Organisation of multicellular animals involves the action of genes that impart “positional information”. All vertebrates are built on a segmental pattern that is most obviously expressed by the appearance of somites during embryonic development.

A common feature of genes that impart individual identity (and therefore positional information) to specific segments is the possession of a “homeobox” DNA binding motif coding for a consensus sequence of 60–63 amino acids that acts as a transcriptional regulator of “downstream” genes. The most widely researched homeobox genes are the so-called homeotic selector genes of the Antp-type (the defining gene is named Antennapaedia). In the fruit fly Drosophila, these are situated on chromosome 3 as part of the HOM cluster. HOM-C genes are strongly conserved during evolution and in mammals have been replicated to appear on separate chromosomes in four paralogous complexes called Hox clusters. They are expressed principally in developing ectodermal and mesodermal tissues and in general terms are responsible for segmental specification of the dermatomes, musculoskeletal, and nervous systems.¹ However, Hox genes are not expressed in the greater part of the gut endoderm but in their place, mammalian members of the Para-Hox genes²—an “evolutionary sister” of the Hox clusters—seem to play an important role in gut patterning. Members of this group are Pdx1 which is required for the correct development of the pancreas and duodenum³ and three homologues of the Drosophila gene Caudal which in mammals are called Cdx1, Cdx2, and Cdx4. In addition to their own unique domains, the Cdx genes exhibit significant topographical overlap of expression during development, as well as in the adult, and it is reasonable to assume that in such areas a degree of compensation may occur in the event of single gene deficiency.

Cdx2 is expressed in the endoderm of the entire postgastric epithelium⁴ from the time of its initiation at the stage of hindgut invagination throughout development and adult life.⁵Cdx1 is not expressed in the early definitive gut endoderm but appears at postsomite stages just before transition of the multilayered intestinal endoderm to a single layered epithelium at 14 days post coitum (dpc).⁶Cdx4 is expressed in the earliest hindgut invagination⁷ but little is known of its distribution after 10 dpc.

It is important to note that in addition to their role in the gut, Cdx genes are active at multiple other sites during early development where, inter alia, they regulate the extent of ecto/mesodermal expression of Hox genes and thus also indirectly influence non-intestinal anteroposterior patterning.⁸ However, this review is principally concerned with their role in the gut.

Much insight into the function of a gene may be obtained by studying the morphology of mice in which it has been inactivated by homologous recombination.⁹ Knockout of the Cdx1 gene¹⁰ produced anterior homeotic shifts (that is, a posterior structure such as a vertebra or rib exhibiting the morphology of a more anterior element) of the axial skeleton but no gut abnormalities. Inactivation of Cdx2 on the other hand produced not only axial homeotic shifts in heterozygotes (one affected allele) but was also found to prevent trophoblast maturation and consequently blastocyst implantation if present in the homozygous (both alleles affected) form. Most importantly in the context of this commentary, Cdx2+/− embryos exhibited multiple polyps in the caecum and adjacent ileum and proximal colon (that is, in the midgut).^11,12 Histological examination showed the presence of forestomach epithelium in the substance of the polyps and this was interpreted as evidence of an anterior homeotic shift involving the gut endoderm analogous to the mesodermal shift involving the axial skeleton. In other words, disturbance of “positional information” has resulted in anterior structures (forestomach epithelium) occurring ectopically at more posterior sites.¹¹ While initially defined as adenomatous polyps, critical histological analysis together with studies on timed development of the polyps established an interesting process of so-called intercalary growth.¹³ It was concluded that the function of Cdx2 was to direct the “default state” forestomach endoderm towards a caudal phenotype and that lower levels of expression of Cdx2 in the developing distal intestine of Cdx2+/− heterozygotes lead to reversion to a more anterior phenotype (that is, to forestomach epithelium). Subsequently, intercalary growth around the gut lesions results in “filling in” of tissue types at the discontinuity between the ectopic gastric and surrounding colonic epithelia and an orderly succession of histologically normal epithelia characteristic of cardia, corpus, antrum, and small intestine, in that order, appeared all around the forestomach epithelium at the junction with colonic epithelium. The molecular basis of the intercalary growth is unclear but probably involves local intercellular communication.

Clonal analysis experiments have been performed to establish the origin of the secondarily generated tissue between ectopic forestomach and colonic epithelia.¹⁴ Y chromosome painting in wild-type/Cdx2−/− chimaeric* mice of opposite sex indicated that once differentiated to express Cdx2, host colonic epithelium can only contribute small intestinal epithelium to the secondarily generated tissue while the Cdx2 mutant chimaeric cells give rise to the succession of gastric-type tissue but never to small intestinal morphology. These findings have interesting implications in the context of intestinal regeneration.

No information concerning the role of Cdx4 in gut development is currently available as knockout or RNAi knockdown experiments have not been performed and no spontaneous mutants are available.

Having shown that Cdx2 is necessary for the establishment of the midgut phenotype, we must now ask whether it is sufficient to convert the “default” stomach phenotype to that of colon. The question has been investigated by making transgenic mice in which Cdx2 is expressed in the stomach (which in wild-type mice is Cdx2 negative). This is achieved by introducing a Cdx2 expressing “transgene” into the genome of wild-type mouse embryos under the control of a promoter that specifically causes expression of Cdx2 in the stomach. The answer appears to be “yes” as Silberg and colleagues,¹⁵ using cis regulatory elements of the Hnf3γ promoter to drive ectopic expression of Cdx2 in the stomach, described the presence of alcian blue positive intestinal-type goblet cells in the gastric mucosa of transgenic mice as well as induced expression of intestine specific genes in the relevant areas. Interestingly, Mutoh and colleagues¹⁶ produced similar results using the promoter of the non-catalytic β-subunit gene of rat H⁺/K⁺ ATPase to drive the Cdx2 transgene. This group described complete replacement of gastric mucosal cells in the body of the stomach at postnatal day 37 by goblet cells, enteroendocrine cells, and absorptive cells expressing alkaline phosphatase together with the establishment of the proliferative zone at the base of the glands rather than at the isthmus. As parietal cell function is established postnatally, it is reasonable to assume that the Cdx2 transgene was expressed in differentiated cells rather than in stem cells; the H⁺/K⁺ ATPase promoter being activated with the onset of proton pump function. This suggests that either some parietal cells retain characteristics of stem cells—unlikely but possible as some parietal cells high up in the gastric glands may not have undergone a terminal differentiative event and may retain proliferative potential¹⁷—or else Cdx2 expression directly or indirectly causes “dedifferentiation” of parietal cells and their establishment as intestinal stem cells in the basal region of the mucosa whence they have migrated. This in turn might cause the overlying normal gastric mucosa to disappear in favour of the newly established intestinal phenotype. Another possibility is that the protein pump promoter used may be active at low levels in cells other than fully differentiated oxyntic cells expressing the proton pump. Clearly, these are speculative suggestions that require rigorous investigation.

Subsequently, Mutoh and colleagues¹⁸ have expressed a Cdx1 transgene in the mouse stomach, again driven by the H⁺/K⁺ ATPase β-subunit promoter, and their study is described in this issue of Gut(see page 1416). Surprisingly perhaps, they again demonstrated intestinal metaplasia similar in many but not all respects to that seen with Cdx2. As Cdx1 knockout has no apparent effect on differentiation of the midgut,¹⁰ the transdifferentiating effect of Cdx1 transgene expression in the stomach is somewhat unexpected. A possible explanation may lie in the overlap of function between cad homologues although clearly Cdx1 does not compensate for Cdx2 in the knockout model of the latter.¹¹ The overlap of effects may only be partial and this would explain differences in proliferating cell distribution and distribution of Paneth cells that exist between the two transgenic models.

Mammalian cad homologues and in particular Cdx2 are multifunctional genes and the picture presented above is simplistic. In the mature gut, Cdx1 is expressed principally in intestinal crypts while Cdx2 is demonstrable in cells clothing the villi where it regulates production of many gut enzymes, such as lactase-phlorizin hydrolase.¹⁹ Apart from its importance in pattern formation during the development of the gut, Cdx2 in conjunction with Cdx1 may well contribute to the balance between differentiation and cell renewal in the mature intestine and recently it has been suggested that in this respect it is controlled by PTEN/phosphatidylinositol 3 kinase signalling and tumour necrosis factor α signalling via nuclear factor κB dependent pathways.²⁰

Furthermore, Cdx2 functions as a tumour suppressor in the adult mouse.^21,22 In one study, Cdx2+/− mice developed adenocarcinomata in the distal colon in response to low doses of azoxymethane compared with a significantly smaller effect of the same dose on wild-type controls. The tumours differed histologically as well as in geographical situation from non-neoplastic heterotopic gastric polyps found in the pericaecal region of these animals.²¹ Another group used Apc mutant/Cdx2+/− compound mice. They found that levels of both Apc and Cdx2 were significantly lower in the distal colon and this was correlated with the development of adenomatous polyps in the distal colon whereas these lesions were predominantly present in the small intestine in mice with Apc lesions only.²² The authors conclude that reduced Cdx2 expression is important in colonic tumorigenesis. Both studies reported lower apoptotic rates in the colonic mucosa thus allowing greater survival of azoxymethane compromised cells or of cells with loss of heterozygosity at Apc.

Clearly, much remains to be done before the role of cad homologues in mammalian gut development and function is fully elucidated. It is certain that they are critically important during development in defining gut pattern and that in the adult they contribute to the complex mechanisms of cell turnover and phenotypic differentiation of stem cells. Future challenges include elucidation of upstream and downstream genes and the possibility of modifying gene expression for therapeutic purposes.