Retinoic acid regulates a subset of Cdx1 function in vivo

The Hox gene products are key players in anteroposterior (AP) patterning of the vertebrate axis (reviewed by Krumlauf,1994; Gellon and Mcginnis,1998; Kessel and Gruss,1991; Burke et al.,1995). The 39 murine Hox genes encode transcription factors possessing a conserved 60 amino acid DNA-binding motif, the homeodomain(Gehring, 1993; Phelan et al., 1995). Mammalian Hox genes are distributed in four clusters, Hoxa-Hoxd, on separate chromosomes and have been proposed to have evolved by duplication of an ancestral complex related to the HOM-C genes of Drosophila(Duboule and Dollé,1989; Duboule,1998; Ferrier and Holland,2001). In the mouse, Hox gene expression is initiated at embryonic day (E)7.5 in the primitive streak, with expression subsequently expanding anteriorly in the neural tube and mesoderm until a predetermined rostral limit is reached (Deschamps and Wijgerde,1993; Oosterveen et al.,2003; Roelen et al.,2002). Both the onset and the rostral limit of expression is generally related to the physical location of a given Hox gene within its complex; more 3′ Hox members are expressed earlier and reach a more rostral limit of expression relative to more 5′ paralogs. This pattern of expression, referred to as co-linearity, results in nested domains of Hox gene expression along the AP axis which is believed to comprise a `Hox code'(Kessel and Gruss, 1991; Burke et al., 1995; Gaunt, 1994).

A wealth of information underscores crucial roles for Hox proteins in AP patterning of derivatives of all germ layers, including the vertebrae. Somites, which arise from segmentation of paraxial mesoderm, differentiate into dermamyotome and sclerotome, the latter of which is the anlage of the vertebrae. Altered Hox expression has well-documented effects on vertebral patterning, typically manifesting as homeotic transformations along the AP axis (Christ and Ordahl, 1995; Chen et al., 1998; Condie and Capecchi, 1993; Favier et al., 1996; Horan et al., 1995). Grafting experiments in the chick embryo indicate that this patterning is imparted before overt segmentation of the paraxial mesoderm into somites, probably during or shortly after gastrulation. In addition, Hox gene expression characteristic of the initial axial position is retained in such transplants,consistent with a crucial role for the Hox code in establishing AP vertebral identity during this window (Kieny et al.,1972; Nowicki and Burke,2000).

Considerable effort has been applied to the elucidation of the mechanisms by which Hox genes are regulated (Di Rocco et al., 2001; Gould et al.,1997; Marshall et al.,1996; Barna et al.,2000). Among such effectors, the vitamin A derivative retinoic acid (RA) plays a crucial role (Marshall et al., 1996). RA can induce Hox genes in embryocarcinoma cells in a manner reminiscent of the normal temporal activation of Hox expression, with 3′ genes from a given cluster responding earlier, and to lower concentrations of RA, than more 5′ members(Boncinelli et al., 1991; Simeone et al., 1990). In vivo, administration of exogenous RA to mouse embryos between E7.5 and E8.5 typically results in anteriorization of a number of Hox genes in a manner that correlates with posterior vertebral homeotic transformations(Conlon and Rossant, 1992; Kessel, 1992; Kessel and Gruss, 1991). Similar effects are also elicited by RA on expression of Hox genes in the CNS and concomitant perturbation of rhombomere patterning(Gavalas and Krumlauf, 2000; Gould et al., 1998; Marshall et al., 1992).

The RA signal is transduced by the RA receptors (RARα, RARβ,RARγ and their isoforms). RARs belong to the family of ligand-inducible nuclear receptors and regulate expression of retinoid target genes as heterodimers with a retinoid X receptor (RXRα, RXRβ, RXRγ)partner. RXR-RAR heterodimers function by binding to cis-acting regulatory sequences (RAREs) present in the promoter region of target genes(Chambon, 1996; Mangelsdorf et al., 1995). Consensus RAREs have been described which consist of direct repeats (DR) of the sequence PuG(G/T)TCA with two or five nucleotides intervening the repeats(a DR2 or a DR5 element, respectively). RAREs are, however, highly polymorphic, and a number of variant motifs have been described(Huang et al., 2002; Glass, 1996). Retinoid signaling is also tightly controlled by the opposing actions of RALDH2, which is essential for the generation of most embryonic RA, and CYP26 members, which catabolizes RA (Swindell et al.,1999; Sakai et al.,2001; Abu-Abed et al.,1998; MacLean et al.,2001; Perlmann,2002).

A role for endogenous retinoid signaling in affecting Hox expression and AP patterning is supported by numerous studies(Gavalas et al., 1998; Huang et al., 1998; Zhang et al., 1997). Vertebral homeosis and hindbrain patterning defects, including altered Hox expression,are observed in a number of RAR-null backgrounds, particularly RAR double null mutant (Lohnes et al., 1995; Lohnes et al., 1993; Dupé et al., 1999; Wendling et al., 2001). Patterning defects of a similar nature are also seen in both Raldh2mutant embryos, which are essentially devoid of RA(Niederreither et al., 2000; Grandel et al., 2002) and CYP26A1 mutants, which have expanded fields of retinoid activity(Abu-Abed et al., 2001; Maden, 1999).

The above data demonstrate a clear relationship between retinoid signalling, Hox expression and AP patterning of both neurectoderm and paraxial mesoderm. RA response elements (RAREs) have been described for a number of Hox genes, demonstrating that they are direct retinoid targets(Dupé et al., 1997; Marshall et al., 1994; Zhang et al., 1997; Frasch et al., 1995; Zhang et al., 2000; Oosterveen et al., 2003). However, gene targeting studies suggest that these RAREs are crucial for directing Hox function in the hindbrain(Marshall et al., 1996; Gavalas et al., 1998; Nolte et al., 2003). Although certain RAREs, such as the Hoxd4 RARE, have been shown to affect expression of a transgenic reporter in the mesoderm(Zhang et al., 1997) the functional significance of this element has not been definitively established,and an RARE crucial for directing Hox function in paraxial mesoderm has not been formally described. These data suggest that RA may regulate Hox expression in paraxial mesoderm either directly, through means such as the aforementioned RAREs, or indirectly; Cdx gene products (Cdx1, 2 and 4) (Gamer and Wright, 1993; Meyer and Gruss, 1993; Beck et al., 1995) are logical candidates for such an intermediary.

Cdx genes encode homeodomain transcription factors, and have been implicated as direct regulators of Hox expression(Subramanian et al., 1995; Charité et al., 1998; Isaacs et al., 1998; van den Akker et al., 2002). Of particular relevance, Cdx1^-/- offspring display homeotic transformations of the axial skeleton reminiscent of defects seen in RARα/γ-null mutants(Subramanian et al., 1995; Allan et al., 2001; Lohnes et al., 1993; Lohnes et al., 1994). We have found that Cdx1 is responsive to excess RA and RAR ablation in vivo,and have documented a functional RARE that regulates expression in tissue culture (Houle et al., 2000). Taken together, these data strongly suggest that Cdx1 is a direct RA target gene and may relay the retinoid signal to contribute to coordinated expression of Hox genes in the paraxial mesoderm.

To establish a more precise relationship between RA, Cdx1, Hox expression and vertebral patterning, we derived mice harbouring a functionally inactive Cdx1 RARE. These RARE mutants exhibit normal onset of Cdx1expression at late gastrulation, although transcript levels were consistently reduced compared with wild type, while expression at later stages was severely compromised. The RARE mutants also present with vertebral defects and altered Hox expression patterns that correlate with a subset of the Cdx1-null phenotype. Although these data underscore a crucial role for retinoid signalling in the regulation of Cdx1 expression, we also found that Cdx1 responded to exogenous RA in the RARE mutant background. Thus,as for several other known RA target genes, Cdx1 may be regulated by several RA-dependent mechanisms.

Cdx1 sequences were isolated from a murine lambda phage genomic library. A 6 kb HindIII fragment encompassing the proximal Cdx1 promoter, the entire first exon and part of the first intron,was subcloned into the KS+ plasmid. The RARE was subsequently disrupted by mutagenesis, and replaced with a unique NruI site, using the Transformer Kit (Clontech). A bifunctional floxed thymidine kinase/neomycin resistance (PGK-TK-Neo) cassette(Iulianella and Lohnes, 2002)was subsequently ligated into the NruI site, generating the targeting vector. R1 embryonic stem cells were cultured on feeder layers under standard conditions (Iulianella and Lohnes,2002). Cells were electorporated with 30 μg of linearized targeting vector and selected with G418 (180 μg/ml) for 10 days. Clones were isolated and homologous recombination assessed by genomic Southern blot from replicate clones using SacI digestion and hybridization with a 1 kb SacI/HindIII fragment immediately 5′ to the targeting sequences depicted in Fig. 1. Positive clones were further assessed for fidelity of recombination by Southern blot analysis following restriction with several different enzymes and hybridization with internal and external probes. Two clones were used to successfully generate germline chimeras by blastocyst injection by standard methods (Hogan et al., 1994)

To determine if the disrupted RARE was functionally inactivated, DNA from the mutant allele (following Cre-recombination as described below) was amplified by PCR and subcloned into a Cdx1-reporter construct described previously (Houle et al.,2000). F9 embryocarcinoma cell transfection and luciferase assays were performed as previously described(Houle et al., 2000).

F1 males from chimera-C57BL/6 crosses bearing the targeted allele were bred with homozygous female CMV-Cre mice(Dupé et al., 1997) and offspring assessed for excision of the floxed selection cassette by genomic Southern blot analysis following EcoRI digestion. All subsequent genotyping was performed by PCR using the primers 5′-GGTACACAATGCAACTCGGTG and 5′-CCTCACACCCGCCACAG which flank the RARE. The wild-type and mutant RARE allele can be distinguished by virtue of the increased size of the PCR product generated from the mutant allele following electorphoresis through a 2% agarose gel. The specificity of PCR analysis was further confirmed by Southern blot analysis of amplification products using oligonucleotide probes specific for wild type (5′GGTCACGACCCTTCGGGTCC) or mutant (5′ CGAAGTTATCCCTGCTTATCG) products.

Lines derived from each ES clone were separately maintained in a 129Sv-C57BL/6 hybrid background. Skeletal defects, assessed as described previously (Allan et al.,2001), were identical with respect to expressivity and penetrance in both lines, and subsequent studies were conducted using only one mutant line. RARE homozygous mutants were crossed with both Cdx1-null mutants and with RARγ heterozygotes. In the former case, double heterozygous offspring were assessed for skeletal defects, whereas in the latter situation, Rare^+/-Rarg^+/-offspring were intercrossed, fetuses collected at term, genotyped and assessed for skeletal anomalies.

Embryos were harvested at E7.5-9.5, with noon of the day of detection of a vaginal plug considered to be E0.5. Embryos were fixed overnight in 4% PFA at 4°C and processed for in situ hybridization as previously described(Allan et al., 2001). Embryos to be compared were processed in parallel to control for variation in signal intensity, and stage matched according to established criteria. Probes for in situ hybridization were generated from previously described plasmids; Hoxa3 (Manley and Capecchi,1995), Hoxb3 (Manley and Capecchi, 1998), Hoxd3(Condie and Capecchi, 1993), Hoxa4 (Wolgemuth et al.,1987), Hoxb4 (Folberg et al., 1999), Hoxd4(Featherstone et al., 1988)and Cdx1 (Houle et al.,2000). Embryo culture, including cycloheximide and RA treatment,was carried out as described (Houle et al., 2000) using 15% FBS in DMEM equilibrated under N2 containing 5% O₂ and 5% CO₂.

The Cdx1 promoter contains an atypical RARE that resembles a palindromic thyroid hormone response element(Houle et al., 2000). In order to functionally inactivate this element, a floxed neomycin selection cassette was embedded in a mutated Cdx1 RARE, yielding a targeting construct comprised of 6 kb of homologous genomic sequences. RI ES cells were electroporated with a linearized targeting construct, and replicate colonies screened for homologous recombination by Southern blot analysis as described in the Materials and methods (Fig. 1 and data not shown). Nine recombinants were identified, out of 300 clones screened, that exhibited the anticipated restriction pattern predicted from homologous recombination(Fig. 1B and data not shown). Chimeras from two independent clones gave germline transmission of the targeted allele. These offspring were subsequently crossed with homozygous CMV-Cre transgenic mice to affect excision of the selectable marker. PCR analysis (Fig. 1C;Materials and methods) showed the predicted RARE mutant allele was represented in 50% of the resultant offspring, and all of these heterozygotes passed the mutant allele to their offspring at the predicted Mendelian ratio, consistent with highly efficient excision of floxed sequences.

Transfection assays were used to verify that the targeted mutation abrogated RA response. To this end, we used PCR to amplify Cre-recombined sequences spanning the mutated RARE and substituted these sequences in a reporter vector comprising 2 kb of Cdx1 genomic DNA directing expression of a luciferase gene. This reporter (Lox mut in Fig. 2) was compared with the wild-type construct and with another previously published RARE mutation (Mut in Fig. 2) which is not capable of RXR/RAR binding or RA response (Houle et al., 2000). In F9 embryocarcinoma cells, the wild-type reporter was induced by 10^-6 M RA, whereas both the Lox mut and Mut reporters were unresponsive (Fig. 2). Thus, the targeted mutation appears to effectively attenuate retinoid regulation through this RARE.

Whole-mount in situ hybridization analysis was used to compare Cdx1 expression between wild-type and RARE mutant embryos. Cdx1 was undetectable until late primitive streak stages, when expression was initiated in both wild-type and mutant embryos. However, at this stage, transcript abundance was consistently reduced in mutant embryos relative to wild-type controls (compare Fig. 3A with 3B). At early somite stages, loss of expression was more pronounced in the RARE mutants compared with controls (compare Fig. 3D with 3C) with expression subsequently nearly extinguished in mutants at later stages (compare Fig. 3F with 3E). In addition, expression of Cdx1 in neurectoderm at E8.0 was also reduced in the RARE mutants, although the normal rostral boundary of expression was not affected (data not shown). These observations suggest that the RARE is critically required to establish normal levels of Cdx1 expression from early stages onwards, and affects expression in both mesoderm and neurectoderm.

The targeted RARE mutation abrogates RA response of a Cdx1reporter in transfection assays in embryocarcinoma cells(Fig. 2). In order to determine if RARE-null embryos were likewise unresponsive to exogenous RA in vivo, we treated pregnant females from E7.0 to E8.5 with vehicle or RA and harvested embryos 4 hours post-gavage. In wild-type embryos treated on E7.0, Cdx1 expression was induced marginally in earlier embryos, with more pronounced transcript abundance in late streak embryos(Fig. 4A,B). Notably, RARE mutants at this stage exhibited a similar profile of responsiveness(Fig. 4C,D), although expression levels were below that of controls in both untreated and RA-exposed mutants. A similar outcome was seen following treatment at E7.5 to E8.5, with both wild type (Fig. 4E,F,I,J,M,N) and RARE mutants(Fig. 4G,H,K,L,O,P) exhibiting induction at all stages assessed.

To determine if the effects of exogenous RA on induction of Cdx1in vivo was a secondary event, we used embryo culture to assess the effects of cycloheximide treatment on retinoid response. RA alone resulted in a pronounced induction of Cdx1 in cultured RARE-null embryos, with a magnitude of response comparable with that seen in vivo(Fig. 5B compared with 5A),although cycloheximide treatment alone had no significant effect (compare Fig. 5C with 5A; the spurious staining seen in one cycloheximide-treated embryo is of unknown basis, and was not reproducible) (Houle et al.,2000). Cycloheximide also did not prevent induction of Cdx1 by RA in the RARE mutant background (compare Fig. 5D with 5A), consistent with an additional, direct, means for RA regulation of Cdx1.

Cdx1-null mutant offspring exhibit vertebral homeosis and altered Hox gene expression. To determine if loss of the RARE affected any of these Cdx1-dependent functions, we compared skeletal patterning between wild-type and mutant backgrounds. The murine axial skeleton is typically comprised of occipital bones, derived from condensation of the four rostralmost somites,and a vertebral column consisting of seven cervical (C1-C7), 13 thoracic(T1-T13), six lumbar (L1-L6), three or four sacral (S1-S4), and over 30 caudal vertebrae. Many vertebrae exhibit distinct morphological features characteristic of their position along the AP axis. The C1, or Atlas, exhibits broad neural arches and a medial ventral process, the anterior arch of the atlas (AAA). The second vertebra, C2, exhibits neural arches of intermediate size, and a second vertebral body, the dens axis that articulates with C1. C6 possesses two ventral anterior tubercles (TA), while the thoracic vertebrae are distinguished by ribs, the first eight of which fuse with the sternum. T2 is also distinguished by the presence of a dorsal spinous process.

RARE heterozygous offspring exhibited a low incidence of defects affecting the first (C1) and second (C2) cervical vertebrae (compare Fig. 6B with control in 6A; Table 1). These consisted of a small posterior extension of the basioccipital bone, slightly narrower C1 neural arches and/or a short cartilaginous AAA on C2; the latter two defects are suggestive of a partial transformation of C2 to a C1 identity. By contrast, RARE homozygotes presented highly penetrant skeletal defects, with∼90% of all specimens examined exhibiting vertebral homeosis and/or other malformations. Defects typically consisted of partial C2 to C1 transformation,as evidenced by the presence of an ectopic AAA and/or broader neural arches on C2 (Fig. 6C; Table 1). Fusions between the AAA and the basioccipital bone, and/or narrow C1 neural arches were also observed, although at a lower frequency(Table 1). Interestingly, the RARE mutants were rarely affected posterior to the C2(Table 1), although Cdx1-null mutants are affected throughout the cervical and anterior thoracic skeleton (Subramanian et al.,1995).

RARE-null mutants exhibit a reduction in Cdx1 expression and a subset of the vertebral malformations seen Cdx1-null offspring. To further investigate these relationships, we assessed the phenotype of Rare^+/-Cdx1^+/- offspring relative to single RARE or Cdx1 heterozygous and homozygous mutants(Fig. 6).

The incidence of vertebral defects exhibited variable sensitivity to RARE and Cdx1 dosage along the axis. Malformations of C1, including reduced NA, enlargement of the basioccipital, or its fusion with the anterior arch of the atlas, were completely penetrant in the Cdx1-null background, as described previously (van den Akker et al., 2002; Allan et al., 2001; Subramanian et al., 1995). These defects were observed at a much lower frequency in Cdx1^+/-, Rare^-/- and Rare^+/-Cdx1^+/- backgrounds(Table 1). The partial C2 to C1 homeotic transformation was also completely penetrant in Cdx1-null mutants and, in contrast to C1 malformations, was observed at a high incidence in Rare^-/- and Rare^+/-Cdx1^+/- offspring.

In contrast to more anterior elements, C3 was never affected in the Rare^-/- or Cdx1^+/- offspring, but exhibited C2 characteristics in all Cdx1^-/- samples. Approximately one third of Rare^+/-Cdx1^+/- offspring also exhibited this transformation. Similarly, transformation of C7 to C6,evidenced by ectopic anterior tuberculi, was absent in Rare^-/- and Cdx1^+/- offspring, but was completely penetrant in Cdx1^-/- mice and observed at an intermediate frequency in Rare^+/-Cdx1^+/- samples. These latter two transformations argue strongly that the RARE mutation is allelic with Cdx1. Additional defects, including fusions between cervical neural arches, were also prevalent in Rare^+/-Cdx1^+/- mutants, and may represent incomplete vertebral homeosis, as previously discussed(Allan et al., 2001).

It is interesting to note that the incidence of defects in both the RARE-null and RARE-Cdx compound mutant backgrounds exhibited a variable frequency along the AP axis: C1 and basioccipital appear to require minimal Cdx1 function, C2 being most sensitive, and more posterior cervical vertebrae exhibit an intermediate level of sensitivity. This suggests a restricted window of function for the RARE, and by extension RA, in affecting Cdx1 expression and function.

To further investigate the relationship between Rare-dependent Cdx function, RAR signalling and vertebral patterning, we assessed the skeletal phenotype of an allelic series of RARE-RARγ compound mutants. Rarg^+/- offspring are essentially normal, whereas RARγ-null mutants exhibit a low frequency of vertebral defects, the most prevalent being a partial C2 to C1 transformation, which is similar to the predominant malformation exhibited by Rare^-/- mutants(compare Fig. 7B with 7A and Fig. 6C; Table 1). Rare^+/-Rarg^+/- compound mutants did not exhibit an increased incidence of vertebral malformations. By contrast, Rare^-/-Rarg^-/- double null offspring showed a marked increase in both the penetrance of defects characteristic of either mutant, as well as additional malformations not observed in either background (Fig. 7 and Table 1). Malformation of C1,including reduction of the neural arches, fusion of the AAA with the basioccipital bone or fusion of the neural arches of C1 and C2, were all observed at a higher incidence in the double mutant background in a manner suggesting a synergistic interaction between these alleles. The C2 to C1 transformation, which was incompletely penetrant in both Rare^-/- and Rarg^-/- backgrounds, was observed in all the double null mutants. The C3 to C2 homeosis, which was absent in RARE and RARγ-null backgrounds, was observed in Rarg^-/-Rare^-/- offspring, albeit with incomplete penetrance (Table 1;data not shown). Consistent with these data, Rarg^-/-Rare^+/- and Rarg^+/-Rare^-/- offspring presented a range of malformations and degree of penetrance that were intermediate between the vertebral phenotype of Rarg^-/-Rare^-/- and Rarg^+/-Rare^+/- compound mutants(Fig. 7D,E; Table 1). Taken together with our prior work (Allan et al.,2001), these finding suggest that retinoid-dependent vertebral patterning is affected by pathways involving both the Cdx1 RARE and other RAR-regulated events.

Both RA and Cdx1 have been shown to regulate the expression of certain Hox genes. Indeed, a number of the vertebral phenotypes observed in RARE-null or RARγ-RARE compound mutants closely phenocopy the defects observed in certain Hox mutant mice (Manley and Capecchi, 1997; Horan et al.,1995). To further understand this relationship, we compared the effect of loss of Cdx1 or the RARE on the expression patterns of Hox group 3 and group 4 genes by in situ hybridization.

In wild-type E9.5 embryos, Hoxa3, Hoxb3 and Hoxd3 are all expressed with a rostral limit between the fourth and the fifth somite(Gaunt, 1988; Sham et al., 1992; Condie and Capecchi, 1993),and null mutants of Hoxb3 and Hoxd3 phenocopy certain aspects of the Cdx1-null phenotype (Manley and Capecchi, 1997; Condie and Capecchi, 1993; Allan et al.,2001; Subramanian et al.,1995). Consistent with this observation, the expression pattern of all three of these Hox genes was consistently posteriorized by one somite in Cdx1^-/- offspring (Fig. 8 and Table 2). By contrast, mutation of the RARE had no discernable effect on the pattern of expression of any of these transcripts. This suggests that although Cdx1 plays a crucial role in establishing the proper rostral expression of these group 3 genes, the RARE is not critically required for this function.

The expression pattern of Hoxa4, Hoxb4 and Hoxd4 was also examined. In wild-type embryos, the anterior margin of Hoxa4distribution extends to somite 7, with somite 6 occasionally exhibiting weak expression (Horan et al.,1994; Gaunt et al.,1989) (Fig. 9A). Expression of Hoxb4 and Hoxd4 was consistently observed with a rostral limit at somite 6, with weak expression in somite 5 sometimes seen(Fig. 9D,G). In the case of Hoxa4, somitic expression was not perturbed by disruption of either the RARE or of Cdx1 (Fig. 9B,C). By contrast, Hoxb4 was posteriorized from somite 6 to somite 7 in the majority of both Cdx1 and RARE-null offspring(Fig. 9E,F and Table 2), although residual expression was often observed in somite 6 in both mutant backgrounds. Hoxd4 was also posteriorized by one somite in Cdx1-null mutants, consistent prior observations(van den Akker et al., 2002). The majority of RARE mutants also displayed a comparable posteriorization of expression (Fig. 9H,I; Table 2). Again, weak residual staining was often detected in somite 6 in both mutant backgrounds. As summarized in Table 2, the altered pattern of Hox expression is consistent with the effect on skeletal patterning, with the RARE-null mutants displaying characteristics similar to those of Hoxb4 and Hoxd4-null mutants. Similar correlations are also seen between Cdx1-null mutants and Hox gene expression(Table 3).

Our prior work demonstrated that Cdx1 is regulated by RA through an atypical RARE that functions at least in tissue culture models(Houle et al., 2000). The present study demonstrates a crucial role for this regulatory element in governing Cdx1 expression in vivo, and underscores a functional relationship between retinoid-dependent regulation of Cdx1 and expression of a subset of those Hox genes normally dependent on Cdx1. The near-complete loss of Cdx1 expression seen at later stages in RARE mutants also suggests that this element is essential for events related to Cdx1 autoregulation (Prinos et al.,2001). Finally, this work also presents findings indicative of a second means for retinoid-dependent regulation of Cdx1.

We chose to inactivate the Cdx1 RARE by insertional mutagenesis,the end result of which was replacement of the RARE sequences with a lox site. This strategy was chosen in order to avoid unknown effects of lox insertion on sequences elsewhere in the locus, and to maximize the frequency of recovery of targeted clones. Indeed, targeting of Cdx1 regulatory sequences using an unlinked selectable marker is much less efficient at this locus (N. Pilon and D.L., unpublished). An identical strategy to the present approach has been used previously to disrupt the Hoxa1 RARE (Dupé et al.,1997).

The reduced Cdx1 expression seen in the RARE-null mutants could be due to non-specific effects resulting from the residual loxsequences, rather than from disruption of the RARE per se, and such a possibility cannot be unequivocally ruled out. However, a number of observations suggest that the targeted disruption of the Cdx1 RARE is both specific and obviates RA-response through this element. First, apart from the RARE, this region of the Cdx1 promoter is poorly conserved between murine and human genomes (Houle et al.,2000), suggesting the absence of other conserved regulatory elements. Second, RARE-null embryos initially exhibit a decrease in expression of Cdx1 at E7.5. This effect is observed in the primitive streak region, which is an active region of retinoid signalling at this stage(Balkan et al., 1992; Rossant et al., 1991). Finally, transfection assays demonstrated that the mutated RARE no longer mediates a retinoid response in tissue culture.

A role for the RARE in other Cdx1-dependent processes is also suggested by the finding that the null mutants exhibit reduced expression of Cdx1in the neurectoderm at E8.0. Although this could occur through a reduction in initial expression at E7.5, RA is also found in the embryonic trunk at this stage (Rossant et al., 1991). The reduced expression of Cdx1 in RARE-null offspring may therefore reflect a role for RA in directly regulating Cdx1 in the neurectoderm per se. A similar retinoid-dependent mechanism has been suggested previously for Hoxb4 gene expression in the neurectoderm(Gould et al., 1998). In should also be noted that, although a role for Cdx1 in the CNS is presently unknown, it is tempting to speculate that some of the effects of RA on Hox expression in this lineage could be conveyed via Cdx1, and that the significance of this relationship may be masked by functional overlap with Cdx2 (van den Akker et al.,2002).

The present study suggests the existence of a second pathway for RA induction of Cdx1 that does not require the RARE and is independent of de novo protein synthesis. Although alternative mechanism cannot be ruled out, this observation is consistent with a second direct means for RA-mediated regulation of Cdx1 expression. This is not without precedent, as disruption of the Hoxa1 RARE does not completely attenuate response of this gene to exogenous RA in vivo(Dupé et al., 1997). Hoxb1 is also subject to regulation by multiple RAREs which function tissue-specifically (Huang et al.,1998). However, despite this selective function, some of the Hoxb1 RAREs can mediate response to exogenous RA in a non-specific manner. For example, disruption of the Hoxb1 RARE that normally affects hindbrain expression in vivo does not completely abrogate retinoid response in the neurectoderm (Huang et al., 2002).

Although the existence and identity of a second Cdx1 RARE is presently speculative, we have documented a DR5-like element in the distal Cdx1 promoter that can mediate RA-response in transfection assays in P19 embryocarcinoma cells (M.H., J.R.S. and D.L., unpublished). A putative DR2 RARE has also been described in first intron of both chick and mouse Cdx1 loci that is necessary for a subset of expression in transgenic reporter assays(Gaunt et al., 2003). Although more definitive analysis is necessary to confirm that either of these elements are bone fide RAREs, a second element would offer a basis for induction of Cdx1 by exogenous RA as seen in the RARE-null mutants. Such a mechanism would also offer an explanation for some of the discrepancies between our present findings and observations from transgenic models of Cdx1 regulation(Lickert and Kemler, 2002). In these latter transgenic models, minimal Cdx1 genomic sequences(containing the proximal RARE mutated in the present study) were found to suffice to recapitulate most of the normal pattern of expression of Cdx1. Mutation of the RARE in this context severely affected expression at E7.5, much more so than the outcome of targeted ablation of this element presented here. As both the putative DR5 and DR2 elements are excluded from this transgenic promoter, they could potentially contribute to both early expression in the primitive streak and/or mediate the response of Cdx1 to exogenous RA seen in the RARE-null mutants.

Previously, we had shown that Cdx1 expression is compromised in RARα1-RARγ double null mutants in the primitive streak region specifically at E7.5, although expression at E8.5 is unaffected relative to controls (Houle et al., 2000). This finding is in accordance with the distribution of bioactive retinoid signalling, which is robust in the primitive streak region at E7.5 but is absent at E8.5 and later in the tail bud(Balkan et al., 1992; Rossant et al., 1991). Conversely, we (Prinos et al.,2001) and others (Ikeya and Takada, 2001), have suggested a role for Wnt signalling in affecting Cdx1 expression at later (E8.5-9.5) stages in the tail bud. Moreover, we have observed that loss of Cdx1 protein leads to eventual failure of Cdx1 expression at E8.5, but not at earlier stages. This model(Prinos et al., 2001) suggests that RA specifically regulates early Cdx1 expression, and that other mechanisms, including Wnt signalling and autoregulation, subsequently maintain expression at later stages.

Although the above observations suggest an exclusive early role for retinoid signalling in regulating Cdx1 in the caudal embryo,expression was severely compromised in RARE-null mutants at E8.5 and later. One interpretation for this finding is that loss of the RARE may led to a reduction of Cdx1 expression below a crucial threshold, resulting in subsequent failure of autoregulation. This observation also suggests that ablation of the RARE leads to a loss of greater than 50% of expression, as Cdx1 heterozygous embryos do not exhibit such a compromise in late expression(Prinos et al., 2001). Such an interaction is not without precedent, as the Hoxa4 RARE and an autoregulatory element genetically interact to maintain a similar loop(Packer et al., 1998). Alternatively, loss of expression at E8.5 may reflect a late, direct, role for the Cdx1 RARE. Such an possibility, however, necessitates a ligand-independent mechanism, as retinoid bioactivity is excluded from the caudal embryo at this stage. In this regard, a role for unliganded RAR has been proposed in the development of both the apendicular skeleton and the anterior embryo (Weston et al.,2002; Koide et al.,2001), albeit in a repressor context. Finally, we cannot exclude the possibility that an effector other than the RARs may function through the RARE to maintain later Cdx1 expression.

RAR-null mutants, in particular RARα-RARγ double mutants,phenocopy some of the vertebral malformations seen in Cdx1 mutants(Lohnes et al., 1995; Subramanian et al., 1995; Allan et al., 2001). The vertebral defects seen in RARE-null mutants are also in agreement with a crucial role for RA in governing a subset of Cdx1 expression and function. It is notable, however, that the vertebral defects in RARE-null mutants were restricted to a subset of the rostral-most region affected by loss of Cdx1,while more caudal elements were not affected. As somites receive patterning information prior to their overt segmentation, the reduction of expression seen in RARE mutants at E8.5 and later does not appear to correlate with vertebral defects, which would be anticipated to occur at more caudal levels. This suggests that the RARE is essential for only a limited, early, function of Cdx1, and, conversely, that Cdx1 is not critically required for AP patterning at E8.5 and later stages. In this regard, Cdx2 has been suggested to overlap functionally with Cdx1 (van den Akker et al., 2002; Charité et al., 1998; Marom et al., 1997), which may mask later roles for Cdx1 in vertebral patterning. It is likewise possible that a second RARE precludes our understanding of the full scope of retinoid signalling on Cdx1 expression and function.

We found that RARγ and RARE-null alleles interact on vertebral patterning. In particular, Rare^+/-Rarg^-/- mutants show high penetrance of the C2 to C1 transformation, as seen in Rare^-/- offspring. This is consistent with our previous finding that RARγ and Cdx1 synergize in vertebral patterning through Hox expression in a manner suggesting that retinoid signalling acts both upstream of, and parallel to, Cdx1 (Allan et al.,2001). An alternative interpretation for the interactions seen between RARE and RARγ-null mutants is that loss of the receptor may affect Cdx1 expression through the proximal RARE, and that a second putative RARE is not affected by the loss of this RAR. This possibility is supported by the phenotype of RARE-RARγ double null mutants, which is not reminiscent of Cdx1^-/- offspring as may have been anticipated if RARγ impacts on Cdx1 expression through multiple,equivalent, RAREs. Thus, it is conceivable that a second Cdx1 RARE may be involved in tissue- or RAR-specific regulation, but is still capable of mediating a response to exogenous RA in paraxial mesoderm analogous to the differential functions of the Hoxb1 RAREs(Huang et al., 2002). Alternatively, it is also possible that RARs can regulate Cdx1expression through a second RARE, but such a role is not seen in RARγ-RARE double null mutants because of functional redundancy among this receptor family (e.g. Lohnes et al.,1994).

Based on the nature of the vertebral defects observed in both Cdx1 and RARE-null backgrounds, we investigated the expression of Hox genes from paralogue groups 3 and 4 as likely candidates for patterning defects at these axial levels. We found that all Hox group 3 genes assessed were posteriorized in Cdx1-null embryos, but were unperturbed in RARE-null offspring. This finding is consistent with the more severe nature of C1 defects associated with both Cdx1 and the respective Hox group 3-null mutants(Subramanian et al., 1995; Chisaka and Capecchi, 1991; Manley and Capecchi, 1997; Condie and Capecchi, 1993)(Table 3).

Although the expression of Hoxa4 was not perceptibly altered in any background examined, the anterior limit of expression of both Hoxb4 and Hoxd4 was posteriorized by one somite in both Cdx1^-/- and Rare^-/- embryos. This is in close agreement with the phenotypes of these particular Hox-null mice (Ramirez-Solis et al.,1993; Horan et al.,1995), which exhibit vertebral defects reminiscent of those seen in RARE-null mutants (summarized in Table 3). Moreover, the frequency of posteriorized Hox expression in RARE-null mutants is similar to the incidence of associated vertebral malformations. Taken together, these data suggest a direct relationship between retinoid signalling and Cdx1 function essential for establishing normal expression of these specific Hox genes.

It is notable that, although Cdx1-null mutants exhibit altered expression of both Hox group 3 and group 4 genes, loss of the RARE affected only Hoxb4 and Hoxd4. One possible reason for this observation is that Cdx1 may not be affected at early stages by RARE disruption, and hence more rostrally-expressed Hox genes are not perturbed. We have not, however, noted such an effect, as Cdx1 expression was uniformly reduced at onset in the RARE mutants. Alternatively, it is conceivable that expression of Hox group 3 genes is reliant on a relatively low threshold of Cdx1 protein, and this value is exceeded in RARE-null mutants. Consistent with this, Cdx1^+/- offspring exhibit a higher frequency of vertebral defects associated with loss of expression of Hox paralogue group 4 genes, relative to group 3 members(Allan et al., 2001). This suggests that Hox genes exhibit differential sensitivity to Cdx dose along the AP axis, as previously discussed (Marom et al., 1997; Charité et al., 1998; Gaunt et al.,2003). It will be of interest to investigate Cdx1-dependent regulatory mechanisms governing expression of these particular Hox genes to determine if this is, indeed, the case.

A relationship between FGF, Cdx members and Hox expression has also been shown in Xenopus and the chick embryo(Isaacs et al., 1998; Pownall et al., 1996; Bel-Vialar et al., 2002), and it has been suggested that this pathway is important in establishing the expression domains of more 5′ Hox gene paralogues, at least in the latter model (Bel-Vialar et al.,2002). This study also suggests a qualitative difference in Hox response to RA versus FGF, with more 3′ paralogs responding to RA, but not FGF. We, and others, have also suggested a similar relationship between Wnt signalling, Cdx1 expression and vertebral patterning(Prinos et al., 2001; Ikeya and Takada, 2001). Based on these observations, it is tempting to speculate that Cdx members may play a general role in conveying posteriorization information from multiple signalling pathways to the Hox genes.

The authors thank Dr Peter Gruss and Barbara Meyer for the Cdx1-null mice and Cdx1 cDNA. Plasmids encoding sequences for in situ hybridization were kindly provided by M. Capecchi, D. Wolgemuth,M. Featherstone and P. Chambon. Suggestions from Dr Mark Featherstone and members of the group are also gratefully acknowledged. We also wish to acknowledge Dr Qinzhang Zhu and Michel Robillard for microinjection and Christian Charbonneau who was most helpful with photography. This work was supported by funding from the CIHR of Canada (#14412). M.H. is supported by a studentship from the CIHR of Canada, and D.L. by the FRSQ.