Non-overlapping expression of CRBP I and CRABP I during pattern formation of limbs and craniofacial structures in the early mouse embryo

Retinoic acid (RA), an acidic derivative of vitamin A, affects several processes and tissues during vertebrate embryonic development, both as a putative morphogen involved in pattern formation, and as a potent teratogen when exposure of developing embryos occurs. The exact role of RA during embryonic development is, however, not well understood. Of the tissues affected by exogenous RA during pattern formation, the limb is by far the most studied (reviewed by Tickle, 1991), but craniofacial structures and CNS seem to be even more sensitive to the teratogenic action of RA (Lammer et al., 1985; Webster et al., 1986; Morriss-Kay, 1991).

In the developing limb, exogenous RA mimics the action of a small group of mesenchymal cells located in the posterior margin of the limb (the polarizing region, ZPA) which can specify the anterior-posterior axis (Saunders and Gasseling, 1968; Tickle et al., 1975, 1982; Summerbell, 1983). These observations, related to the found asymmetrical distribution of endogenous RA, suggested that RA was the morphogen released by the ZPA during limb development (Thaller and Eichele, 1987). Compelling evidence that ZPA generates and releases RA is, however, lacking and the role of RA as a morphogen during limb development has recently been questioned (Noji et al., 1991; Wanek et al., 1991).

The various biological effects of RA are mediated by two classes of nuclear RA-binding receptors, the RARs and the RXRs (Giguére et al., 1987; Petkovich et al., 1987; Benbrook et al., 1988; Brand et al., 1988; Krust et al., 1989; Zelent et al., 1989; Mangelsdorf et al., 1990, 1992). These proteins are ligand-controlled transcriptional enhancer factors related in structure to steroid and thyroid hormone receptors, and a number of genes regulated by the RA receptors have been identified (reviewed by Gudas, 1991). The various RA receptors are differentially expressed in the embryo -both temporally and spatially -in a complex way, which may be instrumental in RA-controlled gene expression (Dollé et al., 1989, 1990; Osumi-Yamashita et al., 1990; Ruberte et al., 1990, 1991; Rowe et al., 1991; Mangelsdorf et al., 1992). The ligand of the RARs is all-trans RA while the preferred ligand of the RXRs is 9-cis RA (Heyman et al., 1992; Levin et al., 1992). The significance of the RXR pathway during early embryonic development is unknown since it is not clear whether the RXRs are expressed before day 10.5 post coitus (p.c.) (Mangelsdorf et al., 1992).

In addition to the nuclear RA receptors, two classes of cytoplasmic retinoid-binding proteins have been identified, namely the cellular retinol-binding proteins (CRBP type I and II) and the cellular retinoic acid-binding proteins (CRABP type I and II) (Chytil and Ong, 1984; Ong, 1984; Giguére et al., 1990). The CRBPs bind retinol as their endogenous ligands. CRBP I has recently been shown to be a substrate carrier in the generation of retinal from retinol, the first oxidative step in the conversion of retinol into RA (Posch et al., 1991). CRBP I should thus be expressed in cells able to generate RA. The CRABPs, in contrast, bind RA and have been suggested to sequester RA in the cytoplasm and prevent it from reaching the nuclear receptors (Maden et al., 1989, Boylan and Gudas, 1991, reviewed by Maden, 1991). Alternatively, the fact that CRABP I is expressed in embryonic tissues and cells sensitive to exogenous RA, and that radiolabelled RA accumulates in these sites, may indicate that cells expressing CRABP I are target cells for RA (Dencker et al., 1987, 1990, 1991). The role of CRABP II is not clear and it binds RA with a significantly higher Kd value than CRABP I (Bailey and Siu, 1988; Siegenthaler et al., 1992)

Similar to most adults cells, the embryo receives retinol as the major source of retinoid. In order to understand the role of RA during embryonic development, it appears essential to reveal the mechanism and the site(s) of RA generation in the developing embryo, and to correlate this to the expression patterns of the nuclear RA receptors and to genes that are regulated by RA. Several investigators have approached these questions by analyzing the expression of endogenous RA-sensitive genes or reporter constructs using transgenic animals (Mendelsohn et al., 1991; Noji et al., 1991; Rossant et al., 1991). However, the used genes may not faithfully reflect endogenous levels of RA and it is not known how the CRABPs affect the RA-induced activation of these genes. As mentioned earlier, expression of high levels of CRABP I has been shown to prevent RA from entering the nucleus and thus activate reporter genes (Boylan and Gudas, 1991). In contrast, embryonic cells expressing CRABP I have the highest levels of RA, at least when measured by accumulation of exogenous RA (Dencker et al., 1987, 1990, 1991).

Of the genes known to be RA sensitive, the Hox genes are the most interesting since they may provide a link between pattern formation and the effects of RA on embryonic development. In teratocarcinoma cells and some embryonic tissues, high concentration of RA can activate transcription of several Hox genes (Ipisúa-Belmonte et al., 1991; Morriss-Kay et al., 1991; Nohno et al., 1991; Sime-one et al., 1991) and ectopic expression of Hox-1.1 in transgenic mice resulted in malformations similar to those obtained in RA-treated embryos (Balling et al., 1989). Thus, if endogenous RA plays a major role in the transcriptional regulation of Hox genes, we would expect the machinery used to synthesize RA in the embryo to be expressed in a complementary pattern.

To approach the question of how and where RA is generated in the early embryo, we have examined the pattern of expression of CRBP I by immunohistochemistry. Furthermore, the sites of accumulation of exogenously added radiolabelled retinol in embryos was examined using an autoradiographic technique. This paper focuses on three structures in the embryo that are known to respond to RA, namely the limbs, the branchial arches and the frontonasal area. In these structures, we found radioactivity and CRBP I to co-localize, mainly in restricted domains of ectoderm and in endoderm.

Three months old C57Bl/6 mice (Alab, Solna Sweden) were kept at 23°C and with a 12/12 hour light/dark cycle. The mice were allowed free access to water and a conventional pellet diet (Ewos, Södertälje, Sweden). Females were put together with males for three hours and mating was confirmed by the presence of a vaginal sperm plug, considered as day 0 p.c. The females were hysterectomized on the appropriate days and the uteri were processed for autoradiography (see below). Alternatively, the embryos were removed and processed for immunohistochemistry or directly frozen at –80°C.

The antibodies used were raised against synthetic peptides, corresponding to residue 67-82 of rat CRBP I and to residues 68-81 and 95-107 of bovine CRABP I (Eriksson et al., 1987; Busch et al., 1990). The specificity of the antibodies to CRABP I is well characterized with no detectable cross-reactivity to CRABP II (Dencker et al., 1990; Maden et al., 1991, 1992; Busch et al., 1993).

Protein extracts from day 10 p.c. mouse embryos were subjected to SDS-PAGE using a linear 15% gel, blotted onto nitrocellulose and incubated overnight with the antipeptide Ig-fraction to CRBP I, using 10 μg of Ig per ml. Following extensive washing, bound antibodies were visualized using alkaline phosphatase-labelled secondary antibodies. The details of the procedure were as previously described (Dencker et al., 1990).

To ensure that the antipeptide Ig fraction to CRBP I recognized CRBP I and not CRBP II, an enzyme immunoassay (EIA) was developed. Microtiter plates were separately coated overnight with 50 μl of a solution of CRBP I and CRBP II. Both proteins were diluted to 10 μg per ml in 100 mM ammonium bicarbonate buffer, pH 7.6. Immunopurified antipeptide Ig to CRBP I and preimmune Ig were serially diluted in the microtiter plates and incubated at room temperature for 2 hours. Following extensive washing, bound Ig was visualized using alkaline phospatase-labelled secondary antibodies and p-nitrophenyl phosphatate was used as substrate. The absorbance at 405 nm was determined using a Titrek microwell reader. The details of the EIA procedure have previously been described (Busch et al., 1990). Rat CRBP II was a kind gift from Dr James Saccettini, New York.

The distribution of CRBP I and CRABP I was determined using a modified avidine-biotin complex (ABC) immunohistochemical technique in sections through mouse embryos (day 9-10 p.c.) (Busch et al., 1990; Dencker et al., 1990). The embryos were fixed in 4% phosphate-buffered formaldehyde and embedded in paraffin. Sections (5-7 μm) were cut through the embryos and collected on microscope slides. The glass-mounted sections were deparaffinized in xylene, passed through an alcohol series to water and finally rinsed in PBS. The embryo sections were then stained overnight. The antibodies to CRABP I and CRBP I were used at concentrations of 500 and 275 ng per ml, respectively. When blocking endogenous peroxidase activity and for washing procedures, PBS containing 0.3% Triton X-100 was used.

To localize specific retinol-binding sites in the embryo, the wholebody autoradiography technique was used as earlier described (Ullberg, 1977; Dencker et al., 1990). Pregnant dams (day 9 and 10 p.c.) were given a single intravenous dose of 15-[¹⁴C]retinyl acetate (spec.act. 1.7 GBq/mol, a kind gift from Hoffmann-La Roche AG, Basel) dissolved in 95% ethanol, using a dose of 4.0 mg per kg. The dams were killed, 4 or 24 hours after administration, and the uteri were removed, quickly frozen and finally embedded in carboxymethyl cellulose. Following sectioning through the whole uteri, with a tape attached to the cut surface, the tape-mounted sections were apposed against X-ray film (Industrex C, Kodak) for several months, depending on survival interval, before they were developed.

To immunolocalize CRBP I, we employed an affinity-purified Ig fraction raised against a peptide derived from rat CRBP I. Two sets of experiments were conducted in order to verify the specificity of this Ig fraction. First, using a protein extract from a day 10 p.c. mouse embryo, we determined by immunoblotting that the Ig fraction specifically recognized mouse CRBP I, shown as a single band migrating with a Mr of 16×10³ in sodium dodecylsulfate poly-acrylamide gel electrophoresis (Fig. 1A). A preimmune Ig fraction analyzed under identical conditions did not show any reactivity, demonstrating the specificity of the technique. Secondly, since CRBP I is highly homologous to CRBP II (Li et al., 1986), we established, by an EIA, that the Ig did not cross-react with CRBP II. Equal amounts of CRBP I and CRBP II were separately coated onto microtiter plates and subsequently incubated with serial dilutions of the Ig fraction to CRBP I or with preimmune Ig. As shown in Fig. 1B, the antipeptide Ig to CRBP I showed reactivity to CRBP I at Ig concentrations above 2-5 ng per ml. In contrast, no reactivity above background was obtained with CRBP II. These two experiments demonstrate that the Ig fraction to CRBP I is specific for CRBP I with no detectable cross-reactivity to CRBP II.

Using the monospecific antibodies directed against CRBP I and CRABP I, we examined by immunohistochemistry, the pattern of CRBP I expression and the relationship between CRBP I- and CRABP I-expressing domains in day 9-10 p.c. mouse embryos. The specific expression of CRABP I in neural crest (NC)-derived cells and CNS observed in this study has been reported earlier (reviewed by Maden, 1991; Maden et al., 1992). The pattern of expression of CRBP I was distinctly different, and apparently mostly non-overlapping with the expression of CRABP I. Thus, CRBP I was found in specific domains of the surface ectoderm, in endoderm, in the myotome (Figs 2, 3), in placodal-derived cranial ganglia, in the heart, in the otic vesicle and in specific cells of the neural tube (not shown). Sections stained with preimmune rabbit Ig lacked specific staining (not shown). This report will focus on the observed expression of CRBP I in surface ectoderm and endoderm, and its relation to CRABP I-expressing neighbouring NC-derived cells in the craniofacial region and along the gut, as well as to limb bud mesenchyme.

CRBP I expression was observed in ectoderm of the fron-tonasal area, the branchial arches and the limb buds, in contrast to squamous-like epithelia, e.g. at the dorsal and ventral aspects of the embryo (Figs 2 and 3, left panels). In the frontonasal area, CRBP I was expressed uniformly in the ectoderm at day 9 p.c. (Fig. 2A). At day 10 (Fig. 2C), the surface ectoderm of the frontonasal area continued to express CRBP I. In the invaginating olfactory placode, the nasal pit, the lateral aspect expressed considerably more CRBP I compared to the medial aspect. In the branchial arches, CRBP I was expressed relatively uniformly by all ectodermal and endodermal cells, both at day 9 and 10 p.c. (Fig. 2E,G). Interestingly, however, some endodermal cells expressed more CRBP I than neighbouring cells (Fig. 2G). The central core of the branchial arches, presumably derived from paraxial mesoderm, also expressed CRBP I (Fig. 2G). Similarly, in limb buds, CRBP I was expressed in anterior and posterior, as well as in ventral and dorsal ectoderm both at day 9 (Fig. 3A) and 10 p.c. (Fig. 3C). The mesenchyme of the frontonasal area, of the branchial arches (except for the central core mentioned above) and of the limb buds did not express CRBP I except for a few scattered cells. In the limb buds, proximally located CRBP I-expressing cells could be muscle cells derived from the likewise CRBP I-expressing myotome of the somites (Fig. 3C) (Chevallier, 1977).

The right panel of Figs 2 and 3 show sections, parallel to those in the left panel, stained for CRABP I. It is obvious that CRABP I is expressed in the NC-derived mesenchyme of the frontonasal area (Fig. 2B,D) and of the branchial arches (Fig. 2F,H), and also in the mesenchyme of the limb buds (Fig. 3B,D). The variations in the staining intensity of the NC-derived mesenchymal cells among the arches relate to the origin of the cells and have previously been recognized (Maden et al., 1992). The NC-derived cells invading the first arch are likely to originate from the 2nd rhombomere of the hindbrain. Both the 2nd rhombomere and the NC-derived mesenchymal cells of the first arch express low levels of CRABP I. The intensively stained NC-derived cells of the 2nd and 3rd arches originate from the also intensively CRABP I-expressing 4th and 6th rhombomeres.

At day 9 p.c., expression of CRABP I was observed in the lateral plate mesoderm at the onset of limb bud formation and in the overlying ectoderm (Fig. 3B). In the day 10 p.c. limb bud, CRABP I was expressed only in the mesenchyme, highest distally in the subectodermal progress zone, whereas proximally located mesenchyme expressed less (Fig. 3D). In fact, this gradient in CRABP I expression, with cells closest to the ectoderm expressing more CRABP I than cells located at a distance, was also observed in the frontonasal area and the branchial arches (Fig. 2D,H).

The observed relation between CRBP I-expressing epithelium and CRABP I-expressing NC-derived cells was also observed along the primitive gut. In this area, endoderm regularly stained for CRBP I (Fig. 3E), some regions more extensively than others, while cells expressing CRABP I were scattered around the endoderm. The distribution of these cells is very similair to that previously reported for NC cells, migrating towards the gut endoderm where they will form enteric ganglia (Fig. 3F) (Le Douarin, 1982).

Studies were conducted to explore the possibility that the CRBP I-expressing embryonic tissues were able to accumulate retinol selectively. The results showed that the distribution of radioactivity in the embryos was more even than we previously had observed for RA and RA analogues (Dencker et al., 1987, 1990, 1991) and the concentration was generally lower. This might be expected for several reasons. First, the specific activity of retinol entering the embryo is probably very low since the injected radiolabelled retinyl acetate is metabolized and mixed with the large pool of endogenous retinol present in the pregnant dams. When employing RA or RA analogues in similar studies (Dencker et al., 1987, 1990, 1991), the dilution of radiolabelled ligand is expected to be negligible since the endogenous levels of RA are very low compared to the levels of retinol. Secondly, some of the injected retinyl acetate had probably been metabolized to RA and other metabolites prior to entering the embryo. Previous studies have shown that oral dosing of pregnant mice with retinol results in a dose-dependent increase in the concentration of retinol and to some extent RA in maternal plasma as well as in the embryo up to 4 hours after administration (Eckhoff et al., 1989), after which RA and its metabolites (but not retinol) decrease in concentration. Thus we observed, both at 4 and 24 hours after dosing the mothers, a selective accumulation of radioactivity in the ectoderm of the frontonasal area, especially in the lateral aspects. Similarly, the ectoderm of the branchial arches and of the limb buds accumulated radioactivity (Fig. 4A-C). Radioactivity was also found in the yolk sac, in the gut, heart, neural tube and in the medial part of the otic vesicle. The low level of accumulation of radioactivity in the CRABP I-expressing NC-derived mesenchyme and in the limb buds, as well as in the hindbrain, suggest that only a small fraction of the injected tracer had been metabolized to RA (cf. retinoid distribution in Dencker et al., 1987, 1990). In summary, these data suggest that cells expressing CRBP I have the ability to accumulate retinol selectively and consequently control the spatial localization of retinol.

A common property of retinoids is their hydrophobic nature and a complex machinery has evolved in order to transport, compartmentalize and metabolize these compounds. In adults, the transport system for retinoids is fairly well char-acterized (for review see Eriksson, 1990). Much less is known about retinoid transport and metabolism in the developing embryo. However, in order to understand the role of RA during embryogenesis, insights into the mechanisms and the possible spatial restrictions in the formation of RA in the embryo are required. The identification of target cells for RA action during normal embryogenesis and how ligand saturation of the nuclear RA receptors is controlled are other important issues to be resolved.

In adults, most cells obtain retinol as their major source of retinoid. In extracellular compartments, retinol is transported bound to the plasma RBP and uptake of retinol into cells is mediated by specific membrane receptors for RBP (Båvik et al., 1991, 1992). It is likely that retinoid transport and metabolism are similar in the embryo. Receptors for RBP are expressed in the placenta (Sivaprasadarao and Findlay, 1988) and the major source of retinoid for the developing embryo is maternal retinol. Furthermore, RBP synthesized by the yolk sac placenta (Makover et al., 1989; A-L. G., L. D. and U. E., unpublished observation) is likely to transport retinol in the embryonic circulation. A RBP-mediated delivery of retinol to certain embryonic cells may also involve receptors for RBP like in adults.

The data presented in this paper are consistent with CRBP I being a cellular acceptor of retinol in the embryo. The largely non-overlapping but spatially closely related domains of expression of CRBP I and CRABP I observed in this study are striking. Since CRBP I has been shown to be a substrate carrier in the formation of RA from retinol (Posch et al., 1991), this expression pattern suggests that RA synthesized in the CRBP I-expressing ectoderm is transferred, by an active mechanism or by passive diffusion, to the underlying CRABP I-expressing mesenchyme. Thus, the role of CRABP I may be to accumulate ectoder-mally derived RA. This is in line with the ability of CRABP I to accumulate exogenous RA (Dencker et al., 1990, 1991). The abundance of CRABP I in the mesenchymal cells suggests that only a narrow zone of subectodermal cells will bind RA and that free RA may form a very steep gradient from the ectoderm. Consistent with this hypothesis, a recent immunohistochemical study showed that antibodies to RA selectively label limb bud ectoderm and a narrow zone of subectodermal mesenchyme (Tamura et al., 1990). Although the antibodies used were not entirely specific for RA and probably do not recognize CRABP I-bound RA, one may interpret the staining pattern as reflecting RA formed in the ectoderm and being transported to the underlying mesenchyme.

The non-overlapping domains of expression of CRBP I and CRABP I, and the corresponding distribution of their ligands, suggest that a retinol/RA pathway has a role in epithelial-mesenchymal interactions. It is well documented that epithelial-mesenchymal interactions are of importance for pattern formation during embryonic development, and several genes and their products have been suggested to be involved (reviewed by Lyons et al., 1991; Jessell and Melton, 1992). Based on the results of this study, it appears likely that RA belongs to this group of molecules. The role of CRABP I is not fully understood (see below), but some alternatives concerning the regulation of RA action in ectoderm and mesenchyme may be considered. RA synthesized in ectodermal cells may act locally in the ectoderm and CRABP I is expressed in mesenchymal cells to prevent action of RA in these cells. Secondly, RA formed in ectoderm, may be rapidly transferred to and act in mesenchymal cells and CRABP I is involved in the regulation of RA action. A third alternative is a combination of the two first, where RA acts both in ectoderm and mesenchymal cells; in the ectoderm via a non CRABP I-mediated pathway while, in mesenchyme, RA action is mediated by CRABP I. In situ hybridization studies have revealed that RARs are expressed in both ectoderm (RARα) and mesenchyme (RARα, RARβ and RARγ), at least in the limb (Mendelsohn et al., 1992). Thus, in principle, both types of tissues would be able to respond to RA. Reconstitution experiments have shown that only the mesenchymal cells but not the overlying ectoderm in craniofacial structures and in the limb buds seem to be competent of responding to RA (Wedden 1987; Tickle et al., 1989). However, it should be kept in mind that excess retinoids were used in these studies and they may not necessarily reflect the normal role of RA. The finding that overexpression of CRABP I prevents activation of RA-induced reporter genes indicates that the role of CRABP I is to prevent RA from reaching the nucleus (Boylan and Gudas, 1991) and it has been suggested that CRABP I has a role in catabolism of RA (Fiorella and Napoli, 1991). However, other roles of the protein cannot be excluded. It is well established that nuclear uptake of many cellular proteins, including several transcription factors, are highly regulated processes often related to the cell cycle or to the state of differentiation of cells (reviewed by Silver, 1991). Since there is limited information regarding the subcellular localization and possible nuclear uptake of CRABP I, it is difficult to draw firm conclusions regarding the role of CRABP I. In a recent study on human epidermis, a tissue-dependent on retinoids for normal growth and differentiation, it was found that CRBP I and CRABP I are co-expressed in epidermal cells (Busch et al., 1993). Based on this observation, it is likely that CRABP I has a role in RA-controlled gene expression and metabolism, including but not limited to a role in deactivation and catabolism of RA.

RA-induced reporter genes in transgenic animals have been used to monitor sites of synthesis and action of endogenous RA during embryogenesis. In day 9-10 p.c. embryos, a largely reciprocal expression of a transgene with CRABP I was observed, but with abundant expression of the transgene in several locations of craniofacial ectoderm and gut endoderm where CRBP I is expressed (Rossant et al., 1991). These results are consistent with the model proposed by us suggesting that RA is generated in these sites. The lack of expression of the transgene in some domains expressing CRABP I may be explained by the fact that expression of CRABP I can efficiently prevent induction of RA-driven genes as discussed above (Boylan and Gudas, 1991). Results partly differing from those of Rossant and coworkers, in which a transgene driven by a RA-inducible RAR-β2 promotor was used (Mendelsohn et al., 1991), are more difficult to interpret, since endogenous promotor elements were used. Interestingly Mendelsohn and coworkers observed expression of the transgene in limb ectoderm but not in craniofacial ectoderm.

In light of the data presented in this paper and the discussion above, it is possible to propose a putative pathway for retinoid transport and metabolism in the embryo (see Fig. 5). Whether this model is valid also for other embryonic structures remains to be shown, but we have observed a non-overlapping distribution of CRBP I and CRABP I in CNS and in several other locations in the developing embryo (A-L. G., L. D. and U. E., unpublished observations). Thus, the role of RA as a mediator in tissue interactions during embryonic development may be a more general one.

A number of investigators have previously examined the expression patterns of CRBP and CRABP in the developing mouse embryo (Dollé et al., 1989, 1990; Perez-Castro et al., 1989; Vaessen et al., 1989, 1990; Dencker et al., 1990; Maden et al., 1990, 1991, 1992; Ruberte et al., 1991). However, CRBP I expression in limb and craniofacial ectoderm has not previously been reported using either in situ hybridization or immunohistochemistry. Generally, in situ hybridization studies have shown a much more widespread distribution of transcripts for both retinoid-binding proteins compared to immunohistochemical studies. For example, frontonasal mesenchyme has been reported to express CRBP transcripts abundantly (Dollé et al., 1990; Ruberte et al., 1991), while we find no expression of CRBP I in the frontonasal mesenchyme. The accumulation of retinol in ectoderm but not in the mesenchyme would argue against an abundant expression of CRBP II in this location. Thus the discrepancies are not easily explained, unless the type II protein is under strong posttranscriptional control.

A comparison of the embryonic expression patterns of several Hox genes, shown to be regulated by exogenous RA, and the sites of RA synthesis proposed by us in this paper, reveals no obvious correlations. One exception is the expression pattern of the Hox-7 and Hox-8 genes in limb and craniofacial mesenchyme. Both genes are expressed in subectodermal mesenchyme and expression of both genes requires a diffusable signal derived from the ectoderm (Coelho et al., 1991; Davidsson et al., 1991; Robert et al., 1991; Takahashi et al., 1991). Retinoic acid is a candidate of this signal.

The striking findings that the spatial distributions of retinol and RA are confined to cells expressing CRBP I and CRABP I, respectively, show that the cellular retinoid-binding proteins play important roles in retinoid metabolism and probably also RA-controlled gene expression in the developing embryo. Of particular interest is that the cellular retinoid-binding proteins have a much more restricted dis-tribution compared to the nuclear RA receptors. For example, RARα and RARγ are expressed throughout the mesenchyme of the developing limb and widely distributed in craniofacial mesenchyme (Dollé et al., 1989, 1990; Osumi-Yamashita et al., 1990; Ruberte et al., 1990, 1991). Thus we conclude that the distributions of retinol and RA in the embryo, respectively, are not linked to the patterns of expression of the RARs. Whether only a subfraction of the RARs, expressed in or close to domains expressing the cellular retinoid-binding proteins, have access to RA remains to be studied. Moreover, if generation of RA is spatially restricted, as our data imply, it is possible that the teratogenic effects of exogenous RA are achieved by inappropriate activation of RA receptors that normally does not receive RA. Thus, the effects on embryonic development using excess RA may only reflect the teratogenic activity of RA and not the role of endogenous RA during embryogenesis. Further work will be required in order to substantiate these hypothesis.

Mrs Raili Engdahl is acknowledged for skilful technical assistance. This study was supported by a grant (No. B92-03X-07899-06B) from the Swedish Medical Research Council. U. E. is a Procordia Research Foundation Scholar.