Cultured microvascular endothelial cells (MVEC) differ in cytoskeleton, expression of cadherins and fibronectin matrix A study under the influence of interferon-γ

The endothelial cells of blood vessels exhibit high plasticity (Lipton et al., 1991). In response to perturbations of the blood flow, the cells are rearranged into either an elongated or a rounded morphology consistent with the physical forces (Gotlieb et al., 1991). The process of neovascularisation during wound healing, inflammation and tumor development is known to depend on the growth of capillaries, in the course of which endothelial cells migrate away from their mother vessels, and at the same time are trans-formed into fibroblast-like cells (Paku and Paweletz, 1991).

In culture, endothelial cells are affected by manipulations of the environment such as a change of culture medium and serum batches, the addition of growth factors or the alter-ation of the extracellular substrata used to coat culture plates. The cells either develop as an epithelioid monolayer, or resemble mesenchymal cells and grow as a multilayer (Vlodavsky et al., 1979; Gospodarowicz and Ill, 1980; Greenburg et al., 1980; Gospodarowicz and Lui, 1981). After transformation into fibroblast-like cells in vitro, endo-thelial cells show a reduction in their circular array of actin filaments, while their centrally located actin filaments become stronger (Hormia et al., 1985). Concomitantly, the number of cell-cell contacts declines, and the substratum proteins by which the endothelial cells achieve their typi-cal apico-basal orientation seem to change in nature (Madri et al., 1988; Yannariello-Brown et al., 1988). Thus, each particular endothelial cell morphology observed depends on the distinct cytoskeleton, the specific cell-cell contacts and the structure of the extracellular matrix. Changes in endo-thelial cell shape are not only governed by extrinsic manipulations of the cultivation design, but also by intrinsic changes in cell behaviour. Repeated cell passages, increasing the age of cell strains during cultivation, or long-term growth in a non-physiological environment, favour the occurrence of changes in intrinsic morphology. Lability is such that any spontaneously occurring morphological changes tend to be accepted as inherent in studies with microvascular endothelial cells (MVEC).

MVEC isolated from diverse organs are generally assumed to be morphologically similar at the start of a cul-ture. However, many morphological studies have shown that in vivo endothelial cells differ according to their origin within the microvascular tree (examples follow), indicating that caution is called for. Small arterioles develop elabo-rate interdigitating myoepithelial junctions between endo-thelial cells and myocytes that are not found in large arte-rioles (Carlson et al., 1982). Endothelial cells from the alveolar capillaries of rat lung contain more micropinocy-totic vesicles than their non-muscular or muscular microvessel counterparts (Defouw, 1988). Antibodies directed against human lung endothelium exclusively rec-ognize venules (Zhu and Pauli, 1991). To our knowledge, this firmly established opinion, that cultured MVEC belong to a morphologically uniform population, has recently been challenged by four groups. Using rat hearts, Diglio et al. (1988) have isolated two phenotypically different MVEC strains whose particular morphology remains preserved after 30 subpassages. Furuya et al. (1990) have obtained endothelial cells with and without intracellular vacuoles from the bovine adrenal medulla. Four MVEC phenotypes that maintained separate morphologies during the period of investigation (4 subpassages) have been isolated from the rat brain by Rupnick et al. (1988). We have established stable cultures of five phenotypically different MVEC strains from the bovine corpus luteum (Spanel-Borowski, 1991; Spanel-Borowski and van der Bosch, 1990). Over a 6-month cultivation period examined up to now, each phenotype has proved to be morphologically independent of the others.

As the conventional morphological observations reported to date do not fully characterize each phenotype, our results are open to criticism. The studies, therefore, are extended here to include an examination of the cytoskeleton, cell-cell adhesion molecules, and the fibronectin matrix for MVEC types 1, 3, 4 and 5. Monolayers of MVEC type 2 were excluded from the present study because they represent co-cultures with desmin-positive cells, which do not seem to express smooth muscle cell myosin. Since tubules were formed under interferon-γ (IFN-γ) treatment, the behaviour of MVEC type 2 has been studied in detail (unpublished data). The results point to the functional individuality of the two different cells, which are positive either for both desmin filaments and Factor VIII-related antigen or for only Factor VIII-related antigen (unpublished data). We specu-late that their occurrence is related to the collagenase-free MVEC isolation technique used.

The procedure used for the isolation and culture of MVEC from developing bovine corpus lutea has already been described in detail (Spanel-Borowski and van der Bosch, 1990). In short, mechanically dislodged cells were separated using a 50% Percoll (Pharmacia, Freiburg, Germany) density gradient and spinning at 16,000 g, 21°C, for 20 minutes. The 4 ml fraction formed above the banded erythrocytes contained a mixed cell population of fibrocytes, myocytes and MVEC. The cell viability of this region was less than 1%, as determined by the trypan blue exclusion test. Samples (0.5 ml) of culture medium containing 5×10⁴ isolated cells were added to each well of 24-well culture plates (Nunc) that had been coated with 1% collagen type 1 solution (Vitrogen 100^R, Celtrix, Santa Clara, USA). Thus, less than 500 live cells were plated and subsequently grew under semiclonal conditions. The culture medium consisted of nutrient mixture F12 (Ham, Gibco) and Dulbecco’s modified Eagle’s medium (DMEM, Gibco) mixed 1 + 1, supplemented with the 0.15 mM Hepes, 0.22 mM NaHCO₃ and 5% fetal calf serum (Myoclone, Gibco). Endothelial mitogen (50 μg/ml culture medium, Paesel) was added once. The development of endothelial cell colonies was observed over 14 days, in the course of which contaminating cells were removed by scraping them off with a bent Eppendorf tip. An inverted phase-contrast microscope used at an objective magnification of ×10 allowed visual control. Once developing clones of the different cell types had covered an area of approximately 3 mm² the colonies were treated with 0.02% trypsin solution for 30 seconds to loosen intercellular junctions. Cells of the colony were then gently scraped off, and dislodged cell aggregates were removed and transferred at a 1:4 split ratio. Colony transfer was repeated once if passaged colonies grown to confluency after another fort-night showed contaminating cell islands. Such islands were clearly demarcated and easily distinguishable from the MVEC types under selection. Cells of pure cultures were enzymically dislodged from 4 or 8 wells and stored at −196°C until use. Both unfrozen cells and those from the frozen stock were employed. When seeding the cells at low density and monitoring their growth rate over a fortnight period, growth was low for MVEC type 5, moderate for MVEC type 1 and high for MVEC types 3 and 4 (unpublished own data). The high growth rate in sublines A and B of MVEC type 3 allowed us to exclude the possibility that the intracellular vacuoles of subline B are signs of approaching senescence.

The four phenotypes were characterized by phase-contrast microscopy. Cell type 1, isomorphic cobblestone-like monolayer; cell type 3, polymorphic monolayer of polygonal and spindle-shaped cells without intracytoplasmic vacuoles (subline A) and with intracytoplasmic vacuoles (subline B); cell type 4, monolayer of round cells and of opaque appearance; cell type 5, monolayer of flat phase-dense cells with no apparent nuclei. Additional crite-ria were based on a positive immunoresponse, allowing the local-ization of Factor VIII-related antigen and of ACE (angiotensin-converting enzyme) antigen. MVEC type 5 exhibited a weak immunoresponse for both factors. Cell type 1 showed a positive immunoreaction for α₂-macroglobulin. Furthermore, the uptake of Dil-Ac-LDL, i.e. fluorochrome-labelled acetylated low-density lipoprotein (Biomedical Technologies, Stoughton, USA), proved to be positive for all cell types, although the intensity differed between MVEC types and seemed to depend on the culture period.

Cells were grown on round collagen-coated glass coverslips mounted in 24-well culture plates. A 0.5 ml portion of culture medium was used per well. Confluent cultures were either left untreated or treated once with 200 units of recombinant bovine IFN-γ (gift from Dr Steiger, Ciba-Geigy, Switzerland) per well for three days. Care was taken to ensure that each MVEC culture was confluent over the whole coverslip for three days, since cytoskeleton filaments are known to change their appearance during the transition from growing to resting cells (Savion et al., 1982; Alexander et al., 1991). All experiments were carried out with cell strains from 3rd or 4th subcultivations. More than 3 cell strains of each MVEC type were studied in all. Both untreated and IFN-γ-treated cultures were fixed at room temperature for 20 minutes with 2% paraformaldehyde in a microtubule-stabilizing buffer (100 mM PIPES, pH 6.8, 1 mM MgSO₄, 1 mM EGTA) containing 0.2% Triton X-100 for cell permeabilization. Cells on coverslips were postfixed with ice-cold 100% ethanol for a fur-ther 20 minutes. Fixation with ethanol was omitted for the localization of cadherins. Subsequently, 0.5 mg sodium borhydride/ml phosphate-buffered saline (PBS) was applied for 5 minutes to break aldehyde cross-links. Between each of the following incubations the cultures were rinsed thrice with PBS containing 1 mg/ml bovine serum albumin (BSA) and 0.1% Triton X-100, and thrice with PBS-BSA.

Primary monoclonal antibodies detecting either tubulin, vimentin or fibronectin were obtained from Sigma (nos T-4026, V-6630, F-7387). The monoclonal antibody Lu-5 against human pan-cytokeratin for both acidic and basic cytokeratins was obtained from Biomedicals, Augst, Switzerland. The polyclonal rabbit antibody directed against mouse N-cadherin was received from Prof. Dr Takeichi, Kyoto, Japan, and the polyclonal rabbit antibody reacting with dog E-cadherin was described by Behrens et al. (1989). The Sigma antibodies were diluted 1:400 in PBS containing 10 mg/ml BSA, prior to use, while the donated anti-bodies required only a 1:100 dilution with the same buffer. All specimens were incubated at room temperature for 40 minutes. Biotin-conjugated anti-mouse antibody or anti-rabbit antibody (no. B-7264; no. B-9642; Sigma) was used as the secondary anti-body, at a dilution of 1:50 in PBS containing 10 mg/ml BSA (40 minutes). Non-specific labelling of nuclei was prevented by incubating with 1 mg poly-D-lysine/ml PBS (no. P-0899; Sigma). Finally, specimens were incubated for 40 minutes with rhodamine-coupled avidin (no. A-3026; Sigma), diluted 1:50 in PBS containing 10 mg/ml BSA. Negative controls were carried out with PBS containing 10 mg/ml BSA in place of the primary antibodies. Following indirect immunofluorescence localization, actin filaments of each sample were stained with phalloidin-FITC (P-5282, Sigma) diluted 1:200 in PBS containing 10 mg/ml BSA. Incubation was carried out for 30 minutes.

Glass coverslips were mounted upside down on object slides using Glycergel^R (Dako) together with 10 μg/ml DAPI (4′,6-diamidino-2-phenylindole. 2HCl; Serva) and 25 μg/ml DABCO (1,4-diazabicyclo [2.2.2] octane; Sigma) to prevent photobleach-ing. Staining remained intact for several weeks if the slides were stored at 4°C. They were examined and photographed under a Polyvar light microscope (Reichert-Jung) equipped with epifluorescence illumination and an oil immersion objective (×100). Pictures were taken on T-MAX 400 film and developed in T-MAX RS (Kodak).

Cell extractions and immunodetection of E-cadherin by west-ern blotting were performed as described (Behrens et al., 1985, 1989) with the omission of the final 100,000 g centrifugation step after cell lysis. Equal amounts of extracted proteins were loaded for gel electrophoresis.

Examining monolayers of the single untreated MVEC types 1, 3, 4 and 5 showed each to have its own characteristic cytoskeleton. Under IFN-γ treatment, cell-specific differences were maintained. Nevertheless, each MVEC type showed modifications to its cytoskeleton. All of the following results were reproduceably obtained from several independent studies carried out for each MVEC type.

Without IFN-γ treatment, cultures of MVEC type 1 formed an epithelioid monolayer of ‘cobblestone’ appearance. Cultures of MVEC type 3 contained polymorphic cells of polygonal or spindle-shaped form. Subline A lacked intra-cellular vacuoles, while subline B contained intracellular vacuoles. Monolayers of MVEC type 4 showed rounded cells. Cultures of MVEC type 5 consisted of polygonal, so-called phase-dense cells of flat appearance. Under IFN-γ treatment the cobblestone-like aspect of MVEC type 1 cultures was less pronounced. Flattened cells were closely apposed in the monolayer and distinct cell borders were visible. There was no widened intercellular gap in the cohesive sheet. The spindle-shaped cells of MVEC type 3, sub-lines A and B, seemed to disappear, giving way to a regular, polygonal monolayer. The small vacuoles predominantly observed in controls of subline B were replaced by strikingly large vacuoles; this was likely to correspond with the onset of cell senescence. MVEC type 4 flattened and became polygonal. MVEC type 5 lacked any clear-cut changes.

In spite of careful handling, untreated cultures of MVEC type 1 began to retract and dislodge during fixation, thereby destroying monolayer confluency. This effect was observed neither for monolayers of IFN-γ-treated MVEC type 1 nor for any of the other treated or untreated monolayers.

The actin filament pattern was uniform for each particular culture studied. Without treatment of IFN-γ MVEC type 1 showed a circular band of fine multiple microfilaments, which was reinforced towards the cell border. Widened intercellular spaces crossed by remnants of cytoplasm were observed, an artifact caused by cell retraction during fixation. Both sublines of MVEC type 3 were characterized by a predominant ‘starburst-like’ pattern of actin filaments frequently situated at the cell centre. MVEC type 4 mainly displayed a delicate network of actin filaments throughout the cytoplasm. In MVEC type 5, thick microfilament bundles formed a prominent circular band from which short fil-aments radiated like tiny ‘spokes’. Under IFN-γ treatment MVEC type 1 exhibited a thin rim of prominent fluorescence delineating cell margins. No intercellular spaces were evident. The cells formed a ‘shoulder to shoulder’ mono-layer with interdigitating cell margins. In some places, thin filaments, whether aligned parallel or perpendicular to the cell borders, appeared to be continuous with similarly arranged filaments of the adjacent cell. MVEC type 3 (sub-lines A and B) possessed nodal intersections at the cell periphery, which were connected by series of actin cables. MVEC type 4 developed knots of actin filaments in the central cell regions. This feature may be compared to the ‘star-burst-like’ pattern of untreated MVEC type 3. In MVEC type 5 straight filaments similar to stress fibers were observed. They crisscrossed the basal cell area and blurred the circular actin band.

Cytokeratin filaments were detected in MVEC type 1. Their dense meshwork was constricted to the central area in untreated cells. Under IFN-γ treatment, a conspicuous meshwork was developed throughout the cytoplasm. Fila-ment bundles radiated towards the cell margin here and there, contacting comparably arranged bundles of the adjacent cells. This feature was reminiscent of cell adhesion plaques like desmosomes.

The vimentin architecture showed two patterns. Pattern 1 occurring in MVEC types 1, 3 and 4 was attributed to a dense vimentin network. Like cytokeratin it was constricted to the central area in untreated MVEC type 1 and appeared to reach cell margins under IFN-γ treatment. The network expanded throughout the cytoplasm in untreated and IFN-γ-treated MVEC types 3 and 4. This network was either feltlike in appearance as for MVEC type 3, or filamentous as for MVEC type 4. Where many intracellular vacuoles had developed in MVEC type 3, filaments assembled around the vacuoles so that a honeycomb-like structure became visible. Thick coiled filament bundles were more striking in MVEC type 4 than 3. The coiling phenomenon became more conspicuous under IFN-γ treatment. Pattern 2 occurring in MVEC type 5 was assigned to a loose net-work with the appearance of fine filaments radiating from the perinuclear area towards the cell periphery similar to the layout of microtubules. Pattern 2 seemed to remain unchanged under IFN-γ treatment.

Microtubules of all untreated MVEC types originated from a single randomly oriented perinuclear region, known as the microtubule organizing centre. From here, wavy micro-tubules spread towards the cell periphery where they tended to bend along the cell margin. Under IFN-γ treatment, no change in the microtubule framework was noticed for MVEC types 1 and 3. In contrast, MVEC type 4 demonstrated much more striking microtubule organizing centres, in addition to intensified peripheral bending. Marked peripheral bending was also shown by MVEC type 5, in this case together with an overlapping microtubule pattern obviously related to overlapping cell borders (not shown).

MVEC type 1 reacted with the antiserum against E-cadherin. Using immunolocalization the cell borders were pictured as continuous fluorescent lines when IFN-γ had been applied, in contrast to the discontinuous labelling that otherwise occurred. MVEC type 5 responded to the anti-serum raised against N-cadherin. The cell borders were delineated as ‘fluorescent punctuation marks’ in both the IFN-γ-treated and untreated samples. In the other MVEC types, cadherins were not detected.

Each examined MVEC culture produced an extracellular fibronectin (FN) matrix at the basal cell side. Two layers were discernable, one being closer to the cell membrane than the other. The FN layer closer to the cell membrane showed globular or patchy deposits, while the more-distant layer was filamentous. The two layers could be connected to one another by anchorage filaments. Although two FN layers were a common feature of all four MVEC cultures, cell-specific variations were apparent. In both untreated and IFN-γ-treated samples, the FN matrix seemed to be less developed in MVEC types 1 and 5 than in the correspond-ing cultures of MVEC types 3 and 4. Under IFN-γ treatment FN matrix of MVEC type 1 looked like an opaque layer. For VEC types 3 and 4 the globular matrix of the layer close to the cell membrane had practically disap-peared. An elaborate architecture of branching and anastomosing filaments of various thickness was noted. Cultures of MVEC type 5 showed no marked changes of the coarse array of the FN matrix.

The induction of E-cadherin by IFN-γ was confirmed by western blotting analysis with the anti-E-cadherin antiserum (Fig. 18). A protein with apparent molecular mass of approximately 135 kDa was detected in detergent extracts of MVEC type 1, but not in extracts of untreated cells (lanes 1 and 2). This band was not observed in extracts of MVEC type 3 (lanes 3 and 4) in agreement with the lack of E-cad-herin immunofluorescence staining in these cells. Further-more, a 95 kDa fragment characteristic of cadherins was detected in trypsin extracts of IFN-γ-treated MVEC type 1 (data not shown). The antiserum also reacted with a band at about 80 kDa, which was preferentially seen in untreated MVEC type 1 (lane 1) and weakly detectable in IFN-γ-treated cells (lane 2) and could represent a digestion prod-uct of E-cadherin. Although the intensity of this band varied between different experiments and was not clearly inversely correlated with the intensity of the 135 kDa band, it is possible that the observed upregulation of E-cadherin after IFN-γ treatment is in part due to increased proteolytic resistance (e.g. through stabilization of the protein at the cell junctions).

The function of MVEC is of interest for both basic and clinical research. Every laboratory starting to work with MVEC has to spend a considerable amount of time before stable and uniform cultures are obtained. If primary cultures assumed to be a homogeneous population in fact con-sist of a mixture of phenotypes the growth of different sub-cultures will not be comparable after serial passages. As a consequence of cell heteromorphology, the end result will be a culture of morphological variability. Our preceding studies of the heteromorphology shown by cultured MVEC were based on visual inspection (Spanel-Borowski, 1991; Spanel-Borowski and van der Bosch, 1990). These were followed by investigations that allowed the characterization of functional differences. The microtubule pattern, the expression of NCAM (neuronal cell adhesion molecule), and the differential adhesion of blood granulocytes were all found to be phenotype-specific (Ley et al., 1992; Mayer-hofer et al., 1992; Wolf and Spanel-Borowski, 1992a,b). The present work substantiates the individual nature of these phenotypes, all of which may be classified as MVEC. Thus, mutual transformation between these MVEC types may no longer be considered a plausible explanation for the experimental observations.

The results presented broaden our knowledge of the actin cytoskeleton detected in confluent MVEC cultures. Only one particular actin cytoskeleton pattern has ever been described in the literature for cultures of resting endothelial cells, i.e. a dense peripheral band similar to that of our MVEC type 1 (Rogers and Kalnins, 1983; Wong and Gotlieb, 1986; Phillips et al., 1988). In contrast, the mono-layer of each separate MVEC type uniformly and reproducibly portrayed its own particular actin filament pattern. This is not related to changes in cell behaviour such as growth rate because each culture appeared to be in a quiescent stage at the time of immunofluorescence localization. MVEC type 3 may be an exception. As soon as it reaches confluency tubules begin to form, probably leading to loss of contact inhibition, which is not detectable at the phase-contrast microscopic level. The observed uniformity of the actin cytoskeleton may be taken as sign of population homogeneity, as described by others (Rogers and Kalnins, 1983; Phillips et al., 1988; Gotlieb et al., 1991). The finding that intracytoplasmic vacuoles may or may not be present in cultures of MVEC type 3 can be ascribed to a subpopulation of cells. That this is indeed the case has been shown by studying the expression of class II MHC antigens by isolated MVEC type 3 cells using fluorescence flow cytometry (Spanel-Borowski and Bein, 1993).

The cytokeratin positivity of MVEC type 1 has been an unexpected finding because endothelial cells, as cells of mesenchymal origin, are known to contain vimentin as their single intermediate filament. There are, however, exceptions to this rule according to reports on cytokeratin-positive endothelial cells in small and large blood vessels of bovine lung and of different human organs (Jahn et al., 1987; Patton et al., 1990). Our laboratory is now able to report on cytokeratin-positive endothelial cells isolated from the bovine aorta and the vena cava (Spanel-Borowski and Fenyves, 1993). This finding allows us to exclude the possibility widely that the cytokeratin-positive MVEC type 1 may turn out to be a cytokeratin-positive mesothelial cell. The diagnosis in favour of an endothelial cell is finally sup-ported by the observation that only MVEC type 1 develops tube-like structures in Matrigel^R, a basement membrane matrix (own unpublished data). This phenomenon is considered to be the decisive criterion for endothelial cells, but is not found for mesothelial cells (Chung-Welch et al., 1989). In considering the basic and acidic proteins of the cytokeratin family, a detailed analysis is under investigation to clarifiy the way in which the two subfamilies are combined in MVEC type 1.

The abundant development of vimentin filaments is a characteristic sign of endothelial cells. The function of these filaments is still uncertain. It is speculated that they associate with and guide the orientation of MT (Blose, 1984). This postulate agrees with the observation that, for MVEC type 5, the vimentin filaments are organized in the same way as the MT. Comparable findings have been described for subconfluent cultures of aortic endothelial cells (Kalnins et al., 1981). In contrast, the vimentin filaments of MVEC types 1, 3 and 4 seem to bear no relation to the course of microtubules. This and the dense vimentin network reported for confluent endothelial cell cultures (Savion et al., 1982; Machi et al., 1990; Alexander et al., 1991), suggest an alter-native or additional function for the vimentin filaments. It may be related to cell-cell contacts because the IFN-γ-treated cytokeratin-positive MVEC type 1 develops desmo-some-like adhesion plaques and maintains a spread vimentin network.

Interferon-γ as well as other cytokines is reported to affect the epithelioid monolayer, leading to a culture of spindle-shaped cells with overlapping growth (Montesano et al., 1984; Sato et al., 1986; Stolpen et al., 1986; Friesel et al., 1987). An increase in stress fibers and cell-substra-tum contacts, together with a decrease in cell-cell contacts and a reduction in the FN matrix, was noted. In contrast, for IFN-γ-treated MVEC type 1 we observed the formation of a striking array of actin and cytokeratin bundles running perpendicular to the fluorescent cell margin. This was coup-led with reinforced E-cadherin expression, which could be demonstrated both by immunofluorescence staining and by western blotting analysis. The perpendicular array of actin filaments may indicate an adherens-type junction as described for cultured endothelial cells (Wong and Gottlieb 1986). The presence of very peripheral cytokeratin bundles may represent a junctional complex similar to the tonofil-aments of desmosomes in epithelial cells.

The expression of cadherins in endothelial cells has been the subject of recent work by various groups. Liaw et al. (1990) cloned cDNAs of homologues of N- and P-cadherin of bovine endothelial cells and V (vascular)-cadherin was identified by Heimark et al. (1990) in a similar cell type. Recently a novel cadherin was isolated from human umbil-ical vein endothelial cells (Lampugnani et al., 1992). Interestingly, treatment of these cells with a combination of IFN-γ and tumour necrosis factor (TNF) resulted in increased permeability of the endothelial monolayer and was accompanied by a punctate redistribution of this particular cadherin. In contrast, in our MVEC type 1 E-cad-herin was upregulated by IFN-γ and, consequently, cells acquired more stable intercellular contacts. Similarly, E-cadherin was found to be upregulated in bovine endothelial cells by cAMP treatment (Rubin et al., 1991). It will be intersting to determine the regulatory mechanisms and the functional consequences of these alterations of E-cadherin expression in endothelial cells.

Under IFN-γ treatment E-cadherin appears at the cell sur-face of MVEC type 1 as a fluorescent line. A similar image is obtained for TNF-treated HUVEC (human umbilical vein endothelial cell) cultures when the cell margins are visual-ized by the immunolocalization of talin, a molecule occur-ring in the cytoplasmic component of adhesion plaques (Molony and Armstrong, 1991). In contrast, labelling N-cadherins reveals the borders of MVEC type 5 cells as a series of fluorescent dots. Thus, either the homotypic inter-action between E-cadherins in MVEC type 1 and N-cad-herins in MVEC type 5 is different, or the framework of adhesion plaques varies between these two phenotypes. Interestingly, in these IFN-γ-treated cells overlapping cell borders were revealed by an overlapping microtubule pat-tern in conjunction with the expanded array of fluorescent dots. It may indicate that cell-cell adhesion between the overlapping cell membranes is mediated by N-cadherins. The phenomenon of overlapping cell borders in treated MVEC type 5 is different from the contact-inhibited appear-ance maintained in the other MVEC types under IFN-γ treatment. In view of the weak labelling of both Factor VIII-related antigen and ACE antigen, further investigation of the true endothelial origin is required for our MVEC type 5.

According to the literature, interferon-γ treatment of con-fluent endothelial cell cultures results in the local down-grading of the FN matrix, to such an extent that this obtains a ‘moth-eaten’ appearance (Stolpen et al., 1986). In our hands, processing of the FN matrix for labelling leads to the destruction of its integrity. Thus, we consider FN cords and areas void of FN to be artifacts. According to our observations, the FN matrix changes its architecture in response to INF-γ treatment. Similar effects have been documented for endothelial cell cultures treated with supernatants of lectin-activated blood monocytes or with transforming growth factor β1 (Montesano et al., 1984; Madri et al., 1991).

To conclude, the basic aim of our work, to accumulate cell characteristics that delineate the existence of separate luteal MVEC types, has been achieved. Working on this basis, other organs and species can now be tested for the existence of endothelial cell heteromorphology. Such het-eromorphology may signify that particular MVEC types are localized beside arterioles, capillaries or venules, according to a strictly segmental distribution pattern. However, the coexistence of a particular MVEC type within a single microvascular segment, resulting in a mosaic-like distribution pattern, is also feasible. Further study is required before any conclusions can be drawn on this point.

We thank Prof. Dr M. Takeichi, Department of Biophysics, Kyoto, Japan, for the gift of cadherin antibodies and also Dr Steiger, Ciba-Geigy, Switzerland, for the generous gift of recom-binant bovine IFN-γ. We are indepted to Mr M. Saxer for his out-standing technical assistance, to Mr H. Stöcklin for the preparation of excellent photographs, to Ms Denise Müller for skilful typing and to Dr Shirley Müller, Maurice Müller Institute, Basel, Switzerland, for amendments to the English text. The study was supported by a grant from the Schweizerischer Nationalfonds (no. 31-320 69.91).