The Cdx (Caudal-type homeobox) group of ParaHox genes (Cdx1, Cdx2 and Cdx4 in the mouse) perform multiple functions in mammalian development. Cdx1 is concerned with axial positional information, and its deletion appears to have no important effect other than a disturbance of axial patterning. In contrast, Cdx2 is required for trophoblast differentiation, axial patterning and extension, as well as for morphological specification (i.e. patterning) of gut endoderm. Cdx4-knockout animals do not present an abnormal phenotype, but, when combined with Cdx2 haploinsufficiency, present a dramatic picture involving abnormal cloacal specification. The latter is probably due in large part to defective paraxial mesodermal development in the caudal region, but may also involve defective endodermal growth. A significant degree of redundancy is apparent between the Cdx genes with respect to caudal extension and possibly also during gut development.
Cdx (Caudal-type homeobox) genes contain the homeobox motif and are classified as ParaHox genes. The group is closely related to the Hox cluster and appears to have arisen from a common ProtoHox cluster by duplication combined with selective gene loss [1,2]. In Cephalochordates, exemplified by Amphioxus, the ParaHox cluster consists of Cdx, Xlox and Gsx. Further cluster duplication and gene loss resulted in a homologous gene cluster in humans and mice comprising Gsh2, Cdx2 and Pdx1, together with one additional Gsh gene and two Cdx genes (Cdx1 and Cdx4) that map to other chromosomal locations. The latter are probably remnants of degraded paralogous gene clusters. In the mouse, we are therefore left with Cdx2 which is linked with Pdx1 and Gsh2 on chromosome 5, Cdx1 on chromosome 18 and Cdx4 which is X-linked. Some degree of spatial co-linearity seems to have been retained in the remaining ParaHox cluster, since Gsh, Pdx and Cdx genes are expressed in an AP (antero–posterior) sequence, but there is no evidence of temporal co-linearity.
Cdx2 is the first of the Cdx genes to be expressed during development. It appears at E (embryonic day) 3.5 in the trophectoderm and persists in the trophoblastic stem cell line, but is extinguished in the mature trophoblast . In embryonic tissues, all three Cdx genes are expressed in the posterior part of the primitive streak, beginning at the late streak stage. They form a nested group, with Cdx1 extending most anteriorly and Cdx4 most posteriorly to the base of the allantois (Figure 1). Expression continues through to the tailbud stage and ends between E12.5 (for Cdx2) and E10.5 (for Cdx4) . Endodermal expression differs from that of the ectomesoderm. Cdx2 is expressed from inception in the endodermal lining of the intestinal epithelium posterior to the stomach and continues throughout life with a maximal level in the paracaecal region of the intestine. Expression is higher in the villous region than in the crypts . Various phosphorylated forms of Cdx2 coexist in the intestine, exerting different effects on cell behaviour . Cdx1 is demonstrable in the post-gastric endoderm from E12.5; expression persists throughout life, is highest in the distal intestine and is expressed primarily in the region of the intestinal crypts [5,7]. Cdx4 is expressed transiently in the hindgut pocket early in development, but, in contrast with Cdx2 and Cdx1, does not persist into later embryonic stages and postnatally .
Expression of (A) Cdx1, (B) Cdx2 and (C) Cdx4 at late gastrulation
The role of Cdx genes in axial development
Subramanian et al.  disabled the Cdx1 gene by homologous recombination. In Cdx1−/− mutants, they reported an anterior homoeotic shift principally involving the occiput and the first three cervical vertebrae, but occasionally extending to the lower cervical and upper thoracic regions . No other phenotypic changes in mutant animals were observed, and, significantly, both the gut and caudal extension of the AP axis developed normally. The changes were accompanied by a posterior shift of Hox expression involving three different clusters, leading to the suggestion that the Cdx1 effects are mediated through Hox genes. Heterozygote Cdx1+/− animals were completely normal.
Cdx2 knockout by homologous recombination presents a much more severe phenotype. Cdx2−/− animals are embryolethal; the blastocyst fails to undergo uterine implantation due to non-development of mature trophoblast from trophectoderm. Blastocysts collapse and die after approx. 30 h (around E3.5) of development . Heterozygotes survive, although there is some pre-natal mortality. At birth, Cdx2+/− animals are growth-retarded and have various degrees of posterior truncation. Skeletal preparations reveal axial anterior homoeotic shifts principally in the lower cervical and upper thoracic regions . Combined Cdx2+/−/Cdx1−/− mutants show a greater degree of posterior truncation, indicating a degree of redundancy between the Cdx genes . The implantation block in Cdx2−/− animals can be overcome by tetraploid fusion , resulting in post-implantation development of a Cdx2−/− animal on a (wild-type) tetraploid trophoblast that differentiates normally to support implantation. Similar results were obtained by Cre/loxP-mediated conditional knockout of Cdx2 in which Cre is expressed under a Sox2 [SRY (sex-determining region Y) box 2] promoter active in the embryoblast only and not in the trophectoderm (J. Deschamps, personal communication). Using either method, embryos are severely truncated posteriorly, little development occurs below the forelimb bud and, from somite 5 onwards, the paraxial mesoderm is underdeveloped (Figure 2). The posterior mesoderm of the allantoic bud fails to develop normally and does not fuse with the chorion. Hence the chorio-allantoic placenta does not develop, and the embryos die at around E10.5. As in Cdx1 mutants, there is a posterior shift of Hox gene expression, again suggesting that Cdx effects are mediated through Hox genes. Cdx4, an X-linked gene, was also inactivated by homologous recombination . Neither Cdx40/− males nor Cdx4−/− females exhibited a significantly abnormal phenotype, but combined Cdx2+/−/Cdx4−/− (Cdx40/−) embryos were grossly affected. Most died before birth because of under-development of the chorio-allantoic placenta. In survivors to term, posterior truncation was again a prominent feature, and the phenotype corresponded with a variety of the clinical condition known as the caudal regression syndrome . There was gross dysmorphology of the lumbosacral vertebrae and almost complete absence of caudal vertebrae. This is additional evidence to support the redundant role of Cdx genes in axial extension. Cloacal development was also compromised, the nature of which is discussed in the next section
Reseue of Cdx2−/− embryos through implantation block
The role of Cdx genes in the gut
Unlike their common function in axial extension, the Cdx genes have distinct roles in gut development, patterning and function, although some degree of redundancy may be operative.
Cdx2 plays a central role in gut morphogenesis. Cdx2+/− and Cdx2−/−/wild-type chimaeras develop polyps of gastric mucosa in the paracaecal region of the midgut . At birth, these frequently present as small areas of stratified squamous epithelium in the terminal ileum, caecum or proximal colon. As the animal matures, histologically normal regions of gastric mucosa are ‘intercalated’ between the stratified epithelium (characteristic of forestomach) and the surrounding colonic mucosa. Eventually, an orderly progression of tissue types consisting of neutral mucus-secreting cells (cardia), gastric glands containing oxyntic and pepsin-secreting cells (stomach body) and branched pyloric-type neutral mucus-secreting glands containing gastrin-producing enteroendocrine cells surround the stratified epithelium. Immunocytochemically, these tissues do not stain for Cdx2 and appear to be histologically normal gastric endoderm. Between the gastric tissue and surrounding colon (in those polyps that develop in the large intestine), an area of Cdx2-positive small-intestinal-type mucosa develops  (Figure 3). These findings are interpreted as suggesting that Cdx2 drives differentiation towards an intestinal phenotype and that, if levels fall below a threshold during development, histogenesis proceeds along a ‘default’ forestomach gastric pathway, followed by the appearance of intercalated tissue to reconstitute the orderly succession of tissue types. The random appearance of polyps in Cdx2+/− animals is explained in stochastic terms as arising in areas where (possibly due to epigenetic factors) Cdx2 levels have fallen locally below threshold levels. These conclusions are supported by the recent work of Gao et al.  in which Cdx2 was conditionally inactivated specifically in the gut from inception, but allowing normal implantation and development to full term. The intestinal mucosa in these animals was entirely converted into a forestomach phenotype, and the distal colon was absent. Thus the whole intestinal mucosa was affected, and no normal intestinal mucosa was present; the (unknown) stimulus driving intercalation was therefore not operative.
Caecal polyp in the paracaecal region of a Cdx2+/− mouse
Further support for the notion that Cdx2 is instrumental in determining the fate of the gut mucosa comes from the findings of Silberg et al.  who made transgenic mice that expressed Cdx2 in the stomach endoderm under a Foxa3 (forkhead box A3) promoter. They demonstrated incomplete intestinal metaplasia in the distal stomach from the fetal stages into adulthood. The predominant intestinal cell type found was the acid mucopolysaccharide-secreting goblet cell rather than the absorptive enterocyte. They provide no evidence for true transdifferentiation rather than for a reprogramming of undifferentiated stem cells. More convincing evidence is provided by Mutoh et al.  who generated mice expressing Cdx2 in the stomach under the promoter for the non-catalytic β-subunit gene of rat H+/K+-ATPase. They convincingly demonstrated complete replacement of the gastric mucosal cells by intestinal-type mucosa consisting of goblet cells, enteroendocrine cells and absorptive enterocytes expressing alkaline phosphatase. Moreover, the transition began at postnatal day 19 when the proton pump first becomes active and was accompanied by a change in position of the proliferative zone from the isthmus (characteristic of gastric glands) to the base of the glands (comparable with the crypt region of the intestine).
Gut morphogenesis involves not only endodermal differentiation, but also concomitant changes in the underlying mesoderm together with complex endodermal–mesodermal cross-talk. Kim et al. [19,20] demonstrated that expression of the homeobox gene BarX1 (BarH-like homeobox 1) is restricted to the stomach mesoderm during gut organogenesis. Extinction of BarX1 by homologous recombination resulted in gross disturbance of gastric mucosal development. They postulated that BarX1 directs mesenchymal cell expression of secreted Wnt antagonists that inhibit canonical Wnt signalling, thus permitting the development of stomach-specific epithelium. In support of this demonstration of epithelial–mesenchymal cross-talk, Stringer et al.  showed that ectopic areas of gastric mucosa in the paracaecal regions of Cdx2+/− and Cdx2−/−/wild-type chimaeric mice expressed BarX1 in the underlying mesoderm while the adjacent mesoderm in the normally developing intestine did not do so .
The multiple actions and expression sites of Cdx2 during development and into adulthood have stimulated enquiry into the responsible promoter elements. Benahmed et al.  used various transgenic genomic fragments covering up to 13 kb of the mouse Cdx2 locus to drive expression of a reporter gene. They demonstrated that the fragments containing up to −5 kb upstream of the start codon recapitulated the axial expression pattern of Cdx2 and were also active in the gut, but became inactive at mid-gestation . Specific persistence of signal in the midgut region of the endoderm into adulthood required genomic fragments extending to −9 kb. The authors defined a 250 bp segment around −8.5 kb that bound and responded to HNF4α (hepatic nuclear factor 4α), GATA6, Tcf4 (T-cell factor 4) and β-catenin. In HeLa cells, these factors activated endogenous expression of Cdx2 synergistically. None of these regions recapitulated Cdx2 expression in the trophectoderm or in other extra-embryonic regions.
Cdx2 expression persists throughout life in the postgastric intestinal mucosa. In this context, it appears to have acquired a role in the control of secretion of specific gut products. Thus it has been shown to bind to a variety of gene promoters, including those for lactase–phlorizin hydrolase , sucrase-isomaltase , carbonic anhydrase1  and calbindin-D9K . Another important action of Cdx2 is its tumour-suppressor function in the distal colon [27,28]. Cdx2+/− mice are more sensitive to the carcinogenic effect of azoxymethane treatment than their wild-type littermates and this is accompanied by a decrease in sensitivity to radiation-induced apoptosis. Aoki et al.  suggest that reduced expression of Cdx2 is important in colon tumorigenesis as a result of mTOR (mammalian target of rapamycin)-mediated chromosome instability.
Cdx1 begins to be expressed in the gut around mid-gestation when the stratified epithelium of the primitive gut begins to be converted into the more mature simple columnar lining. It has been suggested that Cdx2 is required for the establishment of Cdx1 expression , whereas the subsequent expression of Cdx1 may modulate levels of Cdx2 . As previously stated, Cdx1 knockout does not alter gut morphology, and this leads to the suggestion that it is not involved in intestinal development. However, Mutoh et al.  have shown that transgenic expression of Cdx1 in the stomach under their proton pump promoter produced similar effects to those observed with Cdx2. This leads to the conclusion that, as in the case of axial extension, some degree of redundancy exists among Cdx genes with respect to gut morphogenesis.
Cdx4 knockout in isolation produces, as mentioned above, no significant abnormal phenotype. However, when combined with heterozygous mutation of Cdx2, dramatic changes involving the caudal region are observed. The rectum ends as a blind sac and a fistula extends into a grossly dilated bladder outflow region. Additionally, the bladder outflow tract is not patent, although the development of the glans urethra re-establishes this in male embryos late in gestation. The mucosa of the rectum is, nevertheless, normal, and the atresia appears to develop around the area of transition from the columnar epithelium of the rectum to the stratified squamous lining of the anus. Thus there is no disturbance in intestinal histogenesis, and it is possible that the defects seen reflect severe disturbance of the AP ecto/meso-dermal morphogenesis at the caudal end of the embryo. Tam et al.  have demonstrated that gut endoderm, unlike ecto/meso-derm, extends posteriorly by backward growth from more anterior levels. Taken together with the observations by Gao et al.  that conditional Cdx2 knockout resulted in absence of the distal colon, this suggests that Cdx2/Cdx4 compound mutants also exhibit some degree of endodermal growth deficiency. In any case, it is clear that, once again, redundancy among Cdx genes exists with respect to cloacal development.
It seems that Cdx genes have a role in axial growth and positional information, in gut development and in trophoblast maturation. The action on trophoblast seems to be peculiar to Cdx2, although the possibility that other Cdx genes may compensate has not been explored. With respect to axial growth and positional information, it seems that the former is influenced by total Cdx ‘dosage’, whereas the latter is specific to the individual gene. There is little doubt that the effects are mediated, at least in part, through the Hox genes.
Cdx1 has no apparent effect on gut development; in contrast, the effect of Cdx2 on gut morphogenesis is profound. It seems that its action is to drive differentiation towards an intestinal phenotype at least as far as the midgut is concerned. In its absence, the ‘default’ gastric foregut phenotype is manifest. Knockout studies suggest that this effect is specific to Cdx2, but transgenic experiments in which either Cdx1 or Cdx2 is expressed in the stomach suggest that some degree of redundancy may exist. Hox genes are not widely expressed in the gut endoderm, and their orderly pattern of expression in the paraxial mesoderm is not reflected in gut mesoderm. For this reason, it is possible that the action of Cdx2 in determining mucosal type is not directly Hox-mediated. The endodermal–mesodermal cross-talk involving BarX1 indicates that other mechanisms are involved. The phenotype seen in compound Cdx2/Cdx4 mutants is probably largely the result of disturbed pelvic mesodermal development causing secondary effects on cloacal-derived organogenesis. However, the absence of a colon from conditionally inactivated Cdx2−/− mice suggests that there may also be a direct effect on endodermal growth.
The various effects resulting from expression of the Cdx genes is reflected in the complex nature of the promoter partially elucidated for Cdx2. Upstream regulation of the genes is complex and only partially understood. Wnt and Fgf genes are involved in both axial specification and in the gut; a large literature exists, but is beyond the remit of this brief review. With respect to Barrett's oesophagus, a key paper by Kazumori et al.  indicates that Cdx1 is strongly expressed in the oesophageal metaplastic tissue and that bile acids increase promoter activity in cultured oesophageal epithelial cells. This in turn appears to induce production of Cdx2, which, at least in the stomach, is sufficient to cause intestinal metaplasia.
Barrett's Metaplasia: A Biochemical Society Focused Meeting held at University of Bath, Bath, U.K., 17–18 September 2009. Organized and Edited by Mark Farrant (Bath, U.K.), Rebecca Fitzgerald (Cambridge, U.K.) and David Tosh (Bath, U.K.).
This study was supported by a project grant from the Association for International Cancer Research [grant number 08-0199].