|Home | About | Journals | Submit | Contact Us | Français|
We demonstrate that conditional ablation of the homeobox transcription factor Cdx2 from early endoderm results in the replacement of the posterior intestinal epithelium with keratinocytes, a dramatic cell fate conversion caused by ectopic activation of the foregut/esophageal differentiation program. This anterior homeotic transformation is first evident in the early embryonic Cdx2-deficient gut as expression of several key foregut endoderm regulators was shifted caudally. While the intestinal transcriptome was severely affected, Cdx2-deficiency only transiently modified selected posterior Hox genes and the primary enteric Hox code was maintained. Further, we demonstrate that Cdx2-directed intestinal cell fate adoption plays an important role in the establishment of normal epithelial-mesenchymal interactions, as multiple signaling pathways involved in this process were severely affected. We conclude that Cdx2 controls important aspects of intestinal identity and development, and that this function is largely independent of the enteric Hox code.
The mouse endoderm transforms from a two-dimensional epithelial sheet into the primitive gut tube at embryonic day 8.5–9.0 (E8.5–9.0). Subsequent morphological differentiation converts the pseudostratified endoderm layer into a tall columnar epithelium which lines the respiratory and gastrointestinal tracts (Wells and Melton, 1999). The primitive gut appears homogeneous from end to end, with distinct anterior-posterior (AP) regions adopting different fates in subsequent organogenesis. In the gastrointestinal tract, the foregut gives rise to the epithelia of esophagus, stomach, and duodenum, while midgut and hindgut become the small intestine, and the cecum and colon, respectively. Cross-talk between gut mesoderm and endoderm progressively commits the primary endoderm to specific fates (Grapin-Botton and Melton, 2000). Mutations in a number of Hox genes result in malformations in certain gut regions, but do not cause whole-sale AP transformation of the gut (Aubin et al., 1997; Boulet and Capecchi, 1996; Manley and Capecchi, 1995; Warot et al., 1997; Zacchetti et al., 2007), even though these Hox factors play important roles in patterning the mesoderm and neural tube (Deschamps et al., 1999; Krumlauf, 1994; McGinnis and Krumlauf, 1992).
AP asymmetry of the gut endoderm is evident with the onset of Cdx2 expression in the hindgut at its inception (Beck et al., 1995). Cdx2 is the mouse homologue of AmphiCdx in Amphioxus and caudal in Drosophila. It resides in the “ParaHox” gene cluster believed to have evolved from a “ProtoHox” cluster that gave rise to the definitive Hox gene clusters (Brooke et al., 1998). In Drosophila, caudal specifies posterior body segments (Macdonald and Struhl, 1986; Mlodzik et al., 1985; Moreno and Morata, 1999). Morpholino knockdown and overexpression studies in Zebrafish indicated essential roles of caudal orthologues in neural tube and intestinal development (Cheng et al., 2008; Flores et al., 2008; Shimizu et al., 2006; Skromne et al., 2007). Three mouse homologues, Cdx1, Cdx2 and Cdx4 (Duprey et al., 1988; Gamer and Wright, 1993; James and Kazenwadel, 1991), participate in the patterning of the vertebral column (Chawengsaksophak et al., 1997; Subramanian et al., 1995; van Nes et al., 2006) and in embryonic hematopoiesis (Wang et al., 2008), however, their role in endoderm development is less clear. Homozygous Cdx1 or Cdx4 mutants do not display gut defects (Subramanian et al., 1995; van Nes et al., 2006), while Cdx2 null mutants die in utero before the onset of endoderm development (Chawengsaksophak et al., 1997; Tamai et al., 1999).
In the mouse embryo, Cdx2 is expressed in nuclei of cells derived from the late-dividing blastomere, a precursor of trophectoderm (Deb et al., 2006). From E8.5 onward, Cdx2 is activated in the embryo proper, predominantly the posterior gut (Beck et al., 1995). Cdx2 expression subsequently becomes restricted to the intestinal epithelium, with a sharp anterior boundary marking the transition from stomach to duodenum (James et al., 1994; Silberg et al., 2000). Genetic analysis of Cdx2 function in mammalian intestinal development has been limited to Cdx2 heterozygous mice that form multiple colonic polyps (Chawengsaksophak et al., 1997). These polyps contain areas of squamous metaplasia in which the expression of the remaining wild type Cdx2 allele is extinguished through an unknown mechanism (Beck et al., 1999). However, it is still unclear as to: a. what transcriptional programs were altered in these epimorphic lesions; b. how comprehensive the impact of Cdx2-disruption is for cell fate determination; c. through which mechanisms loss of Cdx2 induces squamous metaplasia; d. through which mechanisms Cdx2 promotes intestinal differentiation; and e. what role Cdx2 plays in intestinal epithelial-mesenchymal interactions.
We have previously shown that ectopic expression of Cdx2 in the gastric epithelium induces intestinal metaplasia (Silberg et al., 2002), an example of a posterior homeotic transformation. Here, we demonstrate that Cdx2 is essential for the initial expression and/or subsequent maintenance of a group of pro-intestinal transcription factors, including Cdx1, Isx, HNF1α and HNF4α , which together activate the intestinal transcriptome. The expression of Cdx2 in the posterior gut epithelium antagonizes the foregut differentiation program, which becomes ectopically activated upon Cdx2-disruption, resulting in dramatic cell fate conversion. We further demonstrate that intestinal cell fate establishment by Cdx2 plays a critical role in instructing normal epithelial and mesenchymal interactions, in particular with respect to the integrity of Wnt and Hedgehog signaling.
Cdx2 null mice die before gastrulation (Chawengsaksophak et al., 1997; Tamai et al., 1999). Therefore, we derived a conditional Cdx2 allele to study its role in the gut endoderm(Suppl. Fig. 1A). Correctly targeted embryonic stem cell clones were identified by Southern blot analysis (Suppl. Fig. 1B). After germ line transmission of the targeted allele, the FRT-flanked neomycin resistance gene was removed by crossing to Flp1 deleter mice (Rodriguez et al., 2000). Cdx2loxP/+ mice were then intercrossed, resulting in Cdx2loxP/loxP mice that were viable and fertile (Suppl. Fig. 1C), confirming that the Cdx2loxP allele is functionally wild type. Subsequent Cre-mediated gene ablation results in a null allele that lacks the homeobox domain. To ablate Cdx2 conditionally in the developing gut, we bred Cdx2loxP/+ mice to Foxa3Cre mice (Lee et al., 2005), which delete loxP flanked targets in early endoderm. Using the Rosa26R reporter line, we verified Cre activity in the primitive gut of E9.5 embryos, prior to the onset of gross morphological defects (Fig. 1A, B). Efficient deletion of Cdx2 from mutant (Cdx2loxP/loxP,Foxa3Cre+) gut epithelia was evident with immunohistochemistry using an anti-Cdx2 antibody (Fig. 1C, D). The expression of Foxa1, a pan-endoderm marker, was unaffected (Fig. 1E, F). Examination of mutant embryos at mid and late gestation revealed equal efficiency of Cdx2 ablation throughout the intestinal domain (Fig. 1I–N and Suppl. Fig. 2).
Although the mutant pups were born alive, they did not survive beyond postnatal day one (P1). We examined the gastrointestinal tract of mutant embryos at various developmental stages. The gross abnormalities of the mutant posterior gut region first became evident at E12.5 (Fig. 1G). In contrast to the control intestinal tract that ends with colon and rectum, the mutant intestine developed an abnormal distal structure that terminates in a blind-ended sac (Fig. 1G, arrow). Progressive defects in elongation of the mutant intestine began to appear at E14.5 (Suppl. Fig. 3A) and the mutant gastrointestinal tract developed a malformed cecum at the distal end, with no colon (Fig. 1H, arrow). Cross sections of the E14.5 mutant distal intestine revealed a dilated gut lumen (Fig. 1M–N). All mutant animals examined from E13 to P0 (n=56) demonstrated an absence of the colon, a phenotype reminiscent of the most severe cases of colonic atresia in humans (Etensel et al., 2005; Lau and Caty, 2006). The mutant duodenum was progressively distended and became translucent, likely due to fluid retention caused by distal obstruction (Fig. 2A–B). By E18.5, the duodenum was further dilated with 5–7 fold increase in diameter compared to the control tissue (Fig. 2C–D, D′). Thus, Cdx2-deficiency prevents colon formation and leads to complete intestinal obstruction.
While the mutant proximal and medial intestinal epithelia appeared less organized than the control epithelia, the overall histology at E14.5 differed only subtly, as both mutant and control gut epithelia appeared pseudostratified (Fig. 1I–N). However, defects in differentiation became more apparent later in development. Since the mutant animals die at P1, before the development of Paneth cells, we examined the differentiation of enterocytes (Fig. 2E–F, I–J), goblet cells (Fig. 2G–H, K–N), and enteroendocrine cells (Fig. 2M–N) at different stages using specific markers and found terminal differentiation severely impaired. Instead, the mutant posterior intestinal epithelium expressed a basal epithelial cell marker p63 from E15.5 (Fig. 2H).
Villus hypoplasia was detected from E16.5 throughout the mutant intestinal domain as compared to controls (Fig. 3A–B). Position-matched longitudinal histological sections of E18.5 control and mutant intestines revealed dramatic reductions of intestinal villi (Fig. 3C–J), with severity increasing from anterior to posterior(Suppl. Fig. 3B): The mutant duodenum contained villus-like epithelial foldings (Fig. 3D) that were significantly stunted and broadened (Fig. 3F, and Suppl. Fig. 3C), while the cuboidal epithelia of mutant jejunum and ileum were completely replaced with a flattened epithelium (Fig. 3H, J), and the mutant ileum and cecum lacked villi entirely (Fig. 3J, and Suppl. Fig. 2F for cecum). Intestinal epithelia containing mosaic Cdx2-deletion were observed in a few mutant embryos. Segments of Cdx2-positive epithelium were contiguous to Cdx2-deficient regions (Fig. 3K). Interestingly, cells that retained Cdx2 expression were capable to form villi and elaborate goblet cells normally (Fig. 3L), while adjacent Cdx2-deficient cells failed to do so (Fig. 3M). Thus, Cdx2 is required for initiation of intestinal differentiation and morphogenesis in a cell-autonomous fashion.
Ki67 staining, which marks transit amplifying cells, revealed an expanded proliferative compartment in the mutant duodenal epithelium (Fig. 3N, O). The proliferative index of the mutant duodenal epithelial cells, assayed by BrdU incorporation, was significantly increased in the mutant epithelium following either 1-hr or 24-hr labeling (Suppl. Fig. 4A–C). Interestingly, even after a short labeling period (40–60 min), more than 20% of BrdU+ mutant cells were located at or above position 9 relative to the bottom of the nascent crypts (Fig. 3Q, R, and Suppl. Fig. 4D), while BrdU+ control cells were restricted to the inter-villus space even after 24-hour BrdU labeling (Fig. 3P, R and Suppl. Fig. 4E). BrdU incorporation revealed a continuous proliferative cell layer across the mutant epithelial sheet in the posterior intestine (Suppl. Fig. 5B). We did not, however, detect a significant increase in the apoptotic rate in Cdx2 mutants by either cleaved caspase-3 or TUNEL staining (Suppl. Fig. 5D). Thus, the failure to specify the colon was not caused by a lack of proliferative capacity or enhanced cell death in Cdx2 mutant mice. The loss of terminal differentiation discussed above, and our failure to observe gastric glandular epithelial cell types using specific antibodies (not shown), suggest that the proliferative pattern of the mutant epithelium resembles that of early embryonic stages prior to intestinal differentiation.
To gain insight into the identity of the Cdx2 mutant epithelial cells, especially those in the posterior intestinal epithelium, we examined their ultrastructural features using transmission electron microscopy. Cdx2-deficient epithelial cells failed to develop the brush border typical of enterocytes (Fig. 4A–D). Examination of the posterior intestine revealed multiple layers of flattened epithelial cells, with the axes of the nuclei parallel to the luminal surface (Fig. 4F). Unexpectedly, the mutant posterior intestinal epithelial cells contained abundant tonofilaments (Fig. 4G). Tonofilaments are typical of squamous epithelial cells and are frequently seen in the desmosomal junctions of keratinocytes (Fig. 4H), which contribute to stratified esophageal epithelia but are extremely rare in the normal intestine (Fig. 4E).
Squamous differentiation has been reported in colorectal adenoma(Ouban et al., 2002). To verify whether the Cdx2-deficient posterior intestine has molecular features of squamous epithelia, we performed immunohistochemistry for keratin 13 and p63, markers of the suprabasal and basal squamous epithelial cells in mouse esophagus, respectively (Fig. 5A, B). In contrast to the control ileum where neither gene was expressed (Fig. 5C–D), the Cdx2 mutant epithelium was positive for both markers (Fig. 5E–G). Neither keratin 13 nor p63 was expressed in E10.5 wild type midgut or hindgut endoderm (Suppl. Fig. 6C, D). At E12.5, p63 expression was detected in foregut endoderm cells fated to become forestomach and pharynx (Suppl. Fig. 6E, F). Likewise, a marker of the anterior foregut endoderm, Sox2 (Que et al., 2007), was detected in the Cdx2-deficient ileum at an expression level equivalent to that of normal esophagus (Fig. 5H), confirming that the Cdx2-deficient posterior intestine was indeed anteriorized. These data indicate that the expression of squamous markers in mutant prospective intestine was not due to a developmental delay, but rather due to an ectopically activated foregut differentiation program.
We next performed gene expression profiling using RNA samples extracted from total E18.5 control and mutant ileum as well as normal esophagus. The morphological abnormalities of the mutant ileum precluded separation of the epithelium from mesenchyme. Of the 11,738 significantly changed genes, 268 genes demonstrated fold-changes above 50-fold compared to control ileum. Hierarchical clustering showed that the transcriptome of the mutant ileum resembled that of esophagus far more than that of normal ileum (Suppl. Fig. 7A). The similarity between the mutant ileum and control esophagus is highlighted by a heat map assembled from differentially expressed genes (Fig. 5I). Consistent with the morphological transformation, virtually all intestine-specific genes were downregulated in the mutant ileum (Fig. 5J, Suppl. Table 1).
Next, we compared our microarray results with several previous intestine gene profiling studies (Bates et al., 2002; Li et al., 2007; Schroder et al., 2006). Among genes that show significant enrichment in E18.5 intestinal epithelium over mesenchyme (Li et al., 2007), 35.3% were significantly altered in Cdx2-deficient intestine. Likewise, 39.8% of genes that show enrichment in intestine over stomach (Bates et al., 2002) were significantly affected in our model. Furthermore, genes with highly specific expression patterns in the differentiated intestinal epithelium (Schroder et al., 2006) were all significantly downregulated in Cdx2-deficient mice.
In contrast, many genes involved in keratinocyte differentiation were significantly upregulated in the Cdx2 mutant ileum (Fig. 5K, Suppl. Table 2; Suppl. Fig. 7B). Notably, nine out of the ten most highly upregulated genes in the mutant ileum were enriched in normal esophageal epithelium (Suppl. Table 2). Most of these genes, such as high molecular weight keratins, keratin 5 and 13, small proline-rich protein 3, calmodulin-like 3, cornifelin, plakophilin 1, and dermokine, play important roles in keratinocyte cell envelope formation and desmosome assembly. Thus, gene profiling analysis confirmed the transformation of Cdx2-deficient prospective intestinal endoderm domain to an esophageal cell fate.
To gain a mechanistic insight to the cell fate switch in Cdx2-deficient posterior intestine, we analyzed transcriptional regulators known to be crucial in regulating intestinal differentiation. Along with the striking decrease of Cdx2 mRNA itself, several intestine-enriched transcription factors, including HNF1α , HNF4α , Isx and Cdx1, were dramatically reduced in expression in the mutant intestine (Suppl. Table 1). Next, we performed quantitative reverse transcriptase PCR (Q-RTPCR) analysis on developing gastrointestinal tracts at earlier embryonic time points. A significant reduction in mRNA levels of HNF1α , HNF4α and Isx was already evident in the E12.5 mutant gut (Fig. 6B–D). At E14.5, expression of all these factors as well as Cdx1 was significantly reduced in both proximal and distal intestine (Fig. 6A–D). These data indicate that the reduced expression of these transcription factors at E18.5 was not due to a secondary effect of abnormal development, but due to an impairment of the initial activation of these genes.
The expression of Math1, a basic helix-loop-helix transcription factor that plays a role in the differentiation of intestinal secretory cell types (Yang et al., 2001), was significantly reduced in mutant distal intestines from E12.5 onward (Fig. 6E). Conversely, qRT-PCR of Sox2 and Pax9 demonstrated that these foregut-enriched genes were ectopically activated in the mutant posterior intestine as early as E12.5, at a level equivalent to the stomach (Fig. 6G, H). Activation of Sox2 was detected even in the anterior mutant intestines (Fig. 6G), strongly supporting an early anteriorization event that subsequently drives the ectopic activation of foregut transcriptional program in the mutant gut (Fig. 6J).
qRT-PCR confirmed the decrease of Indian Hedgehog (Ihh) expression and the dramatic activation of Wnt10a in E14.5 mutants (Fig. 6F, I). In the normal gastrointestinal tract, Wnt10a expression is excluded from the intestinal domain from E12.5 (Fig. 6I). These changes in the expression pattern of signaling molecules likely reflect a consequence of early cell fate transition, resulting from the altered transcriptional program in the Cdx2-deficient intestine. We also confirmed the changes of Cdx1, HNF1α , HNF4α , Sox2 and Wnt10 expression at the protein levels using E18.5 tissue lysates (Fig. 5H and Fig. 6K).
To investigate whether Cdx2 directly regulates the expression of HNF1α , Cdx1 and HNF4α in the embryonic intestines, we first examined the regulatory sequences of HNF1α , HNF4α and Cdx1. Among multiple Cdx binding sites within the 5′ upstream region of HNF1α , those located near the transcription initiation site were most conserved from Xenopus to human (Suppl. Fig. 8A). Less conserved Cdx binding sites were identified in the HNF4α and Cdx1 5′ upstream sequence (Suppl. Fig. 8B, C). These Cdx sites are occupied by Cdx2 in vivo, as demonstrated by chromatin immunoprecipitation (ChIP) (Fig. 6L).
In a number of non-endoderm tissues, Cdx factors exert their developmental effect via regulating Hox transcription factors (Charite et al., 1998; Shimizu et al., 2006; Subramanian et al., 1995; Wang et al., 2008), which are key players in the primary AP patterning process of the vertebrate embryo (Krumlauf, 1994). Overexpression or inactivation of specific Hox genes has been shown to affect gastrointestinal development (Aubin et al., 1997; Boulet and Capecchi, 1996; Kondo et al., 1996; Pollock et al., 1992; Wolgemuth et al., 1989), while a cluster of Hoxd genes controls the formation of the ileo-cecal sphincter (Zakany and Duboule, 1999). Our microarray data indicated that a number of intestine-enriched Hox genes, including Hoxa5, Hoxb5, Hoxb6, Hoxa7 and Hoxb7, continue to be expressed in the mutant ileum at a level similar to controls (Suppl. Fig. 9). However, Hoxc9, a gene expressed in the posterior midgut and hindgut (Grapin-Botton and Melton, 2000; Roberts, 2000), was decreased 6.2-fold in the mutant ileum (Suppl. Fig. 9B).
Next we examined the AP distribution of representative Hox genes in the early gut where their expression patterns have well been documented (Choi et al., 2006; Grapin-Botton and Melton, 2000; Pitera et al., 1999; Roberts, 2000). Levels of Hoxa3, Hoxb3, Hoxb4, Hoxc4, Hoxd4, Hoxb5, Hoxc5, Hoxa7 and Hoxb7 mRNA were not significantly changed in the Cdx2 mutant (Fig. 7A–H, Suppl. Fig. 10A). In contrast, at E12.5, Hoxc8, Hoxb9, Hoxc9, Hoxa13 and Hoxd13 mRNA levels were significantly lower in the mutant posterior intestine (Fig. 7I–L, Suppl. Fig. 10B). At E14.5, however, most of these posterior Hox genes had recovered to match the levels of the control intestine (Fig. 7I–L, Suppl. Fig. 10B).
Most Hox genes analyzed are expressed in the gut mesenchyme (Li et al., 2007), while some, such as Hoxa3, Hoxb4, Hoxc5, Hoxb9, Hoxc9, Hoxa13 and Hoxd13, are also active in the epithelium (Grapin-Botton and Melton, 2000; Roberts, 2000). Our results demonstrate that Cdx2-deficiency in the early gut endoderm transiently modifies the expression of selected posterior Hox genes, but had no impact on anterior Hox genes. While the posterior Cdx2-deficient gut was anteriorized as early as E12.5 (Fig. 6G–J), the maintained expression of Hoxc9, Hoxa13 and Hoxd13 in this domain (Fig. 7J–L) indicates that the mutant gut had retained its primary enteric Hox code.
In addition, the Cdx2-deficient gut demonstrated normal AP expression of Pdx1, a second “Parahox” gene, which remained restricted to the duodenum even in the absence of Cdx2 (Suppl Fig. 10E, F). Furthermore, the expression of Barx1, a stomach specific mesenchymal transcription factor (Kim et al., 2005), also maintained its expression domain in Cdx2 mutant embryos (Suppl. Fig. 10C). These data further support the notion that the Cdx2-deficient gut retained certain AP values.
We found dysregulation of Wnt ligand expression in Cdx2-deficient intestine. In addition to the ectopic activation of Wnt10a as a result of anteriorization of the mutant ileum (Fig. 6I, K), multiple other Wnts, as well as the Wnt target genes CD44, cyclin D1, Sox9 and the Tcf factors, were significantly upregulated in the Cdx2-deficient intestine (Fig. 8A, B; Suppl. Fig. 11A–D). In contrast to Wnt, expression of Ihh and Sonic hedgehog (Shh) was significantly reduced in the Cdx2-deficient ileum (Fig. 8A), consistent with the decreased expression of Hedgehog-interacting protein (Hhip), a primary hedgehog target expressed by the intestinal mesenchyme (Li et al., 2007). The severely expanded smooth muscle layer we observed in the mutant duodenum (Fig. 3F) and jejunum (Fig. 8C–F) may reflect the decreased hedgehog signaling activity, as inhibition of Hedgehog signaling in the intestine causes smooth muscle expansion (Madison et al., 2005).
The expression of desmin, a marker of smooth muscle progenitors but not by myofibroblasts (Adegboyega et al., 2002), was increased 6.2-fold in the Cdx2-deficient intestine and accompanied with a significant decrease of several myosin genes (Suppl. Fig. 11E), suggesting an altered myogenic process and terminal differentiation of smooth muscle cells in the Cdx2 mutant gut. The myosin gene expression profile in the Cdx2 mutant ileum highly resembled that of wild-type esophagus (Suppl. Fig. 11E), illustrating a potent epithelial-to-mesenchymal regulatory role that affected the smooth muscle differentiation.
The Cdx2-deficient gut displays severe hindgut abnormalities with a failure of colon development and a complete terminal blockage. Partial or complete colonic atresia has been reported as a human congenital disorder (Etensel et al., 2005). Mutations in PDX1, a neighboring “Parahox” gene, cause pancreatic agenesis in humans (Stoffers et al., 1997). Our findings support the notion that the Parahox genes specify regional identities to the vertebrate gut, and suggest further that mutations in CDX2 or its targets could contribute to colonic atresia in human.
The expression domains of multiple important foregut regulators including Sox2, Pax9 (Grapin-Botton and Melton, 2000), p63 (Glickman et al., 2001), were dramatically extended towards the posterior of the Cdx2 mutant gut tube (Fig. 8G). None of the previously reported mutant mice had such a dramatic impact on AP patterning of the gut (Aubin et al., 1997; Boulet and Capecchi, 1996; Manley and Capecchi, 1995; Warot et al., 1997; Zacchetti et al., 2007). When a cluster of Hoxd genes was deleted, no dramatic disruption of intestinal identity was observed, except that induction of the cecum was affected (Zacchetti et al., 2007). The cecum was correctly induced in Cdx2 mutant embryos, consistent with the observation that the primary enteric Hox code was maintained.
Though Cdx factors have been proposed to function via regulation of Hox gene expression in several non-endoderm tissues (Charite et al., 1998; Subramanian et al., 1995; Wang et al., 2008), the expression of anteriorly localized intestinal Hox genes was independent of Cdx2. Cdx2-deficiency transiently delayed the expression of several posterior intestinal Hox genes at early embryonic stages; however, these genes maintained their relative AP position. The regulation of posterior Hox genes by Cdx factors has been reported in Zebrafish hindbrain (Shimizu et al., 2006), however, functional rescue by downstream Hox factors remains controversial (Skromne et al., 2007). Our findings indicate that Cdx2-deficiency does not profoundly influence the primary enteric Hox code.
Cdx1, whose gut expression pattern resembles that of Cdx2 (Silberg et al., 2000), has the capability to drive intestinal differentiation in a gain-of-function setting (Mutoh et al., 2004). Redundancy between all three Cdx proteins has been reported in a number of non-endoderm tissues (van den Akker et al., 2002; van Nes et al., 2006; Wang et al., 2008). Therefore, it was surprising to see the near-complete homeotic transformation of the Cdx2-deficient intestine, as some compensation was anticipated. We established that Cdx1 activation is directly dependent on Cdx2. This transcriptional hierarchy between the two Cdx genes reflects their sequential expression pattern in the gut endoderm, where Cdx2 precedes Cdx1 by a few days (Hu et al., 1993; Meyer and Gruss, 1993; Silberg et al., 2000). In fact, the expression of Cdx1 starts only when villus morphogenesis and epithelial maturation begin (Hu et al., 1993). Our data provide further evidence for the evolutionary significance of the “Parahox” cluster, where Cdx2, but not Cdx1 or Cdx4, is located. Thus, Cdx1 is controlled by the more ancient caudal orthologue Cdx2 in gut endoderm to facilitate the developmental and anatomical complexity of the organ.
Similar to Cdx1, Isx is another intestine-specific transcription factor whose expression initiates during epithelial differentiation (Choi et al., 2006), consistent with its dependency on Cdx2. In addition, the maintenance of HNF1α and HNF4α expression in the embryonic intestine is directly controlled by Cdx2. Single-gene ablation of Cdx1, Isx, Hnf1α or Hnf4α in mice had no effect on the establishment of the intestinal epithelium (Choi et al., 2006; Garrison et al., 2006; Lee et al., 1998; Pontoglio et al., 1996; Shih et al., 2001; Subramanian et al., 1995). Nevertheless, Cdx1 (Mutoh et al., 2004), Isx (Choi et al., 2006), HNF1α (Martin et al., 2000) and HNF4α (Garrison et al., 2006) regulate the expression of numerous intestinal genes. Our data support the notion that Cdx2 functions upstream of a group of pro-intestinal transcription factors, with which it synergizes to promote intestinal cell fate (Fig. 8G).
Our conditional Cdx2 mutants recapitulate the finding of squamous metaplasia in Cdx2+/− mouse colonic lesions, where Cdx2 expression was diminished (Beck et al., 1999). When Cdx2 was removed in our model, anteriorization was first evident with the caudal extension the of Sox2 and Pax9 expression domains (Fig. 8G). This was followed by squamous differentiation around E14.5–15.5, leading to genome-wide activation of esophageal genes. Due to the lack of pro-intestinal regulators, the prospective intestinal epithelial domain was replaced by keratinocytes. These data provide molecular evidence that Cdx2 normally represses a foregut differentiation program in the posterior gut, explaining the epimorphic changes observed previously (Beck et al., 1999).
We demonstrate that the normal gastrointestinal expression domain of Wnt10a is opposite to that of Cdx2, mimicking the expression pattern of Sox2. Cdx2-deficiency led to ectopic activation of Wnt10a expression in the caudal intestine, possibly as a consequence of the ectopically differentiated squamous cells. Recent findings suggest that ectodermal dysplasia in humans is associated with Wnt10a mutations (Adaimy et al., 2007), while mis-regulation of Wnt10a was found in gastrointestinal cancer (Kirikoshi et al., 2001). We speculate that this gene may involve in keratinocyte differentiation during upper gastrointestinal development. In summary, we have identified Cdx2 as a master regulator in the posterior endoderm, demonstrating that this gene is essential for the establishment of intestinal identity.
Hematoxylin, eosin, and alcian blue staining was performed in the Morphology Core of the Penn Center for Molecular Studies in Digestive and Liver Diseases. Alkaline phosphatase staining was performed using the Vector Red Alkaline Phosphatase Substrate Kit I (Vector Laboratory, SK-5100). The ABC detection system (Vector Laboratory, PK-6100) was used for immunohistochemistry. Cy2- and Cy3-conjugated fluorescent secondary antibodies were purchased from Jackson Laboratory.
For quantification of intestinal length as well as villi number, length and width, images of matched control and mutant intestines were analyzed using ImageJ software (NIH). For cell number quantification, BrdU+ cells within 50 continuous villi were manually counted from three slides of control and mutant intestines. The percentage of BrdU+ cells at each cell position was calculated with the cell located at the bottom of the inter-villus pocket designated as position 0.
Fresh intestinal tissues were washed with PBS and suspended in a fixative solution of 2.5% cacodylate-buffered glutaraldehyde and 4% paraformaldehyde (pH 7.4) for 6 hours. Tissues were rinsed in a cacodylate-buffered solution, post-fixed with 2% cacodylate-buffered OsO4 dehydrated with graded ethanol, clarified in propylene oxide and embedded in Epon. Seventy nm thin sections were obtained with a Leica UCT ultramicrotome using a Diatome diamond knife and placed on 200 mesh copper grids. Sections were stained with an alcoholic solution of uranyl acetate, followed by a solution of bismuth subnitrite. These sections were examined under a JEOL JEM1010 electron microscope and digital images were captured using AMT Advantage HR aided Hamamatsu CCD camera. All EM supplies were purchased from Electron Microscopy Sciences, Fort Washington, PA, and Ted Pella, Redding, CA.
Fresh intestinal tissue lysates were prepared in lysis buffer containing 50 mM Tris (pH 7.5), 150 mM NaCl, 10 mM EDTA, 0.02% NaN3, 50 mM NaF, 1 mM Na3VO4, 1% NP40, 1 mM PMSF, and protease inhibitors (Sigma), from E18.5 mouse intestines. 15 ug total lysates were heated at 70°C for 10 min in 4× LDS buffer (Invitrogen), and loaded on 4%–12% SDS-PAGE (Invitrogen). Proteins were transferred to PVDF membranes (Invitrogen). Membranes were stripped in Western stripping buffer (Pierce) and reprobed sequentially with corresponding antibodies.
E16.5 control and mutant intestinal tissues were finely minced into small pieces followed by 10 min cross-linking with 1% formaldehyde at 37°C and subjected to chromatin purification and immunoprecipitation as previously described (Rubins et al., 2005) using anti-Cdx2 antibodies (Funakoshi et al., 2008). Input chromatin and ChIP DNA were used as template in quantitative genomic PCR using a MX3000 PCR machine (Stratagene). The 28S ribosomal genes were used in as internal reference.
One centimeter of the ileum, immediately above the cecum, was dissected from three E18.5 mutant and three control embryos. 250 ng of total RNA was amplified and labeled with Cy3 using the Low Linear Amplification Kit (Agilent Technologies, CA). This labeling reaction produced 1.75 – 2.0 μg of Cy3-labeled cRNA (anti-sense), by first converting mRNA primed with an oligo (d)T-T7 primer into dsDNA with MMLV-RT and then amplifying the sample using T7 RNA Polymerase in the presence of Cy3-CTP. After purification, 1.65 μg of cRNA was fragmented and hybridized to the Whole Mouse Genome Oligo Microarray (G4122A; Agilent Technologies, CA) array for 17 hr. at 65°C.
Microarray slides were washed and scanned with an Agilent G2565BA Microarray Scanner. Images were analyzed with Feature Extraction 9.5 (Agilent Technologies, CA). Mean foreground intensities were obtained for each spot and imported into the mathematical software package “R”. The Cy3 (green) intensities were corrected for the scanner offset but not further background corrected. The dataset was filtered to remove positive control elements. Using the negative controls on the arrays, the background threshold was determined and all values less than this value were set to the threshold value. Finally, the data were normalized using the Quantile Normalization package in “R” (Bolstad et al., 2003). Complete statistical analysis was then performed in “R” using both the LIMMA and SAM packages to identify statistically significant differential gene expression between the three groups. Microarray data have been deposited in ArrayExpress (www.ebi.ac.uk) under accession number XXX.
Total RNA samples were extracted from E12.5–E18.5 gut tissues. For E12.5 tissues, one biological sample was pooled from 2–3 guts of like genotype. cDNA synthesis and quantitative RT-PCR analysis was performed as described previously (Gao et al., 2007). qRT-PCR primer sequences are available upon request.
Hierarchical clustering was performed on the samples (arrays) using the “R” package “pvclust” (Suzuki and Shimodaira, 2006). Additional hierarchical clustering on differentially expressed genes and generation of heat maps was performed using the TM4 Multiple Experiment Viewer software package (Saeed et al., 2003).
Gene functional classification was performed on differentially expressed genes with at least a 4-fold change between control and Cdx2-deficient ileum. The Refseq_mRNA IDs of these genes were used for analysis by DAVID Bioinformatics resources, NIH (Dennis et al., 2003). Data were also analyzed through the use of Ingenuity Pathways Analysis (Ingenuity System, www.ingenuity.com) as described previously (Phuc Le et al., 2005).
We would like to thank Drs. Michael Pack, Joshua Friedman, John Lynch, Blair Madison, Ben Stanger, and Linda Greenbaum for comments on the manuscript, Dr. Hong Fu for ES cell culture, Drs. Irina Bochkis and Jonathan Schug for help with data analysis and deposition, Ms. Karrie Brondell and Elizabeth Helmbrecht for maintaining the mouse colonies, Alan Fox and Olga Smirnova for performing microarray experiment, Ms. Kelly Kemnetz for assisting with immunohistochemistry, the Morphology Core of the Penn Center for Molecular Studies in Digestive and Liver Diseases (P30DK50306) for tissue embedding and sectioning, and Neelima Shah and Biomedical Image Core for EM analysis. This work was supported by NIH grants (R01-DK053839 and P01-DK049210) to KHK. NG is supported by Juvenile Diabetes Research Foundation fellowship 3-2007-521.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.