Localization of specific retinoid-binding sites and expression of cellular retinoic-acid-binding protein (CRABP) in the early mouse embryo

Retinoic acid (RA), a biologically active form of vitamin A, is known to support embryonic development (Takahashi et al. 1975: Thompson et al. 1964) and to affect differentiation and growth properties of many cell types. (Lotan, 1980; Strickland and Mahdavi, 1978). The presence of RA in limb and its ability to specify the anteroposterior axis during limb development suggests that RA is an endogenous morphogen (Thaller and Eichele, 1987). In addition, RA is a potent teratogen, drastically affecting formation of the craniofacial area, brain and vascular system in exposed humans (Rosa, 1983: Lammer et al. 1985) and experimental animals (Durston et al. 1989; Goulding and Pratt, 1986; Kochhar, 1967; Shenefelt, 1972; Wiley et al. 1983; Webster et al. 1986). There is some evidence suggesting that neural crest-derived cells and neuroepithelial cells are among the primary targets for RA during early embryonic development, since administration of RA early in gestation or to early embryos in vitro, predominantly induces malformations in tissues derived from these cells (Durston et al. 1989; Goulding and Pratt, 1986; Lammer et al. 1985: Webster et al. 1986) Furthermore, a previous study indicated that exogenous RA accumulate in neural-crest-derived cells and in neuroepithelium (Dencker et al. 1987).

The various biological effects caused by RA treatment are likely to be mediated by specific binding proteins. One RA-binding protein, the cellular retinoic-acid-binding protein (CRABP) was identified almost 15 years ago. However, its role in retinoid metabolism remains unclear (for a review see Chytil and Ong, 1984). The close structural similarities to a number of intracellular fatty-acid-binding proteins suggest its possible role as a transport and storage protein for RA (Sundelin et al. 1985a,b). It is conceivable that CRABP may be involved in regulating the supply of RA to the recently identified family of nuclear receptors for RA. The three RA receptors identified in mammals (RAR, α, β and γ) are closely related to steroid and thyroid hormone receptors (Benbrook et al. 1988; Brand et al. 1988; Giguere et al. 1987; Petkovich et al. 1987; Zelent et al. 1989) suggesting that RA may influence the expression of a specific set of genes. It is likely that such genes may directly or indirectly influence pattern formation of embryos and the differentiation processes affected by RA.

To elucidate further the role of RA in embryonic development, experiments were conducted to identify target cells for RA in early mouse embryos using an autoradiographic technique. The specific accumulation of retinoids in some tissues suggested that a retinoid-binding protein was involved. Using biochemical techniques, this protein was identified as CRABP. Consequently, we have also determined the temporal and spatial expression of CRABP in mouse embryos during the early organogenetic period using immuno-chemical techniques.

Three month old C57B1/6 mice (Alab, Solna Sweden) were kept at 23°C and with a 12/12h 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 overnight and mating was confirmed by the presence of a vaginal sperm plug the next morning (considered day 0 of gestation). The females were hysterectomized on the appropriate days and the uteri were processed for autoradiography. Alternatively, the embryos were removed for immunohistochemistry or directly frozen at - 80°C.

To localize specific retinoic-acid-binding sites in the embryo, the whole-body autoradiography technique was used as described (Dencker et al. 1987; Ullberg, 1977). Dams (days 8.5-11.5 of gestation) separately received intravenous injection of 200 μl of mouse serum mixed with the ¹⁴C-labelled retinoid Ro 13-6298 (structure shown in Fig. 3C, specific activity 1.85 GBq mmol^-1, a kind gift from Hoffman-La Roche AG, Basel, Switzerland). The dose was Img of retinoid kg^-1 body weight. After 20min, 1 h and 4h, the dams were killed, the uteri removed, quickly frozen and finally embedded in carboxymethyl cellulose. Following sectioning of the whole uteri with a tape attached to the cut surface, the tape-mounted sections as well as a staircase of isotope standards were apposed against X-ray film (Industrex C, Kodak). The film was developed 6 weeks later.

Mice on day 10.5 of gestation received intravenous injections of 200 μl of mouse serum containing the ¹⁴C-labelled retinoid Ro 13-6298 as described above. A second group of mice on day 10.5 of gestation received similar injections containing 62kBq of 11,12-³H-RA (specific activity 0.85 TBq mmol^-1; a kind gift from Hoffman-La Roche AG, Basel, Switzerland) mixed with unlabelled RA (Imgkg^-1) to get a comparable dose. To correct for non-saturable binding of RA to embryonic components, a third group of animals was given orally 50 mg kg^-1 of unlabelled RA dissolved in soybean oil 2 h prior to the i.v. injection of radioactive RA. This dose has previously been shown sufficient to block accumulation of a following dose of radiolabelled RA in embryonic target tissues (Dencker et al. 1987, 1990). The animals were killed after 4 h and the embryos were carefully dissected and quickly frozen at — 80°C. 13-15 embryos from each group were homogenized using a Dounce homogenizer in 500 gl of 20 mM Tris pH 7.5 containing 150mM NaCl (TBS), 2 % aprotinin and 2mM phenylmethylsulphonylfluoride. Debris was removed by centrifugation at 100000g for Ih at +4°C using a Beckman TL-100. The supernatants (approx. 500 α1) were separately analysed on a Superóse 12 column (lx30cm) connected to an FPLC system (Pharmacia-LKB, Bromma Sweden). The column was equlibrated in 20 mM Tris pH 7.5 containing 150 mM NaCl. The flow rate was 0.5 ml min^-1 and fractions were collected every minute. 0.3ml of each fraction was mixed with 3 ml Instagel (Packard) and counted in an LKB liquid-scintillation counter. Specific binding of radiolabelled RA to embryonic components was calculated by comparing the radioactive profiles obtained from the second and third group of embryos.

The rabbit antibodies used to localize CRABP were raised against two synthetic peptides corresponding to residues 69-84 and 95-107 of bovine CRABP (Sundelin et al. 1985a). Both antibody preparations have previously been characterized in detail (Eriksson et al. 1987; Busch et al. 1990) and gave identical results in immunoblotting and immunohisto-chemistry in the present study.

All manipulations of the embryos were done on ice or at +4°C. The thawed embryos were homogenized in a Potter-Elvehjem homogenizer in TBS containing 2 % (v/v) apronitin and 2mM phenylmethylsulfonylfluoride. Debris was removed by centrifugation at 15000 g for 5 min and the supernatants were subsequently centrifuged at 100000g for 60min. The final supernatants were stored in aliquots at —80°C until use.

Aliquots of the embryo extracts (200 μg of total protein) or separate fractions from the gel chromatography experiments were subjected to SDS-PAGE separation and electrophoretically transferred onto nitrocellulose filters as previously described (Towbin et al. 1979). Following transfer, nonspecific protein binding was blocked by an incubation in TBS containing 5 % (w/v) non-fat dry milk and 0.1 % Tween 20 for lh at room temperature. 20 μg of affinity-purified anti-CRABP Ig per ml was added and incubated overnight at +4°C. The filter was extensively washed in TBS containing 0.1 % Tween 20. Bound Ig was visualized after an incubation of goat anti-rabbit Ig labelled with alkaline phosphatase. Nitrobluetetrazolium and 5-bromo-4-chloro-3-indolyl phosphate were used as the substrates.

Embryos from mice hysterectomized 8.5-11.5 days after breeding were fixed in 4 % phosphate-buffered formaldehyde and embedded in paraffin. Serial, transverse, coronal and sagittal sections (5-7 gm) were cut through the embryos and glass-mounted sections were deparaffinized in xylene, passed through an alcohol series to water and finally rinsed in PBS. The embryo sections were stained using a modified avidinbiotin complex (ABC) technique (Busch et al. 1988; Hsu et al. 1981). Essentially, non-specific peroxidases were blocked by a preincubation with 1 % hydrogen peroxide and non-specific antibody binding to the sections were blocked by a preincubation with 4% (w/v) bovine serum albumin in PBS. The antibodies to CRABP (25 μg of Ig ml^-1) were applied to the sections for 30min. Following extensive rinsing in PBS, a biotinylated swine anti-rabbit Ig (Dakopatts a/s, Glosterup, Denmark), appropriately diluted in 4% (w/v) bovine serum albumin in PBS was applied to the sections. After rinsing, an avidin-biotin complex (Dakopatts a/s, Glosterup, Denmark) was added. Following additional rinsing, the sections were incubated with 3-amino-9-ethyl carbazole (0.42 mg ml^-1) in PBS containing 0.05% hydrogen peroxide. The reaction was stopped after 30 min by extensive rinsing in water and the sections were subsequently mounted under coverslips. In order to allow computer-assisted densitometry analysis, the sections were not counter stained. Microphotographs of the sections were obtained using an Olympus microscope equipped with an OM-2 camera system.

The localization of radioactivity in the embryo sections and the immunohistochemical staining of CRABP were analysed using computer-assisted densitometry. The shades of grey in the autoradiograms and the staining intensities of the immunohistochemical staining were digitalized by means of a Cromenco SDD video digitalizing system and represented in various colours (d’Argy et al. 1989). The relative amounts of radioactivity in the autoradiograms were calculated using isotope standards.

Protein estimations were done according to Bradford (1986) using bovine serum albumin as the standard protein.

To localize specific retinoid-binding sites in embryos, we employed the whole-body autoradiography techniques after dosing pregnant mice at various stages of gestation with a ¹⁴C-labelled synthetic retinoid. The ¹⁴C isotope was used for labelling since ³H-sensitive film emulsions are prone to artifacts using this technique. The retinoid Ro 13-6298, an ester of a biologically active analogue of RA (Ro 13-7410), was employed instead of radiolabelled RA due to better accumulation of radioactivity in some embryonic tissues as compared to the background in the autoradiograms. Ro 13-6298 is likely to be rapidly hydrolysed in vivo yielding the corresponding free acid (Ro 13-7410) as has been shown for other esterified synthetic retinoids (Lofberg et al. 1989). We also present indirect evidence suggesting that, in fact, the radioactivity accumulating in the embryos is the free acid (see following section and Fig. 3A).

Following injection into pregnant mice, increasing amounts of radioactivity accumulated in the embryos from 20min up to 4h. This accumulation, also previously reported for radiolabelled RA, was greatly reduced by pretreatment with a high dose of unlabelled RA (data not shown, Dencker et al. 1987). From day 8.5 of gestation, accumulation of radioactivity was seen in both the neural plate and in neural folds, along which neural crest cells first appear in the embryo (Fig. 1A,B). In later stage embryos (days 9.5-11.5 of gestation), radioactivity accumulated in locations known to be populated by neural crest cells. These locations included the frontonasal mesenchyme (Fig. 1C,D), the visceral arches (Fig. 2A) and the dorsal root ganglia (Fig. 2A). In cross-sectioned visceral arches, accumulation was stronger circumferentially than in the core (Fig. 2A, 2nd and 3rd arches). The scattered cells of neural crest origin in the peripheral nervous system only rarely showed up, most likely due to the limited resolution of the autoradiograms. However, in some cases, peripheral nerves, such as cranial nerves, could be seen to accumulate radioactivity (Fig. 2B). Surface ectoderm did not show up in the autoradiograms, which does not exclude some accumulation, since very thin structures may be under represented and difficult to distinguish from underlying structures.

In the CNS, a distinct pattern of accumulation was observed through the span of our observation. Radioactivity was detected in the midbrain roof (Figs 1C,D, 2C) and in the hindbrain floor and walls, especially in the peripheral zone (Fig. 2A-C). Along the anterioposterior axis, accumulation of radioactivity in the hindbrain was restricted with an anterior boundary rostral to the otic vesicle (Fig. 2B,C). Further caudally, accumulation was observed in the mantle layer of the neural tube (Fig. 2D).

In the limbs, a moderate accumulation could be detected and the distribution was even (Fig. 2E). The yolk sac placenta and the chorioallantoic placenta did not specificially accumulate radioactivity.

The specific and saturable accumulation of the radiolabelled retinoid at certain locations in the developing mouse embryo suggested that a specific retinoid-binding protein was involved. To test this, 100000g fractions were prepared from homogenized embryos previously exposed in utero to ¹⁴C-labelled Ro 13-6298. The generated fractions were subsequently subjected to gel filtration on a Superose 12 column. As shown in Fig. 3A, binding of the radiolabelled retinoid was achieved in a single peak corresponding to a molecular weight of 17000, which indicated that the binding protein might be CRABP. The amount of free retinoid, eluting in the total volume of the column, was very low, which is in accordance with the low general background in the autoradiograms (see previous section). To demonstrate that the binding protein was CRABP, it appeared essential to show that binding of radioactivity was blocked with a pretreatment of the unlabelled compound. The high teratogenicity and embryotoxicity of Ro 13-6298 made it necessary, however, to use RA in this study. Thus 100000g cytosol fractions were prepared from embryos exposed in utero to ³H-RA, with or without a pretreatment with an excess of unlabelled RA and analysed as above. The results, shown in Fig. 3B, demonstrate that specific binding of RA was achieved in a single peak with elution characteristics almost identical to those obtained using the radiolabelled synthetic retinoid as tracer. Immuno-blotting analyses on separate fractions also verified that CRABP comigrated with the radioactive peaks (data not shown). Taken together, our data suggest that the binding protein is CRABP and that retinoids accumulate in various tissues by binding to this protein. Since Ro 13-6298 is esterified and unable to bind to CRABP (Jetten et al. 1987) our data further suggest that it becomes hydrolysed and that it is the corresponding free acid (Ro 13-7410) that accumulates in embryonic tissues.

Since CRABP appeared to be the dominating RA-binding protein in the embryo, it was considered essential to determine its temporal and spatial expression during embryogenesis.

The temporal expression of CRABP was analysed by immunoblotting of extracts derived from embryos of day 10 to day 17 of gestation. The affinity-purified antipeptide Ig, originally raised against synthetic peptide stretches derived from bovine CRABP (Eriksson et al. 1987), specifically recognized a protein migrating close to M_r 16000-17000. Since bovine and murine CRABPs are identical in their primary structures (Vaessen et al. 1989), we conclude that the antibodies specifically recognize the murine counterpart. As shown in Fig. 4, expression of CRABP was temporally regulated with highest expression during the organogenetic period (up to days 12 and 13). In later stages of development, the expression was several-fold lower (days 14 to 17). The apparent lack of immuno-reactive CRABP in days 16 and 17 does not indicate that the protein is not expressed but it appears to occur below the level of detection using the conditions in the blotting experiment.

The sites of CRABP expression during early embryonic development was determined using the ABC-technique (Hsu et al. 1981). Initial control experiments verified that specific staining was only obtained when employing affinity-purified antipeptide Ig to CRABP and not when using preimmune Ig (Fig. 6B).

In accordance with the localization of specific retinoid-binding sites, strong staining using Ig to CRABP was obtained in embryonic tissues known to be colonized by neural crest cells and of cells at specific sites along the neural tube throughout organogenesis. Some staining was also observed in ectoderm and in limb bud mesenchyme. Somites had a relatively weak staining, except for the very earliest stages of development (data not shown), when especially the ventral portion in close apposition to the likewise stained intra-embryonic visceral yolk sac epithelium was stained. The main findings are summarized in Table 1.

At day 8.5, neural crest cells at the tip of the neural folds, where they meet the surface ectoderm, were strongly stained. (Fig. 5A). Later, at day 9, mesen-cephalic neural crest cells migrating to the 1 st visceral arch were stained (Fig. 5B). At days 9.5-10.5, the staining intensity of the resulting mesenchyme of the maxillary and mandibular arches decreased, while the later-developing 2nd, 3rd and 4-6th visceral arches at this stage showed stronger staining (Fig. 5C). The surrounding ectoderm and endoderm was not stained. A close examination revealed that not all cells in the mesenchyme of the visceral arches were stained. It is speculated that the most often strongly stained circumferential cells were of neural crest origin, while the more frequent mesodermally derived mesenchymal cells of the core (mentioned for the chick by Noden, 1987) were unstained. Likewise, neural crest-derived cells, occupying the space between the forebrain and the frontonasal ectoderm and forming the frontonasal processes, and those appearing in the periocular area, were strongly stained days 9 through 11 (Figs 5D,E,6A). Mesodermally derived mesenchyme caudal to this area was not stained above background (Fig. 6A). Dorsal root ganglia of the trunk (Fig. 8A) as well as cranial nerves were strongly stained (data not shown). In addition, scattered cells, presumably of neural crest origin, along the gut and the vascular system stained strongly (Fig. 8A).

At day 8.5, staining was observed variably in the neuroepithelium of the cranial neural plate and neural folds (Fig. 5A), while the closed neural tube in the mid-part of the body axis was not stained. Later (days 9.5-11.5), a characteristic staining pattern was observed. Strongly stained cells were seen in the outer layer of the midbrain roof, while the rest of the neuroepithelium of the fore- and midbrain showed no staining above background (Figs 5D,E,6A). In the floor and the walls of the hindbrain, a different pattern of less distinct staining was observed, often stronger in the peripheral zone (Fig. 6A,C,D). The staining was segmented with marked ventrodorsal bands of immuno-reactivity. These bands were localized to the periodic swellings in the hindbrain known as neuromeres or rhombomeres (Lumsden and Keynes, 1989) and the stained cells are probably neurons. The staining of the different neuromeres appeared to vary in intensity since neuromeres 4 and 6 showed stronger staining than neuromeres 1-3 and 5. The numbering of the neuromeres was confirmed in horizontal sections, where the 4th neuromer was identified as the one just rostral to the otic vesicle (data not shown). Also cells of the thin dorsal roof of the hindbrain were strongly stained (Fig. 6C). Caudal to the hindbrain, in the neural tube, staining was restricted to individual cells in the mantle layer and these cells appeared to be postmitotic early neurons (Fig. 8B). The proliferating neuroepithelial cells adjacent to the lumen of the neural tube were not stained above background. Strong staining could also be seen in axons within the neural tube (Fig. 8B) and of motomeurons leaving the neural tube. Also peripheral nerve bundles were stained (data not shown).

Specific staining was observed in the limb bud although considerably weaker than in the previously described CRABP-positive tissues. Within the limb buds, regional differences in staining intensity were usually insignificant. However, if any diversification was seen, cells in a subectodermal zone were stronger stained than centrally located cells. The ectoderm was less stained than the underlying mesenchyme (data not shown).

A comparison of the CRABP-staining pattern with that of the sites of accumulation of the radiolabelled retinoid was carried out using computer-assisted densitometry. Autoradiograms and stained tissue sections with similar orientations, derived from embryos of the same developmental stage were analysed. As shown in Fig. 7 (computer images of Figs 2C and 6A), retinoid accumulation and CRABP distribution showed a remarkable co-localization. A direct comparison of Fig. IB with Fig. 5B and Fig. 1C with Fig. 5D gives the same impression of the co-localization of radiolabelled retinoids with CRABP in various structures of the embryo. These data support the earlier suggestion that CRABP is involved in accumulation of RA in certain embryonic cells.

It has been shown that vitamin A deficiency in experimental animals cause congenital abnormalities in the offspring (Hale, 1937; Wilson et al. 1953). The defects include brain and eye abnormalities, hydro-cephalus, cleft palate and vascular abnormalities (Kalter and Warkany, 1959). It was later found that RA supports normal development of embryos in vitamin A-deficient animals. However, during the fetal period of development there is a unique requirement for retinol (Thompson et al. 1964; Takahashi et al. 1975). Added to the evidence that RA is an endogenous morphogen, directing anteroposterior pattern formation in the chick limb bud, this may indicate that RA is the active form of vitamin A needed for normal embryonic development (Thaller and Eichele, 1987).

In normal animals, RA at low concentration circulates in plasma bound to albumin while retinol present at higher concentration is bound to the retinol-binding protein (De Leenher et al. 1982; Goodman, 1984; Smith et al. 1973). The intravenous administration of the retinoids used in this study may thus give a physiologic route for RA-delivery to embryos. Our results demonstrate that neural crest cells and several neural crest-derived tissues e.g. visceral arches, frontonasal mesenchyme and dorsal root ganglia, accumulate radioactivity. In addition, certain cells in the neural tube e.g. cells in the roof of the midbrain, the hindbrain and cells in the mantle layer of the spinal cord specifically accumulate RA. The mechanisms by which these tissues accumulate exogenous retinoids are likely to involve CRABP. As determined by immunohisto-chemical techniques, we find expression of this protein specifically in areas of the CNS described above and in cells with a distribution pattern consistent with that described for cranial and truncal neural crest cells (Bronner-Fraser and Fraser, 1988; Le Douarin, 1982; Nichols, 1981, 1986; Noden, 1987). As determined by gel chromatography analysis, CRABP is the major embryonic RA-binding protein. To our knowledge these data are the first to show directly that CRABP is involved in accumulation and possibly in storage of RA in vivo, and that this occurs in specific embryonic cells.

Retinoid-exposed embryos show malformations preferentially in CRABP-expressing and RA-accumulating cells e.g. neural crest derivatives and cells in the CNS (Goulding and Pratt, 1986; Lammer et al. 1985, 1988; Webster et al. 1986). These cells may be target cells for RA during embryogenesis suggesting that the role of CRABP in the metabolism of RA may not only be to bind freely diffusing RA and thus contribute to the generation of RA gradients in embryonic tissues as has been proposed earlier (Maden et al. 1988). Instead, a CRABP-mediated delivery of physiological amounts of RA may be a means of regulating the transcriptional activities of the RARs. Excessive amounts of RA, on the other hand, may exceed the binding capacity of CRABP and thus saturate the much less abundant RARs. This may cause aberrant expression of genes regulated by the RARs and might explain the teratogenic effects caused by an excess of exogenous retinoids. Whether the RARs, individually or in different combinations, are co-expressed with CRABP remains to be established. In the developing mouse limb and epidermis, however, CRABP and RAR expression does not always coincide (Dollé et al. 1989).

A remarkable feature of CRABP in the developing embryo is its high level of expression during the morphogenetic period and its presence in the CNS and neural crest cells during that time. This is consistent with an essensial role of RA in normal development of the CNS and of tissues derived from the neural crest. In CNS, CRABP is expressed behind a relatively distinct anterior boundary in the hindbrain and the expression is extended caudally in the neural tube. A group of genes with a similar spatial expression in the CNS are some of the homeobox genes (for reviews see Holland and Hogan, 1988; Stern and Keynes, 1988). Still other homeobox genes are expressed in neural crest-derived tissues, for example the Hox-7.1 gene (Robert et al. 1989; Hill et al. 1989). Whether the expression of homeobox genes in target cells for RA - revealed in this study - indicate any functional relationship to RA remains to be established. However, the co-expression of CRABP and several homeobox genes in neural crest derivatives and in hindbrain is intriguing considering that exposure to RA (Accutane, 13-cis RA) cause facial dysmorphogenesis and aberrant development of several hindbrain derivatives such as cerebellum and cranial nerves in human embryos (Lammer et al. 1988; E. Lammer, personal communication). Furthermore, among the transcripts shown to be regulated by RA, the ERA-1 transcript is derived from the homeobox gene Hox-1.6 (LaRosa and Gudas, 1988a,/?). It was recently also shown that constitutive and ectopic expression of a homeobox gene, Hox-1.1, in transgenic mice induce craniofacial abnormalities similar to those obtained in retinoid-exposed embryos (Balling et al. 1989).

During the preparation of this manuscript, two reports on the localization of CRABP in chick (Maden et al. 1989) and the corresponding mRNA in mouse embryos were published (Vaessen et al. 1989). In this paper we have extended the analyses to include several additional cell populations that express CRABP during the morphogenetic period. Furthermore, the localization of embryonic tissues that accumulate retinoids have identified putative target cells for RA during embryogenesis and have also revealed a possible mechanism for the selective uptake of RA to these cells.

We thank R. Pensas for technical support, R. d’Argy and G.O. Sperber for help with the computer image analyses and R. Petterson and B. Seliger for valuable comments on this manuscript. This study was supported by a grant from the Swedish Medical Research Council (grant B90/04X/07899/ 04B).