Immunoelectron characterisation of the inter-endothelial junctions of human term placenta

One of the main pathways for transfer of water and hydrophilic solutes across continuous non-brain capillaries is the paracellular cleft. A typical endothelial cleft shows one or more zonulae occludentes or tight junctions interspersed with wide zones of uniform width. The tight junctional regions have been shown to be discontinuous along the axial length of the capillary and contain a separation of 4 nm at the point of apposition of the plasma membranes of adjacent cells (Karnovsky, 1967; Firth et al., 1983; Bundgaard, 1984; Ward et al., 1988; Leach and Firth, 1992). The wide zones are thought to contain a glycocalyx-like extracellular matrix which, in series with the tight junctional regions, determines the selectivity of the cleft (Curry and Michel, 1980). Little information is currently available as to which extracellular components are present within the cleft and how they are connected with the membranes. Three-dimensional modelling of the cleft geometry has led workers to postulate the existence of bridging molecules within the cleft in order to maintain uniform width in the face of changing pressure (Hsiung and Skalak, 1984; Silberberg, 1988). The existence of such “linkers” spanning the clefts in the wide zones has been reported (Firth et al., 1983; Leach and Firth, 1992; Schulze and Firth, 1992), but the nature of their component molecules remains unknown. Knowledge of the molecular composition of endothelial junctions is required to understand the organisation of paracellular junctions and how the paracellular cleft provides a permeo-selective barrier between the blood and the underlying tissue.

So far, a few integral membrane proteins have been described in the paracellular cleft regions of endothelial cells. PECAM-1 (Newman et al., 1990), also called CD31 (Simmons et al., 1990; Muller et al., 1989) or endo-CAM (Albelda et al., 1990); the integrin heterodimers α₅β₁ and α₂β₁ (Lampugnani et al., 1991); V-cadherin (Heimark et al., 1990), N-cadherin/A-CAM (Volk and Geiger, 1984) and pan-cadherin (Geiger, 1990) and most recently an endothelial-specific cadherin called cadherin-5 (Lampugnani et al., 1992) have been identified. Endothelial cadherins, which display Ca²⁺-dependent homophilic binding specificity, are thought to be components of the endothelial plaque or adherens-type junction situated in the wide zones of the paracellular cleft (Heimark et al., 1990). Peripheral constituents of this junction are plakoglobin, vinculin, α-actinin and actin microfilaments (Franke et al., 1988; Magee and Buxton, 1991). Cadherins are therefore thought to be essential for the formation of cell-cell associations (Takeichi, 1991) and regulators of the permeability properties of the vasculature (Lampugnani et al., 1992). Integrins, which usually mediate cell-matrix contacts, have also been shown to mediate intercellular adhesion between endothelial cells; the integrins α₅β₁ and α₂β₁, but not other members of the β₁ subfamily, have been located at cell-cell contact sites (Lampugnani et al., 1991). These authors have also shown that integrins possessing the α_v chain, with the exception of the heterodimer α_vβ₃, are present at cell-cell borders of umbilical vein endothelial cell monolayers and in explanted islets of umbilical vein endothelium. PECAM-1 is an integral membrane glycoprotein found on the surface of platelets, at endothelial intercellular junctions in culture, and on cells of myeloid lineage. It too has been implicated in formation of cell-cell junctions, since cells transfected by PECAM-1 demonstrate calcium-dependent aggregation and possess junctions highly enriched in PECAM-1 (Albelda et al., 1991).

To ascertain whether these molecules are present in the paracellular clefts of continuous non-brain capillaries, we have used an immunoelectron histochemical method applied to ultrathin frozen sections of the term human placenta. Antibodies against vinculin and α-actinin were also used to determine the nature of the intercellular junctions. Double immunolabelling with anti-pan-cadherin and cadherin-5 antibodies was carried out. The human placenta was chosen because of its extensive microvascular bed which is readily accessible to perfusion studies in vitro. The paracellular clefts of the human placental microvessels (Fig. 1) resemble continuous non-brain capillaries, both in structure (Leach and Firth, 1992) and permeability characteristics (Eaton et al., 1993).

Lobules of human term placentae obtained from placentae (n = 4) delivered by Caesarian section were (1) washed thoroughly in cold PBS and immersion fixed in 4% p-formaldehyde for 60 min and (2) extracorporeally perfused for a 20 min equilibration period (Leach and Firth, 1992) and perfusion-fixed for 30 min with 4% p-formaldehyde, excised and immersion-fixed for a further 60 min. After washing in PBS, some tissue was frozen in iso-pentane and cryosectioned (10 μm thickness) for indirect immunofluorescent labelling. For ultracryomicrotomy, tissue pieces were sandwiched in 8% gelatin and cryoprotected by agitating in 2.3 M sucrose for 2 h. They were then frozen on metal stubs by plunging in liquid nitrogen. Ultrathin frozen sections (45-60 nm thickness) were cut using a Reichert-Jung FC4E cryo-ultramicrotome.

Antibodies against PECAM-1 (mouse clone 9G11, undiluted culture supernatant, British Biotechnology, Oxford, UK), A-CAM (mouse ascitic fluid, dilution 1:100; Sigma, UK), pan-cadherin (rabbit polyclonal serum, dilution 1:100, gift from Dr T. Volberg, Weizman Institute, Israel; Geiger et al., 1990), β1 integrin chain (mouse clone Lia 1.2, undiluted culture supernatant; Arroyo et al., 1992), αvβ3 integrin complex (mouse clone LM609 against both subunits, ascitic fluid, dilution 1:100; Cheresh, 1987), vinculin (mouse ascitic fluid, dilution 1:100; Sigma, UK), α-actinin (rabbit polyclonal serum, dilution 1:100; Sigma, UK), human cadherin-5 (both mouse clone 7B4, undiluted culture supernatant; Lampugnani et al. (1992) and clone TEA 1/31, purified IgG, 43 μg/ml, characterised as shown below) were used.

7B4 and TEA recognised the same antigen in human cultured endothelial cells from umbilical cord vein. Indeed (1) both 7B4 and TEA immunoprecipitated a protein of identical molecular mass and (2) TEA could not immunoprecipitate any protein from a cell extract previously immunodepleted with 7B4 (Fig. 2). TEA could still immunoprecipitate the expected band from a cell extract immunodepleted with a negative antibody (Fig. 2). Human umbilical cord vein endothelial cells were cultured, labelled and immunoprecipitated as already described in detail (Lampugnani et al., 1992).

The frozen cryosections were labelled by a modified indirect immunofluorescent method (Coons et al., 1955), where the sections were immersed in 0.01 M glycine for 10 min to eliminate free aldehyde groups, washed in PBS and incubated in normal human serum (1:50 dilution) for 30 min to prevent non-specific activity and block human Fc receptors which may cross-react with the rabbit secondary antibody. The sections were then incubated in the primary antibody for 1 h at 37°C, washed thoroughly in PBS containing 0.1% BSA, and placed for 30 min at 37°C in FITC-labelled rabbit anti-mouse IgG (1:100 dilution, Sigma, UK) for the monoclonals, and FITC-labelled goat anti-rabbit IgG (Sigma) for the rabbit anti-pan-cadherin. The frozen sections were mounted in Mowiol 4-88 (Hoechst, Frankfurt, Germany) viewed under a Zeiss epifluorescence microscope and photographed on Kodak T-Max 400 films.

The ultrathin frozen sections of the human term placentae were labelled by an indirect colloidal gold immunolabelling technique (Geuze et al., 1981; Leach et al., 1989). Briefly, the sections were transferred through drops of 0.1 M glycine (10 min) and 0.1% BSA in PBS before incubation with normal human serum (1:50 dilution, 30 min). They were then incubated in drops of the chosen primary antibody overnight in a humid chamber at 4°C, washed through several drops of PBS containing 0.1 M BSA, and placed on either 10 nm colloidal gold-conjugated goat anti-mouse IgG or goat anti-rabbit IgG (1:25 absorbance) for 30 min. For double immunolabelling experiments the sections were incubated with a mixture of the primary antibodies, TEA and anti-pan-cadherin, in similar conditions as above. The secondary antibodies were a mixture of 5 nm colloidal gold-labelled goat anti-rabbit IgG and 10 nm colloidal gold-labelled goat anti-mouse IgG. The sections were washed throughly, post-fixed in 2% glutaraldehyde, stained by placing on drops of basic uranyl acetate (pH 8, 10 min), followed by 2% aqueous uranyl acetate (10 min), and taken through drops of 1.5% methyl cellulose (Fluka, Switzerland) on ice. The sections were air dried and viewed under a Jeol 100X electron microscope. Photographs were taken at ×30 000, ×50 000, ×100 000 and ×160 000 magnifications.

For controls, sections for both methods were incubated in (1) PBS, (2) rabbit non-immune serum (dilution 1:100, Sigma, UK) and (3) human serum albumin (mouse Ig, dilution 1:100, Sigma, UK) instead of the primary antibody.

The paracellular clefts of fetal microvessels were found to be immunoreactive to mAbs 7B4 and TEA, and anti-PECAM-1 (Fig. 3A,B,C,D). Immunoreactivity to anti-pan-cadherin was localised both in cleft regions and in the endothelium (Fig. 3E). No immunoreactivity to the antibodies against the β₁ and α_vβ₃ integrins was seen in the paracellular clefts, immunofluorescence being localised diffusely in the endothelium and in the abluminal interstitial space (Fig. 3F,G).

Immunogold studies revealed that cadherin-5 (7B4 and TEA antigen), pan-cadherin and PECAM-1 (Fig. 4A-F) were present in or adjoining the wide junctional zones of the paracellular clefts but not in the tight junctional regions. Pan-cadherin and PECAM-1 were also localised in the luminal membrane regions of the endothelium and within the cytoplasm of the endothelium (Fig. 4E,F). A-CAM was found only very occasionally in the wide junctions. It was also seen in the abluminal surface and cytoplasm of the endothelium (Fig. 5A,B).

Only the antigen localised by the mAbs TEA and 7B4 was found to be inter-endothelial cleft specific and was localised in the extracellular space within the clefts, or on or close to the plasma membranes of the wide zones (Fig. 4B,C). Gold labelling was found at the cytoplasmic surface of wide zones with the antibody against pan-cadherin (Fig. 4D). Double immunolabelling with TEA and anti-pan-cadherin showed an extensive labelling with anti-pan-cadherin, whilst the distribution of the TEA antigen was more discrete in the paracellular cleft (Fig. 5C,D). TEA antigen and pan-cadherin were occasionally co-localised in the same wide zones, but were mostly found in different wide zones of the same cleft.

The cytoplasmic surface of the wide zones appeared more electron dense. Both vinculin and a-actinin were found associated with these regions (Fig. 5E,F); the latter was also present near abluminal membrane regions and at sites more distant than vinculin from the wide junctional regions. Placentae which were immersion-fixed immediately after delivery, or which were perfused for equilibriation prior to fixation, both showed similar immunoreactivity; the latter showed less signs of ultrastructural damage.

No staining was seen when PBS or non-immune serum was used instead of a primary antibody (Figs 3H, 5H). When anti-human serum albumin was used, gold particles were seen within cytoplasmic vesicles and at the luminal surface of the endothelium (Fig. 5G). No staining was seen associated with the intercellular clefts.

Our results are the first to show, at the ultrastructural level, that the paracellular clefts of intact microvessels contain cadherins, recognised by anti-pan-cadherin (Geiger et al., 1990), anti-A-CAM/N-cadherin (Volk and Geiger, 1984) and anti-cadherin-5 (Lampugnani et al., 1992), and also the cell adhesion molecule PECAM-1 (Newman et al., 1990). Immunolocalisation at the ultrastructural level established that these molecules, seen throughout the paracellular regions by immunofluorescence, were present in discrete membrane microdomains in the wide junctional regions of the clefts. These wide zones appear similar to zonulae adherentes, both in morphological appearance and by their association with peripheral cytoplasmic molecules such as vinculin and α-actinin.

The integrins, immunolocalised by antibodies against the β₁ chain and the vitronectin receptor α_vβ₃, were not present in the endothelial clefts of human placental capillaries. The absence of α_vβ₃ in the clefts was to be expected, since the α_vβ₃ complex has been reported to be absent in cell-cell contacts of both umbilical vein explants and umbilical vein endothelial cell monolayers (Lampugnani et al., 1991), thus providing a negative control. The absence of staining seen when the antibody against the β1 chain was used is puzzling, since two members of this family, α₅β₁ and α₂β₁, have been located by us at cell-cell contact sites of umbilical vein cells (Lampugnani et al., 1991). The observed negative immunoreactivity could be due to differences between endothelial cells in vitro and in intact capillaries, or to differences between endothelia in the umbilical vein and placental microvessels.

Of all the antigens localised, cadherin-5 (localised by both mAb 7B4 and mAb TEA) appears to be cleft-specific, whilst pan-cadherin, A-CAM and PECAM-1 were also present at the luminal surface and in the cytoplasm of the endothelium. A-CAM showed only occasional labelling at the intercellular junctions compared to the junctional labelling obtained with the pan-cadherin. Salomon et al. (1992) have reported a similar extrajunctional distribution of A-CAM and pan-cadherin and a similar disparity in junctional labelling of the two in human umbilical cord endothelial cells. They suggest that different cadherins, coexpressed in the same endothelial cells, may undergo differential surface distribution. A similar differentiation may be at play in endothelia of intact capillaries. The presence of PECAM-1 in these microvascular junctions suggests that PECAM-1 may also be involved in endothelial cell-cell adhesion in vivo, since PECAM-1 has been shown to influence aggregation of transfected cells in vitro (Albelda et al., 1991) and is present in intercellular junctions of human umbilical vein endothelial cells. PECAM-1 is also present on surfaces of peripheral blood monocytes, neutrophils, platelets and certain T cell subsets, suggesting a role in adhesive events taking place during thrombosis and wound healing. The presence of PECAM-1 on the luminal membrane microdomains of the placental microvessels was therefore to be expected. Since PECAM-1 is expressed on the surface of endothelia and blood cells which do not spontaneously aggregate, cell-cell adhesion mediated by this molecule must require activation. The calcium ion dependence of PECAM-1 cell-cell adhesion (Albelda et al., 1991) also suggests that the interaction is a heterophilic one, though the counter receptor is unknown.

In this study, TEA antigen was co-localised with pan-cadherin in the cleft regions, further enhancing the view that this molecule does belong to the cadherin family (Lampugnani et al., 1992) but it was also seen to occupy different microdomains of the same cleft when compared to pan-cadherin in double immunolabelling procedures. This may be due to the greater avidity of the anti-pan-cadherin antiserum combined with the effect of steric hindrance during competition for the same sites, or because the placental endothelial clefts also contain other cadherins distinct from TEA antigens which are being localised by the pan-cadherin antibody. The antigen localised with the pan-cadherin antibody was typically at the cytoplasmic side of the clefts; the antibody was one directed against the conserved C-terminal cytoplasmic domain of cadherins (Geiger et al., 1990). TEA antigen was localised in the extracellular space or associated with the plasmalemmal leaflets of the wide zones. The monoclonal antibody TEA recognises the extracellular sequence of the isolated transmembrane cadherin-5 (proven by flow cytometry analysis and immunofluorescence, which show that TEA binds and stains all contacts of non-permeabilised endothelial cells; M. G. Lampugnani, unpublished observations). The ultrastructural localisation in our study supports these observations.

Using cultured human umbilical vein endothelial cells, Lampugnani et al. (1992) have shown that addition of a monoclonal antibody to cadherin-5 increases the permeability of confluent layers. The antigen is localised at cell boundaries only where cells are in contact, their distribution being modified by agents which enhance permeability such as tumour necrosis factor, thrombin and elastase. The authors also showed that cadherin-5 was restricted to the vascular endothelial layer of the vessels in a wide range of different tissues examined, and the antigen was not detected in any of the epithelial type tissues tested. Other cadherins, localised by anti-pan-cadherin, have been implicated in the induction of adherens junctions (Salomon et al., 1992; Takeichi, 1988). They have been shown to bind with cytoplasmic anchoring molecules, the catenins (Kemler and Ozawa, 1989), and are associated with vinculin and α-actinin to the F-actin cytoskeleton which influences cell shape and alters permeability (Yuruker and Niggli, 1992). The role of cadherins in regulating permeability of intact capillaries, especially the endothelial cleft-specific cadherin-5, warrants investigation. The exclusive localisation of cadherin-5 in the endothelial adherens-type junctions may reflect specific functional and structural properties of these junctions distinct from those of epithelia.

In this study we have begun to examine the molecular architecture of the wide junctions of the placental inter-endothelial clefts, which appear to be similar to epithelial adherens-type junctions in that they both possess cadherins and associate with cytoskeletal linking molecules. These zones differ from typical epithelial adherens junctions in possessing the Ig superfamily molecule, PECAM-1. PECAM-1 may not be involved directly in cell adhesion at junctional regions, its possible function being to provide, along with other endothelial cell adhesion molecules, an adhesive surface for extravasation of leukocytes via para-cellular pathways. The cadherins investigated in this study may, by virtue of their known adhesive properties, their localisation in the wide junctional regions of the placental microvessels and their functional properties ascertained from experiments with cell cultures (Lampugnani et al., 1992; Salomon et al., 1992), play a role in maintaining junctional organisation and capillary permeability. These molecules may, along with other cell adhesion molecules, such as PECAM-1 and other cadherins, be the bridging molecules postulated to be necessary to maintain the uniform width observed for endothelial wide junctions (Silberberg, 1988). They may also be the anchoring proteins which together with extracellular glycoproteins form the fibre matrix postulated to be necessary for regulating capillary permeability and integrity (Curry and Michel, 1980).

This work was funded by the Wellcome Trust. We wish to thank Professor Richard Beard, Department of Obstetrics and Gynaecology, St. Mary’s Hospital Medical School, London, and the midwives of the Aleck Bourne Labour Ward, St. Mary’s Hospital, for their helpfulness and efficiency in supplying us with freshly delivered human placentae. Fig. 1A was provided by E. D. A. Westcott.