The Cdx4 mutation affects axial development and reveals an essential role of Cdx genes in the ontogenesis of the placental labyrinth in mice

The vertebrate Cdx genes are the homologs of Drosophila caudal(cad) and encode homeodomain containing transcription factors. Together with the Hox genes, they are derived from a common ancestral Proto-Hox complex (Pollard and Holland,2000). During early development the paralogous genes Cdx1,Cdx2 and Cdx4 are initially expressed similarly in the primitive streak, including the base of the allantois(Meyer and Gruss, 1993; Gamer and Wright, 1993; Beck et al., 1995; Chawengsaksophak et al., 2004; Lohnes, 2003). At somite stages, they exhibit nested expression patterns in the posterior axial and paraxial embryonic tissues emerging from the primitive streak and later from the tailbud. The expression of Cdx1 extends most anteriorly and overlaps with the expression domains of Cdx2 and Cdx4posteriorly so that all three Cdx genes are expressed in the tailbud(Meyer and Gruss, 1993; Gamer and Wright, 1993; Beck et al., 1995; Chawengsaksophak et al., 2004; Lohnes, 2003).

The phenotypes of Cdx1 and Cdx2 loss-of-function mutant mice generated by homologous recombination in part reflect the fact that the Cdx proteins are positive regulators of the Hox genes in embryonic tissues(for a review, see Deschamps and van Nes,2005). Cdx1^-/- animals show alterations in vertebral anteroposterior (AP) identity confined to the cervical region(Subramanian et al., 1995). Cdx2^-/- embryos die at E3.5 because of implantation failure (Chawengsaksophak et al.,1997; Strumpf et al.,2005) but Cdx2^+/- embryos implant successfully and, like Cdx1^-/- homozygotes, display an anterior homeotic shift in the axial skeleton, although more subtle and situated more posteriorly. In addition, Cdx2^+/- mice exhibit a slight shortening of the AP axis (Chawengsaksophak et al., 1997). Combined Cdx1^-/-/Cdx2^+/- mutants have an axial phenotype showing abnormalities greater than either mutant separately thus reflecting functional redundancy due to gene co-expression in the presomitic mesoderm (van den Akker et al.,2002). The block to implantation of Cdx2^-/-embryos, that is in keeping with the strong expression of the gene in the trophectoderm at E3.5 (Beck et al.,1995; Chawengsaksophak et al.,1997), can be overcome by tetraploid fusion(Nagy et al., 1993). Development then proceeds to E10.5 but embryos exhibit severe posterior truncation, there being little development posterior to the forelimb buds. The arrest of posterior tissue generation in Cdx2^-/- embryos also concerns the extra-embryonic mesoderm(Chawengsaksophak et al.,2004).

Progenitors of extra-embryonic mesoderm originating from the posteriormost epiblast (Lawson et al., 1991; Kinder et al., 1999) express all three Cdx genes at the late primitive streak stage of development(Meyer and Gruss, 1993; Gamer and Wright, 1993; Beck et al., 1995; Davidson et al., 2003). Cdx2 is the only Cdx gene expressed in the extra-embryonic ectoderm. Expression is demonstrable at E3.5 in the trophectoderm, subsequently in the extra-embryonic ectoderm of the ectoplacental cone(Beck et al., 1995), and eventually in the trophoblastic stem cells(Strumpf et al., 2005). Cdx2 is also expressed in the chorion, though it is downregulated at the late chorionic plate stage. Expression of Cdx2 in the E9.5 placenta is strong in the spongiotrophoblasts but is absent in the adjacent labyrinth (Beck et al., 1995). Thus, the Cdx2-positive cells of the chorion lose expression before they differentiate into labyrinthine syncytiotrophoblast.

We have inactivated the third mouse Cdx gene, the X-linked Cdx4,by homologous recombination in embryonic stem (ES) cells. Cdx4^-/0 mutants are born healthy and are fertile. A modest contribution of Cdx4 in AP patterning, and an important role for this gene in posterior axial extension were revealed in compound Cdx1^-/-/Cdx4^-/0 and Cdx2^+/-/Cdx4^-/0 mutants. In addition, most Cdx2^+/-/Cdx4^-/0 compound mutant embryos die around E10.5, revealing a novel functional involvement of Cdx genes. We show here that a subset of these compound mutants is impaired in chorio-allantoic fusion, even though the allantois reaches a sufficient length to touch the chorion. The remaining majority of the Cdx2^+/-/Cdx4^-/0 compound mutants does undergo successful chorio-allantoic fusion, but subsequently exhibits deficient labyrinthine development, following the failure of allantoic vessels to branch and expand their network in the chorionic plate. This extends the role of the Cdx transcription factors, already known as modulators of embryonic anteroposterior patterning, to include the morphogenesis of the extra-embryonic mesoderm of the chorio-allantoic placenta, the most posterior of the structures originating from the epiblast.

We isolated a 13.2 kb mouse Cdx4 genomic DNA clone from BAC344L11(kind gift of P. Avner, Paris) and used it to construct a targeting vector by standard techniques. A floxed PGK-neo^r cassette was inserted into the EcoRV site upstream of exon 2 and a flipped PGK-hygro^r including one loxp-site was inserted downstream of exon 3 into the HindIII site. We established Cdx4conditional knockout (Cdx4^c) ES cells (129/Ola) by double G418 (200 μg/ml) and Hygromycin (150 μg/ml) selection using standard ES-cell culture protocols and verified positive clones by Southern blot analysis with a unique upstream genomic probe. Chimeric mice were generated by blastocyst injection of two independently targeted Cdx4^cES cells into E3.5 C57BL/6 mouse blastocysts using standard procedures. Chimeric males were mated to FVB females for generation of F1 animals carrying the Cdx4^c mutation. Cdx4^c/+ mice were mated to CMV-Cre to generate Cdx4^-/0 mice. Cdx4^-/0 mice generated from two independently targeted ES cell clones displayed similar phenotypes and both lines were used for further characterization. Further breeding of the mice was in C57BL/6×CBA F1 background. All other mice used were also in C57BL/6×CBA F1 genetic background.

Embryos or visceral yolk sacs were genotyped by PCR using the following primers: NEOF401 5′-GATTGCACGCAGGTTCTCC-3′ and NEOR802 5′-GATGTTTCGCTTGGTGGTC-3′ to detect the 401 bp Cdx4conditional allele. Wild-type Cdx4 was genotyped with primer pair Cdx4FOR4723 5′-TTCCCTAAAAAGACCAAAATCAAA-3′ and Cdx4REV5161 5′-CCCCGGAGAACTGCCTAAC-3′ (438 bp), and the Cdx4 null allele was detected with the Cdx4FOR4723 5′-TTCCCTAAAAAGACCAAAATCAAA-3′ and Cdx4REV7900 5′-AGCGCAAAACCTCACATCA-3′ (400 bp) primer combination. Male(hemizygous) Cdx4-null embryos were confirmed by the presence of the Y-chromosome using primer pair SryFOR 5′-TTATGGTGTGGTCCCGTGGTGAG-3′ and SryREV 5′-TGTGATGGCATGTGGGTTCCTGT-3′ (302 bp). Cdx1 and Cdx2 mutant embryos and mice were genotyped by PCR as described previously (Subramanian et al.,1995; Chawengsaksophak et al.,1997).

Skeletons of E15.5 embryos and newborn mice were stained according to the following procedure: newborn mice were skinned, and both newborn mice and E15.5 embryos were eviscerated, fixed overnight in 96% ethanol containing 1%glacial acetic acid and stained overnight in 0.5 mg/ml Alcian Blue (Sigma)dissolved in 80% ethanol/20% acetic acid. After rinsing twice for 1 hour in 96% ethanol, soft tissue was dissolved in 1.5% KOH for 1 and 6 hours, for E15.5 embryos and newborns, respectively. Bone was then stained overnight in 0.5% KOH containing 0.15 mg/ml Alizarin Red (Sigma). Newborns and embryos were destained in 0.5% KOH/20% glycerol for 1 day or longer and afterwards stored in 20% ethanol/20% glycerol.

Paraffin embedding and processing of the sections, including Hematoxylin and Eosin, and PAS staining, was carried out using standard methods. Methods for in situ hybridization on sections have been described(Moorman et al., 2001). RNA probes used were: Cdx1 (Meyer and Gruss, 1993); Cdx2(Beck et al., 1995); Cdx4 (Gamer and Wright,1993); Mash2(Guillemot and al., 1994); Hand1 (previously eHAND)(Cserjesi et al., 1995); Tpbp (Lescisin et al., 1998). Immunohistochemistry was performed according to standard procedures. The following primary antibodies were used:Pecam (BD Biosciences, 1:50), anti-cleaved caspase 3 (BD Biosciences, 1:300)and anti-Ki67 (BD Biosciences, 1:500). Pecam signal was amplified using Tyramid signal amplification steps (Perkin Elmer). For cleaved caspase 3 and Ki67, Envision+ kit (DAKO) was used as a secondary reagent. Signal was developed with 3,3′ diaminobenzidine (DAB, Sigma) and sections were counterstained with Hematoxylin.

The diameter of Pecam-stained embryonic vessels that had penetrated the chorionic ectoderm and branched in the placental labyrinth was measured at their minimal width on tissue sections with Leica software. Representative sections from the central part of E9.5 placentas were analyzed in eight wild-type controls and eight Cdx2^+/-/Cdx4^-/0compound mutants. Two-tailed t-test assuming equal variance was used to determine P for the total embryonic vessel number per placental labyrinth and for the average embryonic vessel width.

E7.5 and E8.25 allantoises were dissected from 0- to 3- and 5- to 7-somite embryos, respectively. The isolated allantois was fixed in immunohistochemistry-zinc-fixative (BD Biosciences). Fluorescent antibody staining was performed according to standard procedures. Briefly, allantoises were permeabilized with 0.2% Triton X-100 in TBS, blocked for 1 hour in TNB buffer (Amersham) at room temperature and incubated overnight at 4°C with primary antibody against Pecam (BD Biosciences, 1:100) or Flk1 (BD Biosciences, 1:100) in TNB buffer. After three washes with TBS,FITC-conjugated anti-rat IgG (Jackson Immunological, 1:100) secondary antibody in TNB was applied for 1 hour at room temperature. Allantoises were analyzed using a Leica confocal microscope, permitting 3D compilation of individual confocal planes.

We designed a strategy to conditionally inactivate Cdx4 in order to circumvent the risks of creating an embryonic lethal phenotype(Fig. 1A,B). Mice bearing the Cdx4 conditional allele and, more surprisingly, Cdx4^-/0 embryos developed to term and were born healthy,according to Mendelian frequencies. Some 20 adult Cdx4-null males have been used in successful breeding programs, half of them for more than a year. In addition, homozygous Cdx4-null females have been generated,and a homozygous mutant colony has been established, proving that these animals were physiologically healthy. Skeletal preparations of Cdx4-null and wild-type E15.5 embryos (n=5), and of Cdx4-null and wild-type newborns (n=3), established that the mutants have a wild-type vertebral phenotype at all axial levels (not shown). The only exception concerns a difference at the level of vertebra 15 (V15),where one (1/8) mutant had its eighth rib pair attached to the sternum whereas this rib pair was the first free-floating in the other mutants and the controls. This feature might be interpreted as a very mild anterior transformation at the level of V15 with an extremely low penetrance.

Histological analysis of the digestive tract of two 1-year-old adult Cdx4-null mice did not reveal any abnormalities. Given the role of Cdx factors in AP patterning, the complete gastrointestinal (GI) tract was investigated for anatomical and histological organization. Anatomical boundaries between esophagus, stomach, small intestine and colon were clearly defined, and no truncations or homeosis was observed. Stomach, intestine and colon exhibited normal histological structures based on Hematoxylin and Eosin staining. Proliferation was assessed by Ki67 staining and was restricted to the gland isthmus in the stomach (not shown) and to the crypt in the intestine and colon (Fig. 1C), showing no differences to control mice. PAS staining revealed a normal distribution of mucus cells in the stomach and goblet cells in the intestine (not shown). Thus, proliferation and differentiation was normal throughout the GI tract. Finally, unlike earlier observations in Cdx2^+/- mice, no intestinal polyp-like lesions were observed.

In view of early co-expression of Cdx1, Cdx2 and Cdx4,and the functional redundancy between Cdx1 and Cdx2 in axial extension and AP patterning (van den Akker et al., 2002), we set out to investigate whether additional loss of Cdx1 and Cdx2 alleles would reveal functional cooperation between Cdx4 and these other Cdx genes. E15.5 Cdx1^-/-/Cdx4^-/0 double-null mutant embryos exhibited very subtle changes in their vertebral phenotype compared with Cdx1-null littermates. Although 5/9 Cdx1-null embryos carried their first rib on the 9th vertebrae (V9) unilaterally instead of on V8 (thoracic to cervical transformation of V8, with the consequent loss of one rib), 5/7 Cdx1/Cdx4 double null mutants did, of which 3/5 exhibited this transformation bilaterally, manifesting a slight increase in severity and penetrance of the Cdx1 phenotype at this V8 level. In addition, 5/7 double mutant embryos had their 8th rib pair attached to the sternum bilaterally, while only 2/9 Cdx1-null mutants did, and 3/9 unilaterally (not shown). This represents a slight increase in the severity and penetrance of the Cdx1-null phenotype at the V15 axial level. Compared with E15.5 Cdx2 heterozygote mutant embryos (n=4),which did not manifest any vertebral transformations in our genetic background, 4/4 Cdx2^+/-/Cdx4^-/0 littermates had their 8th pair of ribs attached to the sternum, manifesting an anterior homeotic-like transformation at this V15 level. In addition, all four compound mutant embryos had a partial or complete rib attached to V21 bilaterally, at the position normally corresponding to the first lumbar vertebrae. Consequently, the compound Cdx2^+/-/Cdx4^-/0mutants have 14 ribs. We conclude that the Cdx4 mutation has a specific additive effect on the Cdx1 loss of function, and on the loss of one active Cdx2 allele, in patterning the vertebrae at discrete thoracic levels (respectively V8 and V15, and V15 and V21).

Cdx1-, Cdx4-null embryos and Cdx1/Cdx4 double null mutants did not manifest any axial truncation (not shown). By contrast, Cdx2 was found to strongly synergize functionally with Cdx4in posterior tissue generation during axial extension. Cdx2^+/-/Cdx4^-/0 compound mutant embryos exhibited truncation of embryonic structures posterior to the hindlimbs,resulting in a more anteriorly located tail bud, a phenotype much more severe than that of either mutant alone and reminiscent of, though less marked than that of Cdx2-null embryos (Fig. 2A-C).

In addition, and independently from the posterior truncations which are not expected to lead to early embryonic lethality, most compound Cdx2^+/-/Cdx4^-/0 embryos were growth retarded at E10.5 (see Fig. 2A) and died in utero shortly thereafter (Tables 1 and 2). The heartbeat and circulation in the embryo, as well as circulation in the yolk sac were normal at E9.5, but failed around E10.5. Taken together with the expression of Cdx2 and Cdx4 in the allantoic anlage during normal development, these findings suggested chorioallantoic placental failure. A first analysis of the phenotype of embryos from Cdx2^+/-and Cdx4^+/- intercrosses revealed that the allantois of a small number of Cdx2^+/- (three out of 54), Cdx2^+/-/Cdx4^+/- (two out of 28) and Cdx2^+/-/Cdx4^-/0 (seven out of 35) embryos had not fused with the chorion at E9.5 (Fig. 3B,C; data not shown). Measurement of allantoic length around the normal fusion stage of six to eight somites (E8.0)(Downs and Gardner, 1995; Downs, 2002) indicated that the reason for the lack of fusion of the mutant allantoises with the chorion was not a length reduction (see Fig. S1 in the supplementary material). Histological analysis of E9.5 unfused allantoises revealed that the chorio-adhesive mesothelium had formed, and that blood filled vessels were present in the distal allantoic mesenchyme(Fig. 3D). Moreover, in situ hybridization showed that Vcam1 was expressed in the mutant E8.5 distal allantois that had failed to fuse (see Fig. S1 in the supplementary material). The mesodermal layer of the chorion was also present in the mutant,and expressed α4 integrin, the VCAM1 receptor in the chorio-allantoic fusion process. Expression of α4 integrin was still observed in the mutant at a stage when chorio-allantoic fusion and degradation of the chorionic mesoderm layer have occurred in the wild type (see Fig. S1 in the supplementary material). A subset of Cdx2^+/-/Cdx4^-/0 mutant allantoises is thus unsuccessful in fusing with the chorion in spite of the fact that they reach a sufficient length to touch it, and achieve an advanced degree of differentiation.

From a second data set involving exclusively the 80% of E8.5 Cdx2^+/-/Cdx4^-/0 embryos that had undergone successful chorio-allantoic fusion, histological analysis revealed that virtually all (in this case 18/18) Cdx2^+/-/Cdx4^-/0 embryos displayed deficient placental labyrinth development at E10.5(Fig. 4E-H; Table 1).

In order to characterize the most frequent defect causing the death of Cdx2^+/-/Cdx4^-/0 embryos, we selectively recovered compound mutant embryos that had undergone successful chorio-allantoic fusion. At E9.5, blood vessels were present in the Cdx2^+/-/Cdx4^-/0 mutant allantois, and they had initiated penetration of the chorionic ectoderm (Fig. 4B, compare with 4A, and with enlargements in 4C and 4D). We observed some variation in the extent of allantoic vessel penetration into the chorion, and the mildest defects may correspond to the minority of Cdx2^+/-/Cdx4^-/0 mutants that survive(Table 2). In all E10.5 mutant placentas that we analyzed histologically, maternal blood pools and embryonic blood vessels were widely separated in E10.5 mutant placentas (compare Fig. 4F,H with 4E,G),indicating an increasingly more dramatic defect with age, and showing that the mutant phenotype is not a simple developmental delay. These findings are sufficient to explain the high midgestation mortality as being due to placental insufficiency.

We studied the allantoic vasculature that, together with the chorion will generate the placental labyrinth. The allantoic vasculature is known to be initiated by vasculogenesis within the allantois before the latter fuses with the chorion (Downs et al.,1998; Downs,1998). Following chorio-allantoic fusion, allantoic vessels penetrate the chorionic ectoderm and branch extensively to interdigitate between the maternal blood sinuses (reviewed by Cross, 2005; Rossant and Cross, 2001). We followed this process throughout development in mutant and wild-type placentas using several endothelial markers. In situ hybridization of Flk1transcripts (data not shown) and antibody staining for PECAM expression(Fig. 5C-F) both highlighted the limited expansion of allantoic vessels in the chorionic ectoderm in mutants compared with controls at E9.5 and E10.5. Quantification of both the number and the diameter of allantois-derived embryonic vessels that had penetrated the chorionic ectoderm at E9.5 clearly documented that these allantoic vessels branched much less extensively in Cdx mutants(Fig. 5A,B), and almost never came close to the maternal blood sinuses. The thin hemotrichorial membrane(Enders, 1965) normally developing in areas of close contact between fetal and maternal blood circulation is almost completely absent in the mutant(Fig. 4C,D). The defect in establishment of an interdigitated labyrinth was more extreme at E10.5, with maternal and embryonic blood widely separated(Fig. 5E,F and Fig. 4G,H).

Cdx2 is expressed in the placental spongiotrophoblasts but not in the labyrinth, in both wild type and Cdx2^+/-/Cdx4^-/0 mutants at E9.5(Fig. 5G,H). Cdx4 was not expressed in either placenta or labyrinth (data not shown).

We tested the expression of several endothelial markers [Tie1,Tie2 (Tek - Mouse Genome Informatics), Flk1(Kdr - Mouse Genome Informatics), Flt1, Pecam, Vegf] by RT-PCR in the E8.25 allantois before chorio-allantoic fusion. All genes were expressed in compound mutants and controls (see Fig. S2A in the supplementary material). In addition we tested the expression of Angpt1, Angpt2(previously Agpt1 and Agpt2) and Vegf, genes known to be involved in angiogenesis during placental labyrinth development (reviewed by Rossant and Cross, 2001) by semi-quantitative RT-PCR at E9.25 and did not observe consistent changes in gene expression (see Fig. S2B in the supplementary material).

The expression of trophoblast markers was assayed as well in mutants and controls. Mash2 (Ascl2 - Mouse Genome Informatics)(Guillemot et al., 1994) was found to be expressed in the spongiotrophoblasts and in the differentiating labyrinthine trophoblasts at E9.5 (Fig. 5I,J). The expression of Fzd5 expression in labyrinthine trophoblasts and Flt1 in the spongiotrophoblast (reviewed by Rossant and Cross, 2001) were not changed (data not shown). Analysis of the expression of Hand1 in the diploid trophoblasts and trophoblast giant cells(Cserjesi et al., 1995), and Tpbp in spongiotrophoblasts and their precursors in the ectoplacental cone (Lescisin et al., 1998), did not reveal any change between Cdx compound mutants and controls (Fig. 5K-N). These results confirmed the presence of all placental cell types in E9.5 mutants and controls, and only reflected the lack of labyrinthine cell organization in Cdx2^+/-/Cdx4^-/0 mutants, compared with wild types.

We examined the expression of Gcm1, encoding a factor associated with specification of the sites of penetration of allantoic vessels into the chorionic ectoderm, and a marker of differentiating chorionic trophoblasts at later stages (Anson-Cartwright et al.,2000). Although no difference could be detected between the weak Gcm1 expression in E9.5 control and mutant placentas by in situ hybridization (not shown), a slightly reduced expression was observed in semi-quantitative RT-PCR analysis (see Fig. S2B in the supplementary material).

The degree of proliferation of labyrinthine endothelial cells and the possible occurrence of apoptosis were tested at several developmental stages in control and mutant embryos. Ki67 labelling did not reveal impaired proliferation (see Fig. S3 in the supplementary material), and the use of caspase 3 did not indicate increased apoptosis (see Fig. S3 in the supplementary material) at E9.5, the stage at which the allantoic vasculature initiates penetration into the chorionic trophoblast in both mutants and controls. At E10.5, apoptotic cells were more numerous in mutant than control placenta, but this difference was not restricted to the endothelial cells as it concerned chorionic trophoblasts and fetal blood cells as well, and may only reflect developmental arrest around this stage (see Fig. S3 in the supplementary material).

Cdx genes are not expressed in the labyrinth. The only place and stage where progenitors for the allantoic vasculature are expected to express Cdx genes is the early allantoic bud at primitive streak stages. Cdx4,like Cdx2, is widely expressed in the early allantoic bud (E7.5),while expression of Cdx1 is confined to the base of the allantois(Fig. 6A-C). In order to characterize the ontogenesis of the allantoic component of the labyrinth in the Cdx mutants, we analyzed the distribution of Flk1-positive endothelial precursor cells (Downs et al.,1998; Drake and Fleming,2000; Yamashita et al.,2000) at an early stage. Flk1-positive cells are present in both mutants and controls (Fig. 6D,E), in an area and at a stage (E7.5) when Cdx2 and Cdx4 are still fully expressed in the allantois(Fig. 6B,C). This indicates that the precursors of allantoic endothelial cells could be affected by the Cdx mutations.

At E8.0 and E8.5, Pecam-positive endothelial cells are abundant and are organized in a primary vessel network in both mutants and controls. Among the mutant specimens that we examined compared with age-matched controls (nine mutants and eight wild types), we found some variability in the extent of maturation of this network, which may correspond to the variability in the severity of the labyrinth defect at E9.5 described above. Fig. 6F,G shows an example of an E8.25 age-matched pair of allantoises shortly after chorio-allantoic fusion and reveal a lower degree of endothelial organization into a primary vessel network in the mutant compared with the wild type. In the (in this case)slightly smaller mutant allantois, the endothelial cells are locally less well organized than in the control (Fig. 6H,I). Only a minority (two out of nine) of Cdx2^+/-/Cdx4^-/0 mutant allantoises isolated at E8.25 after chorio-allantoic fusion show similar endothelial organization to the wild types (not shown). Thus, these data reveal that mutant allantoic endothelial cells often (seven out of nine) exhibit, to a variable extent, a poorer degree of primary vessel organization around the time of chorio-allantoic fusion. Although endothelial cells are abundant in both mutants and controls, they seem to be less organized in the Cdx2^+/-/Cdx4^-/0 compound mutants than in the wild type, suggesting an incipient defect expressed later as failure to extend and branch into the chorion between E9.5 and E10.5.

Thus, the data altogether reveal that an essential component of the labyrinthine defect observed in E9.5 and E10.5 Cdx mutants resides in the intrinsic property of the allantois to organize and extend its vasculature into the chorion.

The data presented here on the phenotype of Cdx2/Cdx4 double mutant embryos indicate that the Cdx genes are essential for the ontogenesis of a functional placental labyrinth. Their inactivation compromises chorio-allantoic fusion in a subset of the compound mutants. In the majority of these mutants, which successfully unite their allantois with the chorion,loss of Cdx function disables the allantoic vessel extension and branching within the chorionic ectoderm between E8.5 and E10.5. Neither Cdx4deficiency alone, nor, in most cases, Cdx2 heterozygosity,compromises placental development crucially. However, Cdx2heterozygosity together with heterozygous or homozygous loss of Cdx4often result in placental failure. In the Cdx2 heterozygote background, the penetrance of the labyrinth deficiency is dependent on the dose of Cdx4 (Table 1), which is only expressed in the allantoic component of the labyrinth. The combined function of Cdx2 and Cdx4 in the placental labyrinth may appear to be specific. Nevertheless, considering the redundancy between Cdx functions in the embryo, we must await the analysis of more complex genotypes to determine whether Cdx1 is, to some extent,assisting Cdx2 and Cdx4 in this function.

It will be interesting to find out whether Cdx genes exert their effect on the placenta by regulating target Hox genes. This is suggested by the reported biological activity of certain Hox genes to promote migratory behavior of adult endothelial cells in culture(Boudreau et al., 1997) and to induce morphogenesis of new vascular sprouts in chick chorio-allantoic membranes in culture (Myers et al.,2000). Our data suggest that the regulatory function of Cdx genes in placentogenesis is an early one, acting at the level of the progenitors of endothelial cells in the early allantois, as Cdx genes, like Hox genes, are downregulated in the allantois at E8.5.

A possibly related function of Cdx4 has been described in zebrafish (Davidson et al.,2003). Cdx4-null zebrafish embryos manifest a transient deficiency in generating erythroid progenitors(Davidson et al., 2003). The authors suggest an early function of Cdx4 in making the posterior lateral plate mesoderm, containing early hematopoietic progenitors, competent to respond to genes that specify hematopoietic fate. The defect in zebrafish Cdx4 mutants was a selective loss of blood progenitors but not of angioblast specification, thus affecting only one of the hemangioblast derivatives (Huber et al.,2004; Ema et al.,2003). Such a defect in blood development was not evident in the Cdx compound mutant mice examined so far, but a possible slight and/or transient delay in generating erythroid cells at any one of the stages when blood progenitors are born (successively yolk sac blood islands, placenta, AGM and fetal liver) (Gekas et al.,2005; Ottersbach and Dzierzak,2005) might not have been noticed in our experiments. Alternatively, such a function for Cdx4 might be redundant in mice,thus differing from the situation in zebrafish. An insufficiency may remain hidden until the Cdx dose has been lowered below a certain threshold. The fact that Cdx4 rescues the hematopoietic colony-forming potential in mixed lineage leukemia (Mll) mutant ES cell-derived embryoid bodies and enhances the frequency of hematopoietic colonies in control embryoid bodies(Ernst et al., 2004) provides support for this hypothesis.

Vascular lineages were not affected by the Cdx4 mutation in zebrafish but the absence of an allantois in fish leaves the possibility that,after the acquisition of the chorio-allantoic placenta in mammals, the gene became specifically involved in the ontogeny of a functional allantoic vasculature. An implication of these findings is that mutations in Cdx genes might lead to pregnancy failure in humans.

Among the three Cdx family members, Cdx2 is the only one that is expressed in the early extra-embryonic ectoderm, where it ensures establishment and maintenance of the trophectoderm lineage(Strumpf et al., 2005),mediating implantation of the conceptus(Chawengsaksophak et al.,1997). This specific expression and function only concerns Cdx2 and may result from a trophectoderm-specific regulatory influence selectively relating to Cdx2, without counterpart on the other, unlinked, Cdx genes. This early Cdx2 function requires one active allele, and is not relevant in the context of the compound mutants analyzed in the present work.

The common Cdx functions studied here concern epiblast-derived embryonic and extra-embryonic tissues. All three Cdx genes collaborate to pattern the axial skeletal structures [see van den Akker et al.(van den Akker et al., 2002)for Cdx1/Cdx2; see this work for Cdx1/Cdx4 and Cdx2/Cdx4, where the contribution of Cdx4 is modest but observable in the presence of the Cdx1 and Cdx2 mutations]. The distinct AP level of the vertebral changes in the double mutants may relate to the differences in rostral span of the expression domains of Cdx1 and Cdx2. The three Cdx genes contribute to allow completion of posterior axial elongation(van den Akker et al., 2002)(this work). Again, their contributions are unequal, as the effect of inactivating Cdx1 or Cdx4 was visible only in combination with a mutation in Cdx2, whereas Cdx2 heterozygocity alone affects axial elongation (van den Akker et al., 2002) (this work). In this case, we cannot tell whether these differences in contribution result from sequence differences in the proteins or from differences in the spatiotemporal regulation of their expression.

Two of the three Cdx genes (Cdx2 and Cdx4) have been shown in this work to control the ontogenesis of the allantoic part of the placenta, and we cannot rule out that a contribution of Cdx1 in this function might be revealed upon further lowering Cdx dose. All three Cdx genes are expressed in the posterior epiblast, including the most proximal area containing the presumptive extra-embryonic mesoderm. Although loss of function of Cdx1 or Cdx4 does not affect the extra-embryonic derivatives by themselves, Cdx2-null mutations have previously been shown to impair the generation of embryonic and extra-embryonic mesoderm. The Cdx2-null allantois remains extremely short, and does not touch the chorion(Chawengsakhophak et al., 2004). This reveals the earliest Cdx dependence of placental ontogeny, reflected by the fact that one active Cdx2 allele is required for outgrowth of the early allantoic bud. The allantoises in Cdx2^+/-, Cdx2^+/-/Cdx4^+/-and Cdx2^+/-/Cdx4^-/0 mutants reach a normal size. But these three genotypes exhibit specific subsequent defects that compromise the ontogenesis of a proficient chorio-allantoic placenta, with a penetrance that increases with the decrease in Cdx dose. A first defect is manifest by the fact that some of the fully grown allantoises reach the chorion but fail to unite with it. Several mutants in the Wnt and Fgf signaling pathways are impaired in chorio-allantoic fusion [Wnt7b(Parr et al., 2001); Tcf1/Lef1(Galceran et al., 1999); Fgfr2(Xu et al., 1998)]. This could indicate a functional link between these signaling pathways and Cdx genes in the chorio-allantois fusion process, in addition to the links previously documented in the processes of embryonic axial extension and patterning(reviewed by Lohnes, 2003; Deschamps and van Nes, 2005). The majority of mutant allantoises undergo chorio-allantoic fusion but exhibit a later defect, being impaired in the establishment of a functional endothelial network in the labyrinth. The penetrance is again dependent on the dose of Cdx: 10% for Cdx2^+/-, 20% for Cdx2^+/-/Cdx4^+/- and ∼100% for Cdx2^+/-/Cdx4^-/0 embryos. In each of these genotypic classes, the most frequently occurring labyrinth abnormalities are qualitatively indistinguishable and equally severe, suggesting that Cdx2 and Cdx4 function in an identical way, and that the Cdx-dependence operates at the level of reaching a threshold rather than touching different aspects of the events. A small subset of the numerous Cdx2^+/-/Cdx4^-/0 embryos examined were found to be more mildly affected than the majority of them, revealing a certain variability in the severity of this phenotype (these exceptions probably correspond to the occasional survivors among the embryos with this genotype). This variability, which we expect to be proportionally present in each genotypic class (Cdx2^+/-, Cdx2^+/-/Cdx4^+/- and Cdx2^+/-/Cdx4^-/0) would possibly result either from stochastic events among parameters involved in the onset of the phenotype, or from the fact that the genetic background of embryos carrying the new Cdx4 mutation is not homogenous.

In conclusion, our results reveal that the collaborative role of Cdx genes in posterior embryonic development extends to the stepwise establishment of a functional placental labyrinth, from allantois growth and chorio-allantoic fusion to the extension of the allantoic vascular network within the chorionic ectoderm. The data highlight a novel role for these posterior embryonic patterning genes in controlling a vital function that allows the fetus to survive through exchanges with its mother, a function presumably acquired during vertebrate evolution towards placental mammals.

The three Cdx genes exert redundant functions in mediating the generation and AP patterning of posterior embryonic structures. They might also operate redundantly in placentogenesis. In both cases, their functional contribution is quantitatively unequal, owing to either protein or regulatory differences. This suggests that the extra-embryonic mesoderm arising from the posterior part of the primitive streak is controlled by the same genetic system as the posterior axial and paraxial embryonic structures.

We thank P. Avner for the generous gift of the BAC clone containing Cdx4, A. Trumpp for the Neomycin selection cassette, and M. Gessler,F. Guillemot, E. Olson, B. Stecca, C. Wright and J. Yang for probes. We thank B. Meyer and P. Gruss for making the Cdx1-null mice available to us. We are grateful to M. Reijnen and J. Korving for help with generating the Cdx4-null mice, and to J. Korving, H. Begthel and M. van den Born for histological assistance. H. Clevers, K. Downs and A. Gregorieff are gratefully acknowledged for valuable discussions, and F. Meijlink, C. Mummery and K. Lawson for reading the manuscript. The authors were supported by the Dutch Research Organization NWO (ALW grant 810.68.011), the Dutch Bsik Program `Stem Cells in Development and Disease', and the EU Framework 6 Network of excellence `Cells into Organs' (to J.D.).