Mesenchyme-mediated effects of retinoic acid during rat intestinal development

Following association of the visceral endoderm with the splanchnic mesoderm into a closed tube in fetuses, intestinal organogenesis is completed by a progressive reorganization of the endoderm into a single epithelium which lines the growing villi. Basal crypts subsequently form from an epithelial down-growth in the intervillus regions. From this stage onwards, the mucosa is continuously renewed from crypt stem cells, which differentiate in the course of their crypt to villus migration (Henning et al., 1994; Kedinger, 1994). It is worth noting that during the whole developmental period, as well as in the mature gut, mesenchyme-derived cells are closely associated with the epithelial tissue (Richman et al., 1987; Sappino et al., 1989; Kedinger et al., 1990; Valentich and Powell, 1994). Particular attention has been directed towards the sub-epithelial fibroblasts surrounding the crypts; it has been postulated that these fibroblasts, characterized as myofibroblasts, play a role in regulating the epithelial proliferation and/or differentiation programs.

It is well known that, like in several organs, morphogenesis and differentiation in the intestine depend on epithelial-mesenchymal interactions (Birchmeier and Birchmeier, 1993; Simon-Assmann and Kedinger, 1993). The cross-talk between epithelial and mesenchymal cells has been demonstrated using several experimental models, including interspecies or heterotopic epithelial-mesenchymal recombinants (Kedinger et al., 1981; Haffen et al., 1983, 1989; Yasugi, 1993; Duluc et al., 1994). Substantial evidence supports the contention that extracellular matrix (ECM) components, and in particular the basement membrane (BM), which is a specialized ECM sheet located in close vicinity to the epithelium, are involved in this cross-talk (Simon-Assmann et al., 1995). Indeed, the BM is formed from a contribution of both cell types (Simon-Assmann et al., 1995), and several experiments have pointed to its major role in the differentiation process of intestinal epithelial cells in vitro (Simo et al., 1992; Kedinger, 1994; Vachon and Beaulieu, 1995; De Arcangelis et al., 1996). However, despite the function attributed to these BM molecules, little is known about their regulation.

In the present study we analyzed the effects of retinoic acid (RA) on intestinal development, and focused on the role played by the mesenchyme as mediator of RA action. Our interest in studying RA on the intestinal epithelium-mesenchyme unit is based on two general observations. First, RA and its metabolic analogues act as morphogens and as inducers of cell differentiation in many organs (De Luca, 1991; Leid et al., 1992; Gudas, 1994; Hofman and Eichele, 1994). Second, retinoids regulate the expression of BM molecules, of matrix proteases and of integrins (Verrando et al., 1988; Vasios et al., 1989; Nicholson et al., 1990; Rossino et al., 1991; Schule et al., 1991).

The biological role of retinoids is mediated by several classes of binding proteins located inside the cells: the cytoplasmic binding proteins including cellular retinol binding proteins (CRBP I and CRBP II) and cellular retinoic acid binding proteins (CRABP I and CRABP II), and the nuclear receptors that belong to the superfamily of ligand-activated trancription factors (Stunnenberg, 1993; Giguère, 1994). Two classes of nuclear receptors, each consisting of three receptor types, α, β, and γ, have been described: the retinoic acid receptors (RARs) and the retinoid X receptors (RXRs). Both RARs and RXRs control the transcription of target genes by interacting with cis-acting DNA responsive elements (Leid et al., 1992). Expression of CRBPs, CRABPs, RARs and RXRs during embryogenesis and in the adult is spatially and temporally restricted, suggesting that each class of binding protein may play a specific role, which has partly been elucidated in mice lacking receptors (Dollé et al., 1990; Lohnes et al., 1995; Kastner et al., 1995). The intracellular concentration of free RA is critical in determining the extent of activation of RA nuclear receptors, and the cytoplasmic CRABP I and II have been proposed to fine-tune the intracellular concentration of free RA either by sequestering it or by facilitating its catabolism (Boylan and Gudas, 1991; Fiorella and Napoli, 1991; Ruberte et al., 1991). In this context, it is noteworthy that CRBP I (Smith et al., 1991), CRBP II (Mangelsdorf et al., 1992), and CRABP II (Giguère et al., 1990) gene expression is induced by RA.

Sporadic studies have reported the expression of specific subclasses of RARs and RXRs, and of cytoplasmic retinoid binding proteins in the gut, mostly in the epithelial cells and occasionally in the mesenchyme (Dollé et al., 1990; Mangelsdorf et al., 1992; Plateroti et al., 1993). To get more insight into the role of RA in intestinal development and cell interactions, we tried to answer the following questions: first, is the developing gut sensitive to RA? For this purpose we used two experimental models, short-term organ culture and long-term xenograft of rat jejunum anlagen. Second, does RA act on the epithelial-mesenchymal cell interactions? To answer this point, we established and characterized postnatal, subepithelial mesenchyme-derived intestinal cell lines (MIC), and analyzed their role in promoting epithelial cell differentiation upon RA treatment in a co-culture system. Taken together, the results demonstrate that in our gut models RA: (1) enhances the expression of retinoid-specific cytoplasmic and nuclear receptors; (2) acts as a morphogen; (3) induces epithelial polarization only in the presence of mesenchyme-derived cells; and (4) modulates BM molecules, laminin and collagen IV.

Wistar rats were from our breeding colony. Rat fetuses were delivered by caesarean section at indicated days of gestation (the existence of the vaginal plug was designated day 0). Swiss athymic nude mice (nu/nu, Iffa Credo, France) were used as hosts for the grafts.

Fourteen-day fetal proximal jejunum (PJ) anlagen (2-4 mm long) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco BRL, Cergy-Pontoise, France) supplemented with 0.2% heat-inactivated fetal calf serum (FCS, Gibco BRL), 0.25 U/ml insulin (Sigma, Saint-Quintin-Fallavier, France), 10 μg/ml transferrin (Sigma), 20 ng/ml EGF (Sigma) and 40 μg/ml gentamycin (Schering-Plough, Segré, France). The explants were cultured into culture dishes for 3 days under continuous gentle agitation to avoid attachment. After the first 24 hours in culture, 10⁻⁸ M all-trans retinoic acid (RA, Sigma) or the same volume of solvent (1 μl ethanol/ml medium) were added to the medium. At the end of the culture period, the fragments were embedded in Tissue-Tek (Labonord, Villeneuve d’Asq, France) and frozen with isopentane cooled by liquid nitrogen for immunocytochemical analysis on cryosections. Alternatively the explants were immediately frozen in liquid nitrogen for subsequent RNA extraction.

Fourteen-day fetal PJ anlagen were grafted under the skin of the flank of nude mice on both sides of a median line (6 samples/mouse). To evaluate the role of physiological concentrations of RA on intestinal morphogenesis, DEAE-Sephadex A-50 beads (Pharmacia Biotech Orsay, France) were soaked for 30 minutes at room temperature in 10⁻⁸ M RA dissolved in DMSO according to the method of Eichele et al. (1984) and then washed twice in PBS. Samples (100 μl) of PBS containing 20 beads were injected under the skin of nude mice near the grafts every 48 hours. In controls, the beads were soaked in DMSO alone. After 10, 14 and 21 days, the grafts were recovered and processed for cryosectioning and for RNA extraction.

Fourteen-day fetal mesenchyme and endoderm from proximal jejunum (PJ) were separated as previously described (Kedinger et al., 1981; Duluc et al., 1994). Briefly the fetal intestines were incubated in Clostridium histolyticum collagenase (Boehringer Mannheim, Meylan, France) 0.03% in CMRL 1066 medium (Gibco BRL) at 37°C for 1 hour. Intestines were then transferred into Ham’s F-10 (Gibco BRL) supplemented with 50% newborn bovine serum (Gibco BRL) for at least 30 minutes at room temperature to block collagenase activity. The mesenchymes were opened lengthwise with a microscalpel and the endoderms pushed out of the mesenchymal gutter with forceps. For the 17- and 19-day-old fetal intestines, the outer muscular layers were first discarded; the mesenchyme/lamina propria was then separated from the epithelium after a 5 minute incubation of the dissected PJ fragments in 1 mM EDTA solution at 37°C. Three-dayold intestinal lamina propria and epithelium were separated according to the same procedure, by incubating the dissected fragments in 5 mM EDTA solution for 15 minutes at 37°C.

All tissue samples were immediately frozen in liquid nitrogen and stored at −80°C until use for RNA extraction. The purity of the preparations was controlled on cryosections by immunocytochemical staining for either cytokeratin or vimentin.

Eight-day post-natal rat intestinal lamina propria cells derived from PJ were obtained after a 10 minute incubation with 300 U/ml of collagenase XI (Sigma) and 0.1 mg/ml of dispase (Boehringer Mannheim) in Hanks’ balanced salt solution (HBSS-Gibco BRL). The tissues were then cut into small fragments. The fibroblasts tightly linked to the epithelial cells have been preferentially selected according to the method of Evans et al. (1992). After a low speed centrifugation, the supernatant containing isolated mesenchymal cells was discarded; the pellet comprising the large epithelial fragments and the subepithelial fibroblasts was mechanically triturated by pipetting. The explants were washed five times in DMEM-2% sorbitol, seeded in culture dishes and cultured in DMEM supplemented with 10% FCS, 0.25 U/ml insulin, 10 μg/ml transferrin, 20 ng/ml EGF and 40 μg/ml of gentamycin. After four days, mesenchymal cells were passaged using 0.01% Trypsin (Gibco-BRL)-2 mM EDTA treatment; in these conditions epithelial cells do not survive. The subepithelial fibroblast population was cloned twice using the limit dilution technique, and several cell lines named MIC (mesenchyme-derived intestinal cell lines) have been obtained.

MIC lines have been used for co-culture experiments with endodermal microexplants prepared from the PJ of 14-day rat fetuses. Small fragments of endoderm (less than 1 mm²) were seeded over confluent fibroblasts and maintained in co-culture for 3 days in DMEM supplemented with 2.5% FCS, 0.25 U/ml insulin, 10 μg/ml transferrin, 20 ng/ml EGF and 40 μg/ml of gentamycin. After 24 hours in standard conditions, 10⁻⁸ M RA was added to the culture medium; the same volume of solvent was added in control dishes. The areas of epithelium/fibroblasts co-cultures were recovered under a dissecting microscope and processed for cryosectioning and/or RNA extractions. Alternatively they were processed for morphological studies: the cocultures were fixed for 1 hour at 4°C in 0.2 M cacodylate-buffered 2% glutaraldehyde, pH 7.4, postfixed in cacodylate-buffered 1% osmium tetroxide, pH 7.4, for 30 minutes at 4°C, dehydrated and embedded in Araldite. Semithin sections (0.5 μm) were stained with toluidine blue for histological observation.

Frozen sections (5 μm) of tissues or co-cultures as well as cells cultured on glass coverslips were fixed in 2% paraformaldehyde for 15-20 minutes at room temperature, washed in PBS and incubated with the specific primary antibodies for 1 hour. After three washes they were incubated for 30-45 minutes with FITC-labelled sheep antimouse IgG (Sanofi Diagnostic Pasteur, Marnes-la-Coquette, France), or Texas Red-labelled goat anti-mouse IgG (Amersham, Les Ulis, France) or FITC-labelled goat anti-rabbit IgG (Nordic Immunological Laboratories, Capistrano Beach, CA). After mounting under coverslips in phenylene/PBS/diamine buffer, the preparations were observed under a Zeiss Axiophot fluorescence microscope.

The polyclonal anti-laminin antibody, produced in our laboratory, was raised against mouse EHS laminin. On western blots it recognizes the three constituent chains, α1, β1 and γ1 (Simo et al., 1991). By immunofluorescence, it may label any laminin containing α1, β1 or γ1 chain (Timpl, 1996). Polyclonal anti-type IV collagen antibodies, raised against type IV collagen extracted from EHS tumors, prepared in our laboratory, recognize specifically the α1 and α2 type IV collagen chains as tested by immunoblot analysis. The monoclonal antibody (mAb) for cytokeratin, raised against adult rat intestinal crypt cells and recognising a single band of 46 kDa by immunoblot analysis (Plateroti et al., 1993), was a gift from Dr A. Le Bivic (CNRS, Marseille, France). mAb against the brush border enzyme lactase was a generous gift from Dr A. Quaroni (Cornell University, New York, USA; Quaroni, 1983). mAb to desmin was from Dako (Trappes, France), mAb to vimentin from Amersham France and mAb to smooth muscle α-actin from Sigma.

Due to the small size of most samples studied (early developmental stages, isolated epithelial and mesenchymal anlagen, co-cultures and grafts) and thus to the very low amounts of RNA recoverable from our experimental models we analyzed gene expression by RT-PCR.

RNA was extracted from epithelium/mesenchyme dissociated tissues, organ cultures, grafts, mesenchyme cell cultures and co-cultures, by using Trizol reagent (Gibco BRL) as recommended by the supplier. Single-stranded cDNAs were synthesized for 1 hour at 42°C using 3 μg of RNA, 50 pmol oligo (dT)₁₇, 1 mM of all four deoxynucleotide triphosphates, and 15 units of AMV reverse transcriptase (Promega) in 50 mM Tris-HCl, pH 8.3, 50 mM KCl, 10 mM MgCl₂, and 4 mM sodium pyrophosphate. The transcripts analyzed in the study were the cytoplasmic (CRBP I, CRBP II and CRABP I) and nuclear (RAR β and RAR γ) retinoic acid binding proteins, markers of smooth muscle cell and myofibroblastic differentiation (smooth muscle α-actin), BM molecules (α1 and β1 laminin chains), and β actin as control.

The primers used for PCR analysis were synthesized chemically (Eurogentec, Serain, Belgium). They are listed in Table 1.

Polymerase chain reaction was carried out on the cDNA mixture (1/10), using 50 pmoles of each oligonucleotide pairs. The reaction was performed in 100 μl of 75 mM Tris-HCl, pH 9, 20 mM (NH₄)₂SO₄, 0.01% Tween-20, 1 mM MgCl₂, 0.2 mM of each dNTP, and 1 unit of Goldstar DNA polymerase (Eurogentec). PCR was performed in a Thermojet apparatus (Eurogentec) using the following conditions: 45 seconds at 94°C for denaturation, 45 seconds at the appropriate temperature for annealing (see Table 1), 1 minute of elongation at 72°C, and finally 5 minutes at 72°C. Samples (10 μl) of the PCR products were separated by electrophoresis on 3% agarose gels, stained with ethidium bromide and observed under UV illumination. In controls, the cDNA templates were either omitted or replaced by 0.3 μg of RNA. Semi-quantitative RT-PCR analysis has been performed (specific mRNA versus β-actin mRNA as control): for every oligonucleotide pair and for every RNA species, a preliminary analysis was conducted to define the appropriate range of cycles consistent with an exponential increase of the amount of the DNA product (Table 1). Every PCR fragment was inserted into the pGEMT vector (Promega) and sequenced using the T7 Sequencing kit (Pharmacia).

In order to test the role of retinoids on the developing gut, we studied short-term and long-term effects of 10⁻⁸ M RA treatment using two experimental models: organ cultures and xenografts of 14-day fetal proximal jejunum (PJ), in which the endoderm is still undifferentiated and stratified.

Organ cultures of PJ anlagen for three days in the absence of RA resulted in the evolution of the stratified endoderm into a single-layered epithelium like in vivo during the same time interval. Staining with anti-cytokeratin antibody illustrates an almost flat epithelium (Fig. 1A), exhibiting poor cytodifferentiation, as assessed by the absence of lactase at the apical pole of the enterocytes (Fig. 1C) (lactase is the first digestive enzyme localized in the apical membrane of epithelial cells to be expressed during development). During the same period of time, RA treatment provoked the outgrowth of small growing villi (Fig. 1B). In addition, it accelerated the cytodifferentiation process of the epithelium, as demonstrated by the presence of lactase positive epithelial cells in the upper part of the small villi (Fig. 1D).

Fourteen-day fetal rat PJ were implanted under the skin in nude mice and recovered 10, 14 and 21 days later. It has already been reported that such grafting conditions allow the normal ongoing of the ontogenic program of the intestine (Winter et al., 1991; Rubin et al., 1992; Duluc et al., 1994). Control grafts implanted for 10 days in the absence of RA exhibited a typical small intestine morphology characterized by the formation of villi and small intervillus crypt buds (Fig. 2A), and by the functional cytodifferentiation and polarization of the epithelium, demonstrated by the cytokeratin and lactase immunolabelling (Fig. 2C,E). To investigate the long-term effects of RA on intestinal development, DEAE-Sephadex RA-soaked beads were injected every 2 days near the implants. A 10-day RA treatment resulted in an increased number of crypts (Fig. 2B,D), accompanied by an important cellular exfoliation at the apex of villi. Lactase was found at the apical pole of the villus epithelial cells like in the control grafting conditions (Fig. 2E,F). In the RA-treated grafts the depth of the smooth muscle α-actin positive external muscular coat was increased compared to the controls (Fig. 2H versus G). This observation was confirmed by an increase in smooth muscle α-actin mRNA in the RA-treated conditions (Fig. 3, lane d versus c). In intestinal segments which have developed up to 14 and 21 days in control or RA-treated nude mice, the mucosa no longer showed great morphological differences; however, the effect of RA on the depth of the muscular layers remained clearly visible (data not shown).

These data demonstrate that RA acts on gut morphogenesis and differentiation. In short-term organ cultures it accelerates villus outgrowth and epithelial cytodifferentiation, while in long-term xenografts it provokes an increase in the number of crypts and in the depth of muscular layers.

Previous experiments have shown that gut morphogenesis and differentiation depend on epithelial-mesenchymal cell interactions and that the instructive signalling may involve BM components, expressed and secreted by both tissue compartments (Simon-Assmann et al., 1995). Therefore, we analyzed the localization of laminin and type IV collagen by immunofluorescence using polyclonal antibodies recognizing, respectively, the α1, β1, γ1 laminin constituent chains, and the α1 and α2 collagen IV chains. In control organ culture experiments, we observed that both laminin (not shown) and collagen IV (Fig. 1E) staining was localized in the subepithelial basement membrane (BM) as well as in mesenchymal derivatives including the developing external muscular layers and evenlydistributed cellular elements in the connective tissue. In RA-treated cultures, additional positive mesenchymal cells were concentrated in the growing villi (Fig. 1F). The semi-quantitative RT-PCR analysis of the mRNAs encoding the α1 and β1 laminin chains showed that both transcripts were more abundant in the RA-treated samples than in the controls (Fig. 3, lanes a,b). In the xenografts, laminin immunofluorescence staining at the subepithelial basement membrane level as well as in the mucosal connective tissue was similar in both controls and RA-treated grafts (Fig. 2I,J). This was obvious in the 10, 14 and 21-day grafts analyzed. The mRNA analysis of laminin α1 and β1 chains revealed a higher level of expression of these transcripts in the RA-treated implants (Fig. 3, lanes c-f). Note that in the 21-day grafts, the level of both transcripts is very low in the control grafting conditions, confirming their low turnover in mature organs (Simon-Assmann et al., 1995).

The mRNA expression of several retinoid binding proteins was analyzed by RT-PCR; we focused on the expression of the cyto-plasmic proteins CRBP I, CRBP II and CRABP I, and of the nuclear RAR β and RAR γ. We omitted CRABP II, referred to as skin specific protein (Eller et al., 1992), and RAR α because of its ubiquitous expression (Dollé et al., 1990).

The results show that various levels of each mRNA were found in the different control experimental conditions (Fig. 3, lanes a,c,e). In all the controls, the CRBP II and CRABP I were faintly expressed. The RAR β mRNA was not detected in the controls nor in RA-treated samples, except for a very faint expression in xenografts implanted for 10 days in the presence of RA. The CRBP I mRNA was detected in all controls, like RAR γ with the exception of the xenografts maintained for 21 days. It is worth noting that RA treatment (Fig. 3, lanes b,d,f) led to an increase in the expression of most of these genes mainly in organ cultures and in the 10-day grafts.

Having shown that RA acts on the developing gut and in order to further analyze the potential participation of the mesenchymal tissue in this response, we needed to have a more precise knowledge of retinoid binding protein genes expressed in the mesenchyme. For this purpose, an RT-PCR analysis has been performed on separated jejunal epithelium and mesenchyme at various developmental stages. The stages were chosen for their correspondence with the starting (14-day fetal intestines) and end points (17-day fetal and 3-day post natal intestines) of the 3-day organ cultures and the 10-day grafting experiments, respectively. An intermediate stage of intense morphogenesis has also been included in this study (19-day fetal intestines). The purity of the preparations was assessed by immunostaining with antibodies to epithelial and mesenchymal markers, respectively cytokeratin and vimentin. In the four developmental stages analyzed no cross contamination was found in the tissue preparations (data not shown). The results of this study are illustrated in Fig. 4 and summarized in Table 2. Interestingly, CRBP I and CRABP I mRNAs were predominantly expressed in the mesenchymal compartment. Their expression was up-regulated during morphogenesis up to 19-day fetal intestine, and decreased after birth. CRBP II mRNA is well known to be expressed in epithelial tissues (Levin et al., 1987); our results confirm these data and show that the amount of this transcript increased during development. In addition CRBP II mRNA was also detected in the postnatal mesenchyme. The mRNA encoding the nuclear receptor RAR γ was expressed in both mesenchyme and epithelium with a peak at 19 days in the fetal mesenchyme tissue and at 17 days in fetal epithelial preparations. On the contrary, we did not detect any transcript for RAR β whatever the developmental stage (not shown).

On the basis of the results described above in organ cultures and xenografts, we analyzed whether the differentiating effect of RA on intestinal epithelial cells was dependent on the presence of mesenchymal cells. For this purpose, fetal jejunal endoderm was separated from the mesenchyme and cultured alone for 3 days in the presence of RA. This treatment did not provoke any acceleration of differentiation, assessed by RT-PCR analysis of lactase expression (not shown). In addition, when endoderms were cultured alone in RA-complemented medium no significant changes in the faint retinoid binding proteins gene expression were observed (data not shown, see Fig. 4 for the faint expression of retinoid binding proteins in 14-day intestine endoderms). These data suggest that RA does not act directly on epithelial cells but that its effect is mediated by mesenchymal cells.

To approach this hypothesis, an in vitro model of well characterized homogeneous intestinal mesenchymal cells was required to analyze the effects of RA on co-cultures of endoderm and intestinal mesenchymal cell lines. We choose to establish subepithelial mesenchymal cell lines, isolated from the lamina propria of 8-day-old PJ. Two of them (MIC 101-1 and MIC 219) have been selected on the basis of their ability to support epithelial cell growth and differentiation in cocultures and of their homogeneous expression of vimentin and smooth muscle α-actin (Fig. 5A and B). These cell lines have been subcultured at least 10-fold; some of them are now at passage 20.

To evaluate the role of the mesenchymal cell lines in the epithelial response to RA, microexplants of 14-day fetal intestinal endoderms were co-cultured on top of confluent MIC 101-1 or MIC 219 cells. RA was added to the culture medium from days 1 to 3, and the bilayered mesenchyme-endoderm areas were collected for morphological and immunocytochemical studies, and for RNA analysis. The results were similar for both the MIC lines. Fig. 6 shows semithin crosssections of epithelial cells co-cultured onto MIC 101-1 cells. In the presence of RA, the polarization of the epithelial cells was significantly enhanced compared to the controls, as assessed by the regular arrangement of the cells, and by the presence of apical microvillar structures. The immunofluorescence staining with anti-cytokeratin antibody confirmed the enhanced morphological polarization of the endodermal cells upon RA treatment (Fig. 7A,B). In parallel, RA also induced the functional differentiation of the epithelium, as demonstrated by the presence of lactase at the apical pole of the epithelial cells in the treated cultures (Fig. 7C,D). This epithelial response to RA was not obvious when endodermal cells were cultured in conditioned medium from MIC 101-1 RA treated cells (data not shown), suggesting that the effect of RA on epithelial cell differentiation was not, or at least not only, related to soluble factors secreted by the mesenchyme cells.

Because of the major role played by molecules of the basement membrane on intestinal epithelial cytodifferentiation, we have analyzed the effect of RA on laminin and collagen expression in MIC 101-1 cultured alone or in association with fetal intestinal endoderm. The immunofluorescence labelling of laminin and collagen IV in control and RA-treated co-cultures using the clone MIC 101-1 is illustrated in Fig. 7. A staining at the BM level as well as of the fibroblast layer was observed in the control conditions (Fig. 7E,G). However co-cultures in the presence of RA resulted in a more intense staining of laminin and collagen IV at the heterologous cell interface as well as in the fibroblast layer (Fig. 7F,H). In addition a significant accumulation of laminin was seen in the epithelial cell layer (Fig. 7F). The study of the α1 and β1 laminin chains mRNAs in the co-cultures comprising MIC 101-1 cells and endoderms, confirmed the immunohistochemical observations. Indeed the amount of laminin α1 and β1 mRNAs was increased in the RA treated co-cultures compared to the controls (Fig. 8, lanes c,d). It is worth noting that mesenchymal cell lines cultured without endodermal cells displayed an increased α1 and, to a lesser extent, β1 mRNA expression in response to RA (Fig. 8, lanes a,b). In contrast, in endoderms cultured in isolation, the signals were unchanged in the RA-treated versus control cells (Fig. 8e,f).

MIC 101-1 cells cultured alone exhibited some expression of the CRBP I and CRBP II mRNAs, but a very low level of CRABP I and RAR γ mRNAs (Fig. 8, lane a). The addition of RA to the culture medium of the fibroblasts increased the levels of both CRBP I and CRABP I transcripts (Fig. 8, lane b); in contrast, CRBP II and RAR γ remained unchanged. The coculture conditions per se enhanced the levels of the CRBP I, CRABP I and RAR γ transcripts (Fig. 8, lane c), and an additional effect of RA was obvious in the case of CRBP II and RAR γ mRNAs (Fig. 8, lane d).

Taken together, the comparison of the different parameters in the co-cultures and in endodermal or homogeneous mesenchymal cell populations cultured in isolation, indicates that RA induces molecular changes in the mesenchymal cells and that these cells are required to allow RA-dependent epithelial cell differentiation.

In this study, we provide evidence that cytoplasmic and nuclear retinoid binding protein genes are expressed in the intestinal epithelium and mesenchyme, at developmental stages consistent with important steps of the morphogenesis and cytodifferentiation of the digestive mucosa. In addition, we demonstrate that administering RA to the developing gut results in: (i) a morphogenetic induction as assessed by the outgrowth of small villi and by the thickening of the muscular layers; (ii) an accelerated maturation of the epithelium as illustrated by the enlarged crypt compartment and by the expression of lactase at the tip of the growing villi; and (iii) an increased expression of basement membrane components. In addition, by using homogeneous subepithelial mesenchymal cell cultures we report data showing that these cells are required to mediate the differentiating effect of RA on the intestinal epithelium.

The role of RA and its metabolic analogues in development, cell differentiation, and positional information, has been described for many organs (De Luca, 1991; Ragsdale and Brokes, 1991; Stunnenberg, 1993; Mangelsdorf, 1994). However, little is known about the RBPs expression in the intestine. In the present study we report that several RBPs transcripts are to be found in the endodermal and/or mesenchymal tissues during intestinal development. Mainly we show that mRNAs encoding the cytoplasmic proteins CRBP I and CRABP I are predominantly found in the mesenchyme, whereas CRBP II mRNA is mostly expressed in the epithelium. The mesenchymal CRBP I and CRABP I transcripts are the most abundant in 17- to 19-day fetuses, at a stage coinciding with important morphogenetic and cytodifferentiation events in the intestine. On the other hand, the epithelial CRBP II transcript is faintly expressed at early developmental stages and increases after birth. In contrast to previous works, in which CRBP II was described as the intestinal epithelial-specific isoform (Levin et al., 1987), we also detect CRBP II transcripts in the mesenchymal compartment of post-natal animals. This may be due to the higher sensibility of the RT-PCR analysis over other techniques. Interestingly, we demonstrate for the first time that RAR γ mRNA is present in both mesenchyme and epithelium and is developmentally regulated. However, we did not detect any RAR β transcripts although RAR β mRNA has been observed in the intestine of 10.5- to 14.5-day murine embryos (Dollé et al., 1990). The reason for this is not clear at present. Some differences in the distribution of retinoid binding proteins have also been observed along the longitudinal axis of the developing intestinal tract, perhaps suggesting the existence of distinct and specific effects of RA in the jejunum, ileum and colon which display important morphological variations (unpublished data).

The results obtained upon RA administration in organ cultures and xenografts strongly suggest that retinoids play an active role in intestinal development. During the three-day organ cultures of fetal intestinal explants, the consequences of RA treatment include the induction of villus outgrowth and the expression of lactase, an apical digestive enzyme. These effects are consistent with the known morphogenetic and differentiating properties of retinoids (De Luca, 1991; Ragsdale and Brokes, 1991; Stunnenberg, 1993). In the xenografts implanted for 10 to 21 days in the murine hosts, the fetal samples recapitulate normal intestinal development, and in particular they overcome the stage of crypt formation (Winter et al., 1991;Duluc et al., 1994). In this model, RA provokes the appearance of an increased number of crypts as well as the thickening of muscular layers, which may again be considered as morphogenetic effects. Altogether, these results suggest that RA acts at several steps of intestinal development: in fetuses during the re-organization of the stratified endoderm into a single-layered epithelium exhibiting protruding villi, and in neonates during the formation of crypts.

Previous evidence indicates that intestinal epithelial cell lines cultured in isolation were not affected in their differentiation program by retinoids (Plateroti et al., 1993). We confirm here that RA treatment has no effect on 14-day fetal jejunum endoderms maintained in primary cultures. Furthermore, the consequences of RA treatment on co-cultures of endodermal and mesenchymal cells clearly indicate that the effect of RA on differentiation of the epithelial cells is observed only when contact with mesenchymal cells occurs. The homogeneous subepithelial mesenchymal cell lines of the MIC series, established for this study, recapitulate the retinoid binding proteins expression pattern of the freshly isolated tissues, respond to the RA treatment and are able to mediate the differentiating effect of RA on epithelial cells. Therefore, they constitute an efficient tool for in vitro studies of intestinal epithelial-mesenchymal cell interactions. Similar action of RA on mesenchymemediated effects during development have already been reported in other organs: in the chick facial primordia (Wedden, 1987), in developing mouse lung (Schuger et al., 1993), and in tooth (Mitsiadis et al., 1995a). In the latter two tissues, RA may act through mechanisms which involve the regulation of mesenchymal EGF as a paracrine factor and epithelial EGF receptors, or of Notch genes, respectively. Recently, other molecules have been thought to be involved in the cellular cross-talk, such as TGF-β, midkine, hepatocyte growth factor/scatter factor (HGF/SF) and its receptor c-met, or epimorphin (Hirai et al., 1992; Rosen et al., 1994; Mitsiadis et al., 1995b,c). Interestingly, the expression of TGF-β and midkine is modulated by RA (Muramatsu, 1994; Danielpour, 1996). In the intestinal model, conditioned medium from RA-treated mesenchymal cells did not affect the endodermal cells. Furthermore, RA did not alter the expression of HGF/SF in the mesenchymal lines nor in the co-cultures (data not shown). In contrast, in the three experimental models used, intestinal morphogenesis and differentiation are paralleled by changes in the expression of two BM constituents, laminin and collagen IV, upon RA administration. This effect coincides with the up-regulation by RA of most RBPs studied in the various experimental conditions and is consistent with an RA regulation of several genes involved in ECM remodeling. A cis-acting responsive element to RARs has been described in the promoter of the laminin β1 gene (Vasios et al., 1989), and the expression of type IV collagen has been reported to be affected by RA in vitro (Mummery et al., 1990). Interestingly, we obtained a stimulating effect of RA on the expression of laminin α1 gene, whose promoter has not yet been described. It is worth noting that the gene promoters of two proteases, stromelysin 1 and collagenase 1, contain responsive elements to RARs (Nicholson et al., 1990; Schule et al., 1991), and that the expression of laminin-5 and of the α1β1 integrin, changes upon RA administration (Verrando et al., 1988; Rossino et al., 1991). It is noteworthy that previous work in our laboratory allowed us to stress that BM components play a major role in intestinal epithelial-mesenchymal interactions and in epithelial cell differentiation. Indeed several observations using coculture models provided evidence that: (i) deposition of BM at the epithelial-mesenchymal interface precedes the expression of differentiation markers (Kedinger, 1994); and (ii) polyclonal antibodies to laminin inhibit the expression of the brush border enzyme lactase (Simo et al., 1992). More recently the inhibition of the α1 laminin chain expression in colonic cancer cells, by using an antisense RNA strategy, demonstrated its prime importance in mediating epithelial-mesenchymal contacts and epithelial differentiation (De Arcangelis et al., 1996).

The fact that RA enhanced BM molecules in the three experimental models used as well as in isolated mesenchymal cell cultures, but not in endodermal cell cultures, may indicate that RA modulates the functional composition of ECM produced by the mesenchymal cells; ECM constituents in turn may either mediate the effect of RA on epithelial cells, or allow the latter to become responsive to RA. RA is known to act by modulating the expression of nuclear transcription factors. Homeobox-containing genes, which are involved in patterning and cell differentiation, represent a major target of RA, as analyzed in detail in teratocarcinoma stem cells (reviewed by Gudas, 1994). Interestingly, homeobox genes are expressed in the intestine throughout development (James and Kazenwadel, 1991; Freund et al., 1992). It is also worth noting that several genes encoding cell adhesion molecules constitute molecular targets of homeo-proteins (Botas, 1993; Edelman and Jones, 1993). Moreover, we have recently shown that cdx2, an intestinal homeobox gene which triggers epithelial cell differentiation markers and integrin expression varies in parallel to laminin expression (Lorentz et al., submitted).

In conclusion, this study demonstrates that RA acts as a morphogen at important stages of intestinal development, and indicates that RA may be a regulator of the epithelial-mesenchymal unit by modulating the expression of basement membrane molecules. The establishment of homogeneous subepithelial mesenchymal cell lines which support RA-dependent epithelial differentiation, represents an interesting in vitro cellular model which will enable us to analyze the regulation of molecular targets of RA, as well as of additional factors regulating intestinal development, epithelial cell renewal and differentiation.

We gratefully acknowledge the excellent technical support of C. Arnold. We are especially grateful to Dr Patricia Simon-Assmann for ongoing discussion, advice, interest and critical reading of the manuscript, as well as for the gift of the laminin and collagen type IV anti-bodies. We are indebted to Dr A. Quaroni and Dr A. Le Bivic for the gift of anti-lactase and anti-cytokeratin antibodies. We thank Dr I. Duluc for her help at the beginning of this work, Mrs I. Gillot and L. Mathern for the preparation of the manuscript and illustrations. Dr M. Plateroti is currently supported on a postdoctoral fellowship from the European Science Foundation. Previously, she held a postdoctoral fellowship from the Ministère des Affaires Etrangères. Financial support comes from INSERM, ARC (grant N° 1251), Ligue Nationale contre le Cancer and the Fondation Aupetit.