Epithelial-mesenchymal interactions in the production of basement membrane components in the gut

Differentiation of the intestinal epithelium during development is closely associated with ordered sequence of morphogenetic events (for a review, see Kedinger et al. 1986a). Several experimental studies using in vivo transplantation of various tissue recombinants have emphasized that these developmental processes are dependent upon reciprocal epithelialmesenchymal interactions, in which mesenchyme plays a major role in epithelial differentiation (for reviews, see Kedinger et al. 1986a; Haffen et al. 1987).

Among the factors that could be involved in such tissue interactions during morphogenesis and cytodifferentiation, extracellular matrix (ECM) components have been proposed (for reviews, see Hay, 1981; Bissell et al. 1982; Slavkin, 1982; Timpl & Dziadek, 1986). Yet the cellular origin of these components is still controversial. Although the basement membrane is generally considered to be an epithelial product, there are known examples of mesenchymal production of basement membrane components (Kühl et al. 1984; Kimata et al. 1985; for a review, see Timpl & Dziadek, 1986).

In the intestine, some observations suggested a possible role of the ECM molecules in morphogenesis and in the maintenance of epithelial cell polarity. We could show that basement membrane components (laminin, nidogen, type-IV collagen) are present at the epithelial–mesenchymal interface early in embryonic development and that changes in the spatial distribution of some ECM proteins are associated with morphogenetic processes (Simon-Assmann et al. 1986). On the other hand, severe perturbations of proteoglycans, collagens and some glycoproteins have been reported in human colonic tumours in parallel to tissue disorganization including loss of cell polarity and invasion processes (Burtin et al. 1982; lozzo & Wight, 1982; lozzo et al. 1982).

In the present study, we describe attempts to answer two major questions related to ECM production. The first question concerns the relationship between the establishment of the basement membrane and epithelial cell differentiation. For this purpose, we examined immunocytochemically the deposition of ECM components in epithelial-fibroblastic cocultures. Using this model previously, we showed that the fibroblastic population was necessary to support survival, proliferation and morphological as well as functional differentiation of the epithelial cells (Kedinger et al. 1987). The second aim of our work is to define the cellular origin of the ECM components by means of interspecies recombinants and species-specific antibodies.

Wistar rat fetuses were used at 14, 15 and 20 days of gestation and White Leghorn chick embryos at 3 and 5 days of incubation. The developmental stages were counted as days of gestation for the former (the existence of a vaginal plug being taken as evidence of mating and designated as day 0) and as days of incubation at 38°C for the latter.

Cocultures of endodermal cells and fibroblastic cells were performed as previously described (Kedinger et al. 1987). Two types of fibroblastic cell populations were used: intestinal mesenchymal and skin fibroblastic cells. The former were derived from intestinal mesenchymal microexplants after separation of the mesenchyme from the endoderm by collagenase treatment (0·03 % collagenase type IV; 147 i.u. mg^-1, Worthington) of 14- and 15-day fetal rat intestines. Skin fibroblasts were derived from enzymatically (collagenase 0·01 % and trypsin 0·01 % solution) dissociated 20-day fetal rat dermis (Haffen et al. 1983). Both cell types were used at early passages (until passage 3) when they reached confluency, approximately 7 days after plating. Endoderms, separated from their mesenchymes by 0·03 % collagenase treatment of 14- and 15-day fetal intestinal anlagen, were cut into small fragments and seeded on top of confluent monolayers of both types of fibroblastic cells. Each cell population was also cultured in isolation. The culture medium used was composed of a mixture (1:1) of DMEM and Ham’s F12 (Gibco Biocult, Glasgow, Scotland) supplemented with 7·5 % fetal bovine serum (Gibco, containing 2·5 % SerXtend serum substitute, NEN) and with 10⁻⁷ M-dexamethasone (Sigma). Gentamycin (200 μg ml^-1) was added as an antibiotic.

Associations between rat and chick tissue components have been performed using an experimental procedure described previously (Kedinger et al. 1981). Briefly, after collagenase dissociation of 5-day chick embryonic and 14- to 15-day fetal rat intestinal anlagen, chick mesenchymes were recombined with rat endoderms (Cm/Re) and vice versa (Rm/Ce). Both types of recombinants were then grafted into the coelomic cavity of 3-day chick embryos. The grafts were recovered 11–15 days later.

Affinity-purified rabbit antibodies against mouse laminin (Rohde et al. 1979), the globular domain NC₁ of type-IV collagen (von der Mark et al. 1985), nidogen (Timpl et al. 1983), bovine type-III procollagen aminoterminal propeptide (Nowack et al. 1976), and human plasma fibronectin (Fleischmajer & Timpl, 1984) have been kindly provided by Drs R. Timpl and K. von der Mark (Max-Planck Institut für Biochemie, Martinsried, FRG). All antibodies exhibited sufficient cross-reactivity with rat tissues. Rabbit antibodies specific for chick type-IV collagen provided by Dr K. von der Mark were prepared against the pepsin fragment (F1)2F2 of type-IV collagen obtained from chick smooth muscle (Mayne & Zettergren, 1980; Mayne et al. 1982) and their species specificity has been shown in immunochemical and immunohistologic analyses (Sariola et al. 1984b).

Cell sheets of the cocultures and of isolated endodermal or fibroblastic monolayers were detached mechanically from the bottom of the dish under the dissecting microscope after different culture periods (between 2 days and 2 weeks). They were immediately embedded in Tissue-Tek compound and frozen in freon cooled in liquid nitrogen. Transverse sections (5–6/μm thick) were cut at —25°C and placed on gelatin-coated slides. Interspecies recombinants were processed similarly.

Cryostat sections were incubated for 30 min at room temperature with the antibodies diluted in phosphate-buffered saline (PBS) at concentrations of 20–70; μml’¹ (Simon-Assmann et al. 1986). The sections were then washed with PBS and stained for 30 min with fluoresceinisothiocyanate-labelled goat anti-rabbit у-globulin (Nordic) diluted (1:20) in PBS. The preparations were mounted under coverslips in glycerol–PBS–phenylenediamine and observed under an Orthoplan microscope (Leitz). Control sections were processed as above but affinity-purified antibodies were omitted or replaced by normal rabbit serum. None of these controls showed any specific fluorescence.

When isolated intestinal endoderms were cultured directly on plastic culture dishes, they did not exhibit any labelling with the various antibodies (type-IV collagen, nidogen, fibronectin, type-III procollagen) either within the cell layer or at the basal surface of the cells (illustrated for type-IV collagen in Fig. 1A). One exception concerned the occurrence of some fluorescent spots scattered within the epithelial cell sheet with anti-laminin antibodies (Fig. 1B). By contrast, in intestinal and skin fibroblastic cell cultures, a peri- or inter-cellular deposition of interstitial molecules (fibronectin: Fig. 1C, and type-III procollagen), as well as of basement membrane components (type-IV collagen: Fig. 1D, laminin, nidogen), was obvious.

A general feature was that, at any time, none of the antigens were detected immunocytochemically within the epithelial cell layer either intracellularly or in the extracellular spaces (Figs 2, 3). By contrast, positive extracellular sheaths surrounding the cellular elements were detected within the intestinal and skin fibroblastic cell layer with all antibodies tested, the staining of basement membrane components being less intense, however (Fig. 3,versusFig. 2).

At the epithelial– mesenchymal interface, changes in the macromolecular composition of the basement membrane occurred as a function of the culture period. At the time of seeding, just after its separation from the mesenchyme by collagenase, the immunostaining of the endoderm revealed that type-IV collagen had disappeared (Fig. 4A) while nidogen and laminin persisted as a punctuate ring at the periphery of the endoderm (Fig. 4B). After 2 days of coculture, laminin (Fig. 3A), nidogen and type-IV collagen (Fig. 3D) were deposited as a discontinuous sheet at the epithelial–mesenchymal interface. This irregular pattern of labelling shifted towards a continuous bright staining visible as early as 4 days for laminin (Fig. 3B,C) or nidogen, but only at 6 days for type-IV collagen (Fig. 3F versus 3E). Fibronectin and type-III procollagen never delineated the basal surface of the epithelial cell layer at any culture time (Fig. 2).

The chronological appearance of the basement membrane components was similar whether endodermal cells were cultured with intestinal mesenchymal cells or with skin fibroblasts.

Attempts to define the cellular origin of the ECM components that contribute to the formation of the basement membrane have been made by means of interspecies tissue recombinations and speciesspecific antibodies. Of the available antibodies raised against ECM components of mammalian origin, those directed against fibronectin, type-III procollagen, laminin and nidogen also reacted with chick intestinal connective tissue (named lamina propria in the intestine) and/or basement membrane (Fig. 5). On the contrary, the antibody raised against mouse type-IV collagen stained rat tissue (Fig. 6C) but did not cross-react with chick tissue (Fig. 6A). Vice versa, the antibodies produced against chick type-IV collagen labelled the epithelial–mesenchymal interface as well as the muscular layers of the chick intestine (Fig. 6B), but did not stain sections of rat tissue (Fig. 6D). Therefore, type-IV collagen was used as a marker for the study of the cellular origin of the basement membrane in two kinds of chimaeric intestines composed of chick mesenchyme and rat endoderm (Cm/Re) or inversely of rat mesenchyme and chick endoderm (Rm/Ce).

In Cm/Re recombinants, no staining with antimouse type-IV collagen antibodies was found (Fig. 6E), while a bright staining was produced with the anti-chick antibodies at the basement membrane region as well as within the lamina propria and the muscular layers (Fig. 6F). In the reciprocal recombinants Rm/Ce, mouse antibodies labelled the muscular layers, some cellular elements within the connective tissue and the epithelial–mesenchymal interface of the developed grafts (Fig. 6G). It is noteworthy that the anti-chick antibody stained some structures within the lamina propria and underlying the basement membrane region or scattered within the muscular layers (Fig. 6H). This reaction is presumably due to the presence of blood vessels originating from the chick embryonic host which have invaded the grafts. This assumption is supported by a similar labelling pattern in 14-day fetal rat intestines grafted under the same conditions (not illustrated). On the other hand, in these two kinds of chimaeric intestines, nidogen, laminin, fibronectin and type-III procollagen were produced and exhibited their characteristic distribution pattern.

In the present study, an in vitro coculture model system simulating the in vivo ‘epithelial–mesenchymal entity’ was used to study immunocytochemically the temporal aspects of expression and the distribution of extracellular matrix molecules during epithelial differentiation. Furthermore, the tissue source of type-IV collagen was analysed by means of interspecies tissue combinations. The data obtained emphasize, first, that ECM components are produced in cocultures and that basement-membrane-specific molecules are deposited at the epithelial–mesenchymal interface. This deposition does not seem to be organ specific as assessed by the substitution of skin fibroblasts for intestinal mesenchymal cells. The second main result concerns the role of the fibroblastic cells in the production and deposition of type-IV collagen in the basement membrane.

All ECM antigens studied were detected within the fibroblastic cell layer, whether these cells were cultured in isolation or in contact with endodermal cells. By contrast, the deposition of the major basement membrane components (nidogen, laminin and type-IV collagen) in a polar fashion was found only when both epithelial and mesenchymal cell types were associated. Such a cellular cooperation for basement membrane formation has already been demonstrated in various organs (Lipton, 1977; Sariola et al. 1984a and for a review, see Timpl & Dziadek, 1986). Moreover, from the present study it appears that the re-establishment of such a complex interface is gradual and parallels the epithelial cell differentiation described previously (Kedinger et al. 1987). Indeed, as early as 2 days after coculture, a spot-like distribution of these proteins at the basement membrane region occurred. At this stage, endodermal areas had spread; the cells are linked together by junctions, some apical microvilli become obvious but they do not express any digestive enzymes. With increasing coculture time, a gradual assembly of the antigens occurred, leading to a continuously labelled epithelial-mesenchymal interface after one week; morphologically, the epithelial cells resemble the characteristic absorptive cells at this stage with their elongated shape and their apical brush borders. Furthermore, they synthesize microvillar digestive enzymes and are able to respond to glucocorticoids via the mesenchyme (Kedinger et al. 1987). The present data can also be correlated to the previously described ultrastructural observations at the epithelial-mesenchymal junction during morphogenesis (Mathan et al. 1972) and in coculture (Kedinger et al. 1987), which emphasized that contacts between the epithelial cytoplasmic processes and the mesenchymal cells precede the formation of a continuous basement membrane.

Taken together these data point to a parallel development of epithelial differentiation and basement membrane formation, provided that both epithelial and fibroblastic cell layers are present. However, it still remains impossible to determine whether one of these two events acts primarily as trigger. Such a correlation has been investigated by different authors and led to somewhat controversial conclusions. In the kidney, a complete basement membrane seems to be required for terminal differentiation of epithelial cells (Ekblom et al. 1980, 1982). On the other hand, during skin development, it has been suggested that the basement membrane is not a prerequisite but rather an additional criterion of epithelial differentiation (Lavker & Sun, 1983; Bohnert et al. 1986).

Concerning the cellular source of the basement membrane components, we employed as an experimental approach, immunological methods (use of species-specific anti-type-IV collagen antibodies) and embryonic tissue recombinations. The model of in ovo grafted interspecies associations was preferred to the cocultures since it ensured a three-dimensional structure and allowed the differentiation of all epithelial cell types of the mature intestine (Kedinger et al. 1981, 1986b). We could show first, using chick mesenchyme/rat endoderm recombinants, that the epithelial cells did not produce any detectable type-IV collagen and second, in the inverse associations, that the basement membrane was labelled with the antibodies directed against the species from which the mesenchyme was taken. This provided evidence that collagen IV, a component of the basement membrane, is produced by the mesenchymal compartment. In relation to this conclusion, there are known instances of mesenchymal basal lamina formation. Kühl et al. (1984) have emphasized the role of the muscle fibroblasts in the deposition of type-IV collagen in the basal lamina of myotubes. In the developing mammary gland, mesenchymal fat pad precursor cells have been shown to synthesize laminin and proteoheparan sulphate; these components are deposited around individual fat cells in basement membrane-like structures (Kimata et al. 1985). Yet, the finding that mesenchymal cells produce and deposit a basement membrane component at the epithelial–mesenchymal interface is novel. However, from our observations, one cannot exclude an additional participation of the invading host capillaries in the formation of the basement membrane. Such a possibility is supported by the demonstration of a cooperation between interstitial and vascular cells for the formation of embryonic kidney basement membrane (Sariola et al. 1984a; Sariola, 1985).

Finally, the fact that intestinal endodermal cells cultured alone did not reveal any staining with the various ECM antibodies tested is not consistent with data reported for other cultured epithelial cell types that have been shown to produce ECM molecules (for a review, see Timpl & Dziadek, 1986). In particular this was the case for intestinal crypt cell lines (IEC cells), which synthesize and secrete collagens and fibronectin, the latter being detected in regions of cell-cell contacts (Quaroni et al. 1978; Quaroni & Trelstad, 1980). Although it cannot be excluded that low levels of immunocytochemically undetectable ECM molecules are synthesized by the cultured endodermal cells, this discrepancy could be due to the fact that such a production by established cell lines does not necessarily reflect the in vivo situation.

In conclusion, our model systems may be instrumental in further investigations of the cellular origin of extracellular products by in situ hybridization using cDNA probes and also the biosynthetic activity of both cell populations.

We would like to thank Drs R. Timpl and K. von der Mark (Max-Planck-Institut für Biochemie –Münich) for kindly providing antibodies and for their interest in this work. We are very grateful to E. Alexandre and C. Leberquier for skilful technical assistance. Financial support was given by INSERM and CNRS.