Caveolar and intercellular channels provide major transport pathways of macromolecules across vascular endothelial cells

Authors

  • Kazushige Ogawa,

    Corresponding author
    1. Department of Veterinary Anatomy, Graduate School of Agriculture and Biological Sciences, Osaka Prefecture University, Osaka, Japan
    Current affiliation:
    1. Department of Veterinary Anatomy, Faculty of Agriculture, Iwate University, 3-18-8 Ueda, Morioka, Iwate 020-8550, Japan
    • Department of Veterinary Anatomy, Graduate School of Agriculture and Biological Sciences, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka 599-8531, Japan
    Search for more papers by this author
  • Masaki Imai,

    1. Department of Virology I, National Institute of Infectious Diseases, Tokyo, Japan
    Search for more papers by this author
  • Tokiko Ogawa,

    1. Department of Pathology, School of Medicine, Iwate Medical University, Iwate, Japan
    Search for more papers by this author
  • Yasuhiro Tsukamoto,

    1. Department of Veterinary Anatomy, Graduate School of Agriculture and Biological Sciences, Osaka Prefecture University, Osaka, Japan
    Search for more papers by this author
  • Fumihiko Sasaki

    1. Department of Veterinary Anatomy, Graduate School of Agriculture and Biological Sciences, Osaka Prefecture University, Osaka, Japan
    Search for more papers by this author

Abstract

Serum macromolecules are transported through the vascular endothelial layer to the interstitium via the caveolae and interendothelial clefts, but the nature of the permeability of these structures is unknown, and the manner of caveola-vesicle transport is controversial. We have developed a method of detecting macromolecular channels using an in situ HRP perfusion into arteries previously perfused with aldehyde and random conventional sectioning for electron microscopy. Using unbiased morphometry, 4.75% of the abluminal caveolae and 15.13% of the intercellular clefts were the tracer-positive in rat aortic endothelium. In rat aortas treated with N-ethylmaleimide, all caveolae and most free vesicles in the cytoplasm except those around the Golgi area were HRP-positive in the endothelial cells; 1.48% of abluminal caveolae were structurally recognized as caveolar channels through the endothelial layer in a plane of single section. The length density of the abluminal caveolae was decreased to about 80% to the physiological control level whereas the larger invaginations were more frequently observed. Moreover 96.17% of the intercellular clefts were HRP-positive. We suggest that a flexible channel-system functions extensively as a macromolecular transport pathway in the arterial endothelium in vivo because the tracer-labeled abluminal caveolae and intercellular clefts should be opened to the luminal surfaces methodologically. We therefore propose that caveolar channels, rather than transcytosis, provide a mechanism of caveola-vesicle transport in the endothelial cells, because free vesicles involved in transcytosis were few in number. Anat Rec 264:32–42, 2001. © 2001 Wiley-Liss, Inc.

Caveolae have been implicated in transcytosis (Oh et al., 1998; Predescu et al., 1994, 1997; Predescu and Palade, 1993; Schnitzer et al., 1994, 1995a, 1996; Simionescu, 1983), endocytosis (Oh et al., 1998; Schnitzer et al., 1994, 1995a, 1996; Simionescu, 1983), potocytosis (Anderson, 1993; Anderson et al., 1992) and signal transduction (Fujimoto, 1993; Fujimoto et al., 1992; Garcia Cardena et al., 1996; Liu et al., 1997; Schnitzer et al., 1995b) and highlighted in cellular transport in recent years. The former two functions imply a mobile entity as vesicular carriers, whereas the latter two imply a sessile entity, and therefore the functions of caveolae remain controversial (Severs, 1988; van Deurs et al., 1993). Caveolae are 50–90 nm, distinctive omega-shaped invaginations of the plasma membrane present in most cells. They are also known as plasmalemmal vesicles in vascular endothelial cells, where they are most abundant, especially in continuous endothelial cell types. Their functioning in vivo has been most extensively studied in continuous endothelial cells.

The mediation of water and solutes between the circulating blood plasma and the interstitial fluid is recognized as the most important endothelial function. The current findings indicate that vesicular and intercellular systems serve as transport pathways of macromolecules, although the transport mechanism of these systems remains unclear (Clough, 1991; Michel, 1992; Michel and Curry, 1999). Numerous morphological studies have indicated tracers within the luminal and abluminal caveolae and free noncoated vesicles in the cytoplasm of vascular endothelial cells, suggesting caveolae in the transcytosis of macromolecules (Predescu et al., 1994, 1997; Predescu and Palade, 1993; Simionescu, 1983). A “shuttle” mechanism of caveola-vesicle transport (transcytosis) has been proposed; this is the best known and most accepted mechanism, in which vesicles shuttle between the luminal and abluminal plasma membranes by means of fusion and fission events. Three-dimensional analyses based on ultrathin serial sectioning for electron microscopy have revealed that about 99% of the noncoated vesicles are caveolae, i.e., the vesicles open to the cell surfaces (Bundgaard, 1983; Bundgaard et al., 1983; Frøkjer-Jensen et al., 1988; Ogawa and Taniguchi, 1993), whereas conventional sections misleadingly display about 50% of the noncoated vesicle population free in the cytoplasm (Bundgaard, 1983; Bundgaard et al., 1983; Ogawa and Taniguchi, 1993). A three-dimensional analysis makes it possible to follow the vesicular continuity because the diameter of vesicles is equal to the thickness of conventional sections (Bundgaard, 1983), and hence caveolae in endothelial cells are assumed not to be freely mobile in the cytoplasm (Bundgaard, 1983; Bundgaard et al., 1979, 1983; Frøkjer-Jensen et al., 1988; Ogawa and Taniguchi, 1993). Physiological measurements of the permeability properties of continuous endothelium revealed that the endothelial layer serves as a highly selective barrier, and it is regarded as having the same function as an artificial membrane with “small and large pores,” 5–8 nm and 20–25 nm in radius (Taylor and Granger, 1984). This may indicate that the transport pathways for small and large molecules are water-filled channels present in the continuous endothelial cells. We believe that a channel system composed of caveolae through the cells, an alternative mechanism of caveola-vesicle transport, operates extensively for the transport of macromolecules in vascular endothelial cells.

Ultrathin serial sectioning requires excellent skills, and the three-dimensional analysis is time consuming. In addition to these problems, the outcome of such investigations represents only a restricted endothelial segment due to the technical limitations, and therefore previous studies with serial sectioning could not always reveal the vesicular channels successfully (Bundgaard et al., 1983; Frøkjer-Jensen et al., 1988). A simple and reliable method has been sought for the elucidation of the transendothelial transport of macromolecules in vascular endothelial cells. We have developed a new method using in situ perfusion and random conventional sectioning for electron microscopy to detect the channel system. The method is based upon the assumption: if a tracer is perfused through fixed blood vessels in which hydraulically conductive pores through the endothelial cells are faithfully preserved by aldehyde perfusion, the tracer-labeled abluminal caveolae and intercellular clefts indicate transendothelial channels because they should be opened to the luminal surfaces methodologically. The assumption is partially supported by the previous physiological study that showed that aldehyde fixation does not cause changes in the permeability properties of macromolecules in the endothelium (Haraldsson and Johansson, 1985). This study demonstrated that aldehyde fixation of the vascular bed caused reductions of the diffusion and filtration capacities and of albumin clearance in proportion to the reduction of the capillary surface area available for exchange. To test the assumption, we designed the experiment described herein, consequently revealing the caveolar and intercellular channels successfully.

We applied our method to the rat aortic endothelium because the obtained results: 1) can be compared with our previous studies of serial sectioning (Ogawa and Taniguchi, 1993; Ogawa et al., 1993); and 2) can be the basis for the elucidation of a close correlation between abnormal increases in endothelial permeability and atherogenesis (Ross, 1986, 1993). We employed horseradish peroxidase (HRP) in a fluid-phase probe labeling both “small and large pores” due to its small radius of 2.98 nm (Simionescu, 1983)

MATERIALS AND METHODS

Materials

The studies were performed on Cij Wistar male rats weighing 320–360 g kept under standard housing and feeding conditions. Animal experimentation protocols were approved by the university's Animal Research Committee.

In Situ Perfusion

Under sodium pentobarbital anesthesia, the proximal part of the thoracic aorta was cannulated with a needle catheter connected to a peristaltic pump to control the perfusion rate. After flushing the vasculature free of blood with a 10-min perfusion of Hanks' balanced salt solution buffered with 10 mM HEPES and supplemented with 3 g/l glucose (oxygenated and prewarmed to 37°C), the aorta and its continued artery were perfused with a mixture of 2% paraformaldehyde (FA) and 0.5% glutaraldehyde (GLA) buffered with 0.1 M phosphate at a flow rate of 0.7 ml/min for 50 minutes. After the fixative was washed out with the same buffer, the aorta was perfused with 3 mg/ml horseradish peroxidase (HRP: type II, Sigma Chemical Co., St. Louis, MO) dissolved in 0.1 M phosphate at a hydrostatic pressure of 5.5 mm Hg with a flow rate of 0.4 ml/min for 12 min. Control rats were perfused with the same buffer without HRP to evaluate the specificity of reaction products.

N-Ethylmaleimide Treatment

The thoracic aorta was perfused with Hanks' balanced salt solution containing 0.1 mM N-ethylmaleimide (NEM, Sigma) for 10 min under the same conditions described above before the perfusion fixation. The subsequent procedure was the same as that used for the physiological controls.

Microscopy

The thoracic aorta was removed carefully and immersed in 2% FA and 1% GLA buffered with 0.1 M cacodylate for 60 minutes to fix the tracer and the aorta. After a diaminobenzidine reaction to visualize the tracer, the aorta was rinsed, postfixed with 2% OsO4 and routinely embedded in epoxy resin. Semithin cross-sections of the aorta were viewed with a differential-interference-contrast microscope. Random cross-sections of aortic endothelia, 80 nm in thickness, were cut with a diamond knife mounted on an Reichert Ultra-Cut ultramicrotome (Heidelberg, Germany), stained with or without lead citrate and viewed under an electron microscope (JEM-1010, JEOL, Akishima, Japan) at an acceleration voltage of 60 KV. We also examined the conventional ultrastructure of the NEN-treated and untreated aortas without the tracer perfusion.

Morphometry

Five aortas each were used for the morphological measurements of the physiological control and NEM treatment groups, respectively. Five blocks were randomly chosen from each aorta for sectioning. From each block, 10 fields of an endothelial layer on a cross-section were photographed randomly on 8.2 × 11.8 cm film at a magnification of 9,760×. The magnification on the electron microscope was calibrated with a carbon replica of an optical grating (2,160 lines per mm). In each group, a total of 250 micrographs printed at magnifications of 39,040 were used to measure the number of abluminal caveolae, the number of interendothelial junctions and the length of the abluminal plasma membranes. The length was measured on an image processing system photoanalyzer (SPICCA-II, Nippon Avionics Co., Tokyo, Japan). Data were analyzed by Student's t-test with differences considered significant at the P < 0.005 level.

RESULTS

HRP Localization in Control Endothelial Cells

When the perfusion fixation was enough for fixing the whole layers of the aorta and the hydrostatic pressure was adequately held by the fixative and the tracer perfusion, HRP was located on the luminal border of the endothelial cells under a light microscope (Fig. 1a).

Figure 1.

Micrographs by Nomarski optics. a: Rat aorta perfused with the fixative followed by HRP as a tracer. The tracer was localized on the luminal border of the endothelium. b: Rat aorta perfused with 0.1 mM NEM before perfusion fixation. The tracer was accumulated not only in the subendothelial space but also in the media of the luminal side, suggesting the remarkable increase in the transendothelial permeability for macromolecules by NEM. Scale bar = 25 μm.

It was difficult precisely to distinguish plasmalemmal vesicles (vesicles in the transendothelial transport system) from coated vesicle-associated vesicles (vesicles having shed their clathrin coat), and in some cases HRP-positive caveolae (plasmalemmal vesicles open to the vessel lumen or ablumen) from HRP-positive coated pits under an electron microscope. Because coated vesicles/pits usually appear as 100–150 nm vesicles/vesicular invaginations of the plasma membrane (Pearse and Robinson, 1990) and caveolae as 50–90 nm omega-shaped membrane invaginations (Ogawa and Taniguchi, 1993; Rothberg et al., 1992), we therefore defined caveolae as omega-shaped invaginations of the plasma membrane less than 100 nm in the diameter in a plane of single section in the present study. Distinctive coated pits with bristles less than 100 nm in the diameter were properly excluded from that population. We also defined caveolar channels as HRP-positive caveolae that simultaneously open to the vessel lumen and ablumen directly or indirectly in a plane of single section.

Under an electron microscope, HRP was observed on the external side of luminal membranes, in free vesicles and all luminal caveolae (Fig. 2). We observed that a few abluminal caveolae labeled by HRP in the peripheral thin region of endothelial cells (Fig. 2a). They were usually isolated in a plane of a single section. HRP-negative abluminal caveolae overlapped HRP-positive luminal caveolae were occasionally observed (Fig. 2b). This shows that the luminal caveolae were contacted but not fused to the abluminal caveolae, and thus they were tracer-negative. Intensive observation revealed the caveolar channels labeled with HRP through the endothelium in a plane of a single section (Fig. 2c). In addition, HRP-positive abluminal caveolae were located also in the central thick region of the endothelium (Fig. 2d). We observed the HRP-positive endosome-like structures in the central thick region (Fig. 2e, 2f).

Figure 2.

Aortic endothelia perfused with the fixative followed by HRP as a tracer. a–c: Peripheral thin regions. a: Note tracer-positive abluminal caveolae probably opening to the lumen in another plane of the sections. b: The tracer-positive luminal caveolae are contacted but not fused to the abluminal caveolae (arrowheads), and thus they are tracer negative. c: A transendothelial channel composed of the fused caveolae. Scale bar = 0.5 μm. d–f: Central thick regions. d: Tracer-positive abluminal caveolae. e,f: Tracer-positive endosome-like structures and a cluster of fused vesicles probably opening to the lumen in another plane of the sections, thus showing invaginations of the luminal plasma membranes. Note the deep invagination extended beneath the abluminal plasma membranes under the nucleus in (f). Micrographs (e, f) without staining of lead citrate. Scale bar = 0.5 μm.

HRP stopped at intercellular junctions in most cases (Fig. 3a), but was occasionally through junctions and accumulated in the intercellular clefts accordingly seen in Figure 3b. In most cases HRP retained in the HRP-positive abluminal caveolae and intercellular clefts without its diffusion to the subendothelial space in the physiological control.

Figure 3.

Intercellular clefts between aortic endothelial cells perfused with the fixative followed by HRP as a tracer. a: The intercellular cleft is impermeable to the tracer: HRP stops at the intercellular junction. b: HRP is through the intercellular junction and accumulates in the whole intercellular cleft. HRP retains in the intercellular cleft: HRP scarcely diffuses from the exit to the subendothelial space. Arrowhead, coated vesicle; Scale bar = 0.5 μm.

HRP Localization in NEM-Treated Endothelial Cells

In the NEM-treated rats, HRP was accumulated in the subendothelial space, and in the upper part of the tunica media under a light microscope (Fig. 1b).

NEM treatment affected the ultrastructure of the aortic endothelial cells. The endosome/vacuole-like membrane-bounded structures were frequently observed (Fig. 4a) and the peripheral region of the endothelium occasionally appeared in a thin wall like that of capillary endothelium. Large gaps in the intercellular clefts were also occasionally observed (Fig. 4b) and definitive intercellular junctions including tight junctions occasionally disappeared in the NEM-treated endothelial cells.

Figure 4.

NEM-treated aortic endothelial cells without HRP perfusion. a: The vacuole-like membrane-bounded structures containing almost the same substance in appearance as the subendothelial matrix thus suggesting invaginations of the abluminal membrane. b: Note that large gaps in the intercellular cleft, and a vacuole-like structure without substances in its content, suggesting a invagination of the luminal membrane. Scale bar = 0.5 μm.

HRP was observed in all luminal caveolae, all abluminal caveolae and most free vesicles in the cytoplasm except those around the Golgi area in the NEM-treated aortic endothelial cells (Figs. 5,6). HRP-positive caveolar channels appeared in a plane of single section in a low frequency (Fig. 5a). HRP-positive endosome-like structures were frequently observed in NEM-treated endothelial cells. In the serial micrographs of Figure 5b,c, the endosome-like structure (Fig. 5b) opening the abluminal space (Fig. 5c), i.e., the invaginations of abluminal membranes simultaneously opened to the luminal space via a caveola, and thus the transendothelial channel composed of membrane invagination and caveola was recognized. Figure 6a showed a gradient in the density of the accumulated HRP in the subendothelial space. The density was higher beneath the caveolar channel-like structure suggesting the high permeability of HRP in that endothelial region.

Figure 5.

Aortic endothelial cells treated with NEM. The tracer accumulated in the subendothelial space showing a dramatic increase in its permeability. a: A transendothelial channel made up of a single caveola simultaneously opening to the lumen and subendothelial space (arrow) and an intercellular junction permeable to the tracer. Note that only two vesicles (arrowheads) are tracer-negative out of the total of 85 caveolae/plasmalemmal vesicles in the micrograph. b,c: Serial electron micrographs showing a transendothelial channel composed of caveolae and large invaginations. The tracer-positive endosome-like structure in (b) is simultaneously connected with a luminal caveola (arrow) and abluminal plasma membranes in (c). Scale bar = 0.5 μm.

Figure 6.

Aortic endothelial cells treated with NEM. a: The density of accumulated HRP in the subendothelial space shows the gradient: the density is higher beneath the caveolar channel-like structure, suggesting the high permeability of HRP in that endothelial region. b: HRP is accumulated in the intercellular cleft where a large gap appears. c: The junction stops HRP though the intercellular cleft whereas the exit part of the intercellular cleft is HRP-positive probably due to the diffusion of HRP from the subendothelial space. Arrowhead, coated vesicle; Scale bar = 0.5 μm.

HRP was through intercellular junctions in most cases in NEM-treated endothelial cells and was accumulated accordingly in the intercellular clefts showing normal ultrastructure in appearance (Fig. 5a) and occasionally in the large gaps of the intercellular clefts (Fig. 6b). In a few cases, HRP stopped at the junctions whereas the exit part of the intercellular clefts were HRP-positive probably due to the diffusion of HRP from the subendothelial space (Fig. 6c).

Morphometry

To investigate the frequency of caveolar channels, we counted the number of abluminal caveolae; 8,772 and 9,136 total in physiological control and NEM-treated endothelial cells, respectively. We also measured the length of abluminal plasma membranes, 3,703.26 μm and 4,897.34 μm in total in control and NEM-treated endothelial cells, respectively. The length densities of abluminal caveolae were 2.38 ± 0.17 (number/1 μm abluminal plasma membrane, mean ± SD) in physiological control and 1.87 ± 0.22 in NEM-treated endothelial cells (Table 1). There was a significant difference between the two (P = 0.003, t-test). We regarded HRP-positive abluminal caveolae as transendothelial channels because the tracer was perfused after perfusion fixation of the aortas, and then HRP was located in abluminal caveolae. HRP-positive abluminal caveolae would be expected to open to the vascular lumen via other caveolae or the invaginations of plasma membranes in another plane of section. HRP-positive abluminal caveolae methodologically inferring caveolar channels were 4.75 ± 0.84% (mean ± SD) of abluminal caveolae in physiological control. Their length density was 0.11 ± 0.01 (number/1 μm abluminal plasma membranes, mean ± SD). In contrast, HRP-positive abluminal caveolae were 100% in NEM-treated endothelial cells. Moreover 1.48 ± 0.18% (mean ± SD) of abluminal caveolae were structurally recognized as caveolar channels in a plane of single section in NEM-treated endothelial cells, whereas no caveolar channels were present in a plane of single section in physiological control in the samples analyzed morphometrically.

Table 1. Frequencies and length densities of HRP-positive abluminal caveolae and intercellular channels*
 ControlNEMP-value
  • *

    Five aortas each were used for the morphological measurements of the control and NEM treatment groups. Values of the length density and frequency are mean ±SD expressed as number per 1 or 100 μm and percentage, respectively.

  • a

    We defined caveolae as membrane-bounded rounded or elliptic structures less than 100 nm in the diameter that open to the vessel lumen or ablumen in a plane of single section.

  • b

    We defined caveolar channels as HRP-positive caveolae simultaneously open to the vessel lumen and ablumen directly or indirectly in a plane of single section.

  • c

    We defined intercellular channels as HRP-positive intercellular clefts where HRP is present in the whole intercellular space continuously.

Total number of abluminal caveolaea counted (n)87729136
Total number of intercellular clefts counted (n)272308
Total length of abluminal plasma membrane measured (μm)3703.264897.34
Length density of abluminal caveolae (n/μm)2.38 ± 0.171.87 ± 0.220.0033
Frequency of HRP-positive abluminal caveolae (%)4.75 ± 0.84100<0.0001
Frequency of caveolar channelsb (%)01.48 ± 0.18<0.0001
Length density of HRP-positive abluminal caveolae (n/μm)0.11 ± 0.011.87 ± 0.22<0.0001
Length density of intracellular clefts (n/100 μm)7.38 ± 1.346.32 ± 0.710.1574
Frequency of intercellular channelsc (%)15.13 ± 1.5796.17 ± 2.17<0.0001
Length density of intercellular channels (n/100 μm)1.11 ± 0.186.09 ± 0.58<0.0001

To investigate the frequency of intercellular channels, we counted the number of intercellular clefts, 272 and 308 total in physiological control and NEM-treated endothelial cells, respectively (Table 1). The length densities of intercellular clefts were 7.38 ± 1.34 (number/100 μm abluminal plasma membrane, mean ± SD) in physiological control and 6.32 ± 0.71 in NEM-treated endothelial cells. There was no significant difference between the two (P = 0.157, t-test). Interendothelial clefts can be divided into two regions (Predescu and Palade, 1993), luminal intercellular cleft (LIC, the intercellular space between the lumen and junctional elements) and abluminal intercellular cleft (AIC, the intercellular space between the junction proper and exit into the subendothelial spaces). We defined HRP-positive intercellular clefts where HRP is present in the whole intercellular space from LIC though junctional elements to AIC continuously as intercellular channels in the present study. We did not count discontinuously HRP-positive intercellular clefts seen on Figure 6c as intercellular channels because AIC was HRP-positive probably due to the diffusion of HRP from the subendothelial space. In most cases HRP was restricted in LIC in physiological control; intercellular channels occurred at the frequency of 15.13 ± 1.57% (mean ± SD, Table 1). Their length density was 1.11 ± 0.18 per 100 μm of abluminal plasma membranes (mean ± SD). On the other hand in NEM-treated endothelial cells, intercellular channels was present at the frequency of 96.17 ± 2.17% (mean ± SD) and at the density of 6.09 ± 0.58 per 100 μm of abluminal plasma membranes. A significant difference of the frequency or the length density was recognized between physiological control and NEM-treated endothelial cells (P < 0.0001, t-test).

DISCUSSION

Methodological Validity

Because caveolae/plasmalemmal vesicles in vascular endothelium have been assumed to be engaged in the transendothelial transport of molecules, their arrangement and property in the endothelium are important to elucidate the transport mechanism in detail. For the morphological observation of the caveolae/vesicles, chemical fixation (aldehyde fixation) or cryofixation (rapid freezing and freeze substitution) is indispensable at present. Studies proposed that only cryofixation could deduce the precise arrangement of caveolae/vesicles in the endothelium (Casley-Smith, 1981; Mazzone and Kornblau, 1981). Cryofixation, however, did not change the proportion of free vesicles to vesicular invaginations (caveolae) in conventional thin sections of aldehyde-fixed endothelium (Wagner and Andrews, 1985). Frøkjaer-Jensen et al. (1988) and Noguchi et al. (1987) showed by the use of cryofixation and ultrathin serial sectioning into three-dimensional reconstruction of vesicles that more than 98% of the vesicles connected directly or indirectly with the surface of the capillary endothelium. These results are in agreement with those from the three-dimensional reconstruction of vesicles in the aldehyde-fixed endothelium (Bundgaard et al., 1983; Coomber and Stewart, 1986; Frøkjaer-Jensen, 1980, 1984). In these respects, we think that aldehyde-fixation could closely preserve the arrangement of caveolae/vesicles in the aortic endothelium in vivo. When applied to aortic endothelium, cryofixation may lead to morphological artifacts during the procedures of tissue preparation before cryofixation because the aortic endothelium located on the surface of the vessel has a high risk of becoming dry and can be damaged mechanically. It is necessary to make the luminal surface of aortic endothelium flat for excellent cryofixation by metal-contact freezing. We believe that cryofixation may be generally unsuitable for fixation of the aortic endothelium.

We found that the perfusion fixation for fixing the whole aorta before the tracer perfusion was essential for the proper localization of HRP on the endothelial cells. The perfusion fixation for a few minuets is enough for fixing the endothelial layer. When we employed the shorter fixation, however, HRP was frequently located in the whole cytoplasm of endothelial cells diffusely. The partially fixed aortas combine certain hardness in a fixed inner layer and certain flexibility in an unfixed outer layer. We suggest that mechanical damages of endothelial cell membranes may occur easily in the fixed endothelial cells of the partially fixed aortas during the tracer perfusion followed by the tissues dissection. We preferred tissue dissection after the tracer perfusion followed by immersion fixation because perfusion fixation after the tracer perfusion tended to wash out HRP from the lumen, LIC and luminal caveolae (Florey and Sheppard, 1970; Hüttner et al., 1973; Ogawa, unpublished data).

When the hydrostatic pressure was adequately held by the fixative and the tracer perfusion, the HRP was clearly diffused into the places outside of endothelial cells such as luminal openings of caveolae and intercellular clefts. This implies that HRP has a molecular dimension suitable for the detection of the endothelial channels by its perfusion after fixation. Under such conditions, a few abluminal caveolae were labeled by HRP, although they were generally isolated in a plane of a single section. These HRP-positive caveolae may indicate transendothelial channels, because HRP-positive abluminal caveolae would be expected to open to the vascular lumen via other caveolae in another plane of sections methodologically.

Caveola-Vesicle Transport

HRP-positive abluminal caveolae methodologically inferring caveolar channels were 4.75% of the abluminal caveolae. This value is much higher than we expected. Our previous study using three-dimensional reconstruction from ultrathin serial sections showed that caveolar channels appear in 1.71% of abluminal caveolae in the rat aortic endothelial cells (Ogawa and Taniguchi, 1993). The present data reflect the frequency much closer to a physiological condition because the earlier study excluded all channels having ambiguous entities in vesicular continuity due to the section thickness of 25 nm, which was far from the ideal thickness (∼15 nm) (Bundgaard, 1983). Of the total HRP-positive abluminal caveolae (0.1% of the total abluminal caveolae) counted, 2.2% were located near by the intercellular clefts where HRP was present in the whole intercellular space continuously, that is, intercellular channels. This may indicate that some abluminal caveolae became HRP-positive due to the diffusion of HRP from the exit of the intercellular clefts not from the lumen. We may slightly overestimate the frequency of caveolar channels in the present study.

In situ studies have indicated that NEM significantly inhibits macromolecular transport by caveolae in the capillary endothelial cells of the myocardium (Predescu et al., 1994, 1997). Recent studies have also shown that a caveolin-rich endothelial membrane-fraction from rat lung tissue fissions into vesicles under certain conditions (Oh et al., 1998; Schnitzer et al., 1996), and contains NSF, SNAP and vSNARE, molecules constituting the machinery for the fission, docking, and fusion of vesicles (Schnitzer et al., 1995a). The authors postulated that caveolae are involved in transcytosis on the vascular endothelium according to the SNARE mechanism (Rothman, 1994; Rothman and Orci, 1992), although they have not demonstrated a considerable number of free vesicles in the cytoplasm, a prerequisite for a transcytotic mechanism. Thus, we examined the effect of NEM on the tracer labeling.

The present results indicated the opposite effect by NEM. NEM caused a significant increase in the endothelial permeability to the macromolecule as seen in Figures 5 and 6. In the NEM-treated rats, HRP was accumulated in the subendothelial space. Not only all luminal caveolae but also all abluminal caveolae and most vesicles free in the cytoplasm (except those around the Golgi area) were HRP-positive. Moreover, we could observe caveolar channels in a plane of single section much more occasionally in NEM-treated endothelial cells than those in physiological controls. These results may suggest that NEM induces the de novo formation of caveolar channels. If caveolar channels are transient entities, NEM may promote to extend the life span of the channels. We occasionally observed that HRP was concentrated more in the subendothelial space beneath caveolar channels and channel-like structures in the endothelium treated with NEM (Fig. 6a) whereas HRP usually retained in HRP-positive abluminal caveolae without its diffusion to the subendothelial space in the physiological control. Thus, NEM may also promote the permeability of caveolar channels. Because the HRP-density in the subendothelium was highly correlated with that in the neighboring caveolae, we think that a considerable number of abluminal caveolae became HRP-positive due to the diffusion of HRP from the subendothelium, not from the lumen through caveolar channels. In such a case or not, HRP-positive vesicles free in a plane of a single section imply the entity of caveolae, and consequently most vesicles were caveolae in the NEM-treated endothelial cells. If transcytosis occurs in vascular endothelial cells according to the SNARE mechanism (Rothman, 1994; Rothman and Orci, 1992), NEM makes free vesicles to increase in number or to be stable because of its inhibitory effect on vesicle fusion to the plasma membranes. The present data therefore imply that the transcytosis supported by the SNARE mechanism is unsuitable for the caveola-vesicle transport in the endothelial cells, because transcytosis requires many free vesicles. We suggest that caveolar channels function extensively in vascular endothelial cells because a considerable number of these channels were also observed in the capillary, arteriolar, and venular endothelial cells of the rat myocardium (Ogawa, unpublished data). Vascular biologists may consider caveolar channels to be the principal mechanism of macromolecular transport via caveolae in vascular endothelial cells.

HRP-positive abluminal caveolae, i.e., methodologically inferring caveolar channels were located not only in the peripheral thin region but also in the central thick region of the endothelial wall. Previous studies have shown all of the transendothelial channels in the peripheral region of capillary (Simionescu, 1983) and arterial (Ogawa and Taniguchi, 1993; Ogawa et al., 1993) endothelial cells that are characterized by large vesicle populations. To our knowledge, no investigators of vascular biology have ever previously described transcytosis and caveolar channels in the central thick region of the endothelium, probably because physical hindrances by organelles or the nucleus and longer distances between the lumen and ablumen exist in this region. It is thus understandable that caveolae have been assumed to function in endocytosis, potocytosis, or signal transduction at least in the central region. We found large invaginations associated with caveolae forming transendothelial channels in the NEM-treated serial sections. We may, therefore, reasonably conclude that the caveolar channels occur everywhere in the endothelial cells and in complicated profiles associated with large invaginations in their formation in the thick region. Recent transport studies have indicated that vesicle-tubule or vesicle-vacuole structures are responsible for the macromolecular transport in tumor-associated microvascular endothelial cells (Dvorak et al., 1996; Feng et al., 1999; Kohn et al., 1992) and VEGF-, histamine- or serotonin-treated venular endothelial cells (Feng et al., 1996, 1999). In this study and in a previous study (Ogawa et al., 1993), we observed HRP-positive deep invaginations. Most of them appeared to be tubules and large vesicles or vacuoles similar to endosomes in a plane of a single section, probably opening to the vascular lumen out of the section plane. We suggest that deep invaginations are common to vascular endothelial cells, and appear usually as vesicle-tubule or vesicle-vacuole structures like endosomes in a random conventional section under electron microscopy.

Using unbiased morphometry, we demonstrated a significant difference in the length density of abluminal caveolae between NEM-treated and physiological control endothelial cells. These data are indicative of a specific effect of NEM on the population of abluminal caveolae. Concomitantly with the decrease, the NEM treatment increased virtually the entire population of the larger invaginations in the endothelial cells, although we did not measure their density morphometrically. We may suggest from these data that NEM changes some caveolae into larger invaginations in arterial endothelial cells, but further studies are needed to elucidate our suggestion.

Intercellular Cleft Transport

It is well accepted that the discontinuous properties of the tight junctions are responsible for the formation of paracellular channels, and that interendothelial clefts are thus permeable to macromolecules due to the bypassing of the tight junction in the physiological condition (Crone and Levitt, 1984; Michel and Curry, 1999).

HRP diffusion probably occurred from the subendothelium to the exit of the intercellular clefts in the NEM-treated aortas because HRP was accumulated in the subendothelial space. In this case, intercellular clefts may show discontinuously HRP-positive as seen in Figure 6c. Because these intercellular clefts were excluded from the population of intercellular channels, we probably identified intercellular channels properly. Using unbiased morphometry, HRP-positive AIC occurred at the frequency of 15.1% in the physiological controls, and at 96.2% in the NEM-treated animals, and there was a significant difference between the two. In the control, HRP usually retained in intercellular channels without its diffusion from the exit to the subendothelial space whereas HRP was accumulated in the subendothelial spaces around intercellular channels as well as the other endothelial spaces in the NEM-treated aortas. These may indicate that: 1) paracellular channels function extensively; 2) they are a flexible entity in arterial endothelium; and 3) NEM promote the permeability of intercellular channels. Previous studies have shown no correlation between NEM treatment and the paracellular permeability to macromolecules in the capillary endothelium (Predescu et al., 1994, 1997). NEM is a thioalkylating reagent that has broad effects on cellular activities in addition to the inhibition of exocytic and endocytic vesicular transport. It was found that the disorganization of perijunctional F-actin caused the striking fragmentation of tight junctional distribution between the endothelial cells (Blum et al., 1997), and NEM indirectly decreased the cell-marginal F-actin via the inhibition of ADP-ribosylation on certain G proteins (Wu et al., 1992). We suspect that NEM indirectly modulates the endothelial permeability via the disordering of tight junction integrity, because tight junctions occasionally disappeared in the NEM-treated endothelial cells. We occasionally observed large gaps in the intercellular clefts of the NEM-treated endothelium as seen in Figure 4b and 6b. There is a hypothesis stating that a fiber matrix in the intercellular clefts of capillary endothelium retains their low permeability of macromolecules (Michel and Curry, 1999). NEM may also break the integrity of the fiber matrix and promote the permeability of HRP accordingly. Further studies are necessary to elucidate the effect of NEM on the permeability of intercellular clefts in aortic endothelium.

In conclusion, our method successfully revealed the caveolar and intercellular channels in relatively high frequencies, the de novo formation of the channels due to the application of a certain reagent, and few free vesicles in the arterial endothelial cells. Based on these data, we propose that caveolar channels rather than transcytosis provide the mechanism of caveola-vesicle transport, and that a flexible channel system functions extensively for macromolecular transport in vascular endothelium.

Acknowledgements

We thank Dr. T. Fujimoto for critically reading the manuscript. Supported by grants from the Japanese Ministry of Education, Science and Culture (09660313) to K. Ogawa, and from the Hokusuikai Foundation to K. Ogawa.

Ancillary