelial gaps were determined in confocal microscopic images of tracheal venules in mice injected with bradykinin (1 mg/kg) or saline, followed 1 min later by 500-nm microspheres. The vasculature was perfused with fixative 2 min after the bradykinin, and tracheal whole mounts were stained with the two anti-CD31 antibodies. Confocal images were obtained from the tracheas of four or five mice in each of four groups: two Ad-Ang-1* groups and two Ad-GFP groups, one of each challenged with bradykinin and the other with saline. A total of 15–17 confocal image projections were examined in each group (×40 oil objective, ×3 zoom). Each projection represented a stack of 28–64 optical sections, each 0.39–0.58 μm in thickness, through the entire vessel wall. Scanning parameters were adjusted for each image to optimize gap identification. At this high magnification, intercellular gaps were clearly visible at endothelial cell borders marked by the hamster CD31 antibody.

All optical sections in the image stacks were examined to identify and encircle each gap with the Line Drawing and Measuring function of the Zeiss LSM Image Browser software ( http://128.40.242.20/LSMContent.htm). Gap borders thus drawn were used to determine the number, perimeter, and area of gaps, which in turn were used to calculate the gap minor and major axes. The number of gaps was expressed per millimeter of venule midline length. Venular length, measured with a digitizing tablet (Digi-Pad; GTCO CalComp, Scottsdale, AZ), rather than surface area was used as the reference because prolonged exposure to Ang-1* can result in circumferential enlargement of venules due to endothelial cell proliferation (2, 7, 36). In our experiments, the enlargement was minimal at 3 days after Ad-Ang-1*.

Perimeters of endothelial gaps in confocal images were measured with the Closed Polyline Overlay measuring tool of the Zeiss LSM Image Browser software. Gap areas were calculated from the number of pixels contained within these perimeters.

The minor axis of gaps was calculated from gap perimeter (P) and area (A), based on an approximation formula for the circumference of an ellipse (15)

The major axis of gaps was calculated from gap area, perimeter, and minor axis (15)

The proportion of venular surface occupied by gaps in each image projection was calculated from the total area of gaps and vessel surface area by treating venules as cylinders. Average vessel diameter and length were measured on each image projection, and endothelial surface area was calculated with the formula for surface area of a cylinder (2πrl, where r is the radius and l is the length). Fractional gap area was expressed as square micrometers of gap openings per square millimeter of venular surface.

Statistical Analysis

Values are expressed as means ± SE. Experimental groups consisted of three to nine mice unless otherwise indicated. The significance of differences among groups was assessed by ANOVA followed by the Bonferroni-Dunn test for multiple comparisons. The relationship between area density of extravasated microspheres and proportion of venular surface occupied by gaps was assessed by linear regression. Differences between groups were considered significant when P < 0.05.

RESULTS

Bradykinin-Induced Leakage of Evans Blue

Before examining the effect of Ang-1* on vascular leakiness, we examined the dose response of bradykinin-induced leakage of Evans blue in the mouse trachea. Mice received Evans blue followed 2 min later by an injection of saline or bradykinin. Dye accumulation in the trachea was measured 30 min after injection of Evans blue and expressed as a percentage of the baseline value. After injection of saline, the baseline concentration of Evans blue at 30 min was 21.2 ± 3.5 ng/mg trachea (Fig. 1A). Doses of bradykinin from 0.01 to 0.1 mg/kg did not increase the value above this baseline, but 1 mg/kg caused a 38% increase and 10 mg/kg produced an 86% increase (Fig. 1A). Larger doses of bradykinin (20 or 50 mg/kg) produced somewhat more leakage (Fig. 1A) but were lethal in some mice.

We next examined the time course of the leakage. Injection of Evans blue was followed 2 min later by an injection of bradykinin (10 mg/kg) or saline. The dye was flushed from the vasculature by perfusion of acidified fixative 5, 30, or 60 min after the dye injection (Fig. 1B). Saline-injected groups provided a reference for baseline Evans blue accumulation at the three time points. At 5 min after bradykinin challenge, tracheal Evans blue content was 91% higher than the 5-min baseline value. Dye content was still increased at 30 min (123% above the 30-min baseline) but decreased thereafter, and at 60 min was only 25% above the 60-min baseline (Fig. 1B).

Bradykinin-Induced Leakage of Fluorescent Microspheres

Evans blue dye studies helped to define the dose response and time course of leakage after bradykinin, but high baseline values made the approach relatively insensitive for monitoring changes in blood vessel leakiness in the trachea. Therefore, we used 25- to 1,000-nm fluorescent microspheres as tracers in subsequent experiments. Particles of this size do not cross the endothelium of tracheal vessels in the absence of increased permeability, and little extravasation occurs under baseline conditions. Extravasated microspheres had the additional attribute of marking sites of leakage and providing an estimate of the pore cutoff size of the leaky sites.

Location and pore size of leakage sites.

Microspheres were injected 1 min after saline or bradykinin (10 mg/kg) and were removed from the bloodstream by vascular perfusion 4 min later. Each mouse received microspheres of only one size. Extravasated microspheres were rare after saline challenge but were abundant after bradykinin. Most extravasated microspheres, regardless of the size injected, were located in the wall of venules (Fig. 2, A–D). The number of extravasated microspheres in venular walls varied with the size injected (Fig. 2, A–D). The smallest microspheres (25 nm) were most abundant (Fig. 2A), consistent with their greater extravasation propensity and abundance due to logfold differences in number injected (see materials and methods). Even the largest microspheres extravasated, indicating the presence of openings as large as 1,000 nm (Fig. 2D). CD31-positive leukocytes were present in some venules after bradykinin (Fig. 2E). Few microspheres extravasated from arterioles or capillaries, as found in previous studies (25, 26).

Time course of leakage.

To determine the time course of microsphere leakage, 100-nm microspheres were injected with bradykinin at time 0 or 1–9 min thereafter. In all cases, the microspheres circulated for 1 min and were then flushed from the bloodstream by vascular perfusion of fixative. Baseline leakage was assessed after injection of saline followed by microspheres and 1 min later by perfusion of fixative. Area density measurements showed that accumulation of 100-nm microspheres after a 10 mg/kg dose of bradykinin was 3-fold greater than baseline during minute 1, 10-fold greater during minute 2, and 15-fold greater during minute 3 and was back to baseline by minute 10 (Fig. 3, A–E and G). Bradykinin at 1 mg/kg resulted in a similar amount of leakage during minute 2 (Fig. 3, F and H). Because microsphere circulation time was held constant, where perfusion times ranged from 1 to 10 min after bradykinin injection, the amount of extravasated microspheres reflected leakage rather than clearance of microspheres from the circulation.

Magnitude of Ang-1* Effect on Blood Vessel Leakiness

Effect of Ang-1* on Evans blue leakage.

On the basis of data for the dose response of bradykinin and the time course of Evans blue leakage, we next examined the antileak effect of Ang-1*. Mice received Ad-Ang-1* or Ad-GFP, and 3 days later were challenged with saline or bradykinin (10 mg/kg) followed by Evans blue that circulated for 30 min. In the Ad-GFP group bradykinin increased tracheal dye accumulation by 60% above the saline baseline (Fig. 4A), but in the Ad-Ang-1* group bradykinin produced no detectable increase above baseline (Fig. 4A). In the absence of the bradykinin stimulus, baseline leakage over 30 min was about the same in both groups (Fig. 4A).

Effect of Ang-1* on microsphere leakage.

Multiple factors, including blood flow, vascular surface area, and endothelial barrier function, determine the amount of Evans blue accumulation in tissues. Increased tissue content of a tracer such as Evans blue that can cross the normal endothelium is not proof of increased endothelial permeability. Therefore, we repeated the Ang-1* studies with a particulate tracer, fluorescent 500-nm microspheres, which are too large to extravasate from tracheal vessels without a defect in endothelial barrier function.

With the goal of determining whether the antileak effect of Ang-1* involves a permeability decrease consistent with restriction of endothelial gap formation, we asked whether Ang-1* reduces leakage of 500-nm microspheres. Mice received Ad-Ang-1* or Ad-GFP and 3 days later were challenged with bradykinin (1 mg/kg) or saline. Leakage from tracheal blood vessels was assessed by quantifying the accumulation of 500-nm microspheres. Leakage of microspheres at 2 min after bradykinin was 39 times the saline baseline value in the Ad-GFP group but 12 times baseline in the Ad-Ang-1* group, reflecting a 69% reduction compared with the Ad-GFP group (Fig. 4B). This change could be explained by a reduction in gap number, size, or both.

Endothelial Gap Formation After Bradykinin

The results of studies with fluorescent microspheres led us to learn more about changes in the endothelium of venules after bradykinin challenge. In control mice, most endothelial cell borders of tracheal venules were smooth (Fig. 5, A and A′), and few or no endothelial gaps were visible by confocal microscopy after immunohistochemical staining with the two anti-CD31 antibodies. Because the hamster anti-mouse CD31 antibody stained endothelial cell borders and the rat anti-mouse CD31 antibody uniformly stained nonjunctional endothelial cell plasma membranes, gaps could be seen as unstained regions between endothelial cells. Similar observations were made when endothelial cell junctions were stained for VE-cadherin or ZO-1 immunoreactivity (data not shown). At 1 min after bradykinin (1 mg/kg), gaps were visible between many endothelial cells of venules (Fig. 5B). The number and size of gaps increased between 1 and 4 min (Fig. 5, B and C). Extravasated microspheres were present near some gaps at 1 min (Fig. 5B) and were more abundant and widely distributed on venules at 4 min (Fig. 5C). Fingerlike processes (filopodia) partitioned some of the larger gaps (Fig. 5, C and C′). At 10 min, gaps and extravasated microspheres were rare, and endothelial cell borders again had a smooth contour with few filopodia (Fig. 5D).

Effect of Ang-1* on Gap Number and Size

Consistent with its effect on microsphere leakage, Ang-1* reduced both the number and size of endothelial gaps in tracheal venules after bradykinin challenge. In the Ad-GFP group examined 2 min after bradykinin (1 mg/kg), gaps of various size were visible at the borders of many endothelial cells and 500-nm microspheres were abundant in the wall of venules (Fig. 6, A–C). Measurements showed on average 836 gaps/mm venule length (Fig. 6G). By comparison, endothelial gaps in the Ad-Ang-1* group were 38% less numerous (516 gaps/mm) and were accompanied by fewer extravasated microspheres (Fig. 6, D–G). Gaps were rare in venules of saline-challenged mice in either group (Fig. 6G).

In addition, gaps were 38% smaller in the Ad-Ang-1* group than in the Ad-GFP group (gap area: 0.45 vs. 0.73 μm²; Fig. 6H). The mean major axis of gaps in the Ad-Ang-1* group was reduced 22% (0.76 vs. 0.98 μm; Fig. 6I). Mean minor axis was 20% smaller in the Ad-Ang-1* group (0.16 vs. 0.20 μm; Fig. 6I). The aspect ratio (minor axis as a proportion of major axis) was almost identical in the two groups (0.22 vs. 0.23), indicating that gaps tended to be highly elongated regardless of treatment. Together, these measurements indicate a 61% reduction in venular surface area occupied by gaps after Ad-Ang-1* (0.28% of endothelial surface) compared with the value in the Ad-GFP group (0.71% of endothelial surface; Fig. 6J). Corresponding values for saline-challenged mice were 0.02% and 0.03% (Fig. 6J), which was at the limit of gap detection.

Comparison by linear regression showed that the proportion of vessel wall occupied by gaps had a significant positive correlation with the amount of extravasated 500-nm microspheres on venules (P < 0.0001, correlation coefficient r² = 0.89; Fig. 7).

DISCUSSION

The present study sought to determine the magnitude and cellular mechanism of the antileak effect of Ang-1 in airway inflammation. Our approach was to inject mice intravenously with adenovirus encoding Ang-1* or GFP, which transduced hepatocytes and produced systemic levels of the protein in the bloodstream. Three days later we examined the leakiness of blood vessels in the tracheal mucosa after intravenous challenge with bradykinin. We found that Ang-1* reduced the leakage of Evans blue to baseline and the leakage of 500-nm fluorescent microspheres by 69%. Confocal microscopic analysis revealed that, after Ang-1*, gaps between endothelial cells of mucosal venules were smaller and less numerous. Together, these changes decreased by 61% the fraction of venular surface occupied by endothelial gaps after bradykinin challenge. Ang-1* thus reduces leakage produced by inflammatory mediators in part by decreasing the number and size of gaps that form at intercellular junctions. This antileak action of Ang-1* is probably not limited to the response to bradykinin because similar effects are found with a wide range of inflammatory stimuli (14, 23, 29, 34, 35, 38).

Plasma Leakage in Airways

Bradykinin, a member of the kinin family of vasodilator peptides, was used as the stimulus to test the antileak effect of Ang-1* because it is known to produce rapid but transient plasma extravasation from venules in the respiratory tract (5). This action of bradykinin is mediated via B₂-receptor signaling in endothelial cells. Leakage is restricted to postcapillary venules and is thought to result from the formation of gaps between endothelial cells (1, 12, 17).

To optimize the model for studying the effects of Ang-1*, we first determined the dose response and time course of Evans blue accumulation in the mouse trachea after bradykinin challenge. An intravenous dose of 10 mg/kg of bradykinin nearly doubled the Evans blue content of the trachea. Higher doses produced somewhat more leakage but were lethal in some mice. Dye accumulation was increased at 5 min, still elevated at 30 min, and almost back to baseline by 60 min.

High baseline Evans blue values indicated that appreciable dye extravasated from tracheal blood vessels even without bradykinin. Therefore, we sought to confirm that bradykinin did indeed increase permeability in this system. For this purpose we used fluorescent microspheres that were too large to extravasate without a change in endothelial barrier function. Baseline leakage of 25- to 1,000-nm microspheres was very low. Bradykinin challenge was followed by extensive leakage of microspheres, which was transient, peaked at ∼3 min, and returned to the near-zero baseline by 10 min. By keeping the microsphere circulation time constant at 1 min, it was possible to distinguish transience of leakage from rapid clearance of microspheres from the bloodstream. Because, unlike Evans blue, 500-nm microspheres do not leak from normal tracheal vessels, the increase in microsphere leakage after bradykinin (∼39-fold) was much larger than for Evans blue (∼1.6-fold).

Unlike extravasated Evans blue that moved throughout the tissue, most extravasated microspheres were trapped in the wall of leaky vessels by the endothelial/pericyte basement membrane. This is a well-documented property of particulate tracers (25, 26). Extravasated microspheres thus marked the leaky vessels, which were easily identified as postcapillary venules or collecting venules in the simple architecture of the tracheal microcirculation. Restriction of the leakage to this segment of the vasculature led us to explore the role of endothelial gaps (24–26).

Role of Endothelial Gaps in Plasma Leakage

Multiple mechanisms have been proposed, discussed, and debated to explain the movement of macromolecules across the vascular endothelium (8, 13, 26, 27). Intercellular gaps, transcellular holes, fused endothelial vesicles, vesiculovacuolar organelles, transcytosis, and fenestrations have all been considered potential contributors. The classic studies of Majno and Palade (24) implicated endothelial gaps in plasma leakage from inflamed venules. Results of many subsequent studies are consistent with that view (1, 3, 16, 17). The presence of endothelial gaps at sites of particulate tracer leakage and the extravasation of tracers from inflamed vessels after fixation provide further evidence for openings in the endothelium that permit extravasation driven by hydraulic pressure gradients rather than transcellular transport (26). The present studies showed that microspheres from 25 to 1,000 nm in diameter extravasate after bradykinin challenge. As endothelial gaps are reported to have an average size of ∼500 nm (26), they appear to be a reasonable pathway for extravasation of microspheres in the system we studied.

A conventional method for studying endothelial gaps involves electron microscopy (1, 3, 16, 17, 24). This approach has the attribute of high resolution but may be limited by the small regions sampled. To explore the involvement of endothelial gaps in our system, we developed a confocal microscopic approach for determining the number and size of endothelial gaps in leaky venules of mouse airways. For this purpose, tracheal whole mounts were stained by immunohistochemistry using two different antibodies to the endothelial cell junctional molecule CD31/PECAM-1. Because the hamster anti-mouse CD31 antibody selectively labeled endothelial cell junctions and the rat anti-mouse CD31 antibody uniformly stained nonjunctional endothelial cell plasma membranes, gaps could be seen as unstained holes surrounded by intensely stained cell junctions. Without the rat anti-mouse CD31 antibody, gaps could not always be distinguished from nonjunctional regions of plasma membrane.

Using this approach, we examined endothelial gap formation in bradykinin-induced leakage. Within only 1 min of bradykinin injection, gaps were visible at focal regions of intercellular junctions marked by CD31 staining where endothelial cells appeared to retract from one another. Extravasated microspheres accumulated near gaps.

With its potential for broad sampling, the confocal microscopic approach was more efficient for determining the size, shape, number, and distribution of endothelial gaps than transmission or scanning electron microscopy as used in previous studies (3, 16). The lower resolution of confocal microscopy made gap size measurements not as accurate as with electron microscopy, and counts of gap numbers probably underestimated the actual values because gaps smaller than ∼100 nm may not have been visible. However, the confocal microscopic approach made it possible to measure a large population of gaps with a precision adequate to detect changes in gap number and size under the conditions of our experiments.

After bradykinin, endothelial gaps in airway venules tended to be slitlike, with an average length (980 nm) fivefold the width (200 nm). By 10 min, gaps were no longer visible, and junctions appeared normal. These findings are consistent with what is known about the time course of endothelial gap formation and leakage after exposure to mediators in several models of acute inflammation (3, 16, 25, 26).

Effect of Ang-1* on Number and Size of Endothelial Gaps

Ang-1* completely blocked the accumulation of Evans blue dye and significantly reduced the leakage of 500-nm microspheres in the trachea after bradykinin challenge. Similar results have been obtained in mouse skin, in which adenoviral Ang-1* or transgenically overexpressed Ang-1 inhibits Evans blue leakage induced by local application of mustard oil, VEGF, serotonin, or platelet-activating factor (34, 35). In cultured endothelial cell monolayers, Ang-1 reduces baseline permeability and response to bradykinin, histamine, VEGF, thrombin, and TNF-α (14, 23, 29, 38). It is unknown whether Ang-1 modifies bradykinin receptor expression or its downstream signaling, but the antileakage action against a wide range of mediators favors a more generalized effect on endothelial barrier function.

Confocal microscopic studies revealed that Ang-1* reduced both the number (38% reduction) and the size (38% reduction) of endothelial gaps that formed in airway venules after bradykinin. Together these changes represented a 61% decrease in the area of venular surface occupied by endothelial gaps. This change fit with the 69% decrease in extravasation of 500-nm microspheres.

The restriction in endothelial gap formation by Ang-1* is likely to involve changes in cell-cell junctional molecules. Ang-1 increases expression of the tight junction molecule occludin, decreases phosphorylation of CD31 and VE-cadherin, and increases CD31 at endothelial cell-cell junctions, which may contribute to improved barrier function (14, 18). Ang-1 increases the interaction of VE-cadherin and β-catenin at cell-cell junctions and the activation of PKC-β associated with VEGF-induced increased monolayer permeability (38). Ang-1 also inhibits thrombin-induced activation of PKC-ζ (23).

Further studies will be needed to determine whether the effects on endothelial gaps of adenovirally delivered Ang-1* can be reproduced by administering recombinant Ang-1 in vivo, where Ang-1 stability and pharmacokinetic and pharmacodynamic issues come into play. Future experiments should also address whether Ang-1* limits gap number and size by limiting gap opening or speeding gap closure. Indeed, our studies of gap formation focused on the first minutes after bradykinin with the thought that Ang-1 may hasten gap closure, but the speed of gap opening and closure made this a challenging question to address in vivo with current methods.

Under baseline conditions, junctional molecule signaling helps maintain the quiescent phenotype of endothelial cells (10, 37). Inflammatory mediators change this in part through actions on the cytoskeleton linked to junctional molecules (39, 40). Contraction of cytoskeletal elements seems not to be essential for gap formation. Rather, gaps form by elastic recoil of endothelial cell borders when junctional and cytoskeletal attachments are focally weakened or disrupted (39, 41).

Can changes in number and size of endothelial gaps explain all effects of bradykinin and Ad-Ang-1* on microsphere extravasation? Although participation of vesiculovacuolar organelles or transcellular holes near endothelial cell borders cannot be excluded, Ad-Ang-1* caused similar reductions in amount of venular surface occupied by gaps and in leakage of 500-nm microspheres. Gap number and size, expressed as a proportion of endothelial surface under the various conditions, were strongly correlated with amount of microsphere extravasation (correlation coefficient r² = 0.89). Furthermore, the proportion of venular surface occupied by gaps in our study is in the range of earlier estimates based on different methods (16, 25, 26). Determination of the quantitative relationship between gap number/size and amount of microsphere extravasation will require more information on the contribution of Starling forces to transvascular fluid flux in our system.

The inhibitory action of Ang-1 on endothelial gap formation and plasma leakage in the airways has potential clinical significance because mucosal edema is a common feature of inflammatory diseases such as asthma and chronic bronchitis (5, 42). Ang-1-like therapeutic agents may be clinically beneficial by reducing airflow obstruction associated with airway edema.

GRANTS

This work was supported in part by National Heart, Lung, and Blood Institute Grants HL-59157 and HL-24136.

FOOTNOTES

• 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.