ABSTRACT
Retinoic add (RA) is a signaling molecule apparently involved in a variety of morphogenetic processes, such as patterning of developing and regenerating vertebrate limbs. RA binds to specific intracellular receptors that constitute a multigene family. RA receptors (RAR) bind to the regulatory region of specific target genes and thereby control the expression of these genes. Here we report the sequence and spatiotemporal expression pattern of RAR-β from chick. Northern blots of RNA from whole embryos and from limb buds reveal the presence of transcripts of 3.2, 3.4, and 4.6 kb in size. Using two riboprobes, one that hybridizes to all three RAR-β mRNAs and a second one, specific for the 4.6 kb transcript, we found by in situ hybridization a differential distribution of RAR-β transcripts in limb bud mesenchyme, in craniofacial mesenchyme and in hindbrain neuroectoderm. In the hindbrain the 4.6 kb mRNA exhibits an anterior boundary of expression at the level of the constriction between rhombomeres 5 and 6. Examination of neural plate stage embryos by in situ hybridization indicates that this boundary of expression is already defined by this stage. In addition to having several RA receptors that are expressed with distinct spatial patterns in the embryo, our data indicate that the expression pattern of transcripts derived from a single receptor gene can also be differentially expressed, thus providing another level for regulating RA action.
Introduction
Retinoids, the family of compounds that includes vitamin A and its metabolites, are potent regulators of cell differentiation and of embryonic development (reviewed by Roberts and Sporn, 1984; Brockes, 1989). Retinoids support growth and differentiation of many cell types, including epidermal (Fuchs and Green, 1981), chondrogenic (Lewis et al. 1978), lymphoid (Carman et al. 1989) and neuronal lineages (Haussier et al. 1983). Retinoids also appear to act as morphogenetic signaling molecules that, when exogenously provided, will influence pattern formation in vertebrate embryos and in regenerating urodele limbs. Specifically, retinoic acid will induce pattern duplications in the developing chick limb (Tickle et al. 1982, 1985; Summerbell, 1983) and it will respecify the pattern of regenerating urodele limbs (Maden, 1982; Thoms and Stocum, 1984). Moreover, in Xenopus laevis as well as in chick, retinoic acid can apparently transform anterior tissues to more posterior ones (Durston et al. 1989; Sive et al. 1990; Sundin and Eichele, unpublished observations, in chick). The finding that retinoic acid is endogenously present in chick limb buds (Thaller and Eichele, 1987) and in Xenopus embryos (Durston et al. 1989) at concentrations that affect morphogenesis when exogenously applied, support the hypothesis that retinoic acid is an endogenous morphogenetic signaling molecule.
It has now become clear that retinoids mediate their effects by binding to specific nuclear receptors (Petko-vich et al. 1987; Giguère et al. 1987), which are members of the steroid/thyroid nuclear receptor superfamily (Evans, 1988). Two families of retinoic acid receptors have been identified and are known as RARs and RXRs. Each receptor is encoded by a distinct gene referred to as RAR-a (Giguère et al. 1987; Petkovich et al. 1987; Zelent et al. 1989), RAR-β (Benbrook et al. 1988; Brand et al. 1988; Zelent et al. 1989), RAR-γ (Krusteta/. 1989; Zelent et al. 1989), RAR-δ (Ragsdale et al. 1989), RXR-α (Mangelsdorf et al. 1990) and RXR-β(Hamada et al. 1989; Mangelsdorf et al. 1990). Like the other members of the nuclear receptor superfamily, RAR and RXR molecules can be subdivided in several regions (A, B, C, D, E and F regions, see Fig. 1). The A and F regions (amino and carboxy terminus, respectively) are divergent between the various receptor forms, whereas the C region (DNA-binding domain) and E region (ligand-binding domain) is conserved amongst the RARs. Each RAR gene can give rise to several transcripts. In the case of mammalian RAR-y, for example, seven transcripts have been identified that are splice variants in the 5’ region (Krust et al. 1989; Kastner et al. 1990; Giguère et al. 1990).
In addition to RARs and RXRs there is a second class of retinoid-binding proteins known as cellular retinol-binding protein (CRBP; Bashor et al. 1973) and cellular retinoic-acid-binding protein (CRABP; Sani and Hill, 1974; Ong and Chytil, 1975; Kitamoto et al. 1989). CRBP and CRABP are structurally unrelated to RARs and RXRs, and are expressed in a broad variety of cell types (reviewed by Chytil and Ong, 1984), as well as in developing vertebrate embryos (Perez-Castro et al. 1989; Maden et al. 1989a,b, 1990). The role of CRBP and CRABP is not entirely clear, but they may function in retinoid metabolism (Kakkad and Ong, 1988; Ong et al. 1988) and/or in the regulation of retinoid homeostasis (Robertson, 1987; Smith et al. 1989).
At present, it is not known what specific task each of the receptors and each of the binding proteins has in embryonic development. A clue is likely to come from examining in detail the spatiotemporal expression pattern of RARs, RXRs, CRABPs and CRBPs in embryos. For example, in situ hybridization studies have revealed transcripts of RAR-α RAR-β RAR-γ and CRABP in mouse limb buds (Dollé et al. 1989), and limb regeneration blastema were found to express RAR-α, RAR-β and RAR-γ (Giguère et al. 1989; Ragsdale et al. 1989). The presence of such transcripts in these tissues is consistent with the proposed role of retinoids in limb development and regeneration. Here we report the cloning and tissue localization of RAR-/3 in the developing chick embryo. We were particularly interested in the distribution in the limb bud and face, in light of the known effects of retinoic acid on the development of these tissues in the chick (Tickle et al. 1982, 1985; reviewed by Wedden et al. 1988 and Eichele, 1989). In the course of these studies, we discovered that RAR-β is expressed at significant levels in the developing central nervous system (CNS). Expression in the CNS is detected at the time of neurulation (Hamburger-Hamilton (1951) stage 6), and continues well into the stages of neuronal outgrowth (stage 20 and later). The presence of RAR-β mRNA indicates that retinoic acid is likely to have a physiological role in CNS development.
Materials and methods
Isolation of cDNA clones
Hamburger-Hamilton stage 20 (3.5 day, Hamburger and Hamilton, 1951) chick embryo  ZAP cDNA library (Stratagene) was prepared from oligo(dT)-primed poly(A)+ RNA. Blunted double-stranded cDNA was ligated to blunt end EcoRI linkers, and inserted into the EcoRI site of the  ZAP vector. In vitro packaged phage was transfected into XL-1 Blue cells and plated. Screens were performed as described (Wedden et al. 1989) using a random primed 1135 bp Xhol/ BamHI fragment of human RAR-/J. Duplicate filters were hybridized in 35% formamide, 6xSSC, 20mM sodium-EDTA and 0.1 % SDS at 42°C and washed in 2xSSC, 0.1 % SDS at 55°C. Seven out of 2.4X106 plaques were purified, and the inserts were sequenced using the dideoxy method.
Northern analysis
Poly(A)+ RNA was isolated and Northern blots were prepared as described (Wedden et al. 1989). Random-primed 32P-labelled probes were derived from the entire gRAR-β cDNA (nucleotides 1-1593), the carboxy-terminal F region and 3’ untranslated DNA (nt 1279-1593), or the aminoterminal A region (nt 1-97). The filters were hybridized in 40% formamide and washed in 0.2× SSC (long probe), 2× SSC (short probes), 0.1% SDS at 60°C. As an internal control, blots were hybridized with chick /1-actin (Cleveland et al. 1980).
In situ analysis
In situ hybridization was performed as previously described (Wedden et al. 1989) with the following modifications. Alternating 6zzm thick sections were hybridized with antisense riboprobes derived from either the A region (nt 1 –97) or F region (nt 1279-1593), or with the corresponding sense probes. Following acid treatment, sections were incubated in 20 μgml-1 proteinase K (in 50mM Tris-HCl, pH7.5, 5 HIM sodium-EDTA) at room temperature. Proteolysis was quenched with 0.2% glycine. Sections were hybridized with 0.01 nggl-1 riboprobe in 35% formamide (A region; specific activity 5.0mCinmol-1) or in 45% formamide (F region; specific activity 2.8 mCi nmol-1) for 18h at 50°C. Slides were washed in 50% formamide, 2–SSC, 20mM 2-mercapto-ethanol at 63°C, and then treated with RNAase A. The NTB-2 emulsion was exposed for 23 days. Slides were developed, stained with Hoechst dye (5figml-1) and the sections were viewed and photographed in a fluorescence microscope equipped with a red light dark-field illumination (Sundin et al. 1990).
Results
A cDNA library of 3.5-day-old chick embryos was screened using a hRAR-β (h, human) probe. The sequence of the longest insert revealed a 1347 bp ORF encoding a 448 amino acid protein (Fig. 1). The predicted sequence of this protein is 95 % identical with that of hRAR-β (Benbrook et al. 1988; Brand et al 1988)and mRAR-β (Zelent et al. 1989; m, mouse). In contrast, the identity with the a, y and <5 retinoic acid receptors from human (Giguère et al. 1987; Petkovich et al. 1987; Krust et al. 1989), mouse (Zelent et al. 1989) and newt (Giguère et al. 1989; Ragsdale et al. 1989) is only about 70 %, and the identity with RXR-a and -β is <25 % (Hamada et al. 1989; Mangelsdorf et al. 1990). It had previously been noted that the A and F regions (amino and carboxy termini) of a, fi, y and ô RARs have largely diverged and thus earmark a particular receptor form (Brand et al. 1988). We found that within the A and F regions the sequence of the chick clone is >95 % identical with that of hRAR-/J (Benbrook et al. 1988; Brand et al. 1988) and mRAR-/3 (Zelent et al. Therefore we conclude that the chick protein sequence in Fig. 1 is chick RAR-β (gRAR-βg for Gallus’).
To determine transcript size and the overall temporal expression pattern of gRAR-β, poly(A)+ RNA from whole chick embryos (Hamburger-Hamilton stages 16 to 34), and of limb rudiments (stages 22–34) was analyzed by Northern blotting. When probed with the complete gRAR-β cDNA (Fig. 2A), stage 16 embryos and limb tissues revealed bands at 3.2 and 3.4 kb. In addition there was a weaker band at 4.6 kb that has, however, essentially disappeared in more developed limbs. This indicates that the 4.6 kb mRNA is under a developmental control distinct from the other two transcripts. Identical results were obtained with poly(A)+ RNA derived from embryonic heads and trunks from stages 18-34 (not shown).
These Northern blots, when hybridized with probes of the B, C/D, E and F regions, revealed the same set of three bands (hybridization with F region is shown in Fig. 2B; B, C/D, E data are not presented). In contrast, a probe derived from the A region revealed only the 4.6kb band (Fig. 2C). Thus, the cDNA depicted in Fig. 1 is likely to represent the 4.6 kb transcript. Possibly, the 3.2 and 3.4 kb mRNAs lack the A region of this cDNA or, like hRAR-y (Krust et al. 1989; Giguère et al. 1990; Kastner et al. 1990), have undergone alternative splicing in the 5’ coding region. Additional experimental support for this interpretation comes from RNAase protection experiments. An antisense riboprobe derived from nt 1–409 (A, B and partial C regions) is full-length protected, corresponding to the 4.6 kb mRNA. In addition there are two partial protections, one of which corresponds to a site of divergence at or near the junction of the A and B regions, and the other to a mRNA that diverges within the A domain (Smith and Eichele, unpublished observations).
Because the developmental regulation of the 4.6 kb transcript is different from that of the other RAR-β mRNAs, we sought to determine whether tissue distributions were àlso different. Therefore, serial sections of stage 20 chick embryos were hybridized in situ with one of two probes, a ‘pan-β probe’ (nt 1279–1593, the prefix ‘pan’ denotes that this probe is specific for RAR-β but reveals all three bands on Northern blots; Fig. 2B) that hybridized on Northern blots with all three transcripts (Fig. 2B) or a probe consisting of nts 1–97 that is specific for the 4.6 kb band (Fig. 2C). In craniofacial mesenchyme, limb buds and CNS, the 4.6 kb transcript is found only in a subregion of tissues that hybridized with the pan-β probe. In the developing face, the 4.6 kb transcript was detected in the frontonasal mass and in the anterior portion of the maxillary process, but was absent or very weakly expressed in the posterior region of the maxillary process, and in the mandibular and hyoid arches (Fig. 3A). The pan-β probe also detects transcripts in the frontonasal mass (not shown), but in addition hybridized to the entire maxillary process and to the mandibular and hyoid arches (Fig. 3B). The finding that the 4.6 kb mRNA is unevenly distributed across the anteroposterior axis of the maxillary process is unexpected, since there is no known morphological boundary within this tissue.
In the chick limb bud, the 4.6kb mRNA-specific signal is graded along the proximodistal axis (Fig. 3C), with a 4.3-fold higher silver grain density in the proximal region (grain counts are given in the legend of Fig. 3). In contrast, pan-β signal is detected throughout the limb bud mesenchyme, with only a 1.8-fold difference along the proximodistal axis (Fig. 3D). The slightly higher proximal grain density in Fig. 3D may be caused, in part or entirely, by the uneven distribution of the 4.6 kb transcript (Fig. 3C). gRAR-β mRNAs are absent in ectoderm and apical ectodermal ridge. The presence of RAR-β transcripts in chick limb bud mesenchyme, but not in ectoderm or AER, is consistent with the finding by Tickle et al. (1989) that retinoic acid primarily affects the limb bud mesenchyme.
In the CNS of stage 20 embryos, the 4.6 kb mRNA is found in the spinal cord and hindbrain (Fig. 3E and F), with a marked anterior boundary of expression (Fig. 3F). This boundary is situated at the level of the otic vesicle and approximately follows the constriction between rhombomeres 5 and 6. Grain counts in rhombomeres 6-8 are 4 times higher than in anterior rhombomeres, where the signal approaches the level of background (see Fig. 3F legend), pan-β signal is also seen in spinal cord and hindbrain (Fig. 3G) and continues anteriorly into the midbrain. There was at most a 1.5-fold difference in transcript levels between rhombomeres posterior and anterior to the otic vesicle. Transverse sections through the spinal cord, hybridized with either the pan-β probe (Fig. 3E or the 4.6 kb mRNA-specific probe, not shown) revealed essentially uniform labeling throughout the neurectoderm. This markedly differs from the expression pattern of cellular retinoic acid binding protein, which, at a similar stage, is restricted to commissural axons, mantle layer and dorsal roof plate (Maden et al. 1989a,b).
Inspection of neural plate stage embryos reveals that the expression border of the 4.6 kb mRNA in the hindbrain is established before metamerization of the presumptive hindbrain neurectoderm, which begins around Hamburger-Hamilton stage 9 (Vaage, 1969). Fig. 3H shows a sagittal section passing through the midline of a head-fold stage embryo (stage 6). In the presumptive trunk, silver grains are present in ectoderm, mesoderm and endoderm. Hence, the 4.6kb mRNA of RAR-β is clearly not restricted to a particular germ layer. Anterior to Hensen’s node we detect elevated levels of signal and a distinct expression boundary (arrow), beyond which neither mesoderm nor
Discussion
We have cloned and sequenced (Fig. 1) beta retinoic acid receptor (RAR-β) from chick and characterized its spatiotemporal expression pattern in the chick embryo, with particular emphasis on the first three days of development. We found that RAR-β is expressed at least as early as neurulation (Fig. 3H,I) and continues to be transcribed throughout the period of organogenesis (Fig. 2). The pattern of expression is regional in nature, and in its complexity is reminiscent of, but different from, that of the RAR-y recently described in mouse embryos (Noji et al. 1989; Osumi-Yamashita et al. 1990; Ruberte et al. 1990). Two new and possibly fundamental aspects have emerged from our study. First, we found that the multiple transcripts of RAR-/β are differentially regulated, both in time (Fig. 2) and space (Fig. 3). A second significant finding is that RAR-is prominently expressed in the developing CNS.
Northern blots show that the gRAR-β gene gives rise to at least three different transcripts of 3.2, 3.4 and 4.6 kb in length (Fig. 2). The molecular nature of these mRNAs remains to be determined, but Northern analyses and RNAase protection experiments using probes corresponding to distinct regions of the cDNA in Fig. 1, suggest that the 4.6 kb mRNA differs from the 3.2 and 3.4kb transcripts in the /V-terminal A region. Such a difference could be due to differential splicing and/or alternative promoter usage. A precedence for splice variants in the /V-terminal region is provided by mammalian RAR-γ(Krust et al. 1989; Giguère et al. 1990; Kastner et al. 1990).
When comparing the spatial expression pattern revealed by a probe that hybridizes to all gRAR-/β mRNAs (referred to as pan-β probe) with that of a probe specific for the 4.6 kb mRNA, we found that the
4.6 kb mRNA is expressed only in a subset of cells that hybridize with the pan-β probe. As has been pointed out by Dollé et al. (1989), the existence of distinct RARs and retinoid-binding proteins that exhibit different patterns of expression could account for the diverse regulatory effects of retinoids on cell differentiation and morphogenesis. Our in situ hybridization data suggest that an additional level of regulation, that of differential spatial expression of transcripts, that are derived from a single RAR gene. Whether the different transcripts we observe encode functional receptor proteins remains an open question.
Our finding that RAR-β is abundantly expressed in the chick CNS supports a role for retinoids in CNS development. It has been known for some time that high doses of retinoic acid have deleterious effects on CNS development (Kochhar, 1967; Lammer et al. 1985). However, the presence of CRJBP and CRABP, and as illustrated in this study, of RAR-βtranscripts, indicates an undetermined physiological role of retinoids in the CNS. It should be noted that the distributions of CRBP, CRABP and RAR-β mRNA are overlapping, but are not identical. For example, at stages 20– 22, RAR-β mRNA is expressed essentially uniformly in thespinal cord (Fig. 3E). In contrast, CRBP and CRABP are highly localized: CRBP to the ventral floor plate, and CRABP to the dorsal roof plate and the mantle layer and commissural axons (Maden et al. 1989a,b). Additional evidence for a physiological function of retinoids in CNS development comes from the recent demonstration that CNS tissue contains the enzymes that generate retinoic acid from its precursor retinol (Wagner et al. 1990).
In situ hybridization studies of definitive streak stage chick embryos (stage 4) reveal a uniform distribution of signal along the anteroposterior axis of the embryo for both the A region-specific probe and the pan-β probe (unpublished data). However, during the subsequent head-process and head-fold stages (stages 5 and 6) when the neural plate appears, one observes a distinct regionalization along the anteroposterior axis. Both the
4.6 kb mRNA-specific probe and the pan-β probe hybridize to presumptive trunk tissue, including presumptive spinal cord, and the posterior neural plate (Fig. 3H,I). Fate mapping studies by Spratt (1952) show that the region of the neural plate that is RAR-β positive is destined to form posterior hindbrain. As hindbrain development continues, the 4.6 kb mRNA expression boundary persists and eventually lies near or at the anterior border of rhombomere 6 (Fig. 3F). In contrast, pan-/3-derived signal extends well into the midbrain. Because of the rather distinct boundary of expression in the hindbrain, the 4.6kb mRNA of RAR/β falls into a growing collection of transcription factorencoding mRNAs whose expression terminates or is confined to specific rhombomeres. Examples are HOX genes (Wilkinson et al. 1989; Murphy et al. 1989; Frohman et al. 1990; Sundin and Eichele, 1990) and the Krox 20 gene (Wilkinson et al. 1989). It has been suggested that these genes are involved in establishing the metameric pattern of the hindbrain and/or impart segment identity (reviewed by Keynes and Lumsden, 1990).
An interesting feature is that the domains of expression of several HOX genes and RAR-/β overlap partially or fully. This is intriguing since both HOX genes and retinoic acid have been implicated in positional signaling. There are numerous in vitro studies demonstrating that retinoic acid can influence the expression of HOX genes (La Rosa and Gudas, 1988; Simeone et al. 1990). This raises the possibility that retinoids may influence positional signaling in the CNS by affecting the expression of HOX gene products.
To date, the spatial distribution of RAR-α -, -β, and -γ have been examined in mouse embryos (Dollé et al. 1989; Noji et al. 1989 Osumi-Yamashita et al. 1990; Ruberte et al. 1990). These receptors differ in their spatial distributions, indicating that they may mediate different functions. Interestingly, the RAR-β distributions in chick and in mouse limb buds and face are not identical. In the chick limb, the 4.6 kb mRNA is present in the proximal two-thirds of the limb mesenchyme, while pan-/β transcripts are detected throughout the limb. In contrast, Dollé et al. (1989) reported that in mouse limb buds of similar stage, RAR-β transcripts are restricted to the most proximal portion of the limb bud. Thus, RAR-β has a more extensive distribution in the chick limb than in the mouse limb. RAR-β distributions also differ in the facial primordia. In the chick, RAR-β transcripts are abundant in the anterior maxillary process and frontonasal mass, with lower but clearly detectable levels in the mandibular and hyoid arches (Fig. 3A,B). In the mouse, RAR-β transcripts are found in the frontonasal mass, but are apparently absent from the maxilla and mandible (Osumi-Yamashita et al. 1990). It is not yet known whether the multiple transcripts of murine RAR-β (Zelent et al. 1989) are differentially expressed in space. Moreover, it is not clear whether the observed differences in expression pattern reflect variations between chick and mouse development, or are due to difficulties in detecting possibly low levels of RAR-β mRNA by in situ hybridization.
ACKNOWLEDGEMENTS
We thank Professor P. Chambón for providing us with human RAR-β probe, Drs O. Sundin and S. E. Wedden for their advice, and Dr T. Jessell for comments on the manuscript. This work was supported by the American Cancer Society and the Lucille P. Markey Charitable Trust. S.M.S. is a fellow of the Muscular Dystrophy Association.