Retinoic acid signalling centres in the avian embryo identified by sites of expression of synthesising and catabolising enzymes

Authors

  • Aida Blentic,

    1. MRC Centre for Developmental Neurobiology, King's College London, Guy's Campus, London Bridge, London, United Kingdom
    Search for more papers by this author
  • Emily Gale,

    1. MRC Centre for Developmental Neurobiology, King's College London, Guy's Campus, London Bridge, London, United Kingdom
    Search for more papers by this author
  • Malcolm Maden

    Corresponding author
    1. MRC Centre for Developmental Neurobiology, King's College London, Guy's Campus, London Bridge, London, United Kingdom
    • MRC Centre for Developmental Neurobiology, 4th floor New Hunt's House, King's College London, Guy's Campus, London Bridge, London SE1 1UL UK
    Search for more papers by this author

Abstract

Retinoic acid is an important signalling molecule in the developing embryo, but its precise distribution throughout development is very difficult to determine by available techniques. Examining the distribution of the enzymes by which it is synthesised by using in situ hybridisation is an alternative strategy. Here, we describe the distribution of three retinoic acid synthesising enzymes and one retinoic acid catabolic enzyme during the early stages of chick embryogenesis with the intention of identifying localized retinoic acid signalling regions. The enzymes involved are Raldh1, Raldh2, Raldh3, and Cyp26A1. Although some of these distributions have been described before, here we assemble them all in one species and several novel sites of enzyme expression are identified, including Hensen's node, the cardiac endoderm, the presumptive pancreatic endoderm, and the dorsal lens. This study emphasizes the dynamic pattern of expression of the enzymes that control the availability of retinoic acid as well as the role that retinoic acid plays in the development of many regions of the embryo throughout embryogenesis. This strategy provides a basis for understanding the phenotypes of retinoic acid teratology and retinoic acid–deficiency syndromes. Developmental Dynamics 227:114–127, 2003. © 2003 Wiley-Liss, Inc.

INTRODUCTION

Retinoic acid (RA) is an important signalling molecule in the developing embryo, being involved in establishing patterns of gene activity in several systems, for example, the nervous system (Maden et al., 1996; Gale et al., 1999; Niederreither et al., 2000), the lung (Malpel et al., 2000), the limb (Stratford et al., 1996, 1999; Lu et al., 1997), the kidney (Mendelsohn et al., 1999), and the eye (McCaffery et al., 1999). RA performs the task of activating or repressing genes by binding to transcription factors, the retinoic acid receptors (RARs) and the retinoid X receptors (RXRs), which heterodimerise and bind to retinoic acid–response elements (RAREs) in the regulatory sequences of RA-responsive genes (Chambon, 1996).

There are two obvious ways in which the regulation of RA signalling could be controlled: regulate the transcription of the RARs and RXRs or regulate the synthesis of the ligand, RA. In the former case, the RARs and RXRs are widely expressed in the embryo, and although there is some differential distribution, it is difficult to find a tissue in which some combination of RARs and RXRs are not expressed. Regulation of signalling at this level would seem to be a very complex arrangement. On the other hand, synthesising and metabolising RA itself is a much more direct way of regulating signalling. Thus, it is of interest to consider where and when such enzymes are expressed in the embryo.

RA is generated from vitamin A (retinol) supplied in the diet by the action of two classes of enzymes. The first class, the retinol/alcohol dehydrogenases, oxidise retinol to retinal, and the second class, the retinaldehyde dehydrogenases (RALDHs), oxidise retinal to all-trans-RA and 9-cis-RA (Duester, 2000). All-trans-RA is further metabolised by the action of two cytochrome P450 enzymes, to products such as 4-oxo-RA, 4-OH-RA, and 18-OH-RA (Fujii et al., 1997; White et al., 1997, 2000a). To date, three RALDHs have been identified in embryos, RALDH1, RALDH2, and RALDH3 and two CYPs, CYP26A1, and CYP26B1. The coordinated expression of these enzymes throughout development, therefore, may be crucial for the local generation of RA. Indeed, the idea has arisen (Swindell et al., 1999; Maden, 1999) that an RALDH generating RA and a CYP breaking it down could between them establish a morphogen gradient of RA in a classic source sink system of positional information (Wolpert, 1969). We have identified RA itself in the developing chick embryo in distinct patterns and at different levels in various regions of the embryo by a combination of reporter cell analyses and HPLC (Maden et al., 1998), but it may be that there are more subtle patterns of RA synthesis that we cannot detect by these methods. The analysis of enzyme expression domains may provide us with such subtlety.

A considerable amount of work has already been performed on the expression of these enzymes in development. For example, Raldh1 is expressed relatively late in development and in very restricted regions such as the dorsal retina of the day 9 mouse embryo (McCaffery et al., 1991) and the stage 16 chick embryo (Godbout et al., 1996; Grun et al., 2000); in the mesonephros of the mouse embryo (Haselbeck et al., 1999); and in the olfactory placode of the Xenopus embryo (Ang and Duester, 1999).

Raldh2 has been the most intensively studied enzyme. It seems to be the earliest one expressed as it appears soon after gastrulation, at the primitive streak stage, when it is expressed in the newly generated mesenchyme (Niederreither et al., 1997; Berggren et al., 1999; Swindell et al., 1999). As development proceeds, Raldh2 remains in the mesenchyme of the somites and lateral plate and additionally is expressed in the heart (Moss et al., 1998) and nephrogenic mesenchyme. There is little expression of Raldh2 in the head of the embryo. Later, it is expressed in the meninges around the spinal cord and in the motorneurons and the roof plate within the spinal cord (Zhao et al., 1996).

Raldh3 is expressed after Raldh2, but before Raldh1, at stage 10 in chicks and day 9.5 in mouse. Initially, Raldh3 is present in the surface epithelium rostral to the optic vesicles, which becomes the olfactory placode, but later at stage 18, it is localised to the ventral half of the developing eye, the dorsal part of the otic vesicle, Rathke's pouch, and a stripe at the midbrain/hindbrain border (Li et al., 2000; Mic et al., 2000; Grun et al., 2000).

Cyp1A1 has a fascinating expression pattern which in some regions is complementary to Raldh2. It begins to be expressed at the same time as Raldh2, but in a region at the anterior end of the embryo, which is the presumptive forebrain and midbrain neurectoderm (Fujii et al., 1997; de Roos et al., 1999; Swindell et al., 1999). It then becomes restricted to the anterior neural folds and rhombomere 2 in the hindbrain. At the posterior end of the embryo where Raldh2 is not expressed, there is another domain of Cyp26A1 in the tail bud and hindgut. A similar pattern is seen in Xenopus embryos (Hollemann et al., 1998; de Roos et al., 1999). This complementarity continues into the developing eye with a stripe of Cyp26A1 around the equator of the eye, sandwiched between a dorsal domain of Raldh1 and a ventral domain of Raldh3 (McCaffery et al., 1999). The chick limb bud seems to show something of a complementarity with Cyp26A1 in the ectoderm at the distal tip of the limb and Raldh2 at the proximal base (Swindell et al., 1999), although in the mouse limb bud, the other CYP, Cyp26B1 is strongly expressed in the distal mesenchyme of the progress zone (MacLean et al., 2001).

These enzyme distributions, particularly the Raldhs, have usually been studied in isolation from each other and in differing organisms. Here, we put them all together in a temporal sequence during the early embryonic development of one organism, the chick embryo, and describe novel domains of expression of some of these enzymes, for example, in and around Hensen's node. The node previously has been suggested to synthesise RA (Hogan et al., 1992), but no enzymes have been seen there. The temporally and spatially dynamic nature of the expression patterns becomes apparent when the enzymes are considered together, and as a result, we can identify novel regions that may be patterned by local RA signalling.

RESULTS

Stages 3–6

Previous data on chick embryos (Swindell et al., 1999) and mouse embryos (Ulven et al., 2000) has suggested that Raldh2 is first expressed at stage 4 (chick) or mid-primitive streak stage (mouse) slightly preceded by Cyp26A1 in the anterior epiblast and that these are the only two enzymes expressed at these stages. However, it is possible to detect these enzymes slightly earlier, at stage 3+ (although this detection may be a staging issue, because it is difficult to stage early embryos accurately), but more importantly we find another enzyme, Raldh3, is expressed in Hensen's node and that there is a mesodermal Cyp26A1 domain around the node.

The temporal expression of enzymes that we have observed is, therefore, the following. First, Raldh2 at around stage 3+ in the newly involuted mesenchyme makes a butterfly wing pattern (Fig. 1A shows a stage 4 embryo). It is absent from the midline and transverse sections, confirming that the expression is only in the mesenchyme (Fig. 1F). Longitudinal sections show that Raldh2 has a rostral border exactly where Hensen's node begins (Fig. 1B). Second, Cyp26A1, also at stage 3+, is expressed anterior to the node (Fig. 1C), and sections confirm that it is only in the epiblast layer (Fig. 1G). Third, Raldh3 is expressed in the node at stage 4 (Fig. 1I), and sections show that it is only in the epiblast of the node (Fig. 1J). Fourth, an additional small domain of Cyp26A1 appears around the node at the same time that Raldh3 appears (Fig. 1D), and sections reveal that it is expressed in the mesoderm below the node (Fig. 1E). This finding means that the expression of Cyp26A1 switches from the epiblast to the mesoderm at the level of the node. These domains of nodal Raldh3 and supranodal Cyp26A1 are very temporarily restricted as they have disappeared by stage 5. The expression domains of these three genes around the node and possible signalling interactions are summarised in Figure 6A.

Figure 1.

Expression of retinoic acid (RA) synthesising and catabolising enzymes in stages 3+ to 8 chick embryos. The stage is shown in the left-hand corner, the enzyme in the right-hand corner of each panel. A:Raldh2 in a stage 4 embryo showing wing-shaped expression in the mesoderm and no expression at the anterior of the embryo or in the primitive streak. A sagittal section of this is shown in B and a transverse section at the level marked F is shown in F. B: Sagittal section of A showing that Raldh2 expression in the mesoderm extends up to the node. C:Cyp26A1 expression in a stage 3+ embryo showing expression in the anterior epiblast. A section at the level of G is shown in G. D: Expression of Cyp26A1 in a stage 4+ embryo showing an extra expression domain surrounding the node. A sagittal section of this embryo is shown in E. E: Sagittal section of D showing Cyp26A1 expression in the anterior epiblast and in the mesoderm of Hensen's node. F: Transverse section of A showing Raldh2 expression in the mesoderm. G: Transverse section of C showing Cyp26A1 expression in the epiblast (presumptive neural plate). H: Transverse section of K showing Raldh3 expression in the lateral endoderm. I:Raldh3 expression in Hensen's node of a stage 4 embryo. A sagittal section of this is shown in J. J: Sagittal section through Hensen's node of a stage 4 embryo showing Raldh3 expression in the epiblast of the node. K:Raldh3 expression in the lateral edges of a stage 5 embryo. A section at the level of H is shown in H. L:Raldh3 expression in the endoderm of the heart tubes of a stage 8 embryo. A section at the level of P is shown in P. M:Cyp26A1 expression in a stage 6 embryo showing the absence of expression in the midline of the anterior neural plate and the spreading posteriorly of expression at the lateral edges of the neural plate to meet at the posterior end in a W-shape. A section of this stage embryo is shown in Q. N:Cyp26A1 (purple)/Raldh2 (turquoise) double in situ of a stage 6 embryo revealing the gap (white arrowheads) between the two domains where the hindbrain will develop. O:Cyp26A1 (purple)/Raldh2 (turquoise) double in situ of a stage 8 embryo revealing a larger gap between the two domains and that the lateral line of Cyp26A1 (white arrowhead) now lies in the middle of the lateral stripe of Raldh2 (yellow arrowhead). P: Section through L showing diminished expression of Raldh3 in the endoderm of the heart tubes of a stage 8 embryo. Q: Section of a stage 6 Cyp26A1 embryo showing expression in two lateral domains at the edge of the neural plate. R: Section through the tail bud of a stage 8 Cyp26A1 embryo showing that expression in the neural plate (as in G and Q) moves into the mesoderm at the posterior end of the embryo. ps, primitive streak; hn, Hensen's node; ne, neuroepithelium; m, mesoderm; en, endoderm.

Figure 6.

Summary diagrams of the expression domains of the four enzymes studied here. Red, Cyp26A1; blue, Raldh3; green, Raldh2; yellow, Raldh1; purple arrows, potential directions of RA signalling. A: Sagittal section, stage 4 embryo. B: Transverse section, stage 5 embryo. C: Transverse section, stage 6 embryo. D: Sagittal section, stage 6 embryo. E: Transverse section, stage 12 embryo. F: Sagittal section, stage 9 embryo. G: Transverse embryo, stage 11 embryo. H: Transverse section, stage 18 embryo. I: Head of a stage 18 embryo showing a summary of the various expression domains. ANT, anterior; POST, posterior; LAT, lateral; fb, forebrain; mb, midbrain.

At stage 5, Cyp26A1 and Raldh2 continue to be expressed in their characteristic domains with Cyp26A1 marking out the anterior neural plate and Raldh2 in the mesenchyme posterior to the node. However, Raldh3 suddenly appears in a horseshoe shape around the edges of the anterior neural plate (Fig. 1K). Sections reveal that this expression domain is in the endoderm (Fig. 1H), and it is directly below the heart-forming region of the mesoderm (Yutzey and Kirby, 2002). Indeed, expression is most intense in precisely the location of the cardiogenic precursors located on the lateral edges (Fig. 1K). Although the expression level of Raldh3 decreases, it can still just be detected at stage 8 (Fig. 1L), when it is located in the endoderm below the two primitive heart tubes (Fig. 1P). Possible signalling interactions at this stage regarding the heart are summarised in Figure 6B.

At stages 6 and 7, the expression of Cyp26A1 begins to change. As the notochord differentiates in an anterior direction from the node, the expression of Cyp26A1 disappears from the midline of the anterior neural plate and at the same time spreads posteriorly in two lateral stripes to meet at the posterior end of the embryo in a Wshape (Fig. 1M). This posterior expression is in the upper epiblast layer as is the anterior expression, and sections show the lateral stripes are at the edge of the neural plate in the region where the neural crest arises (Fig. 1Q). A Cyp26A1/Raldh2 double in situ (Fig. 1N) reveals that the lateral stripes of Cyp26A1 overlay the edges of the Raldh2 expression in the mesenchyme below and also emphasises the gap between the mesodermal Raldh2 expression and the anterior epiblast expression of Cyp26A1 (white arrowheads in Fig. 1N). This finding has been described previously (Swindell et al., 1999), and the gap is where the presumptive hindbrain is forming. It has been hypothesised that a gradient of RA across this region could be responsible for hindbrain patterning (Swindell et al., 1999; Maden, 1999). A similar Raldh2/Cyp26A1 juxtaposition exists at the posterior end of the embryo, seen more clearly in Figure 1O. There the expression of Cyp26A1 has by stage 8 become intense in the tail bud and regressing node (Fig. 1O). The expression in the regressing node itself changes from being epiblastic to mesodermal in the node (Fig. 1R). Thus, there are three possible regions of localised RA signalling between the Raldh2 and Cyp26A1 expressing zones at this stage: first, between the lateral plate Raldh2 and the lateral stripe of Cyp26A1 in the overlying epiblast (Fig. 6C) perhaps involved in neural crest formation; second, between the anterior border of expression of Raldh2 at the level of the presumptive first somite and the anterior Cyp26A1 expression in the epiblast where the hindbrain will form (Fig. 6D, anterior end); third, between the posterior edge of Raldh2 expression in the mesoderm and the Cyp26A1 expression in the mesoderm of the regressing node (Fig. 6D, posterior end).

Stages 8–11

At stage 8, Raldh3 is fading but just detectable in the endoderm below the inflow tract in the developing heart (Fig. 1L,P). By stage 9, the fading Raldh3 is replaced in the developing heart by strong expression of Raldh2 in the splanchnic mesoderm (Fig. 3C). By stage 8, Cyp26A1 is expressed strongly in the anterior neural folds and the tail bud and regressing node at the posterior end of the embryo (Fig. 1O). The Cyp26A1/Raldh2 double in situ shown in Figure 1O reveals that the two lateral lines of Cyp26A1 expression, which were previously at the edges of the neural plate (Fig. 1M,N), move towards the midline as neurulation takes place because the Cyp26A1 is in the rising edges of the neural folds. Thus, the Cyp26A1 expression (white arrowhead in Fig. 1O) now overlays the Raldh2 expression (yellow arrowhead in Fig. 1O).

Figure 3.

Moving borders of gene expression of Raldh2 (A–F) and Cyp26A1. The stage is shown in the left-hand corner, the enzyme in the right-hand corner of each panel. A:Raldh2 in a stage 7+ embryo showing the anterior border of expression at the level of somite 1. B: Sagittal section through A. C:Raldh2 expression in a stage 9 embryo showing the down-regulation of Raldh2 in somites 1 and 2. D: Sagittal section through C. E:Raldh2 expression in a stage 11 embryo showing the down-regulation of Raldh2 in somites 1, 2, and 3. F: Sagittal section through E. G: Double in situ showing Cyp26A1 (turquoise) and Hoxb-1 (purple) and the gap in between them (red arrowheads) at stage 7. H: By stage 8, the gap between Cyp26A1 and Hoxb-1 has narrowed (red arrowheads). Because Hoxb-1 does not move anteriorly, this shows that the border of Cyp26A1 must spread posteriorly. I: By stage 9, the two expression domains have met (red arrowhead) but not overlapped. s1, somite 1; s2, somite 2; s3, somite 3; s4, somite 4; s5, somite 5.

At stage 8+, Cyp26A1 expression remains strong in the tail bud, regressing node, and the neural folds at the posterior end of the embryo and begins to be down-regulated in the anterior neurectoderm in an anterior to posterior direction (Fig. 2A). As the anterior border of Cyp26A1 regresses posteriorly, so does the posterior border, but more slowly. The posterior border is initially around the midbrain/hindbrain border (Fig. 1O), and this moves posteriorly to the level of the rhombomere 3/4 border. We determined this finding by double in situ studies using Cyp26A1 and Hoxb-1 on carefully staged embryos (see next section). Because the anterior border regresses faster than the posterior border, there is a period at stages 9 and 10 when there is only a thin stripe of Cyp26A1 left, which ends up in rhombomere 3 (Fig. 2B, red arrow). The stripe then disappears.

Figure 2.

Expression of retinoic acid (RA) synthesising and catabolising enzymes in stages 8+ to 11 chick embryos. The stage is shown in the left-hand corner, the enzyme in the right-hand corner of each panel. A:Cyp26A1 expression at stage 8+ in the anterior neural folds and the neural tube adjacent to the somites and the tail bud. B:Cyp26A1 expression at stage 10 in rhombomere 3 (a section of this is shown in C), anterior neural tube (a section of this is shown in D), and the tail bud (a section of this is shown in E). C: Section through rhombomere 3 of a stage 10 embryo showing Cyp26A1 expression throughout the neuroepithelium. D: Section through the anterior neural tube of a stage 10 embryo showing expression of Cyp26A1 in the dorsal half. E: Section through the tail bud of a stage 10 embryo showing Cyp26A1 expression in the mesoderm. F: Anterior end of a stage 9 embryo showing Raldh2 expression at the anterior end of the infolding foregut. A section of this is shown in G. G: Sagittal section of F. H:Raldh3 expression in a stage 9 embryo showing a domain at the very anterior end of the embryo. I: Sagittal section through H, showing Raldh3 expression in the epithelium at the anterior end of the embryo, which abuts directly onto the forebrain. J: By stage 11, this Raldh3 domain has widened and moved slightly posteriorly. K: Sagittal section through J showing the posterior movement of the epithelial domain of Raldh3. L:Raldh2 expression in the heart inflow tracts and somites of a stage 11 embryo. Sections at three levels are marked. M,N,O: Sections through L at progressively more posterior levels showing expression of Raldh2 in the splanchnic mesoderm. P:Cyp26A1 expression in a stage 11 embryo showing expression in the neural tube and heart inflow tracts. Sections at three levels are marked. Q,R,S: Sections through P at progressively more posterior levels showing expression of Cyp26A1 in the endoderm. fb, forebrain; mb, midbrain; hb, hindbrain; e, epithelium.

At the same time as it disappears in the anterior central nervous system, Cyp26A1 becomes intense in the developing spinal cord posteriorly from the hindbrain/spinal cord junction (Fig. 2B). Sections reveal that, whereas the r3 stripe is throughout the neuroepithelium (Fig. 2C), the expression in the anterior spinal cord is only in the dorsal half (Fig. 2D), and in the tail bud, it is in the mesenchyme (Fig. 2E). The region where Cyp26A1 is in the dorsal half of the spinal cord is the region that has somites adjacent to it. These somites express Raldh2 (Figs. 2L, 3A–F), and this may explain the proposed role of RA in signalling to the spinal cord to induce the appropriate LIM code (Ensini et al., 1998), but the Cyp26A1 expression may also suggest a role for RA in dorsoventral patterning in the spinal cord (Fig. 6E,H).

At stage 9 Raldh3, which was previously in the endocardial tubes, now comes to be expressed in the most anterior epithelium in an arc shape in front of the neural tube (Fig. 2H). Sections reveal that this expression domain is in the epithelium directly abutting the forebrain (Fig. 2I), and it remains in this location for a prolonged period, although it slowly moves ventrally relative to the overlying neuroepithelium. This finding can be seen by comparing the sections in Figure 2I (stage 9) and Figure 2K (stage 11). As this ectodermal domain of Raldh3 moves posteriorly, it widens and then separates into three: the ventral third of the eye, Rathke's pouch, and the nasal placode (see below). The domain of Raldh3 expression at stage 9 is a signalling region (Fig. 6F) and has been demonstrated to be involved in the development of the frontal region of the head, activating Fgf-8 and Shh (Schneider et al., 2001).

Also at stage 9, Raldh2 appears in two domains in addition to the somites. These are the mesoderm of the inflow tract of the developing heart (Fig. 3C) and the rostral end of the anterior intestinal portal (Fig. 2F,G). The heart expression remains throughout development (Moss et al., 1998; Xavier-Neto et al., 2000), and at stage 11, sections reveal that it is in the splanchnic mesoderm, extending laterally into the somatic mesoderm (Fig. 2L–O). At the same time, Cyp26A1 appears in the inflow tract (Fig. 2P) and sections show that it is complementary to Raldh2, being in the endoderm of this region (Fig. 2Q–S). As has been suggested before, in the mouse (Moss et al., 1998) this area could be another signalling region (Fig. 6G).

Moving Borders of Gene Expression

Even after studying these few early stages of expression of these enzymes, it was clear that they were very dynamic both in terms of on/off and in terms of movements relative to anatomic structures. One movement already described is a gradual ventral shift of Raldh3 (Fig. 2I,K) and another is the spread of Cyp26A1 posteriorly along the edges of the neural plate in a W-shape (Fig. 1D,M). We investigated two other movements in more detail. One of the movements concerned Raldh2 and the other Cyp26A1.

As somitogenesis progressed, the border of Raldh2 expression that was initially at the level of the anterior border of the first somite (Fig. 3A,B) regresses somite by somite. By stage 9+, it has left somite 1 and is beginning to fade in somite 2 (Fig. 3C,D), and at stage 11, it is beginning to fade in somites 3 and 4 (Fig. 3F). Revealingly, this progressive loss of Raldh2 expression precisely mimics the loss of gene inductive power of transplanted somites (Itasaki et al., 1996), reinforcing the idea that RA production by the somites may be responsible for some aspects of patterning in the neural tube.

It is also clear that, within the cranial neuroepithelium, the posterior border of Cyp26A1 moves—but how far and to what extent? To answer this question we performed a series of double in situ studies with Cyp26A1 and Hoxb-1. The latter has an anterior expression border at the rhombomere 3/4 border, which does not move. Thus, by combining these two genes, we could see how Cyp26A1 moves relative to the presumptive r3/4 border. Comparing stages 7, 8, and 9 shows that the posterior border of Cyp26A1 moves posteriorly back to the r3/4 boundary to meet the anterior border of Hoxb-1 but does not go beyond that point (Fig. 3G–I). As the anterior border of Cyp26A1 regresses posteriorly, a thin stripe of expression is generated, which ends up in rhombomere 3 by stage 10 (Fig. 2B).

Stages 12–17

At stage 12, Raldh1 makes its first appearance in the embryo. It is expressed in two domains; one in a patch of endoderm beneath somites 5–10 (Fig. 4B) and the second domain is in the eye (arrowhead in Fig. 4A). This enzyme continues to be expressed in these two domains, and by stage 16, it is clear that Raldh1 is localised to the dorsal third of the eye (Fig. 4D). It is only the neural retina layer that expresses this enzyme, not the pigmented retina layer (Fig. 4E). The endodermal domain becomes the endodermal lining of the posterior foregut (Fig. 4F,G). An endodermal fate map suggests that this region of endoderm will form the pancreas and duodenum (Matsushita et al., 2002). At stage 12, the arc of Raldh3 expression in the anterior ectoderm has widened and has become associated with the ventral half of the developing eye (Fig. 4H). By stage 15, Raldh3 becomes additionally expressed in Rathke's pouch (Fig. 4J,K) and in the isthmus (Fig. 4J,K). Sections reveal that both the neural retina and the pigmented retina express Raldh3 (Fig. 4I). These localisations are summarised in Figure 6I.

Figure 4.

Expression of RA synthesising and catabolising enzymes in stages 12 to 20 chick embryos. The stage is shown in the left-hand corner, the enzyme in the right-hand corner of each panel. A,B:Raldh1 in a stage 12 embryo showing expression in the dorsal part of the eye (red arrow in A) and in the endoderm in the middle of the embryo (B). The level of the section in C is marked. C: Transverse section through B showing expression of Raldh1 in the endoderm. D,E: Expression of Raldh1 in the eye of a stage 18 embryo showing the dorsal location and that it is expressed in the neural retina, not the pigmented retina. F:Raldh1 in a stage 15 embryo showing expression in the endoderm of the foregut. G: Transverse section of a stage 20 embryo showing Raldh1 expression in the mesonephros and gut endoderm. H:Raldh3 in a stage 13 embryo showing that the expression has moved ventrally from the anterior end of the embryo to the epithelium below the eye. I:Raldh3 in a stage 18 embryo showing expression in the ventral half of the neural retina and pigmented retina. J:Raldh3 in a stage 15 embryo showing that the anterior epithelial expression has now moved back to Rathke's pouch. The other two domains at this stage are the isthmus and the ventral half of the eye. K: Section through a stage 18 embryo showing expression of Raldh3 in the isthmus and Rathke's pouch. L:Raldh2 in a stage 16 embryo showing expression in the dorsal pigmented retina and in a chevron-shaped piece of mesoderm posterior to the eye (red arrow). M:Raldh2 in a stage 18 embryo showing the domains in L at a higher power, namely the pigmented retina and the mesoderm beside the eye. pr, pigmented retina; nr, neural retina; ln, lens; i, isthmus; rp, Rathke's pouch; e, eye.

Raldh2 also has several new domains of expression during these stages in addition to the somites and developing heart. The somitic expression is present at these stages in the dermamyotome (Fig. 5B). One of these domains is the expansion from the paraxial mesoderm of the somites into the lateral plate but excluding the limb bud and tail bud mesoderm, which remains Raldh2 negative (Fig. 5A). A second domain is in the mesonephros (Fig. 5B). A third domain is a forked region of mesenchyme adjacent to the developing eye, which first appears at stage 14 and increases in intensity throughout subsequent stages (Fig. 4L,M). A fourth domain is in the dorsal half of the eye (Fig. 4L,M), which appears at stage 15 and is in the pigmented retina layer.

Figure 5.

Expression of retinoic acid (RA) synthesising and catabolising enzymes in stages 14–20 chick embryos. The stage is shown in the left-hand corner, the enzyme in the right-hand corner of each panel. A:Raldh2 in a stage 15 embryo showing that expression which was formerly in the somites has now moved into the lateral plate but is not present in the regions of the lateral plate, which will give rise to the limb buds. B: Transverse section of A showing Raldh2 in the dermamyotome (black arrow d) of the somites, the lateral plate, and the mesonephros (red arrow). C:Raldh2 in a stage 18 embryo showing expression in the somites, lateral plate, and mesonephros of the trunk but also new domains in the head region. These are the dorsal pigmented retina, the mesoderm adjacent to the eye (black arrow), and the dorsal epithelium of the first branchial groove (red arrow). D:Cyp26A1 in a stage 14 embryo showing expression in the tail bud, the dorsal half of the spinal cord, and the lateral plate endoderm (red arrow). E: Transverse section of a stage 20 spinal cord showing Raldh2 expression in the roof plate (red arrow) and developing motor neurons (black arrow). F: Transverse section of a stage 18 embryo through the developing limb bud showing Cyp26A1 expression in the limb bud epithelium. G:Raldh3 expression in the head of a stage 20 embryo. These domains are the nasal placode, the ventral half of the eye, Rathke's pouch, the isthmus, and the dorsal half of the otic vesicle. H:Cyp26A1 in the head of a stage 18 embryo showing expression in the dorsal half of the lens, the epithelium of the first branchial groove, and the anterior end of the lateral plate endoderm. I: High-power view of H, showing that Cyp26A1 is in the dorsal part of the lens and not in the neural retina or pigmented retina. d, dermamyotome; pr, pigmented retina; np, nasal placode; rp, Rathke's pouch; i, isthmus; ov, otic vesicle; bg, branchial groove; endo, endoderm; le, lens; nr, neural retina.

Cyp26A1 continues to be expressed in the dorsal third of the spinal cord (Fig. 5D) and in the tail bud during these stages. In addition, there are two further sites of expression. One is the lateral plate endoderm extending posteriorly from the heart to the most recently formed somite (Fig. 5D). A second is in the ectoderm covering the limb bud, which appears at stage 15 and expands as the limb bud enlarges (Fig. 5F). These domains are summarised in Figure 6H.

Stages 18–20

At stage 18, a new domain of expression of Raldh1 has appeared in addition to the dorsal eye and foregut endoderm. This domain is the mesonephros (Fig. 4G). At the same stage, a new domain of Raldh3 appears in addition to the isthmus, Rathke's pouch, and the ventral eye. This domain is the dorsal part of the otic vesicle (Fig. 5G). The expression of Raldh3 in the epithelium ventral to the eye has by stage 18 expanded anteriorly, and by stage 20 (Fig. 5G), it is clear that this new domain is the nasal placode.

Three new domains of Raldh2 appear at these stages. One is the epithelium at the top of the first branchial groove, which appears at stage 18 (Fig. 5C). The second is the motor neurons of the spinal cord, which appear at stage 19 (Fig. 5E) and only at limb levels, as has been described previously (Sockanathan and Jessell, 1998). The third is the roof plate (Fig. 5E).

Cyp26A1 continues at these stages to be expressed in the dorsal spinal cord and the tail bud while in the lateral plate endoderm it declines at the posterior end, but remains at the anterior end around the heart. In the expanding dorsal ectoderm of the limb bud Cyp26A1 continues expression eventually becoming concentrated in two stripes and the base of the apical ectodermal ridge (Swindell et al., 1999). Two new domains of expression appear. One is in the eye (Fig. 5H), and sections reveal that, whereas the synthesising enzymes are expressed in the neural or pigmented retina, Cyp26A1 is expressed in a dorsal segment of the lens (Fig. 5I). The second domain is in the epithelium of branchial arch 1 and branchial arch 2 along the first branchial groove (Fig. 5H) in an apparently complementary domain to Raldh2 at the top of the pouch. These domains are summarised in Figure 6I.

DISCUSSION

We describe the expression domains of three RA synthesising enzymes, Raldh1, Raldh2 and Raldh3, and one RA catabolising enzyme, Cyp26A1, during the early stages of chick embryogenesis. Although some of these data have been reported before in descriptions of individual enzymes on different species (see Introduction section), our intention here was to assemble all the enzyme data together in one species so that they might provide a source of data and also that we might possibly identify novel RA signalling domains.

The data, which are summarised in Figure 6 and Table 1, justify the latter aim as several novel regions of enzyme expression have been identified. These are Hensen's node (Raldh3 and Cyp26A1), the cardiac endoderm (Raldh3), the presumptive pancreatic endoderm (Raldh1), the dorsal lens (Cyp26A1), the first branchial groove (Cyp26A1 and Raldh2), and the dynamic spread of Cyp26A1 in the anterior hindbrain.

Table 1. Summary of the Expression Domains of the Three Retinoic Acid Synthesising Enzymes (Raldh1, Raldh2, Raldh3) and One Retinoic Acid Catabolising Enzyme (Cyp26A1) in Stages 4–20 of Chick Development
StageRaldh1Raldh2Raldh3Cyp26A1
3+Mid to posterior mesenchymeAnterior epiblast
4Mid to posterior mesenchymeHensen's nodeAnterior epiblast; mesenchmye around Hensen's node
5Mid to posterior mesenchymeEndoderm below edges of anterior neural plateAnterior neural plate; leaves midline
6Mid to posterior mesenchymeEndoderm below edges of anterior neural plateAnterior neural plate; spreads posteriorly at edges of neural plate meets at tail bud
7Somites with anterior border at 1st somiteEndoderm below inflow tracts of heartAnterior neural plate; neural folds; mesoderm of regressing node; tail bud
8Somites with anterior border at 1st somiteEndoderm below inflow tracts of heartAnterior neural plate; neural folds; mesoderm of regressing node; tail bud
9Somites, regresses from somite 1 & 2; mesoderm of heart inflow tracts; rostral end of anterior intestinal portalEpithelium abutting anterior forebrainClosing neural folds anteriorly regressing in an anterior direction; open neural folds posteriorly; tail bud
10Somites, regresses to somite 4/5 border; mesoderm of heart inflow tractsEpithelium abutting anterior forebrainr3 stripe; dorsal half of anterior spinal cord; tail bud
11Somites from 4/5 border; mesoderm of heart inflow tractsEpithelium abutting anterior forebrainr3 stripe; dorsal half of anterior spinal cord; tail bud; endoderm of heart inflow tract
12Endoderm beneath somites 5–10; dorsal neural retinaSomites from 4/5 border; mesoderm of heart inflow tractsEpithelium below eye; spreading into ventral third of eyeDorsal third of spinal cord; tail bud
13Endoderm beneath somites 5–10; dorsal neural retinaSomite from 4/5 border; mesoderm of heart inflow tractsEpithelium below eye; spreading into ventral third of eyeDorsal third of spinal cord; tail bud
14Endoderm beneath somites 5–10; dorsal neural retinaDermamyotome of somites; spreading into lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; mesonephrosEpithelium below eye; ventral third of eyeDorsal third of spinal cord; tail bud; lateral plate endoderm
15Posterior foregut endoderm; dorsal neural retinaDermamyotome of somites; lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; dorsal pigmented retina; mesonephrosVentral third of eye; Rathke's pouch; isthmusDorsal third of spinal cord; tail bud; lateral plate endoderm; limb bud ectoderm
16Endoderm of pancreas and duodenum; dorsal neural retinaDermamyotome of somites; lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; dorsal pigmented retina; mesonephrosVentral third of eye; Rathke's pouch; isthmusDorsal third of spinal cord; tail bud; lateral plate endoderm; limb bud ectoderm
17Endoderm of pancreas and duodenum; dorsal neural retinaDermamyotome of somites; lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; dorsal pigmented retina; mesonephrosVentral third of eye; Rathke's pouch; isthmusDorsal third of spinal cord; tail bud; lateral plate endoderm; limb bud ectoderm
18Endoderm of pancreas and duodenum; dorsal neural retina; mesonephrosDermamyotome of somites; lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; dorsal pigmented retina; mesonephros; epithelium of 1st branchial groove; roof plateVentral third of eye; Rathke's pouch; isthmus; dorsal third of otic vesicle; nasal placodeDorsal third of spinal cord; tail bud; anterior lateral plate endoderm; limb bud ectoderm; epithelium of 1st branchial groove; dorsal lens
19Endoderm of pancreas and duodenum; dorsal neural retina; mesonephrosDermamyotome of somites; lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; dorsal pigmented retina; mesonephros; epithelium of 1st branchial groove; roof plate; motor neuronsVentral third of eye; Rathke's pouch; isthmus; dorsal third of otic vesicle; nasal placodeDorsal third of spinal cord; tail bud; anterior lateral plate endoderm; limb bud ectoderm; epithelium of 1st branchial groove; dorsal lens
20Endoderm of pancreas and duodenum; dorsal neural retina; mesonephrosDermamyotome of somites; lateral plate; mesoderm of heart inflow tracts; mesenchyme adjacent to eye; dorsal pigmented retina; mesonephros; epithelium of 1st branchial groove; roof plate; motor neuronsVentral third of eye; Rathke's pouch; isthmus; dorsal third of otic vesicle; nasal placodeDorsal third of spinal cord; tail bud; anterior lateral plate endoderm; limb bud ectoderm; epithelium of 1st branchial groove; dorsal lens

Although this work was intended as a summary, it is likely that there will be more enzymes to be taken into consideration. For example, the interneuronal region of the neural tube and the telencephalon express an as yet unidentified enzyme (Mic et al., 2002; Niederreither et al., 2002), and the mouse embryo expresses and additional CYP, Cyp26B1 (Maclean et al., 2001), which presumably is likely to be present in the chick embryo as well. So further additions, therefore, are likely in the future.

Early Hensen's Node

A new domain of gene expression, and the earliest in the embryo that we have identified is the stage 4 Hensen's node, which expresses three enzymes: Raldh3 in the epiblast of the node, Cyp26A1 in the epiblast anterior to the node and in the mesoderm of the node, and Raldh2 in the mesoderm posterior to the node (Fig. 6A). This finding surely resolves the issue of whether the node synthesises RA. The node was reported to synthesise RA at a much higher rate than the surrounding tissues (Chen et al., 1992; Hogan et al., 1992) and to be capable of inducing duplications when grafted to the anterior side of the chick limb bud (Hornbruch and Wolpert, 1986; Wagner et al., 1990), and a bead soaked in RA can substitute for Hensen's node in the induction of a secondary axis (Chen and Solursh, 1992). The reason why the node was not detected as a site of RA synthesis in our earlier studies using a F9 RARElacZ reporter cell line (Maden et al., 1998) could be that the presence of Cyp26A1 in the mesoderm of the node might have rapidly catabolised any RA generated by Raldh3 into inactive metabolites, thus preventing activation of the F9 cells below.

What might be the function of this RA generated by the node? It is unlikely that it is involved in establishing the anteroposterior pattern of the main body axis as was originally proposed (Hogan et al., 1992; Chen et al., 1992), because the expression of Raldh3 is so brief, rather than continuing as the node regresses. Furthermore, vitamin A–deficient embryos do not have a completely disrupted anteroposterior axis, only a missing posterior hindbrain (Maden et al., 1996; Gale et al., 1999; White et al., 2000b). This finding suggests an involvement in a more discrete patterning event and one such possibility is the determination of left/right asymmetry, which is also a function of the node.

Although rarely mentioned in conventional descriptions of the gene cascades that are thought to give rise to left/right asymmetry (e.g., Capdevila et al., 2000), there is considerable evidence that RA plays a role in this process. For example, a bead soaked in RA placed on the right of Hensen's node results in randomisation of heart looping, ectopic expression of Nodal and Pitx2, and two extracellular matrix proteins, hLAMP1 and a fibrillin-related protein (Smith et al., 1997; Tsukui et al., 1999). Application of an RA antagonist to the left side of Hensen's node abolishes endogenous Nodal and Pitx2 expression and results in randomisation of heart looping (Tsukui et al., 1999). In the mouse, RA treatment at headfold stages perturbs cardiac looping and induces bilateral expression of the normally left-sided genes Nodal, lefty-1, lefty-2, Nkx3.2, and Pitx-2 (Chazaud et al., 1999; Schneider et al., 1999; Wasiak and Lohnes, 1999); conversely, treatment of mice embryos with an RAR antagonist results in the randomisation of heart looping and the down-regulation of these laterality genes (Chazaud et al., 1999; Wasiak and Lohnes, 1999). Vitamin A–deprived quail embryos have abnormal heart looping (Heine et al., 1985; Dersch and Zile, 1993), 72% of which have a reversal of cardiac asymmetry (Zile et al., 2000). RA is considered to act in parallel with Shh, both of which act to establish Pitx2 expression on the left side of the embryo. An asymmetric distribution of a RA synthesising enzyme or a RA catabolising enzyme, therefore, would be an ideal situation, but this does not seem to be the case. Thus, if RA is involved in left/right determination, then we need some other way of explaining its asymmetrical distribution.

Cardiac Endoderm

Another novel signalling region we identified was the endoderm at the anterior lateral edges of the embryo, below the cardiac mesoderm, which expresses Raldh3 (Fig. 6B). RA is well known to be involved in the complex processes of heart development subsequent to the establishment of left/right asymmetry, which manifests itself as heart looping. Both an excess and a deficiency of RA causes a wide array of heart defects, including transposition of the great arteries, double aortic arch, aortic and pulmonary trunk hypoplasia, valve hypoplasia, and spongy ventricular muscle (Wilson and Warkany, 1949; Shenfelt, 1972; Fantel et al., 1977; Lammer et al., 1985; Pexieder et al., 1992; Yasui et al., 1997). Continued RA signalling is required throughout heart formation (Moss et al., 1998; Xavier-Neto et al., 1999, 2000; Niederreither et al., 2001) generated by Raldh2 and transduced by the RARs and RXRα (Mendelsohn et al., 1994; Sucov et al., 1994; Kastner et al., 1995, 1997; Chen et al., 1998). Later signalling in the heart region is summarised in Figure 6G, involving Raldh2 in the splanchnic mesoderm and Cyp26A1 in the endoderm.

The RA-deficient quail embryo displays multiple heart defects, the inflow tract fails to develop, and there is a hypoplastic, nonseptated heart tube in addition to the abnormal looping referred to above (Heine et al., 1985; Dersch and Zile, 1993). It has been discovered that grafts of stage 4 normal anterior endoderm can completely rescue the RA-deficient heart phenotype (Ghatpande et al., 2000), resulting in the restoration of expression of GATA-4. Indeed, the anterior lateral endoderm is known to play a role in cardiac lineage determination and molecules such as FGF and BMP-2, which are expressed there induce Nkx2-5, GATA-4, and Tbx5 in the overlying cardiac mesoderm (review, Yutzey and Kirby, 2002). We hypothesise, therefore, that RA generated by Raldh3 in the anterior lateral endoderm is involved in this inductive cascade (Fig. 6B).

Neural Crest

A further potential signalling region could be an involvement in neural crest generation as Cyp26A1 marks the line of neural crest at the edges of the neural plate (Fig. 6C). RA is known to be involved in this process (Villaneuva et al., 2002), perhaps by means of the role that BMP-4 plays, because the expression pattern of Cyp26A1 coincides with that of BMP-4, msx-1, and Sox-3 (Streit and Stern, 1999). It could be involved in stabilising the boundary between BMP-4 expressing (non-neural ectoderm) and BMP-4 nonexpressing (neural ectoderm) regions where the crest forms and where the local regulation of BMP activity generates a ‘border’ state (Streit and Stern, 1999).

Presumptive Hindbrain and the Tail Bud

In the stage 6–8 embryo, two potential signalling regions between Cyp26A1 and Raldh2 are identified one at each end of the embryo (Fig. 6D). At the anterior end, the gap between Raldh2 and Cyp26A1 is 100–200 μm, between which the hindbrain develops and across which a gradient of RA could be generated. However, this hypothetical gradient has not been demonstrated, and the Cyp26A1 knockout mouse does not display anterior abnormalities consistent with a loss of this gradient (Abu-Abed et al., 2001; Sakai et al., 2001). Nevertheless, RA generated by Raldh2 is crucial to hindbrain development, because without Raldh2 (Niederreither et al., 2000) or without RA (Maden et al., 1996; Gale et al., 1999), the posterior hindbrain fails to develop. A scheme whereby RA performs this function without necessarily forming a gradient has been proposed recently (Gavalas, 2002; Maden, 2002). This scheme involves the interaction of mesodermally derived RA with an expanding neuroepithelium where the function of RA is to sequentially generate rhombomere boundaries. But the possibility of an additional RA synthesising enzyme within the neural tube itself (Mic et al., 2002; Niederreither et al., 2002) will undoubtedly alter our ideas on these issues. At the posterior end of the embryo, however, the Cyp26A1/Raldh2 interaction is crucial as the Cyp26A1 knockout mice have severe posterior defects, including sirenomelia (Abu-Abed et al., 2001; Sakai et al., 2001).

Somites and Neural Tube Patterning

It is already known that Raldh2 expressed in the somites provides a source of RA for patterning the neural tube (Fig. 6E). When somites are grafted adjacent to the hindbrain, then various Hox genes are ectopically induced (Grapin-Botton et al., 1995, 1997; Itasaki et al., 1996; Gould et al., 1998) but not when RA signalling is prevented (Gould et al., 1998). The gradual loss of this inductive ability in somites 1–4 over stages 8–11 has been demonstrated (Itasaki et al., 1996), and this finding is exactly the same time period over which Raldh2 disappears from the same somites (Fig. 3A–F). The grafting of brachial somites to thoracic neural tube levels induces a brachial motor neuron identity in thoracic ventral horn neurons (Ensini et al., 1998). This role for RA is concerned with establishing a CNS pattern along the anteroposterior axis, but there is also a suggestion that RA is involved in patterning the dorsoventral axis of the neural tube. For example when naive neural plate tissue is cultured in the presence of retinol, certain subsets of interneurons are induced (Pierani et al., 1999) and several classes of interneurons are absent from the RA-deficient quail neural tube (Wilson and Maden, unpublished data). The new RA generating activity in the neural tube (Mic et al., 2002; Niederreither et al., 2002) fits perfectly into this scenario, as it is expressed in the interneuronal region. This enzyme may function with Cyp26A1 in the dorsal half of the neural tube to generate a local signalling region (Fig. 6E).

Pancreas

During stages 12–15, the endoderm shows a discrete area of Raldh1 expression in a region below somites 5–10 (Fig. 6E). According to fate maps of the endoderm, this region will generate the pancreas and duodenum (Matsushita et al., 2002), suggesting an involvement of RA signalling in the patterning of these structures. Indeed, a role for RA in the formation of the pancreas has been elaborated very recently (Stafford and Prince, 2002), and the expression of Raldh1 in this location can provide a mechanism for localised RA signalling.

Anterior Forebrain and Face

The region of Raldh3 expression in the anterior ectoderm abutting the forebrain in the stage 9 embryo (Fig. 6F) already has been identified as a signalling region (Schneider et al., 2001). When beads soaked in RAR and RXR antagonists to inhibit RA signalling are grafted adjacent to this Raldh3 expressing region then Shh, Fgf-8 in the forebrain are down-regulated and the result is cyclopic embryos. As in the case of the pancreas described above, a very discrete enzyme localisation highlights the temporal and spatial requirement for RA.

Other Regions

In the later trunk (Fig. 6H) and head (Fig. 6I), there are many sites of enzyme expression that reveal a potential role for RA signalling. Some already have been established such as Raldh2 in the motor neurons (Sockanathan and Jessell, 1998) and the multiple domains in the eye (McCaffery et al., 1999). With regard to the latter, we note that Cyp26A1 is not expressed around the equator of the eye as it is in the mouse embryo, but in the chick, it is expressed in the dorsal lens. This location would not be available, therefore, to generate an RA-free zone in the middle of the developing chick eye. Four other regions where RA could play a role include the following: the dermamyotome, suggesting RA might be involved in somite patterning; the roof plate, possibly involving interactions with BMP gradients; the lateral plate, where one function of RA is to up-regulate EphA4 (Schmidt et al., 2001); and the mesonephros, where RA is known to play a role in kidney development (Mendelsohn et al., 1999). The mesonephros is unusual in expressing two synthesising enzymes Raldh1 and Raldh2. In the head, additional regions include the following: the nasal placode; a region of mesoderm behind the eye that might be responsible for guiding the trigeminal axons as it precisely overlays the division of the trigeminal nerve into the ophthalmic and maxillary branches; the isthmus, where RA may play a role in establishing the midbrain/hindbrain border; Rathke's pouch, suggesting a role for RA in the development of the pituitary gland; the dorsal otic vesicle, where RA may play a role in the development of the vestibular apparatus; and the epithelium of the first branchial pouch, suggesting a role in patterning the first or second branchial arch (Wendling et al., 2000; Quinlan et al., 2002). With regard to the latter, it seems that ectopic RA, interacting with noggin respecifies the pattern of the maxillary prominence into that of the frontonasal mass, rather than more posterior branchial arches (Lee et al., 2001).

In conclusion, having highlighted all these potential regions, some of which are novel, it is clear that we should now design experiments to test the role that RA may play in their development in addition to answering the question of whether individual enzymes perform distinct functions or do they all just make the same RA product?

Experimental Procedures

Fertilised chicken eggs were purchased from local suppliers. They were incubated until the appropriate stages (Hamburger and Hamilton, 1951) and fixed in 4% paraformaldehyde. Whole-mount in situ hybridisation was performed according to standard protocols using digoxigenin-labelled probes colour reacted with BMP purple to give a purple colour reaction. For double in situ hybridization, a digoxigenin-labelled probe was cohybridised with a fluorescein-labelled probe, then sequentially exposed to antibody and colour reaction. The fluorescein reaction was performed first with a fluorescein antibody, then colour reacted with BCIP (5-bromo-4-chloro-3-indolyl phosphate, toluidine salt). After quenching the alkaline phosphatase from the first antibody with 4% paraformldehyde, embryos were then exposed to the digoxigenin antibody and colour reacted with BMP purple. Probes were kindly given by R. Godbout (Raldh1), F. Grun (Raldh3), and E. Swindell (Cyp26A1). Raldh2 was cloned by PCR using chick somite RNA with primers taken from the published sequence, accession no. AF064253 (Sockanathan and Jessell, 1998). For sectioning, embryos were embedded in gelatin and sectioned on a Vibratome at 60 μm.

Acknowledgements

We thank the BBSRC for financial support for this project.

Ancillary