Expression of the murine retinol dehydrogenase 10 (Rdh10) gene correlates with many sites of retinoid signalling during embryogenesis and organ differentiation

Authors

  • Laura Cammas,

    1. IGBMC (Institut de Génétique et de Biologie Moléculaire et Cellulaire), Inserm, U596, CNRS, UMR 7104, Illkirch, and Université Louis Pasteur, Faculté de Médecine, Strasbourg, France
    Search for more papers by this author
  • Raymond Romand,

    1. IGBMC (Institut de Génétique et de Biologie Moléculaire et Cellulaire), Inserm, U596, CNRS, UMR 7104, Illkirch, and Université Louis Pasteur, Faculté de Médecine, Strasbourg, France
    Search for more papers by this author
  • Valérie Fraulob,

    1. IGBMC (Institut de Génétique et de Biologie Moléculaire et Cellulaire), Inserm, U596, CNRS, UMR 7104, Illkirch, and Université Louis Pasteur, Faculté de Médecine, Strasbourg, France
    Search for more papers by this author
  • Carole Mura,

    1. Institut Clinique de la Souris (Mouse Clinical Institute), Illkirch, France
    Search for more papers by this author
  • Pascal Dollé

    Corresponding author
    1. IGBMC (Institut de Génétique et de Biologie Moléculaire et Cellulaire), Inserm, U596, CNRS, UMR 7104, Illkirch, and Université Louis Pasteur, Faculté de Médecine, Strasbourg, France
    • Université Louis Pasteur, Faculté de Médecine, Strasbourg, F-67000 France
    Search for more papers by this author

Abstract

Retinoic acid acts as a signalling molecule regulating many developmental events in vertebrates. As this molecule directly influences gene expression by activating nuclear receptors, its patterns of synthesis have to be tightly regulated, and it is well established that at least three retinaldehyde dehydrogenases (RALDHs) are involved in such tissue-specific synthesis. Whereas embryos from oviparous species can obtain retinaldehyde by metabolizing carotenoids stored in the yolk, placental embryos rely on retinol transferred from the maternal circulation. Here, we show that the gene encoding one of the murine retinol dehydrogenases, Rdh10, is expressed according to complex profiles both during early embryogenesis and organ differentiation. Many of its expression sites correlate with regions of active retinoid signalling and Raldh gene expression, especially with Raldh2 in the early presomitic and somitic mesoderm, retrocardiac and posterior branchial arch region, or later in the pleural mesothelium and kidney cortical region. Rdh10 also shows cell-type and/or regional specificity during development of the palate, teeth, and olfactory system. During limb bud development, it may participate in retinoic acid production in proximal/posterior cells, and eventually in interdigital mesenchyme. These data implicate the retinol to retinaldehyde conversion as the first step in the tissue-specific regulation of retinoic acid synthesis, at least in mammalian embryos. Developmental Dynamics 236:2899–2908, 2007. © 2007 Wiley-Liss, Inc.

INTRODUCTION

Retinoic acid (RA), the active derivative of retinol (the circulating form of vitamin A), is a signalling molecule that plays pleiotropic roles during embryogenesis and organ differentiation (for reviews, see Ross et al.,2000; Clagett-Dame and DeLuca,2002; Niederreither and Dolle,2006). This molecule acts as a ligand for nuclear receptors, the RA receptors α, β, and γ, that act as heterodimers with RXRs to regulate the expression of target genes. RA is a diffusible, lipophilic compound, and there is evidence that it can act by diffusion across tissue layers in certain developmental processes (e.g., Diez del Corral et al.,2003; Matt et al.,2005; Molotkov et al.,2005). Thus, the distribution of RA must be tightly controlled in embryonic tissues, and this control is achieved by an interplay between enzymes involved in its synthesis (see below) and its metabolism into inactive compounds (for a review, see Pennimpede et al.,2006).

In oviparous species, retinoids are stored in the egg yolk mainly as β-carotenoids that will be metabolized to retinaldehde by the BCOX enzyme (Lampert et al.,2003, and references therein), whereas in placental species retinol is transferred by an active mechanism from maternal to embryonic and fetal blood circulations (Ward et al.,1997). Two oxidative reactions are required to sequentially transform retinol into retinaldehyde and RA. A large body of work has established that the retinaldehyde to RA conversion step is temporally and spatially regulated in embryonic tissues. Three retinaldehyde dehydrogenases (RALDHs), each with a distinct expression pattern, operate during development. Gene knockout studies in mice have shown that RALDH2 is involved in several morphogenetic processes in the early embryo (Niederreither et al.,1999,2001,2002b; Mic et al.,2002,2004), whereas RALDH3 acts more selectively during nasal (Dupe et al.,2003), ocular (Matt et al.,2005; Molotkov et al.,2006), and kidney (P. Dollé and C. Mendelsohn, unpublished observations) development and RALDH1, although dispensable on its own, acts redundantly with RALDH2 and 3 during craniofacial development (Matt et al.,2005; Molotkov et al.,2006; Halilagic et al.,2007). The expression patterns of Raldh genes are highly conserved across species, including in oviparous organisms (e.g., Begemann et al.,2001; Blentic et al.,2003).

Unlike the retinaldehyde to RA conversion, until recently, there was no evidence suggesting that the retinol to retinaldehyde step needs to be spatiotemporally regulated in the mammalian embryo. Two classes of enzymes, the cytosolic alcohol dehydrogenases (ADHs) and the microsomal short-chain dehydrogenase/reductases known as retinol dehydrogenases (RDHs), can catalyze the oxidation of retinol (Duester,2000; Zhang et al.,2007, and references therein). At least two Adh genes are expressed in a tissue-specific manner during mouse development (Vonesch et al.,1994; Ang et al.,1996), whereas the Adh3 gene is ubiquitously expressed (Molotkov et al.,2002). Gene knockout studies revealed that none of these genes is indispensable for normal development to proceed; some compound mutant genotypes lead to a reduced survival under conditions of vitamin A deficiency, although no tissue-specific abnormalities were observed (Duester et al.,2003; Zhang et al.,2007, and references therein).

In the mouse, approximately a dozen RDHs have been characterized, most of which acting as all-trans or 11-cis retinol dehydrogenases (or retinaldehyde reductases) during the visual cycle (Liden and Eriksson,2006; Zhang et al.,2007, and references therein). Recently, a screen for murine ENU-induced alleles generating recessive embryonic phenotypes has identified a Rdh10 mutation as being responsible for a spectrum of abnormalities reminiscent of RA-deficiency phenotypes (Sandell et al.,2007). It was further demonstrated using a reporter transgene and phenotypic rescue by maternal RA supplementation that the Rdh10/trex (Tyrannosaurus rex, named after the forelimb bud-specific reduction defects) phenotype does indeed result from inefficient embryonic RA signalling (Sandell et al.,2007). These authors provided initial data on tissue-specific expression of Rdh10 in embryos from embryonic day (E) 8–E10.5, for example, in somitic and lateral plate mesoderm, neural groove, lung buds, and optic and otic vesicles (Sandell et al.,2007).

These results showed for the first time that an enzyme involved in the retinol to retinaldehyde conversion participates to the tissue specificity of RA synthesis in the mouse embryo. Given the complex expression patterns of the retinaldehyde dehydrogenase genes, a detailed characterization of Rdh10 expression features throughout development is important, as its induction in a given tissue may indicate how RA synthesis is primed, and its coexpression with one or several Raldh genes may indicate “hot spots” of such synthesis. Here, we have performed a detailed analysis of Rdh10 expression during embryonic and fetal mouse development, using both whole-mount in situ hybridization (ISH) and ISH on cryosections. These results revealed striking similarities with the known expression patterns of Raldh genes, although in some instances there were differences in the timing of activation and/or the exact tissue distribution of Rdh10. Some of its expression features, however, could not be correlated with expression of known Raldh genes. The significance of these similarities or discrepancies is discussed.

RESULTS

Regionalized Expression at Presomite and Early Somite Stages

Rdh10 expression was first detected in headfold stage embryos (E7.5; Fig. 1A). Expression was not consistently seen among the youngest embryos at early headfold stage (Fig. 1A, inset), whereas slightly older embryos displayed weak signal among mesodermal cells (main panel). This expression intensified at the late headfold stage (Fig. 1B), and section analysis showed strong signal in the lateral mesodermal wings, which extended at lower levels toward the base of the headfolds and was excluded from the primitive streak and node (Fig. 1B, inset, and data not shown). Later on in presomite stage embryos (E8), the ISH signal was clearly regionalized from the base of the headfolds to the node region (Fig. 1C). Analysis of transverse sections confirmed mesodermal expression, whereas no signal was seen in the neurectoderm, epiblast, and node proper (Fig. 1D, inset). The striped pattern observed rostrally to the node (Fig. 1D, main panel) may reflect segregation of mesodermal cells engaged in somite formation.

Figure 1.

Whole-mount in situ hybridization analysis of Rdh10 expression in presomite and early somite stage embryos. A,B: Frontal views of two embryos at slightly successive stages of headfold development. Inset in A shows a profile view of a younger embryo with no visible Rdh10 expression. Inset in B is a frontal section of the same embryo. C: Profile view of a late headfold stage embryo, collected shortly before formation of the first somite pair. D: View of the same embryo centered on the node. Inset: Transverse section, detail of the node. E: Six-somite stage embryo, profile view. Insets: ventral view of the heart (lower panel). Frontal section of the midbrain region (upper panel). F: Dorsal view of the mid–hindbrain region. G: A 10-somite stage embryo, profile view. Inset: Section at the level of the cloacal–allantois junction. H: Detail of the caudal region (the view is dorsal, hence, the ventral spinal cord appears on top). I–L: Four transverse sections of an eight-somite stage embryo, from upper cervical to trunk levels. fg, foregut; hb, hindbrain; hf, headfold; im, intermediate mesoderm; no, node; ht, heart; lpm, lateral plate mesoderm; mb, midbrain; sc, spinal cord; sl, splanchnopleure; sm, somatopleure; so, somite.

Rdh10 expression was most conspicuous in mesoderm derivatives in early somite stage embryos (E8.5–E9; Fig. 1E–H). Strong signal was observed in the somites themselves (Fig. 1F), and while somitic differentiation proceeded, it became concentrated dorsally (Fig. 1G,I,K,L). Analysis of transverse sections showed expression in all three components of the mesoderm, that is, the somitic, intermediate, and lateral mesoderm, forming the somatopleure and splanchnopleure layers (e.g., Fig. 1L). In the posterior region of the embryo where somites are being formed, Rdh10 was induced in the newly formed somites undergoing epithelialization (Fig. 1H, arrowhead). Expression was absent from presomitic mesoderm (PSM) and extended more caudally in the lateral plate mesoderm than the somitic mesoderm (Fig. 1H). Toward the heart region, expression was found up to the level of the posteriormost chambers, the developing sinus venosa (Fig. 1E, inset). Dorsally to the heart, strong signal was found in lateral mesoderm up to the level of the posterior branchial region, surrounding the foregut pocket (Fig. 1G,J,K). Actually a few cells of the lateral foregut epithelium were also positive (Fig. 1J).

Expression of Rdh10 was regionally restricted in the developing neuroepithelium. A stripe of expression was seen in the ventralmost cells (floor plate) of the posterior hindbrain (Fig. 1F) and spinal cord (Fig. 1H,I–L). Weaker expression was also seen in the midbrain region (Fig. 1E), which appeared to localize rostrally to the mid–hindbrain isthmus (Figs. 1G, 2A). Thus, contrasting with the strong expression found at prospective posterior cervical and trunk levels, many of the head tissues as well as the caudal region of the embryo were devoid of Rdh10 transcripts at these stages (Fig. 1E–H). Weak expression was seen, however, in a patch of cells at the base of the allantois (Fig. 1G, insert).

Figure 2.

Whole-mount in situ hybridization analysis of Rdh10 expression in embryonic day (E) 9.5 embryos. A–E: Overall profile view (A) and details of the otocyst and branchial region (B), posterior trunk (C), and dorsal views of the forelimb bud(s) (D,E) of E9.5 embryos. Insets in A show details of the mid–hindbrain region in whole-mounts and sections, respectively. F–I: Four serial transverse sections from the level of the second branchial arches to the posterior foregut near the putative lung buds. J,K: Sections through the optic vesicle and otocysts, respectively. L: Section through the foregut at the level of the prospective stomach pocket. M: A more posterior section (caudal to forelimb level) showing the somitic derivatives, mesonephric columns and midgut. The insert is a different section plane showing the hindgut. N: Detail of the forelimb bud and mesonephros. ao, dorsal aorta; b1, b2, branchial arches; dm, dermomyotome; fg, foregut; flb, forelimb bud; hb, hindbrain; ht, heart; lpm, lateral plate mesoderm; mg, midgut; mn, mesonephros; ot, otocyst; ov, optic vesicle; sc, spinal cord; sl, splanchnopleure; so, somite.

Altogether, the Rdh10 expression pattern in pre- and early somite stage embryos shares strong similarities with that of the retinaldehyde dehydrogenase gene Raldh2 (Niederreither et al.,1997; Moss et al.,1998). However, there are some discrepancies both in the timing of induction and in the exact distribution of both gene transcripts in mesodermal derivatives. Rdh10 also displays expression features, for example, in the floor plate neuroepithelium, which have no counterpart for Raldh2 or any of the known Raldh enzymes (see the Discussion section).

Specific Expression During Foregut, Mesonephros, and Limb Bud Morphogenesis

Rdh10 expression became further regionalized by E9.5. Transcripts were still detected within the differentiating somites. However, the somites posterior to the forelimb level were only weakly labeled (Fig. 2A,C), suggesting that Rdh10 induction declines after formation of approximately the first 10 somite pairs. With respect to differentiation of somitic mesoderm, transverse sections at trunk levels showed that only the dorsal edge of the dermomyotome, together with cells surrounding the dorsal spinal cord, were labeled (Fig. 2M). Similarly at posterior hindbrain levels only the dorsal superficial mesenchyme was positive (Fig. 2F–I).

At this stage, Rdh10 transcripts were especially concentrated around the foregut in the posterior pharyngeal region (Fig. 2A,B). Section analysis showed that the highest labeling was in the mesenchyme surrounding the foregut and dorsal aortae, dorsally to the atrial chambers of the developing heart (Fig. 2H,I). Discrete areas of foregut endoderm also expressed Rdh10 (Fig. 2G–I), including at the level of the second branchial arch (Fig. 2F). At more posterior levels, section analysis clearly showed endoderm-specific expression at the prospective stomach level (Fig. 2L).

Another region of strong Rdh10 expression was in intermediate (mesonephric) and lateral plate mesoderm, from forelimb to allantoic levels (Fig. 2A). Analysis of whole-mounts and transverse sections (Fig. 2C and M,N, respectively) showed that highest expression was specific to the mesonephric ducts within the intermediate mesoderm. With respect to lateral plate mesoderm, both the somatopleure layer forming the outer body wall and the splanchnopleure layer surrounding the mid- and hindgut were labeled (Fig. 2M, main panel and insert, respectively). The Rdh10 expression pattern was interesting regarding the developing forelimb buds. Labeled cells were seen along the entire proximal margin of the buds (Fig. 2D,E). Additionally, Rdh10-expressing cells extended along both the anterior and posterior margins of the growing buds, and labeling was most intense in the posterior area known as the “zone of polarizing activity” (Tickle et al.,1985; Sanz-Ezquerro and Tickle,2000, and references therein; Fig. 2D,E, brackets).

At E9.5 Rdh10 transcripts had mainly retreated from the floor plate cells of the neural tube, although they could still be seen in some regions of the hindbrain (Fig. 2F) and posterior spinal cord (Fig. 2M). Expression also weakly persisted in the mid–hindbrain boundary area (Fig. 2A, insets). Regionalized expression appeared in two precursors of sensory organs: (1) the optic vesicles, where transcripts were dorsally restricted (Fig. 2J); and (2) the otocysts, in which transcripts were restricted to cells along the hindbrain surface (Fig. 2K).

Rdh10 Expression During Craniofacial Development

As whole-mount ISH does not label inner tissues in embryos aged E10.5 or beyond, our analysis was further conducted at late embryonic and fetal stages by hybridization of serial cryosections. This analysis confirmed the persistence of Rdh10 expression in discrete areas of the developing branchial arches. Unlike at E8.5–E9.5, when most conspicuous expression occurred throughout the mesenchyme surrounding the posterior foregut (Figs. 1G, 2A,B), at E10.5 labeling was mainly found in both the ectodermal and endodermal components of the branchial pouches (Fig. 3A–C). Region-specific mesenchymal labeling was also seen, especially at the level of the developing maxillomandibular cleft (data not shown).

Figure 3.

Rdh10 expression in differentiating facial components. A–P: Selected sagittal (A,F,H,J,N–P) and frontal (B–E,G,I,K–M) section planes from embryonic day (E) 10.5 (A–C), E12.5 (D,E,H), E14.5 (F,G,I,K–N), E16.5 (J), and E18.5 (O,P) embryos and fetuses are shown. See main text for details. The arrowheads in A indicate specific regions of expression in branchial pouches. b1–b3, branchial arches; ec, ectoderm; en, endoderm; ep, epithelium; fg, foregut; in, incisor; mo, molar; md, mandibular process; me, mesenchyme; mx, maxillary process; ng, nasal gland; np, nasal pit; ns, nasal septum; oe, olfactory epithelium; ps, palatal shelf; td, thyroid duct; to, tongue.

The developing nasal epithelium is a major site of RA synthesis during embryogenesis (Li et al.,2000; Rawson and LaMantia,2006).Whereas the early olfactory placodal epithelium was not obviously labeled before and at E9.5 (Figs. 1G, 2A), region-specific expression of Rdh10 was seen from E10.5 onward (Fig. 3D,E, and data not shown). Intriguingly, most of the nasal placodal epithelium was negative (Fig. 3D), although strong expression was found distally at the level of the nasal pits (Fig. 3E). Additionally, expression was seen in the distal mesenchyme of the growing lateral nasal processes (Fig. 3D). Regionalized Rdh10 expression was a hallmark of early olfactory epithelium differentiation, and by E14.5 specific regions of the epithelium expressed the gene, with highest levels toward the recesses of the developing olfactory (ethmoturbinate) cavities (Fig. 3F,G). There was also region-specific mesenchymal expression toward the distal tips of both the lateral and medial nasal processes (Fig. 3G). Eventually Rdh10 expression decreased in the olfactory epithelium by E16.5, whereas it was strongly induced in cells of the developing nasal serous glands (Fig. 3J).

Several additional regions and/or tissues of the developing face specifically expressed Rdh10 expression during fetal development. Thus, expression was seen in cells surrounding the developing thyroid duct at E12.5 (Fig. 3H). Between E12.5 and E16.5, expression was seen in specific mesenchymal areas. These were usually located toward the extremities of growing structures, such as the maxillary and mandibular processes (Fig. 3M,N). Another trend consisted of cells surrounding grooves or clefts, for example, at the base of the tongue (Fig. 3I). Expression was also specific to the mesenchyme of the dental sac and papilla of the tooth buds at E14.5–E16.5 (Fig. 3I, and data not shown). Specific expression was detected in the palatal shelves, which are known to be highly sensitive to both excess and deficient RA signalling during development (Lohnes et al.,1994; Cuervo et al.,2002). Rdh10 was expressed toward the extremities of the palatal mesenchyme in regions where the palatal shelves had not yet fused (Fig. 3L). Highest expression was found, however, in the palatal epithelium near the area of fusion of the palatal shelves (Fig. 3K,L).

As a general trend, many of the facial regions or structures that expressed Rdh10 at fetal stages were weakly, or no longer labeled by E18.5 (data not shown). A noteworthy exception was seen in the developing tooth germs, where Rdh10 expression increased and became specific of the dental papilla mesenchyme (Fig. 3O,P). In molar mesenchyme, expression was concentrated toward the tips of the cusps facing the inner dental epithelium (Fig. 3P). This finding suggests that Rdh10 may be involved, together with Raldh enzymes, in producing RA that may be required in a sustained manner for tooth development (see the Discussion section). Another notable exception was the eye, in which Rdh10 expression in retinal pigment epithelium and Müller cells is known to be required throughout adult life (Wu et al.,2002,2004). Additional details about Rdh10 expression during differentiation of ocular, auditory, and brain structures will be described in a separate report (Romand et al., manuscript in preparation).

Rdh10 Expression and Organogenesis

Analysis of thoracoabdominal structures also revealed several sites of sustained Rdh10 expression. The prospective pleural mesothelium strongly expressed the gene by E12.5 (Fig. 4A). Intriguingly there was no (or comparatively much weaker) labeling of the pericardium (Fig. 4A), whereas both the pleural and pericardial layers do express the Raldh2 enzyme (Niederreither et al.,1997). Rdh10 expression persisted, albeit at lower levels, in the pleural mesothelium at E14.5, whereas it was also detected in the mesenchyme of the proximal bronchi (Fig. 4B,C). Weak expression was also seen in lateral regions of the prospective diaphragm (Fig. 4B). At fetal stages, no expression was detected within heart tissues, except for weak signal in some regions of the epicardium (Fig. 4B, and data not shown). Rdh10 was expressed in a patchy manner in the developing thymus (Fig. 4D) and spleen (Fig. 4E), two lymphoid organs that have not been reported to express Raldh enzymes. Whether this expression is in lymphoid or stromal cells awaits further investigation.

Figure 4.

Expression of Rdh10 during organ and limb differentiation. A–M: Selected sagittal (A–E,G,H,K–M) and frontal (F,I,J) section planes from embryonic day (E) 10.5 (G), E12.5 (A,K,L), E14.5 (B–E,H,I,M), and E16.5 (F,J) embryos and fetuses are shown. See main text for details. The inset in B is a detail of the epicardium of the atrial region. ad, adrenal gland; br, bronchi; di, diaphragm; ec, epicardium; gt, genital tubercle; ht, heart; in, intestine; ki, definitive kidney; li, liver; lu, lung; md, Müllerian duct; mt, mesonephric tubules; pl, pleura; re, rectum; sp, spleen; st, stomach; th, thymus; tr, trachea; ur, urethra; wd, Wolffian duct.

Rdh10 expression persisted in the mesenchymal layer of the E12.5 intestine (Fig. 4K), and by E14.5–E16.5, it became localized to two concentric layers (Fig. 4F; see also the rectum in Fig. 4I,J). Upon stomach differentiation, Rdh10 transcripts were no longer seen in the epithelium, but were present in the underlying mesenchyme (Fig. 4E). Rdh10 was specifically expressed in the segmented mesonephric tubules by E10.5 (Fig. 4G). Within the definitive kidney (metanephros), Rdh10 expression became concentrated in the outer cortical region where the sustained generation of new nephrons occurs (Fig. 4H). There was also region-specific expression in the lower genital tract. This finding was seen at the base of the genital tubercle at E12.5 (Fig. 4K). At E14.5, expression was found around the distal portions of the Wolffian and Müllerian ducts (the precursors of the male and female genital tracts, respectively; Fig. 4I). At E16.5, once sexual dimorphism has taken place, Rdh10 expression was seen around the uterine and vaginal primordia in females (data not shown), and in mesenchyme of the urethral and prostatic area in males (Fig. 4J, and data not shown). Altogether the Rdh10 expression pattern in the metanephros and prospective genital tract is reminiscent of that of Raldh2 (Niederreither et al.,1997,2002a). However, unlike Raldh2 which is expressed in the testis seminiferous tubules by E18.5 (Niederreither et al.,1997), Rdh10 transcripts were not detected within the developing testis (data not shown).

Within the developing limbs, Rdh10 expression was specifically seen in the prospective interdigital mesenchyme (Fig. 4L). This expression was highest at E14.5, while digit separation is under way (Fig. 4M) and was detected until E16.5 (data not shown).

DISCUSSION

The retinol dehydrogenase RDH10 has first been identified as an all-trans-retinol oxidizing enzyme in the pigmented epithelium and Müller cells of the retina (Wu et al.,2002,2004). The finding that an ENU-generated mutation of murine Rdh10 leads to a severe, recessive embryonic malformative phenotype has implicated this enzyme as participating in the tissue-specific production of RA during development, and indeed initial data showing a tissue-specific expression have been presented (Sandell et al.,2007). Our present expression analysis clearly shows that the Rdh10 gene is expressed in several tissues known to be targets of RA signalling, both during early morphogenesis and organ differentiation.

The Rdh10/trex mouse mutant (Sandell et al.,2007) has been shown to share several common phenotypic traits (albeit not the full spectrum) with those resulting from gene disruption of Raldh2, the first retinaldehyde dehydrogenase to act during embryogenesis (Niederreither et al.,1999). There are clear similarities between the early expression profiles of Rdh10 and Raldh2 (Niederreither et al.,1997). In headfold stage embryos, expression of both genes is mesodermal specific and is regionally restricted from the base of the headfolds to the node region. However, there is a temporal delay in Rdh10 activation, Raldh2 being already expressed at E7.0 in primitive streak stage embryos (Niederreither et al.,1997) and Rdh10 becoming detectable shortly before the onset of somitogenesis (Fig. 1A–D). Such a difference may correlate with a major phenotypic discrepancy between both mutants, only the Raldh2 mutants exhibiting an abnormal compaction of the somites and an eventual left–right desynchronization of somitogenesis (Vermot et al.,2005; Sirbu and Duester,2006). It is possible that the function of Raldh2 in generating RA required to control somite size and left–right symmetry occurs quite early, before Rdh10 is even induced. At early stages, RALDH2 may thus use retinaldehyde produced by one of the ADHs expressed in the early embryo (Ang et al.,1996; Molotkov et al.,2002). Arguments for such an early critical function of RALDH2 have been presented, including that the somitic defects can mainly be rescued when mutants are provided exogenous RA through maternal supplementation as early as E7.5 (Sirbu and Duester,2006; J. Gallego and P. Dollé, unpublished data). This explanation does not exclude, however, that RDH10 may normally participate to a burst in RA production when it is induced at presomitic stages, although its absence can be functionally compensated for by the presence of ADHs.

Clear similarities between Rdh10 and Raldh2 expression patterns persist at later stages of embryogenesis; thus at E8–E8.5, both genes are strongly expressed in the somitic mesoderm and the adjacent lateral plate mesoderm, with a rather sharp boundary toward the presomitic and caudal region (Fig. 1H). This reinforces the model (Diez del Corral et al.,2003; Vermot et al.,2005) stating that RA production is regulated in an anterior to posterior direction during embryonic axis extension, acting in particular as a paracrine signal required for differentiation and patterning of the spinal cord (for reviews, see Diez del Corral and Storey,2004; Wilson and Maden,2005). While somitic differentiation proceeds by E8.5–E9, both Rdh10 (this study) and Raldh2 (Niederreither et al.,1997) transcripts are concentrated toward the dorsal somitic mesoderm. Raldh2 loss of function mutants display molecular evidence of an abnormal differentiation of somitic derivatives, especially into myotomes (J. Gallego, V. Ribes, P. Dollé, unpublished data). It will be interesting to further investigate in the Rdh10/trex mutants, which survive in utero several days beyond Raldh2−/− mutants, whether differentiation of somitic derivatives is similarly affected.

Regarding possible roles in neural tube differentiation, Rdh10 has a peculiarity not shared with any of the known Raldh genes: it is transiently expressed from ∼E8–E9 in the ventral cells (floor plate) of the developing hindbrain and spinal cord. Functional assays in chick embryos, as well as analysis of retinoid-deficient quail embryos, have shown that RA signalling is important for proper expression of ventral-patterning genes in the developing spinal cord, in a Sonic hedgehog-independent manner (Diez del Corral et al.,2003). This function may not solely rely on RA signalling from the somitic mesoderm. Indeed, RA may additionally be produced in floor plate cells due to RDH10 activity, from which it could influence ventral spinal cord gene expression. The question remains open as to which retinaldehyde dehydrogenase(s) would then catalyze the final step to generate RA. The same question applies to the prospective mid–hindbrain boundary within the forebrain neuroepithelium, a signalling region (Echevarria et al.,2003, for a review) that specifically expresses Rdh10 at early stages, and was not previously thought to be a site of RA signalling. Interestingly, the chick CYP1B1 enzyme has recently been shown to catalyze both the retinol–retinaldehyde and retinaldehyde–RA conversion, and the corresponding gene is expressed both at the level of the mid–hindbrain boundary region and ventral neural tube in avian embryos (Chambers et al.,2007). Whether mouse Cyp1b1 is similarly expressed and, thus, could mediate the RA conversion in these regions remains to be investigated.

Another common feature of Rdh10 and Raldh2 expression is seen in the posterior branchial arch and retrocardiac region. There is a clear similar spatial restriction, both enzymes being restricted to the posteriormost segment (sinus venosus) of the heart tube (Moss et al.,1998). Furthermore, both of the loss of function mutants display similar defects of posterior pharyngeal and foregut derivatives, including the morphological absence of posterior branchial arches (Niederreither et al.,2003; Sandell et al.,2007) and agenesis or hypoplasia of the lung buds (Desai et al.,2006; Wang et al.,2006). This finding demonstrates that the posterior branchial arches and foregut region require optimal levels of RA synthesis, that need to be controlled at the level of both enzymatic reactions. It is striking, however, that the Rdh10 mutants do not exhibit another major feature of the Raldh2−/− embryos, the heart tube looping defect (Niederreither et al.,1999,2001). As discussed above and in Sandell at al. (2007), one cannot exclude that lack of certain phenotypic features in Rdh10/trex mutants are explained by some residual activity of the mutated enzyme, and/or functional compensation by ADH enzyme(s). Another explanation would be that RALDH2 already acts very early in the embryonic heart field, before Rdh10 is significantly induced in mesodermal cells (Hochgreb et al.,2003). This explanation is supported by the fact that the looping defect is corrected in Raldh2−/− mutants by short-term overnight maternal RA supplementation from E7.5–E8.5 (Niederreither et al.,2001). Additional studies of Rdh10/trex mutants would be of interest to see whether other aspects of RA function, for example, for proper development of posterior heart chambers or septation of the outflow tract, critically depend on RDH10.

Unlike Raldh2, the Raldh3 enzyme is mainly expressed in embryonic head structures, namely at the level of the optic vesicle, the otocyst and the prospective nasal epithelium (Li et al.,2000; Mic et al.,2000). Although the eye is a major site of RA synthesis, expression of Rdh10 appears relatively late at E9.5 within the developing optic vesicle, and details of its expression during eye differentiation will be described elsewhere (Romand et al., manuscript in preparation). There is ample evidence that RA is important for differentiation of the olfactory system (for a review, see Rawson and LaMantia,2006). Although Rdh10 expression appears to be delayed with respect to that of Raldh3, expression of both genes in the distal epithelium of the developing olfactory pits is particularly striking. Rdh10/trex mutants and Raldh3−/− knockout mutants both have hypoplasia of the nasal chambers (Dupe et al.,2003; Sandell et al.,2007). We show here that Rdh10 is expressed in specific areas of the differentiating olfactory epithelium, as well as in the distal mesenchyme of the nasal processes. All three RALDH enzymes are differentially expressed during differentiation of the nasal system (Niederreither et al.,2002a), and there is evidence that Raldh2 and Raldh3 functionally cooperate for proper nasal morphogenesis (Halilagic et al.,2007). Our present observations suggest that RDH10 is responsible for generating retinaldehyde in specific epithelial and mesenchymal populations, that will be used by one or several RALDH enzymes in a region-specific manner.

A similar conclusion could be made for two other processes during facial development, palate formation, and odontogenesis. Both conditions of RA excess and defective signalling lead to cleft palate in mouse (Lohnes et al.,1994; Cuervo et al.,2002; Halilagic et al.,2007, and references therein). We show that Rdh10 is expressed at E14.5 in two cell populations of the palatal shelves: the distal mesenchyme and the epithelial margin. Interestingly, this region of the epithelium undergoes programmed cell death necessary for fusion to proceed, and interference with RA signalling, including by using a retinol dehydrogenase inhibitor, reduces this apoptosis (Cuervo et al.,2002). Our data clearly suggest that RDH10 participates in the local synthesis of RA necessary for palatal shelf fusion. The involvement of endogenous RA in tooth development is less clear. All three Raldh genes are differentially expressed during tooth bud development (Niederreither et al.,2002a), and excess RA has marked effects on tooth development both in vivo and in organotypic culture (Bloch-Zupan et al.,1994; Kronmiller et al.,1995). We identified the dental papilla mesenchyme as one of the few tissues that strongly expresses Rdh10 at late gestational stages. This finding strengthens the idea that RA is one of the signals involved in tooth bud differentiation.

Finally, our study has shown several interesting aspects of Rdh10 expression with respect to limb development. Before and during forelimb budding, Rdh10 is strongly expressed in the somatopleure mesenchyme and along the proximal margin of the buds. Rdh10 expression also extends within the limb buds, where it is especially concentrated along the posterior margin, the so-called zone of polarizing activity (ZPA). Expression of the Raldh2 enzyme is partly overlapping, as it is present in the flank and limb bud proximal margin, without extending in the bud itself (Niederreither et al.,1997; Mic et al.,2004). RDH10 may thus provide an enriched source of retinaldehyde within the ZPA, which would be converted into RA by RALDH2 at the posterior/proximal interface. As Rdh10 is the first murine RA-generating enzyme shown to be expressed in the ZPA mesenchyme, it provides further support to the model postulating a posterior source of RA within the limb buds (Tickle et al.,1985; Thaller and Eichele,1987). That RDH10 and RALDH2 sequentially act to produce RA in the early limb bud is fully supported by the comparable forelimb reduction defects seen in both loss of function mutants, and that both mutants similarly fail to express the ZPA determinant Sonic hedgehog in the correct posterior domain (Niederreither et al.,1997; Mic et al.,2004; Sandell et al.,2007). Whether RA plays a similar role during hindlimb bud development is less clear, as both the Rdh10/trex and Raldh2 mutants exhibit normal hindlimbs by E13.5–E14.5 (Niederreither et al.,2002b; Sandell et al.,2007). Another shared feature of Rdh10 and Raldh2 expression is found at later stages, when both genes are coexpressed in prospective interdigital mesenchyme (this study; Niederreither et al.,2002b). The corresponding function of RA cannot be satisfactorily studied in the loss of function mutants, which die in utero before complete digit formation even when maternally RA rescued. Expression of both genes occurs in both the forelimb and hindlimb interdigital mesenchyme (e.g., Fig. 4L,M). Remarkably, the RA receptor Rarb gene is also coexpressed in interdigital areas (Niederreither et al.,1997), and Rarb;Rarg compound mutant mice display severe interdigital webbing (Dupe et al.,1999). Tissue-specific mutants targeted for the limb mesenchyme (e.g., Logan et al.,2002) will have to be generated to investigate this aspect of RDH10 and RALDH2 function in the future.

In conclusion, this study has demonstrated temporal and spatial regulation of Rdh10 expression during early embryogenesis, as well as in several differentiating organs and tissues known to be targets of retinoid signalling. These data clearly indicate that the distribution of RA in embryonic tissues needs to be controlled at the level of the two enzymatic steps leading to its conversion from retinol. These findings may have implications in human populations; heterozygous mutations of Rdh10 and one of the Raldh genes, even if asymptomatic on their own, may lead when combined (or in conditions of poor vitaminic uptake) to defects in development of the systems discussed above. The available Rdh10/trex mouse allele is lethal at an early fetal stage (Sandell et al.,2007). Of interest, the relatively discrete sites of Rdh10 expression should allow strategies to be designed for generating conditional mutants allowing to characterize the consequences of impaired RA synthesis in specific tissues and in viable animals.

EXPERIMENTAL PROCEDURES

This expression study was performed on mouse embryos and fetuses of the CD1 outbred strain. Fetuses from the C57Bl/6j strain were also analyzed at E14.5, which led to comparable data. ISH was performed with digoxigenin-labeled riboprobes synthesized from a template cDNA fragment generated and kindly provided by the EURExpress (www.eurexpress.org) consortium. Whole-mount ISH was performed on E7.5–E9.5 embryos using an Intavis In Situ Pro robot. The detailed procedure can be found at http://www.eumorphia.org/EMPReSS/servlet/EMPReSS.Frameset (gene expression section). After whole-mount analysis, some of the embryos were sectioned with a Vibratome (Leica VT1000S) and the sections were examined extemporaneously. ISH on cryosections was performed on serial sections of E10.5, 12.5, 14.5, 16.5, and 18.5 embryos and fetuses, using a Tecan GenePaint robot (for details on the procedure, see www.eurexpress.org and www.genepaint.org). All expression patterns were documented using a macroscope (Leica M420) or microscope (DM4000B), both connected to a Photometrics camera with the CoolSNAP (version1.2) software (Roger Scientific Inc).

Acknowledgements

We thank Drs. C. Thaller and G. Eichele, and the EURExpress consortium, for implementing the robotic ISH procedure and providing the Rdh10 template DNA. We thank M. Philipps, V. Alunni, and M. Battaini for help with the ISH procedure. L.C. and C.M. were supported by the EVI-GENORET and EURExpress grants, respectively.

Ancillary