Developing with lethal RA levels: genetic ablation of Rarg can restore the viability of mice lacking Cyp26a1

Retinoic acid (RA), a biologically active metabolite of vitamin A(retinol), is a signaling molecule required for normal embryogenesis. RA signaling is mediated by nuclear RA receptors (RARs — α, β,γ — and their isoforms; and RXRs — α, β, γ)(reviewed by Chambon, 1996) and depends on the coordinated expression of specific synthesizing (RALDH1, RALDH2 and RALDH3) (Mic et al., 2000;Niederreither et al., 2002b)and catabolizing (CYP26A1, CYP26B1)(Abu-Abed et al., 1998;Fujii et al., 1997;MacLean et al., 2001) enzymes,whose actions specify RA distribution patterns throughout the developing embryo. Gastrulation, a process whereby the early embryo is converted from two germ layers into three (ectoderm, mesoderm and endoderm) and the anteroposterior axis is defined (Beddington and Robertson, 1999), is vulnerable to perturbations in RA signaling caused by either genetic or nutritional means. Raldh2 andCyp26a1 knockout and expression studies indicate that the tightly regulated synthesis and distribution of RA is especially crucial during late stages of gastrulation (Abu-Abed et al.,2001; Fujii et al.,1997; Niederreither et al.,1997; Niederreither et al.,1999; Sakai et al.,2001).

Although RA is undetectable during the early stages of gastrulation(E6.5-E7.5) (Rossant et al.,1991; Ulven et al.,2000), Cyp26a1 is expressed in the primitive streak and its newly formed mesoderm (Fujii et al.,1997). The onset of Raldh2 expression at E7.5 complements that of Cyp26a1, which shifts from the posterior embryo to all three germ layers of the anterior half. At this stage, ingressing epiblast cells may encounter a pool of RA, generated by Raldh2, in newly formed mesoderm that surrounds the primitive streak and node region(Niederreither et al., 1997;Rossant et al., 1991), and predominantly contributes to somite formation(Kinder et al., 1999). A state of RA deficiency generated by loss of RALDH2 activity leads to severe developmental defects that affect late-streak-derived structures, such as the trunk somites, heart and limbs(Niederreither et al., 1999). By contrast, exposure to teratogenic doses of RA at E7.5 affects development of more anterior structures, such as the face and brain(Iulianella and Lohnes, 1997;Simeone et al., 1995). During this stage, CYP26A1 may protect anterior structures from RA exposure, as someCyp26a1-null animals exhibit exencephaly(Abu-Abed et al., 2001).

During late gastrulation (E8.5-E10), Cyp26a1 expression shifts posteriorly, becoming mainly restricted to the open neuropore, hindgut endoderm and tail bud mesoderm. Thus, the domain of Cyp26a1expression in the tail bud abuts that of Raldh2 in the trunk mesoderm, a complementarity that persists until somitogenesis is complete(∼E13). Tissues derived from the tail bud at this stage include paraxial mesoderm of the lumbosacral region, and, to a lesser extent, lateral plate mesoderm, notochord, posterior neural tube and hindgut(Kinder et al., 1999). In the absence of CYP26A1 function, assays involving a RA-responsive lacZreporter transgene reveal that tail bud tissues normally devoid of RA stain positively for RA activity (Sakai et al.,2001). Loss of CYP26A1 function is embryonic lethal, with mutant embryos dying from caudal regression associated with spina bifida, imperforate anus, agenesis of the caudal portions of the digestive and urogenital tracts,malformed lumbosacral skeletal elements, and lack of caudal tail vertebrae(Abu-Abed et al., 2001;Sakai et al., 2001). This phenotype indicates that CYP26A1 is also essential for maintaining morphogenetic events during late gastrulation, probably by protecting tail bud tissues from exposure to RA.

The caudal regression phenotype of Cyp26a1 mutants closely resembles that of wild-type embryos treated with teratogenic doses of RA at E8.5-E9.5 (Alles and Sulik,1990; Kessel,1992; Padmanabhan,1998). Such RA-treated wild-type embryos show elevated levels of apoptosis, develop ectopic neural tubes and exhibit severe downregulation ofWnt3a expression in tail bud tissues(Alles and Sulik, 1990;Iulianella et al., 1999;Shum et al., 1999;Yasuda et al., 1990). Interestingly, loss of WNT3A function leads to a caudal regression syndrome similar to that observed in Cyp26a1 mutants and RA-treated wild-type embryos (Takada et al., 1994). Finally, resistance to the teratogenic effects of RA has been observed in mice that lack Rarg, which develop caudal tissues normally and show no change in Wnt3a expression after RA treatment during late gastrulation (Lohnes et al.,1993; Iulianella et al.,1999). Based on these observations, we were interested in testing whether we could rescue Cyp26a1 mutants from endogenous RA teratogenicity by disrupting the Rarg gene.

We report that loss of Rarg rescues Cyp26a1 mutants from embryonic lethality and suppresses their caudal regression phenotype. Furthermore, ablation of Rarg restores normal gene expression patterns in the tail bud ofCyp26a1^-/-Rarg^-/- double mutants. Our results suggest that RA signaling mediated by RARγ downregulates WNT3A and FGF8 signaling activities, leading to defects in caudal mesoderm and definitive endoderm formation. Using this double mutant model, not only do we better dissect the biological role of CYP26A1 in late gastrulation and tail bud development, but we also show that it is possible to suppress the lethal effects of one null mutation by introducing another, a phenomenon as yet rarely observed in mouse.

The Cyp26a1- and Rarg-null mice used in this study have been previously described (Abu-Abed et al.,2001; Iulianella and Lohnes,1997). Cyp26a1^+/-Rarg^+/-animals were mated overnight and females examined the following morning for the presence of a vaginal plug; noon of the day of plug was considered as embryonic day (E) 0.5. Pregnant females were sacrificed by cervical dislocation and embryos or fetuses were collected in sterile phosphate-buffered saline (PBS). Yolk sacs were used for genotyping by PCR. The following three primers were used to amplify 200 and 350 bp fragments from the Cyp26a1 wild-type and mutant alleles, respectively: 5′CTGGGCTGAAGGTGGGATAAGTGTCCTC 3′, 5′CCACGTCCACTCTGCTATGCTGTTGAGCATCAGCTG 3′ and 5′GAAGATATCTGCCACATCGTGGGTGTCACAGATAC 3′. The following two primers were used to amplify a 120 bp fragment from the Rarg wild-type allele:5′ CAGAGCACCAGCTCGGAGGA 3′ and 5′ CTTCACAGGAGCTGACCCCA 3′. In a separate reaction, a primer complementary to the 5′coding region of the Neo gene, 5′ GGCCGGAGAACCTGCGTGCAATCC 3′, was used in combination with the first Rarg primer to amplify ∼700 bp fragment from the mutant allele. The PCR conditions used were as follows: 94°C for 30 seconds, 55°C (Cyp26a1) or 60°C (Rarg) for 30 seconds, and 72°C for 1 minute for 30 cycles.

Fetuses were skinned, eviscerated, and dehydrated in 100% ethanol. Carcasses were stained with 0.03% Alcian Blue in 80% ethanol:20% glacial acetic acid to detect cartilaginous elements. Specimens were then dehydrated in 100% ethanol, cleared overnight in 1% potassium hydroxide, and stained overnight in 0.1% Alizarin Red S in 1% potassium hydroxide. Clearing was achieved by several changes of 20% glycerol in 1% potassium hydroxide over 1 week, followed by 50% glycerol:50% ethanol for 2-3 weeks. Specimens were scored under a dissecting microscope and photographed with a digital camera.

Staged embryos were pooled by genotype and rehydrated in a methanol series.Brachyury, Cdx4, Fgf8, Hnf3b, Hoxd11, Tbx6 and Wnt3a probes were kindly provided by Drs B. Herrmann, J. Deschamps, G. Martin, S.-L. Ang,D. Duboule, V. Papaioannou and A. McMahon, respectively. Whole-mount in situ hybridization was performed according to Wilkinson(Wilkinson, 1992). After hybridization, embryos were cleared and photographed under a dissecting scope(Leica MZ 9.5) with a digital camera.

The relationship between Cyp26a1 and Rarg expression domains was analyzed by comparative whole-mount in situ hybridization of wild-type E9 embryos (Fig. 1A-D). Cyp26a1 exhibits strong and specific expression within the tail bud neuroepithelium and mesoderm(Fig. 1A,B) has been previously reported (Fujii et al., 1997;Abu-Abed et al., 2001;MacLean et al., 2001).Rarg transcripts were also markedly expressed in the caudal region of the embryo (Fig. 1C-E). However, these transcripts were not as sharply restricted as those ofCyp26a1 and extended up to slightly more rostral levels, both within the posterior neuroepithelium and the lateral mesoderm (compareFig. 1B with 1D). In E9.0 embryos, RARγ expression was not compromised by Cyp26a1ablation (compare Fig. 1E with 1F). At this stage, there were no obvious changes in tissue morphology in the Cyp26a1^-/- mutants when compared with wild-type littermates. Because we have previously observed that reducingRaldh2 expression can result in rescue of theCyp26a1^-/- caudal defect(Niederreither et al., 2002a),we also wanted to establish whether Raldh2 expression was normal inRarg mutants. Fig. 1G,H shows that Raldh2 expression levels are not decreased in Rarg-null embryos.

Cyp26a1^+/-Rarg^+/- double heterozygous mutant mice were intercrossed to generate compound mutants. For simplicity, Cyp26a1 and Rarg are abbreviated A1 andγ, respectively (e.g. A1^-/- andA1^-/-γ^-/- representCyp26a1^-/- andCyp26a1^-/-Rarg^-/- mutants,respectively). Embryos at E8.5-9.5, E18.5 fetuses and 10-day old newborn pups collected from A1^+/-γ^+/-intercrosses were considered for analysis of Mendelian ratios; specimens were also collected from A1^+/-γ^-/-females mated with A1^+/-γ^+/-males, in order to increase the yield of double mutant embryos and fetuses. At E8.5-9.5, A1^-/-γ^-/- embryos were collected at approximately a third of the expected ratio, while other genotypes were observed at the expected Mendelian ratios(Table 1). At E18.5, theA1^-/-γ^-/- andA1^-/- genotypes were observed at two-thirds and one-third of the expected frequencies, respectively. At 10 days post-partum,A1^-/-γ^-/- mutants were observed at approx. half the expected ratio, distinguished in some cases by a shortened and/or kinked tail (Fig. 2A-C). No A1^-/- orA1^-/-γ^+/- newborns were recovered 10 days post-partum. Altogether, these results suggest that aRarg-null mutant background rescues approximately half of theA1^-/- mutants from lethality. AsA1^-/-γ^-/- mutants obtained at embryonic and fetal stages were externally indistinguishable from their wild-type counterparts, it is likely that theA1^-/-γ^-/- embryos not rescued by the Rarg-null background were resorbed before E8.5. Previously, we have recovered Cyp26a1^-/- embryos exhibiting early lethality (Abu-Abed et al.,2001). We have also found that exposingCyp26a1^-/- embryos to subteratogenic doses of RA at E7.5 increases early lethality (S. A.-A., P. D., D. M., C. W., G. M., P. C. and M. P., unpublished), suggesting that variations in maternal RA status results in variability in the expressivity of the Cyp26a1^-/-phenotype. Lack of RARγ may make embryos more susceptible to this phenotype, possibly by potentiating inappropriate signaling through RARαor RARβ. We are currently attempting to characterize the cause of this early lethality. All generatedA1^-/-γ^-/- animals continued to gain weight at a rate similar to that of their wild-type offspring. However,A1^-/-γ^-/- females were apparently sterile, even after a 3-month exposure to fertile males. The cause of this infertility is under investigation.

Previous analysis of Cyp26a1-null mutant embryos revealed that, in addition to malformed lumbosacral vertebrae and lack of caudal tail vertebrae,A1^-/- mutants suffer from an open posterior neuropore(spina bifida), as well as agenesis of the posterior urogenital and digestive systems (Abu-Abed et al., 2001;Sakai et al., 2001). The most severely affected A1^-/- mutants lack the entire caudal body and/or show various degrees of hindlimb fusions (sirenomelia). Skeletal phenotypes for both A1^-/- and γ^-/- null mutants have been previously described(Abu-Abed et al., 2001;Iulianella and Lohnes, 1997;Lohnes et al., 1993;Sakai et al., 2001) (seeTable 2). Apart from the lumbosacral truncation and lack of tail vertebrae(Fig. 2F),A1^-/- mutants exhibit consistent abnormalities of cervical vertebrae, most of which correspond to posterior homeotic transformations(Table 2;Fig. 3B). On the other hand, a fraction of the γ^-/- mutants exhibit anterior transformations of cervical and thoracic vertebrae (Table 2, and data not shown).

We analyzed the vertebral patterns ofA1^-/-γ^-/- double mutants at E18.5 [for a detailed description of the wild-type vertebral patterns, see Kessel(Kessel, 1992)]. TheRarg^-/- background almost fully rescuedCyp26a1-null mutants from the caudal regression syndrome. In contrast to A1^-/- mutants (Fig. 2F), A1^-/-γ^-/- skeletons(Fig. 2C,E) exhibited six normal, evenly spaced lumbar, three normal sacral and 15 ossified caudal vertebrae, whereas wild-type littermates(Fig. 2B,D) had 18 ossified tail vertebrae. Furthermore, A1^-/-γ^-/-mutants had a normal pelvic bone and correctly set hindlimbs (compareFig. 2B,D with 2C,E). We also examined the effects of Rarg haploinsufficiency by analyzing two E18.5 A1^-/-γ^+/- mutants. Although one of these A1^-/-γ^+/- skeletons(Fig. 2G) showed an intermediate phenotype with lumbosacral vertebrae less deformed than inA1^-/- mutants and a truncated tail rudiment, the other showed fusion and twisting of the hindlimbs, as well as malformed lumbosacral region and no caudal tail vertebrae (data not shown). These results indicate that the rescue of A1^-/- animals by Rarghaploinsufficiency is partial and not fully penetrant, possibly accounting for the absence of viable A1^-/-γ^+/- animals in our crosses.

We also examined whether loss of RARγ function may revert theA1^-/- cervical phenotype(Table 2;Fig. 3). The first and second cervical vertebrae (C1 and C2) developed normally in allA1^-/-γ^-/- mutants examined(Fig. 3C; n=4). By contrast, one of the A1^-/-γ^+/- skeletons showed fusion between the exoccipital bone and neural arch of C1, whereas bothA1^-/- skeletons exhibited this malformation (compareFig. 3B with 3D). In addition,A1^-/- skeletons generally showed bifidus of the neural arch on C2 (Fig. 3B). This latter phenotype is also observed in some of the γ^-/- mutants(Lohnes et al., 1993;Lohnes et al., 1994). None of the A1^-/-γ^-/- orA1^-/-γ^+/- double mutants exhibited such malformations (Fig. 3C,D).

The A1^-/-γ^-/-,A1^-/-γ^+/- and A1^-/-mutants showed the following posterior homeotic transformations within the C5-T1 region: a C5 to C6 transformation, which was demonstrated by the presence of unilateral or bilateral anterior tuberculi on C5, structures normally characteristic of C6 (2/4A1^-/-γ^-/-, 1/2A1^-/-γ^+/-, 2/2A1^-/-); a C7 to T1 transformation, as evidenced by the presence of ectopic ribs on C7 that were fused to either T1 or to the sternum(4/4 A1^-/-γ^-/-, 2/2A1^-/-γ^+/-, 2/2A1^-/-); and a T1 to T2 transformation, indicated by the presence of a prominent ectopic spinous process on T1 in addition to the one normally observed on T2 (2/4 A1^-/-γ^-/-,1/2 A1^-/-γ^+/-, 2/2A1^-/-) (compare Fig. 3A-D; see Table 2for a summary). Thus, in contrast to the rescue of the C1-C2 abnormalities, it appears that ablation of Rarg does not rescue the posterior transformations observed within the C5-T1 region of A1^-/-mutants. Furthermore, the A1^-/-γ^-/- andA1^-/-γ^+/- compound mutants did not display anterior homeotic transformations normally observed in the cervicothoracic region of Rarg^-/- mutants(Iulianella and Lohnes, 1997;Lohnes et al., 1993). SomeA1^-/- mutants showed an L1 to T13 transformation due to the development of unilateral or bilateral ectopic ribs on L1 (data not shown;Table 2).A1^-/-γ^+/-, but notA1^-/-γ^-/- compound mutants, also exhibited this anterior transformation. Finally, the tracheal rings were ventrally fused and disrupted in allA1^-/-γ^-/- skeletons examined, a malformation also observed in all Rarg^-/- mutants(Lohnes et al., 1993)(Table 2). Thus, not all defects resulting from Cyp26a1 ablation are rescued by loss of RARγ function, suggesting that other RARs may compensate for ectopic RA signaling in the C5-T1 region.

To characterize the molecular defects resulting from loss of CYP26A1 function, and their possible rescue by ablation of Rarg, we analyzed genes known for their specific expression in tail bud tissues and/or whose mutation generates developmental defects in the caudal region of the embryo. In situ hybridization experiments were carried out at E9.5, which corresponds to an early stage of tail bud growth in wild-type embryos, at which mild caudal defects are seen in the A1^-/- embryos Each expression pattern was analyzed on oneA1^-/-γ^-/- and twoA1^-/-γ^+/- compound mutants, in comparison with A1^-/- and γ^-/- single mutants, as well as wild-type control embryos processed in the same experimental series. As none of the expression patterns found in γ^-/- mutant embryos differed from wild-type patterns, these mutants are not discussed further.

Wnt3a transcripts are expressed in the primitive streak, nascent mesoderm and neuroepithelium of the developing tail bud(Parr et al., 1993;Roelink and Nusse, 1991).Wnt3a-null embryos show defects caudal to the forelimbs in that they lack trunk somites, have a disrupted notochord and fail to form tail bud tissues (Takada et al., 1994). We (Abu-Abed et al., 2001) and others (Sakai et al., 2001)have previously examined the expression of Wnt3a inCyp26a1-null mutants and found its expression to be downregulated within the nascent mesoderm and neuroepithelium. By contrast, the E9.5A1^-/-γ^-/- embryo expressed Wnt3anormally (data not shown). To substantiate these findings, we chose to examine a downstream target of WNT3A signaling, the T-box gene Brachyury(Herrmann et al., 1990).

As reported (Abu-Abed et al.,2001; Sakai et al.,2001), A1^-/- embryos showed reducedBrachyury expression in the tail bud nascent mesoderm and neuroepithelium, whereas expression in the notochord indicated abnormal development of this structure (compare Fig. 4A with 4D). By contrast, Brachyury expression levels in the tail bud tissues of A1^-/-γ^-/-(Fig. 4B) andA1^-/-γ^+/-(Fig. 4C) compound mutants were comparable with those observed in wild-type littermates(Fig. 4A). Interestingly, while the A1^-/-γ^-/- mutant displayed normal morphology (Fig. 4B), theA1^-/-γ^+/- embryo was malformed caudally,showing eversion and twisting of tail bud tissues(Fig. 4C, inset). This malformation was also observed, albeit more severely, in theA1^-/- mutant (Fig. 4D), indicating that restoration of Brachyury expression may not be sufficient to achieve normal development of tail bud tissues in compound mutants. These results confirm that down-regulation of WNT3A signaling in A1^-/- mutants affects expression of a WNT3A downstream target.

Fgf8-null mutants also show severe gastrulation defects and lack several mesoderm- and endoderm-derived structures(Sun et al., 1999).Fgf8 is expressed in the primitive streak, dorsal half of the nascent mesoderm, neuroepithelium and hindgut endoderm of the wild-type tail bud(Fig. 5A)(Crossley and Martin, 1995).Fgf8 expression was virtually absent from the tail bud ofA1^-/- and A1^-/-γ^+/-embryos, except for weak expression within the nascent mesoderm(Fig. 5B; inset, and data not shown). Furthermore, abnormal tail bud morphology of theA1^-/- mutant suggested minimal tissue development caudal to the posterior neuropore (PNP) in this particular embryo(Fig. 5B, inset). By contrast,Fgf8 expression in the A1^-/-γ^-/-embryo was comparable with that of the wild-type embryo (compareFig. 5A with 5C).

We examined the expression of an additional hindgut endoderm marker,Hnf3b, which is also expressed in the notochord and floor plate of the neural tube (Fig. 5D)(Ang et al., 1993;Monaghan et al., 1993;Sasaki and Hogan, 1993).Hnf3b expression persisted in the notochord and floor plate ofA1^-/- and A1^-/-γ^+/-embryos (Fig. 5E; data not shown), although labeling indicated that both structures developed abnormally(Fig. 5E, inset; data not shown). In tail bud tissues caudal to the PNP, Hnf3b labeling also indicated a reduction in the size of the hindgut pocket, whose development appeared to stop prematurely in both A1^-/- andA1^-/-γ^+/- mutants(Fig. 5E; data not shown). Abnormal patterns of expression were not detected in theA1^-/-γ^-/- embryo(Fig. 5F; data not shown).

Loss of FGF signaling has been shown to disrupt expression ofcaudal-related genes in Xenopus(Northrop and Kimelman, 1994). As Fgf8 was abnormally expressed in A1^-/- andA1^-/-γ^+/- embryos, we examined the expression of Cdx4, one of three known caudal-related genes in mouse (Gamer and Wright,1993). Cdx4 shows a rostrocaudal gradient of expression within tail bud tissues of E9.5 wild-type embryos, with its expression being most intense caudally (Fig. 6A). In A1^-/- mutant embryos, Cdx4expression was downregulated in both the neuroepithelium and nascent mesoderm of the tail bud, having an expression boundary that altogether shifted posteriorly (compare Fig. 6A with 6B, insets); A1^-/-γ^+/-mutants showed a less pronounced caudal shift (data not shown). Cdx4expression in the A1^-/-γ^-/- embryo was similar to that of wild-type littermates (compareFig. 6A with 6C).

Pownall et al. (Pownall et al.,1996) have demonstrated that caudal-related transcription factors regulate the expression of Hox genes in Xenopus; similar observations have been made in the case of Cdx1-null mice(Subramanian et al., 1995). We chose to analyze the expression of Hoxd11, which together with other 5′-located Hox genes, is involved in setting up regional identities along the lumbosacral column and posterior urogenital and digestive structures(Favier and Dollé,1997). Hoxd11 is strongly expressed at E9.5 throughout most of the tail bud tissues, its expression being stronger towards the caudal extremity (Fig. 6D).Hoxd11 expression was downregulated in the tail bud tissues ofA1^-/- embryos, especially within the hindgut, presomitic and lateral mesoderm, and to a lesser extent in the neuroepithelium(Fig. 6E). Moreover,A1^-/- mutants showed comparably reduced expression domains of Cdx4 and Hoxd11 expression (compareFig. 6B with 6E, insets),indicating that CDX4 may indeed regulate Hox gene expression. By contrast,Hoxd11 expression levels inA1^-/-γ^-/- andA1^-/-γ^+/- compound mutants were comparable with those seen in wild-type embryos(Fig. 6F; and data not shown),although expression was expanded along the ventral mesoderm (compareFig. 6D wit 6F, insets; see Discussion).

CYP26A1 is a highly conserved component of the RA signaling pathway,showing tail bud-specific expression in several vertebrate species, including mouse, chicken, Xenopus and zebrafish(de Roos et al., 1999;Fujii et al., 1997;Hollemann et al., 1998;Iulianella et al., 1999;MacLean et al., 2001;Swindell et al., 1999).Cyp26a1-null mutant mouse fetuses die from severe caudal defects that mimic RA teratogenicity (Alles and Sulik,1990; Durston et al.,1997; Griffith and Wiley,1991; Ruiz i Altaba and Jessell, 1991; Yasuda et al.,1990), suggesting that CYP26A1 protects the developing tail bud from ectopic RA signaling (Fig. 7A) (Abu-Abed et al.,2001; Sakai et al.,2001). Disruption of CYP26A1 activity allows RA generated in the differentiating trunk somites to diffuse into the tail bud, which ultimately disrupts mesoderm and definitive endoderm formation(Niederreither et al., 2002a;Sakai et al., 2001). We show that caudal development is rescued inCyp26a1^-/-Rarg^-/- double-mutant animals,substantiated by the normal expression of molecular markers involved in tail bud morphogenesis. This provides evidence that the catabolic activity of CYP26A1 suppresses the teratogenic effects of ectopic RA signaling. Specifically, we have also demonstrated that RARγ mediates the deleterious effects of RA on caudal development by downregulating the critical signaling activities of WNT3A and FGF8(Sun et al., 1999;Takada et al., 1994). Therefore, our results reveal the existence of an epistatic relationship between the Cyp26a1 and Rarg gene functions.Fig. 7B summarizes a genetic model based on our present work and that of others.

Although Cyp26a1 and Rarg tail bud-expression patterns overlap in a spatiotemporal manner (Fujii et al., 1997; MacLean et al.,2001; Ruberte et al.,1990) (see Fig. 1),CYP26A1 is critical for normal caudal development whereas RARγ is dispensable (Abu-Abed et al.,2001; Lohnes et al.,1993; Look et al.,1995; Sakai et al.,2001). Rarg-null animals are viable, exhibiting vertebral transformations and malformations in the cervical region, in addition to squamous metaplasia and/or keratinization of various glandular epithelia. Moreover, Rarg^-/- embryos resist the effects of RA toxicity on caudal development, suggesting that RARγ mediates ectopic RA signaling in the tail bud (Lohnes et al.,1993). However, although caudal development proceeds normally inCyp26a1^-/-Rarg^-/- animals, ablation ofRarg does not rescue all defects characterized inCyp26a1^-/- mutants. In particular, the double mutants retain several vertebral posterior transformations in the cervical region(Fig. 2;Table 2), thus implicating other RARs in mediating the effects of ectopic RA after loss of CYP26A1 function.

Removal of RARγ function rescues Wnt3a and Fgf8expression, as well as normal tail bud development, inCyp26a1^-/- mutants. Both Fgf8 and Wnt3agenes exhibit markedly reduced expression domains in the caudal region ofCyp26a1^-/- embryos(Fig. 4)(Abu-Abed et al., 2001;Sakai et al., 2001). These abnormal transcript patterns could result from a combination of two events:(1) an impaired development of the tail bud tissues that normally express these genes; and (2) a downregulation of their expression levels (seeFig. 5B for Fgf8) [see Sakai et al. (Sakai et al.,2001) for Wnt3a]. As a result, impaired FGF8 and WNT3A signaling is likely to affect the tail bud gastrulation movements inCyp26a1^-/- embryos. In Wnt3a^-/- andFgf8^-/- mutants, epiblast cells appear to undergo the gastrulation epithelial-to-mesenchymal transition, but fail to migrate laterally or to contribute to mesoderm and endoderm formation(Sun et al., 1999;Takada et al., 1994;Yamaguchi et al., 1999). Furthermore, similar to what we observed in Cyp26a1^-/-embryos, Wnt3a^-/- and Fgf8^-/- mutants show decreased Brachyury (Fig. 4) and Tbx6 (data not shown) expression(Sun et al., 1999;Yamaguchi et al., 1999). TheFgf8^-/-, Brachyury^-/- andTbx6^-/- abnormal phenotypes are more severe than those observed in Wnt3a^-/- and Cyp26a1^-/-mutant embryos, with severity of the null mutant phenotypes appearing to correspond to onset of their respective gene expression in the tail bud(Chapman and Papaioannou,1998; Yamaguchi et al.,1999; Yoshikawa et al.,1997). As tail bud defects in CYP26A1 mutants are mediated by RARγ, which only first appears in the caudal embryonic germ layers at E8.0, they should accordingly be less severe than those observed inWnt3a^-/- embryos. Cyp26a1^-/- fetuses indeed develop thoracic vertebrae, as well as some lumbosacral vertebral remnants, whereas Wnt3a^-/- animals show loss of all vertebrae and tissues caudal to the forelimbs. Thus, our results show that, in the absence of CYP26A1, RARγ downregulates WNT3A and FGF8 signaling(Fig. 7B).

WNT3A and FGF8 are thought to affect the expression of other common targets, such as caudal-related genes, which regulate Hox expression inDrosophila, Xenopus and mouse(Ikeya and Takada, 2001;Isaacs et al., 1998;Lickert et al., 2000;Northrop and Kimelman, 1994;Prinos et al., 2001). Cdx4 expression is strongest in the tail bud at a time when 5′ Hox genes,which are believed to define identity of more caudal body segments, are first expressed (Burke et al., 1995;Gamer and Wright, 1993). Therefore, we investigated Cdx4 and Hoxd11 expression. Our results demonstrate that loss of WNT3A and FGF8 signaling inCyp26a1^-/- mutants coincides with downregulation of bothCdx4 and Hoxd11. Interestingly, Hoxd11 expression expanded ventrally in mutants lacking Rarg. A similar phenomenon was observed in animals harboring a mutation in a 3′ repressor region of theHoxd11 gene (Gérard et al., 1996). Together with our results, these observations suggest that in the absence of RA, RARγ may repress Hoxd11 gene activity in the ventral mesoderm by binding to its 3′ repressor element,thereby defining its rostral boundary of expression. As almost normalCdx4 and Hoxd11 expression is restored inCyp26a1^-/-Rarg^-/- double mutants,ectopic RA signaling through RARγ in the tail bud appears to affect directly and/or indirectly the expression of additional downstream targets of WNT3A and FGF8 signaling, such as the Cdx and 5′ Hox genes(Fig. 7).

Ablation of Rarg restored normal gene expression patterns and essentially alleviated caudal regression inCyp26a1^-/-Rarg^-/- double mutants. What then, besides the possible setting of the rostral boundary of expression ofHoxd11, is the physiological function of RARγ in the caudal embryo, and why are Cyp26a1 and Rarg co-expressed in the tail bud? Loss of RARγ function severely compromises male fertility, by disrupting normal differentiation of glandular epithelia in the seminal vesicles and prostate glands (Lohnes et al., 1993; Look et al.,1995). We note thatCyp26a1^-/-Rarg^-/- double mutant males and females also fail to reproduce normally (data not shown). Thus, asCyp26a1^-/- males and females in which one Raldh2allele has been ablated are fertile(Niederreither et al., 2002a),RARγ-mediated RA-signaling is likely to perform an essential role at some stage in the ontogeny of reproductive organs.

RALDH2 and CYP26A1, respectively, show complementary RA-synthesizing and-catabolizing activities throughout embryonic development in several tissues undergoing morphogenesis (Abu-Abed et al.,2002; Fujii et al.,1997; MacLean et al.,2001; Niederreither et al.,1997), enzymatic activities which are conserved in the trunk and tail bud of both mouse and chick embryos(Berggren et al., 1999;Swindell et al., 1999). Our present data indicate that by catabolizing RA, which diffuses from neighboring trunk somites, CYP26A1 activity prevents ectopic RARγ-mediated RA signaling in the tail bud. Interestingly, using a lacZ-RA-reporter transgene it has been shown that under Raldh2-haploinsufficient conditions the extent of RA diffusion into caudal tissues is reduced, which in turn rescues Cyp26a1-null mutant mouse fetuses from lethality(Niederreither et al., 2002a). These observations suggest that below a certain threshold, RA synthesis in the trunk does not interfere with tail bud development(Perlmann, 2002), even in the presence of RARγ. In this respect we note that, although RA treatments downregulate Raldh2 expression in the trunk(Niederreither et al., 1997),Cyp26a1 gene expression is induced by RA in the tail bud and craniofacial regions (Iulianella et al.,1999), indicating that feedback responses may have evolved to regulate local RA concentrations. Moreover, we have previously shown that RA-induced Cyp26a1 expression is mediated by RARγ(Abu-Abed et al., 1998), and,therefore, RARγ may be involved in regulating Cyp26a1expression in the tail bud when endogenous RA levels exceed a certain threshold. Thus, it is only in the absence of CYP26A1 or following administration of excess RA, that RARγ-mediated RA signaling becomes lethal by suppressing the expression of genes that control tail bud morphogenesis.

Previously, several mechanisms of partial genetic suppression of lethal null mutations have been reported in the mouse(Kodera et al., 2002;Litingtung and Chiang, 2000;Niederreither et al., 2002a;Tsai et al., 1998). Our present double mutants, in which a Rarg-null background rescues caudal regression and neural tube defects in Cyp26a1^-/-mutants, provides a clear illustration of genetic suppression of a lethal mutation, as the ablation of Rarg^-/- allowsCyp26a1^-/- mutants to survive lethal RA levels. Finally,it is interesting to note that in their characterization of mouse models with neural tube defects, Juriloff and Harris(Juriloff and Harris, 2000)identified several genes whose mutational phenotypes are affected by nutritional factors. Supplementation of curly tail mice with inositol or of splotch (Pax3) mutants with folic acid reduced the incidence of spina bifida in both types of animals. Our previous(Niederreither et al., 2002a)and present doublemutant mouse models similarly demonstrate that alterations in RA availability have marked effects on the incidence of spina bifida, and defects of caudal morphogenesis and neural tube formation. Whether similar human congenital defects could result from dysregulation of mechanisms that control the levels of RA remains to be determined.

We thank Drs K. Niederreither, D. Lohnes and A. Tahayato, and Mr K. Young for useful discussions. This work was supported by the Canadian Institutes of Health Research (CIHR) and the National Cancer Institute of Canada (M. P.), as well as institutional funds by CNRS, INSERM, Collège de France,Hôpitaux Universitaires de Strasbourg, Association pour la Recherche sur le Cancer, Fondation pour la Recherche Médicale and Bristol-Myers Squibb (P. D., D. M. and P. C.). S. A.-A. holds a studentship from CIHR.