COUP-TFII is essential for radial and anteroposterior patterning of the stomach

Gut development depends upon crosstalk between endoderm and splanchnic mesenchyme cell layers along the anteroposterior axis, resulting in the subdivision of the gut into foregut, midgut and hindgut(Kedinger et al., 1998b; Kedinger et al., 1998a; De Santa Barbara et al., 2002). The stomach becomes specified at the caudal end of foregut and emerges as a bulge of the developing GI tract (Kaufman and Bard, 1999). The stomach then undergoes progressive remodeling and differentiation in a regionally specific fashion. The detailed genetic and molecular pathways that direct gastric organogenesis are still unknown, but it is clear that patterning of the GI tract requires elaborate cellular communication between the epithelium and mesenchyme cell layers(Fukuda and Yasugi, 2002).

Sonic hedgehog (Shh) patterns a variety of embryonic tissues,including the fore-stomach epithelium by signaling to the mesenchyme prior to organ regionalization (Chiang et al.,1996). Shh-null mice exhibit inappropriate expression of alkaline-phosphatase (EAP) in the epithelium of the hind-stomach(Ramalho-Santos et al., 2000),which is consistent with the stomach epithelium acquiring an intestinal or`posteriorized' character. Conversely, the stomach of the activin receptor IIA and B compound mutant (ActRIIA/B mutant; Acvr2a/Acvr2b–Mouse Genome Informatics) exhibits posterior extension of the fore-stomach epithelium with expansion of Shhexpression (Kim et al., 2000),a process referred to as `anteriorization'. These data suggest that Shh expression in the fore-stomach acts to induce and/or maintain non-intestinal character, resulting in the development of gastric epithelium(Ramalho-Santos et al.,2000).

COUP-TF proteins are nuclear orphan receptors, highly conserved across species. Two members have been identified in mice, COUP-TFI(Nr2f1–Mouse Genome Informatics) and COUP-TFII(Nr2f2–Mouse Genome Informatics). The temporal and spatial expression pattern of COUP-TFII in mesenchyme led us to hypothesize that COUP-TFII plays a role in mesenchymal-epithelial interactions during organogenesis (Tsai and Tsai,1997). In the developing neural tube, Shh has been shown to regulate COUP-TFII expression during the differentiation of motoneurons (Lutz et al.,1994). We have also identified a 5′-regulatory element that mediates Shh stimulation of COUP-TFII expression(Krishnan et al., 1997a; Krishnan et al., 1997b). COUP-TFII is likely to be a downstream target of Shhsignaling, and the requirement for Shh in gastric organogenesis led us to infer that COUP-TFII may play a role in stomach development.

To circumvent the early embryonic lethality of the COUP-TFII-null mutation and to investigate its function during gastric organogenesis, we generated a conditional null mutant of COUP-TFII using the Cre/LoxP system. Nkx3-2 (Bapx1) is a homeobox-containing gene (Tribioli et al.,1997) that is coexpressed with COUP-TFII in the stomach primordium; thus, Nkx3-2^Cre knock-in recombinase was used to ablate COUP-TFII in the gastric mesenchyme. The stomachs of conditional mutant mice exhibited dysmorphogenesis accompanied by abnormalities of both compartmentalization and radial patterning,demonstrating a functional link between Shh and COUP-TFII.

Genomic DNA fragments, 7.2 kb BamHI-SalI, 1.4 kb SalI-XbaI and 1.4 kb XbaI-EcoRI from Q#10 clone and a 2.6 kb EcoRI-BamHI fragment from D-3.6RI clone containing COUP-TFII locus as described(Qiu et al., 1995) were subcloned into pBlueScriptII KS(+). The first loxP site was inserted into BglI site of 7.2 kb BamHI-SalI fragment and the NsiI site on the 2.6 kb EcoRI-BamHI fragment was modified with XhoI linker. These four genomic fragments were reconstructed, and XhoI site on the 2.6 kb EcoRI-BamHI region was modified with linker containing ClaI site at the 3′ end. pNeoTKLoxP, a positive-negative selection cassette, was modified and ligated with lacZ from pDP46.21. This NeoTKLoxP-lacZ construct was inserted into XhoI-ClaI linker at 3′-UTR. 5′-SalI site of NeoTKLoxP-lacZ was destroyed by ligation with XhoI,and ClaI site and NotI site were blunt-end-ligated after fill-in reaction. Targeting construct consists of 4 kb 5′-arm, 6.5 kb floxed COUP-TFII locus, 8.5 kb NeoTKLoxP-lacZ and 2 kb 3′-arm,a total size of ∼24 kb was linearized with NotI at the 3′end. Homologous recombination was performed in AB1.2 cells(Qiu et al., 1997) and ES cell clones were screened by two rounds of Southern blot analysis. Correctly targeted clones were first identified by screening XbaI digest with 5′-external and 3′-external probes (5′-ext and 3′-ext,respectively) and subsequently transfected with a Cre recombinase expression vector followed by screening HindIII digests with a 5′-internal probe (5′-int) and XbaI digest with a 3′-external probe. Chimeric animals were generated by microinjection and germline transmission was achieved by crossing with C57Bl/6 wild-type female. Thereafter, mutant mice carrying floxed allele were maintained in 129/B6 mixed background. Specific primers were designed for floxed COUP-TFII genotyping by PCR. The primer sequences were NT1 5′-CAGTCGCCTCTCCTTCTCCTTCTCC-3′, NT2 5′-CATCCGGGATATGTTACTGTCCGG-3′, NT3A 5′-TTCTGGTCTTCACCCACCGGTACC-3′ and NT6A 5′-TGGGGAAGCTAAGTGTTGTAGTGATTCC-3′. The sizes of the PCR product are 785 bp for wild type and 394 bp for floxed allele.

Nkx3-2^Cre knock-in animal was generated by homologous recombination in ES cells (K.M., W.E.Z. and R.J.S., unpublished). Cre recombinase was inserted in-frame into the exon1 of Nkx3-2 locus. No overt phenotype was found in heterozygous animals.

X-gal staining was performed according to the published methods(Hogan et al., 1994). For whole-mount staining, embryos (up to E10.5) were dissected and fixed in 4%paraformaldehyde, then stained in X-gal staining solution at room temperature. Histological sections of whole-mount stained embryos were processed in Histoclear, embedded in paraffin, and sectioned at 10-12 μm. Sections were cleared with Histoclear, rehydrated and counterstained with Eosin, and mounted with Permount. Larger embryos (∼E12.5 or later) and postnatal samples were cryoprotected with 20% sucrose/PBS solution after fixation with 2%paraformaldehyde and embedded in OCT compound. Cryostat sections (16 μm)were briefly fixed in 2% paraformaldehyde, stained in X-gal staining solution and counterstained with Eosin.

Tissues were fixed in 4% paraformaldehyde, embedded in paraffin and sectioned at 5 or 7 μm. Hematoxylin and Eosin staining was carried out according to the regular protocol. For immunohistochemistry, paraffin sections were dewaxed, rehydrated and incubated with primary antibodies. Antibodies against human CK14 (1:500) was a gift from Dr Roop (Baylor College of Medicine, Houston, TX). Polyclonal anti-serum against PGP9.5 was from Chemicon(1:1000). Antibodies against TUJ1 (Convance, 1:1000), GATA4 (Santa Crutz,1:400), H⁺/K⁺-ATPase β-subunit (Affinity BioReagents, 1:500) were used in immunostaining. Texas Red-tagged Griffonia simplicifolia II (GSII) lectin was used to stain the stomach tissue. Positive staining for CK14 and PGP9.5 was visualized by using biotinylated secondary antibody and streptavidin-horseraddish peroxidase conjugate, and NovaRed (Vector) as a chromogen. SMA-α, TUJ1, GATA4 and H⁺/K⁺-ATPase β-subunit were visualized by using biotinylated secondary antibody and streptavidin-horseraddish peroxidase conjugate, and DAB (Vector) as a chromogen. Monoclonal anti-α-smooth muscle actin (SMA-α, 1A4, 1:1000) was purchased from Sigma-Aldrich and immunostaining was performed using Mouse-On-Mouse kit (Vector) according to manufacturers instruction, and detected with AlxaFluor488 (Molecular Probes). For ultrastructural study of the disorganized smooth muscle layer, 0.5 μm semi-thin sections were stained with Toluidine Blue. Endogenous alkaline-phosphatase staining was performed as previously described(Ramalho-Santos et al., 2000; Aubin et al., 2002) and counterstained with Methyl Green. The glycoconjugated production in parietal cells of the stomach was marked by DBA lectin. Whole-mount in situ hybridization for COUP-TFII was carried out as described(Qiu et al., 1997; Pereira et al., 1999). Probe for Shh was a gift from Dr M. P. Scott (Stanford University, CA). For section in situ hybridization of Shh, E11.5 embryos were cross-sectioned and Shh expressions in the fore-stomach were detected using ³⁵S-UTP labeled probes. Positive signals were pseudo-colored(red) and overlaid on bright field image of Hematoxylin staining as described(Pereira et al., 1999).

Foreguts were dissected from E10.5 lacZ knock-in heterozygous embryos and cultured on tissue-culture insert (Corning) with defined serum-free medium (Opti-MEMI, Invitrogen Life Technologies). Explants were incubated with vehicle (DMSO), 1, 5 and 10 μM of cyclopamine (Toronto Research Chemicals) for 48 hours, and fixed with 2% paraformaldehyde for whole-mount lacZ staining. Explants from wild-type embryos were fixed with 4% paraformaldehyde and used for whole-mount in situ hybridization.

To facilitate the profiling of COUP-TFII gene expression during development, we generated a targeted vector according to a scheme diagramed in Fig. 1A. This targeting vector was introduced into ES cells together with Cre recombinase. Upon recombination between the first and the last LoxP sites, COUP-TFII was deleted and the expression of lacZ reporter, inserted into the 5′ untranslated region of COUP-TFII locus, was activated. The ES cells were subsequently injected into blastocysts to generate lacZknock-in alleles in mice. lacZ gene expression is under the control of COUP-TFII promoter and its expression recapitulates the expression of the endogenous COUP-TFII gene. Using this lacZ knock-in allele, we examined the expression pattern of COUP-TFII during gut development.

At embryonic day (E) 12.5, X-gal staining was detected in both the epithelium and mesenchyme of the stomach. Brief staining of cryostat sections revealed that mesenchymal cells adjacent to the endoderm had higher expression of COUP-TFII than did distal mesenchymal cells, suggesting that a potential diffusible molecule, secreted from the epithelium might regulate COUP-TFII expression in the mesenchyme(Fig. 1B). At postnatal day 3, COUP-TFII/lacZ was detected throughout the stomach(Fig. 1C), and was sharply downregulated in duodenum (red arrowhead indicates junction between stomach and duodenum). COUP-TFII/lacZ was also expressed in the glandular epithelium of adult stomach. Adult glandular epithelium forms tubular structures, each of which is referred to as a zymogenic unit, which can be subdivided into four regions; a surface pit, an isthmus, a neck and a base that is located proximal to the sub mucosa(Karam and Leblond, 1993). Strong lacZ staining was evident at the base, and much lower at surface pit layer (Fig. 1D). To ensure COUP-TFII is also expressed in the parietal cells, we used DBA lectin for immunostaining. As shown in Fig. 1D, X-gal and lectin-positive green cells colocalized, indicating that COUP-TFII is expressed in the parietal cells of the Zymogenic unit. The expression pattern of COUP-TFII in the Zymogenic unit mimics the reported expression pattern of Shh in mouse stomach(van den Brink et al., 2001). Thus, COUP-TFII is abundantly expressed from an early developmental stage in both epithelium and mesenchyme of the stomach in parallel with Shh.

To examine the role of COUP-TFII in gastric development, we chose to ablate the COUP-TFII gene in a tissue-specific manner using the Cre/LoxP system. The targeting vector depicted in Fig. 1A, was introduced into ES cells. After excision of the Neo-TK cassette by Cre-based recombination, ES cells harboring the conditional allele were generated as depicted in Fig. 1A and used to generate floxed COUP-TFII mice. An Nkx3-2^Cre mouse line was also generated (K.M., W.E.Z. and R.J.S., unpublished), which was crossed with the floxed COUP-TFII mice. As lacZ is only turned on in cells with COUP-TFII locus recombined or deleted, X-gal staining can be used as a marker for successful COUP-TFII recombined or deleted cells. As shown in Fig. 2B,X-gal staining of an E12.5 stomach from Nkx3-2^Cre/+; COUP-TFII^flox/+ embryo demonstrates that COUP-TFII was ablated specifically in the gastric mesenchyme (Fig. 2C). This is expected because X-gal staining of lacZ knock-in mice is detected only in the stomach primordium (Fig. 2A) which is similar to Nkx3-2-Cre expression at E9.5, as demonstrated by the product of intercrossing the Nkx3-2^Crewith ROSA26 reporter strain (Fig. 2B).

The size of the mutant stomach is slightly smaller when compared with the control littermate at E12.5. Histological analysis of Hematoxylin and Eosin-stained sagittal sections of the mutant stomach showed that the epithelium is considerably thicker than the controls at E12.5(Fig. 3A,B). In addition, the circular smooth layer, which is stained with α-smooth muscle actin antibody, is disorganized in comparison with the controls at E13.5(Fig. 3C,D). These abnormal morphological changes were again seen at E14.5 (data not shown), suggesting radial patterning of the stomach might be perturbed.

To further assess the potential radial patterning defects exhibited by the conditional COUP-TFII-null mutants, we examined the differentiation of smooth muscle and enteric neurons of the mutant stomach at later development. Immunostaining of α-smooth muscle actin confirmed the presence of a thickened circular smooth muscle layer formation in both the fore-stomach (Fig. 3E,F) and hind-stomach of conditional mutant mice(Fig. 3G,H). Moreover,submucosal extension of α-smooth muscle actin positive cells in the mutant fore-stomach (Fig. 3F,white arrows) indicated precocious differentiation of smooth muscle cells. To further demonstrate that the smooth muscle layer in the conditional mutant stomach is disorganized, semi-thin sections(Fig. 3I,J) and the ultrastructure (data not shown) of the stomachs from E15.5 embryos were examined. Again, circular smooth muscle layer of the conditional mutant stomach is disorganized (Fig. 3J) in comparison with the controls(Fig. 3I) at E15.5 as revealed by semi-thin sections. However, cellular defects, other than the disorganization of cell layers, have not been observed (data not shown). Immunostaining for PGP9.5, an enteric neuron marker, showed an increased number of positively stained cells in the conditional mutant mice(Fig. 3L). To ensure there is expansion of enteric neurons in the mutant stomach, an additional neuronal marker, TUJ1 (neuronal class III tubulin-β) was employed in immunostaining. An increase of TUJ1-positive cells is shown in the conditional mutant (Fig. 3N) in comparison with the littermate control (Fig. 3M). Taken together, conditional ablation of COUP-TFII in the mesenchyme resulted in at least two patterning abnormalities, epithelium outgrowth and expansion of circular smooth muscle and enteric neuron, both radial-patterning defects of the stomach. The fact that COUP-TFII is only deleted in the mesenchyme compartment but the epithelium of the mutant stomach is expanded, suggesting signals from the mesenchyme is essential for proper growth of the epithelium.

The abnormal radial patterning of the stomach manifested by the COUP-TFII mutants prompted us to assess if other patterning defects were also present in the developing mutant stomach. A whole-mount view revealed dysmorphogenesis in the stomach of conditional mutants(Fig. 4A,B). The anatomical demarcation between the stratifying fore-stomach epithelium and the columnar hind-stomach epithelium, referred to as `the limiting ridge' as indicated by broken line, begins to develop, forming a presumptive margin between the fore-stomach and hind-stomach. The size of the fore-stomach compartment and of the entire organ was noticeably reduced in the conditional mutants(Fig. 4C,D, white arrowheads indicate the junction of fore- and hind-stomach). Detailed histological examination reveals abnormalities in both the epithelium and mesenchyme. The glandular epithelium was thicker in mutants when compared with controls(Fig. 4E-H). In control mice,the epithelium was thickest at the pyloric region (to the left of Fig. 4E), and the thickness of the epithelium progressively decreased anteriorly (towards the right). By contrast, the epithelium of conditional mutant mice remained thick even in more anterior regions (Fig. 4F). The mesenchyme of the conditional mutant mice was also thickened. The hind-stomach epithelium showed several invaginations in the control mice (Fig. 4G), while more extensive invagination associated with the thickened epithelial layer in the conditional mutant mice (Fig. 4H). The morphological reduction of the fore-stomach and the extension of the hind-stomach suggest aberrant compartmentation of the stomach, a possible patterning defect in the anteroposterior axis.

Histological analysis of whole embryo sagittal sections were carried out to examine if indeed patterning of anteroposterior axis in the conditional mutant is perturbed, Analysis of E16.5 conditional mutants revealed an alteration in epithelial characteristics along the AP axis(Fig. 5A-H). The proximal duodenum exhibited typical characteristics for this stage in both control(Fig. 5A) and conditional mutant mice (Fig. 5E). However,the pylorus of conditional mutant mice exhibited a disorganized thickened circular smooth muscle layer, with an epithelium that resembled the epithelium of the duodenum and was devoid of vacuoles(Fig. 5F), unlike the highly vacuolated control (Fig. 5B). The hind-stomach of controls had a simple columnar epithelium with few invaginations, and a tight circular smooth muscle layer(Fig. 5C). By contrast, the hind-stomach of conditional mutant mice exhibited a vacuolated epithelium(Fig. 5G), resembling the pyloric region of control mice (Fig. 5B), with numerous invaginations and a thickened but disorganized circular smooth muscle layer. In addition, a cell layer located distal to the circular smooth muscle layer (Fig. 5G, indicated by small arrows at the bottom), where enteric neurons are normally located, was expanded in the conditional mutants. Finally, the fore-stomach epithelium begins to stratify at E16.5, with characteristic rough and deep infolding that was evident in the control mice(Fig. 5D). The conditional mutant mice showed similar infolding, but epithelium did not show clear signs of stratification (Fig. 5H). In fact, the fore-stomach epithelium of conditional mutant mice more closely resembled hind-stomach epithelium (Fig. 5C). These observable morphological changes in the compartmentation of different regions of the stomach are consistent with the notion that the stomach is posteriorized.

If AP patterning of the stomach is indeed disrupted in the conditional mutant, it is anticipated that the specification/differentiation of gastric epithelium might be affected. To assess if the fore-stomach is altered, we examined the fore-stomach regions using the stratified epithelial cell marker cytokeratin 14 (CK14). Immunostaining using CK14-specific antibody showed intense staining in the E16.5 control fore-stomach and the very anterior region of the mutant fore-stomach (data not shown). CK14 staining remained as intense throughout the fore-stomach of the control(Fig. 6A). However, it was barely detectable in the more caudal regions of the presumptive fore-stomach of the mutant (Fig. 6B),suggesting that the mutant stomach is posteriorized in which only the very anterior region of the stomach epithelium is stratified and expresses CK14,while the hind-stomach is morphologically expanded and is non-stratified.

Shh has been shown to be highly expressed in the fore-stomach and can serve as a fore-stomach marker. In the controls, Shh is highly expressed in the fore-stomach and the expression extends into the hind-stomach(Fig. 6C). By contrast, the expression of Shh is anteriorly restricted in the conditional mutant(Fig. 6D). The anterior restriction of Shh expression observed in the conditional mutants is similar to that detected in the stomachs of Hoxa5 mutant, indicating posteriorization of the fore-stomach (Aubin et al., 2002).

Similar result for CK14 staining was observed at the later stage of development. CK14 staining remained as intense throughout the fore-stomach of the control and mutant at E18.5 (Fig. 6E,F). Again, it was barely detectable in the more caudal regions of the stomach of the mutant (Fig. 6G), in comparison with similar anatomical position of fore-stomach of the control (Fig. 6E). This result further support the notion that the mutant stomach is posteriorized in which only the very anterior region of the stomach epithelium is stratified and expresses CK14, whereas the hind-stomach is expanded and is non-stratified.

As shown earlier, the radial patterning and the AP patterning of the stomach are altered in the mutants at E13.5, leading to the respective thickening of the epithelium and the expansion of the hind-stomach (Figs 3, 4, 5). To examine if these phenotypes persist at later development, we examined the mutant stomach at E18.5. Similar to early development, it is clear that the epithelium is thickened and the hind-stomach of the mutant is considerably expanded(Fig. 7B) in comparison with the controls (Fig. 7A). To examine whether differentiation of the glandular epithelium is perturbed in the mutant, the expression of parietal cell marker H⁺/K⁺-ATPase β-subunit and of the glandular gastric epithelium marker GATA4 were analyzed at E18.5. Although H⁺/K⁺-ATPase β-subunit positive parietal cells are present in the glandular gastric epithelium of both controls(Fig. 7C) and conditional mutant mice (Fig. 7D), the number of the parietal cells were increased in the thickened epithelial layer of the conditional mutant mice (Fig. 7D). The expression of GATA4 in the thickened glandular epithelial layer of the mutant (Fig. 7F)is similar to that of the controls (Fig. 7E), suggesting differentiation of the glandular epithelium is unchanged in the mutant.

To further substantiate that differentiation of the hind-stomach is largely intact in the mutant, we examined the mutant hind-stomach at adulthood. Histological analysis of the stomachs from 35-day-old of mutant mice and littermate controls showed no obvious difference in the hind-stomachs at the regions of Zymogenic zone (Fig. 7G,H), mucoparietal zone (Fig. 7 I,J) and pure mucus zone(Fig. 7K,L) by Hematoxylin and Eosin. Mucins secretion was analyzed by periodic acid-Schiff (PAS) staining and similar PAS staining was observed in both control(Fig. 7M) and mutant(Fig. 7N) stomachs. To further examine if there is any changes in cell specification in the mutant hind-stomach, we employed H⁺/K⁺-ATPase β-subunit as the parietal cells marker (Fig. 7O,P), GSII as the neck and pre-neck cell marker(Fig. 7Q,R) and GATA4 as marker in the base of the pure mucus unit (Fig. 7S,T) in the analysis. The expression profiles of all these markers are similar in the mutants when compared with the controls. These results, taken together, support the notion that there are no cell type changes in the conditional mutant in comparison with littermate control throughout development even though there are observable patterning defects in the stomach.

As histological analysis indicated that the very caudal part of the hind-stomach might have acquired pylorus characteristics, we used alkaline phosphatase (EAP) to determine whether hind-stomach is indeed posteriorized. As shown in Fig. 8A, EAP is mainly expressed in the duodenum and pylorus region of the controls. By contrast, high level of EAP expression not just found in the duodenum and pylorus, but it was also found in the hind-stomach of conditional mutant mice using longitudinally sectioned E16.5 dissected stomach(Fig. 8B). Clear anterior extension of EAP-positive epithelium, as indicated by brackets, was found in the hind-stomach of conditional mutant mice. Although expression of the markers H⁺/K⁺-ATPase β-subunit(Fig. 8C,D) and GATA4(Fig. 8E,F) is no different in the regions anterior to the junction of the pylorus and duodenum from E18.5 control (Fig. 8C,E) and mutant(Fig. 8D,F) stomachs, the higher level of EAP activity was again found in the mutant hind-stomach(Fig. 8H,J). Taken together,the extended and high levels of EAP expression in the hind-stomach again indicate that patterning of the AP axis is abnormal in the conditional mutants.

The radial and AP patterning defects elicited by the COUP-TFII-null mutants bear resemblance to the dysmorphogenesis of the stomach in Shh-null mutant mice(Ramalho-Santos et al., 2000). Because of this similarity, we asked if COUP-TFII expression is altered in the Shh-null mutant. Whole-mount in situ hybridization showed that COUP-TFII expression in the caudal region of the stomach of the Shh-null mutant is downregulated in comparison with the control littermate. However, the expression of COUP-TFII in other regions of the stomach is only slightly downregulated(Fig. 9A,B). An anterior shifting of the caudal margin of COUP-TFII expression at the gastro-duodenal junction (indicated by broken lines; presumptive gastro-duodenal junction is indicated by arrowheads based on morphology and location of the dorsal mesogastrium) was evident in the Shh-null mutant. In addition, a significant size reduction, particularly in the fore-stomach was noted in the Shh-null mutant stomachs, even as early as E12.5 stage, using the esophagus as a guide(Fig. 9B, indicated by white lines).

As the presence of Ihh in the Shh-null mutant stomach(Bitgood and McMahon, 1995)might functionally compensate for Shh and influence COUP-TFII expression, we used cyclopamine to inhibit all Hhsignals (Chen et al., 2002) and asked if COUP-TFII expression is affected in the explants organ culture. Explants dissected from E10.5 lacZ knock-in embryos were cultured for 48 hours in the presence of increasing amounts of cyclopamine. Whole-mount X-gal staining of COUP-TFII illustrated a significant dose-dependent reduction of COUP-TFII/lacZ expression, especially in the mesenchyme between lung bud and fore-stomach (indicated by arrowheads)(Fig. 9C-F). At 1 μM of cyclopamine, a slight reduction of COUP-TFII mRNA, as represented by X-gal staining, was detected in the stomach explants from E11.5 embryos. At 5μM and 10 μM of cyclopamine, the reduction of X-gal staining was more pronounced in the stomach when compared with the controls(Fig. 9E,F), while COUP-TFII mRNA was not affected in lung explants (not shown). The downregulation of COUP-TFII expression when Hh signaling is inhibited by cyclopamine substantiates the notion that Hh signaling regulates COUP-TFII expression in the stomach.

The lacZ knock-in mice allow us to follow the expression of COUP-TFII during gastric development. It is apparent that COUP-TFII is expressed in both the mesenchyme and the epithelium of the developing stomach. The graded mesenchymal expression profile, with highest levels in mesenchyme subjacent to the epithelium, suggests a diffusible molecule secreted by the epithelium may regulate the expression of COUP-TFII. COUP-TFII is also expressed in the glandular epithelium of the stomach,with highest expression in the base where the zymogenic cells are localized,but is not expressed in the surface pit cells. The expression patterns in the zymogenic unit resemble that of Shh (van den Brink et al., 2001). Taken together of the above findings, we speculate that Shh might regulate COUP-TFII in the stomach in a similar manner as previously shown in the motoneurons(Krishnan et al., 1997a).

Conditional ablation of COUP-TFII in the mesenchyme of the developing stomach by Nkx3.2 Cre recombinase firmly established that COUP-TFII is essential for radial patterning of the stomach. At E12.5, it is already apparent that the circular smooth muscles layers and the enteric neurons layers are expanded and disorganized in the conditional mutants. These defects persist at later stages of development. Although COUP-TFII in the epithelial compartment has not been deleted, the epithelium is considerably thicker in comparison with the control littermates,indicating signals from the mesenchyme are required for appropriate epithelium growth. The defects display by the COUP-TFII conditional mutants indicate that COUP-TFII is essential for radial patterning of the developing stomach and these defects are not seen in the compound heterozygote of Nkx3.2 and COUP-TFII. Instead, the inappropriate differentiation of the smooth muscles cells and enteric neurons in the mesenchyme has been shown when Hh signaling is disrupted by treatment with cyclopamine (Sukegawa et al.,2000; van den Brink et al.,2001). Shh is believed to be the epithelial signal that inhibits the gut mesenchyme to differentiate into smooth cells and the neural crest cells to differentiate into enteric neurons. Our findings indicate that COUP-TFII participates in the negative regulation of differentiation of smooth muscle cells and enteric neurons in the gut mesenchyme. COUP-TFII can either serve as a downstream target of Hh to mediate its negative function in the gut mesenchyme or exert its effect in a pathway parallel to Hh to regulate mesenchymal differentiation.

In addition to radial patterning of the stomach, COUP-TFII is also required for AP patterning of the stomach. The anterior shift of the limited ridge that divides the fore- and hind-stomach, the reduced size of the fore-stomach, the expansion of the hind-stomach and the expansion of EAP expression into the hind-stomach are all consistent with disruption of the AP axis patterning in the stomach. However, the differentiation of the epithelium of the fore- and hind-stomach in the adult mutant remains unchanged; with all the molecular markers analyzed, even the mutant stomach is posteriorized. The posteriorization and ectopic extended expression of EAP in the hind-stomach of the conditional mutant have been demonstrated in the Shh-null mutant mice (Ramalho-Santos et al.,2000). The striking similarity of the phenotypes exhibited by conditional COUP-TFII-null mutants and animals, chick and mouse, when Hh signaling is disrupted further implicates that Hh and COUP-TFII act in the same and/or parallel pathways.

Although the general slight reduction of expression of COUP-TFIIin the Shh-null mutant is a little surprising, the high expression of Ihh in the expanded hind-stomach of the Shh-null mutant could have compensated for the missing Shh to induce COUP-TFII expression. Indeed, the downregulation of COUP-TFII expression in the mesenchyme of the stomach is more pronounced in the explants treated with cyclopamine,supporting the notion that COUP-TFII is a downstream target of Hh signaling. In addition, Shh signaling can directly activate COUP-TFIIexpression and Shh-N activates COUP-TFII expression in P19 cells without de novo protein synthesis(Krishnan et al., 1997a; Krishnan et al., 1997b). Furthermore, analysis of the COUP-TFII promoter has identified an ShhRE that is distinct from the defined Gli-binding site(Krishnan et al., 1997a). Thus, all the above results indicate that COUP-TFII mediates Hh signaling in the stomach. Taken together, our results established that COUP-TFII is essential for radial and AP patterning of the stomach. Restriction of anterior Shh expression and attenuation of Shh action in the epithelium of the COUP-TFII conditional mutant, in turn, suggests a potential role for COUP-TFII in the stimulation or maintenance of Shhexpression.

Shh derived from the epithelium signals the mesenchyme of the stomach, but how mesenchymal factors influence endodermally expressed Shh is unclear. The Hoxa5-null mutant phenotype provides evidence for mesenchyme-mediated regulation of Shh expression in the developing stomach (Aubin et al., 2002). Similarly in chick, it was suggested that adjacent mesenchyme regulates Shh expression (Narita et al.,1998). In our study, COUP-TFII was ablated in the mesenchyme and the observed dysmorphogenesis suggests that COUP-TFIIpotentially affects such mesenchymal factor(s). One such candidate, Bmp4, which belongs to the TGFβ superfamily, is expressed in the mesenchyme of developing murine stomach, while the chick ortholog has been demonstrated to play a role in gizzard patterning. Bmp4 is important for mesoderm development, but because of early lethality in Bmp4-null mutants, no specific role for Bmp4 in gastric development has been elucidated (Winnier et al.,1995). Interestingly, it has been shown that COUP-TFIIbinds to the mouse Bmp4 promoter and regulates Bmp4 promoter activity in transient transfection assays(Feng et al., 1995), suggesting COUP-TFII might function through BMP signaling. Another member of TGFβsuper family, activin, governs embryonic axial patterning, and restricts Shh expression in Hensen's node(Monsoro-Burq and Le Douarin,2001). Furthermore, the expansion of Shh expression domain within the stomach of Acvr2a/Acvr2b mutant mice indicates a restriction of Shh in the foregut that is potentially regulated by TGFβ/Bmp signaling. As Acvr2 is able to interact with BMPs(Kim et al., 2000),mesenchymal expressed Bmp4 may restrict Shh expression. Alternatively, it is possible that COUP-TFII may modulate the expression of activin/activin receptor themselves. Enhanced activin signaling may restrict the expression domain of Shh to the anterior border and perturb the epithelial growth, as seen in the conditional mutant stomach. Further analysis of TGFβ/Bmp/activin signaling in the COUP-TFII conditional mutant stomach may shed light on which mesenchymal factor(s) that regulate Shh expression are affected by the absence of COUP-TFII.

We thank Xiaoyan Huang, Kikue Takamoto and Wei Qian for excellent technical help; Dr Milton Finegold, Mr James Barrish and the Core Morphology laboratory of the NIH Gulf Coast Digestive Disease Center for the thin-section and the EM analysis; Drs Zijian Lan, Hong Wu and Dennis Roop for various suggestion and discussion; Drs Kimi Araki, Takeshi Yagi and Matthew Scott for plasmid DNA;and Kurt Schillinger, Debra Bramblett and Peter Tsai for preparation of manuscript. NIH grants DK55636 to S.Y.T. and DK45641 to M.J.T. supported this work.