Independent functions and mechanisms for homeobox gene Barx1 in patterning mouse stomach and spleen

Lower metazoans meet nutritional needs with a rudimentary intestine that is lined by a simple absorptive epithelium. The more complex nutritional demands in vertebrate animals require additional organs such as the stomach, pancreas and liver, which connect with the small bowel. Early in vertebrate embryogenesis, the visceral endoderm and splanchnic mesoderm combine to create a tube that is subsequently patterned into specialized segments: esophagus,stomach, the intestine and its evaginated derivatives, the liver and pancreas. A component common to all these structures is the mesothelium, a connective tissue that envelops the gut and tethers organs to the body wall. The dorsolateral mesothelium of the stomach (mesogastrium) provides a compartment for early development of the spleen and dorsal pancreas, which appear initially as a confluent primordium near the greater curvature of the stomach(Brendolan et al., 2007; Hecksher-Sorensen et al.,2004; Thiel and Downey,1921). Stomach rotation and leftward movement of the dorsal pancreas subsequently juxtapose the dorsal and ventral pancreatic buds, which fuse and come to lie near the duodenum, whereas the spleen remains associated with the lateral stomach wall, near its site of origin. Patterning of the rostral gut and its neighboring organs is poorly understood and little is known about the role of mesothelium in their development.

Stomach mesenchymal expression of the homeobox gene Barx1 seems to be required to suppress regional Wnt activity in prospective gastric endoderm and thus allows stomach-specific epithelial differentiation(Kim et al., 2005). During the period of gastric morphogenesis and gut endoderm specification, Barx1is expressed selectively in stomach mesenchyme(Kim et al., 2005; Tissier-Seta et al., 1995). Small-interfering (si) RNA-induced loss of Barx1 in recombinant cultures of embryonic day (E) 12 mouse fetal tissues profoundly affects differentiation of overlying stomach endoderm: intestinal marker genes are robustly activated at the expense of stomach epithelial transcripts(Kim et al., 2005). Barx1-null E12 embryos have normal intestines and a small, aberrantly shaped stomach with atypical endodermal lining; Cdx2, a specific marker of intestinal epithelium (Silberg et al.,2000), is expressed ectopically in the distal stomach. Levels of the secreted frizzled-related proteins Sfrp1 and Sfrp2, soluble antagonists of Wnt signaling (Finch et al.,1997; Rattner et al.,1997), are reduced in the absence of Barx1, and forced Sfrp expression in Barx1-deficient stomach mesoderm restored gastric markers in co-cultured endoderm (Kim et al.,2005).

On the inbred 129/Sv genetic background, Barx1^-/- mouse embryos die at E13 of unknown causes and so we could not study their subsequent development. Moreover, recombinant fetal tissue cultures convey information about molecular markers but not about histomorphology. Breeding the Barx1 mutation into a mixed genetic background with contribution from the C57BL/6 strain circumvented embryonic lethality and allowed us to elucidate an unprecedented and completely penetrant patterning defect of the stomach. In addition to this homeotic aberration, Barx1 loss causes a unique defect in development of the spleen, which is consistently mislocalized and severely hypoplastic. As Barx1 is never present in the spleen primordium but highly expressed in surrounding mesogastrium, its effects on spleen development, like those on stomach epithelial specification, must also occur across tissue planes. We confirmed the role of Barx1 in suppressing stomach endodermal Wnt activity, but our studies suggest that its role in spleen development is exerted through a different mechanism. In particular, absence of Barx1 specifically reduces mesothelial expression of Wt1, a transcription factor known to be required for spleen morphogenesis. These findings help define the basis for the diverse functions of a homeodomain transcription factor in the development of abdominal organs.

Barx1^+/- males from the 129/Sv strain were back-crossed repeatedly with heterozygote animals on the C57BL/6 genetic background, and we studied most Barx1 mutants after at least five back-crosses. TOPGAL transgenic mice and strain-matched CD1 controls were purchased from Jackson Laboratories (Bar Harbor, ME); Barx1^-/-;TOPGAL^Tg mice were generated by interbreeding. Shh^+/Cre mice originated by targeted insertion of a GFP-Cre fusion cDNA into the Shh locus(Harfe et al., 2004). Catnb^+/lox(ex3) mice carry an allele with loxP sites flanking exon 3 of the β-catenin (Catnb; Ctnnb1 - Mouse Genome Informatics) gene (Harada et al.,1999) and were generously provided by Mark Taketo (Kyoto University, Japan). Axin2^lacZ mice have lacZ cDNA embedded in the Axin2 locus (Yu et al., 2005) and were kindly provided by Walter Birchmeier(Max-Delbrück Center, Berlin, Germany). Animals were handled according to protocols approved by an institutional committee. The morning following vaginal plugging was regarded as day 0.5 of gestation.

After overnight fixation in Bouin's solution or 4% paraformaldehyde, whole embryos or isolated organs were dehydrated, embedded in paraffin and sections of 5-6 μm were prepared. Hematoxylin and Eosin (H&E), PAS and Alcian Blue staining were performed using routine methods. For antigen retrieval prior to immunostaining, specimens were heated in 10 mM Na citrate buffer (pH 6.0) in a decloaking chamber (Biocare Medical, Concord, CA), then cooled for 60 minutes at room temperature. To eliminate endogenous peroxidases, tissues were treated in methanol containing 0.5% H₂O₂ for 30 minutes. After blocking with normal goat serum, samples were incubated for 24 hours at 4°C with one of the following monoclonal antibodies (Ab): Cdx2(1:20; Biogenex, San Ramon, CA), activated β-catenin (1:500; Upstate Millipore, Charlottesville, VA), Ter119, B220 (Ly76 and Ptprc, respectively -Mouse Genome Informatics) (1:100; B-D Pharmingen, Franklin Lakes, NJ),H⁺/K⁺-ATPase (2B6, 1:1000; MBL, Nagoya, Japan), smooth muscle actin (1A4, 1:3000; Biogenex) and Muc5ac (45M1, 1:500; Novocastra,Newcastle, UK), or rabbit antisera against gastrin (1:1000; Novocastra), Pdx1(1:6000; gift of Christopher Wright, Vanderbilt University, TN), insulin(1:1000; Santa Cruz Biotech, Santa Cruz, CA), Barx1 [1:9000(Kim et al., 2005)], Wt1(1:3000; Santa Cruz) or Sox2 (1:1000; Chemicon, Temecula, CA). Samples were washed, incubated with biotinylated goat anti-mouse, anti-rabbit or anti-rat IgG and treated with avidin-biotin-peroxidase complex (Vector Laboratories,Burlingame, CA). Color reactions were developed with diaminobenzidine hydrochloride solution (Sigma, St Louis, MO).

Pregnant dams were sacrificed at various stages and embryos exposed to aβ-galactosidase (β-gal) staining protocol that yielded no background in non-transgenic animals (Kim et al.,2005). Briefly, mouse embryos or organs were isolated in Ca²⁺- and Mg²⁺-free Hanks' Balanced Salt Solution(Invitrogen, Carlsbad, CA), fixed for 15 minutes with 4% paraformaldehyde in PBS, washed three times in PBS, and incubated in staining solution [PBS (pH 7.2), 1 mg/ml 5-bromo-4-chloro-3-indoyl-β-D-galactoside, 5 mM K₃Fe(CN)₆, 5 mM K₄Fe(CN)₆·3H₂O, 1 mM MgCl₂,0.01% sodium deoxycholate, 0.02 % NP40] for 9-10 hours at 37°C.

Sections (6 μm) were cut and mounted on SuperFrost Plus slides (Fisher Scientific, Kalamazoo, MI), deparaffinized, rehydrated, washed in PBS and treated with 1 μg/ml proteinase K (Roche, Indianapolis, IN) for 10 minutes. After acetylation with 0.25% acetic anhydride in 0.1 M triethanolamine (pH 8.0), slides were washed in 2×SSC and air-dried. Hybridization was performed overnight at 60°C with digoxigenin-labeled antisense riboprobes in 50% formamide, 5×SSC, 2×Denhardt's solution, 0.02% bovine serum albumin, 0.1% Tween-20, 0.25% sodium dodecyl sulfate, 5 mM EDTA (pH 8.0) and 50 μg/ml yeast tRNA. Slides were subsequently washed in 2× or 0.2× SSC between 60 and 65°C and again in PBS, followed by incubation for 90 minutes with 20% sheep serum. The hybridized probe was detected by incubating tissue sections overnight at 4°C with alkaline phosphatase-conjugated sheep anti-digoxigenin Ab diluted 1:2000 in PBS supplemented with 5% sheep serum and 5% fetal bovine serum. Color reactions were developed with nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate(Roche); slides were monitored until color development was observed and the reaction was terminated with distilled water. In situ hybridization with radioactively labeled Barx1 probe was performed as described previously (Kim et al.,2005).

Embryonic stomachs were fixed overnight at 4°C in a solution containing 2.5% paraformaldehyde, 5% glutaraldehyde, 0.06% picric acid, 0.1 M Na cacodylate, and 0.06% CaCl₂, post-fixed in OsO₄, and embedded in Epon 812. Thin (0.1 μm) sections were stained with uranyl acetate and lead citrate and examined in a JEOL 1200 electron microscope at an accelerating voltage of 80 kV.

Total RNA was extracted using Trizol (Invitrogen), treated with RNase-free DNase (Ambion, Austin, TX) and reverse transcribed using oligo-(dT) primers. Wt1 mRNA levels were assessed by conventional and SYBR Green real-time quantitative RT-PCR (Applied Biosystems, Foster City, CA) using a common forward primer (5′-GCCTTCACCTTGCACTTCTC-3′) and the reverse primers 5′-CATTCAAGCTGGGAGGTCAT-3′ and 5′-GACCGTGCTGTATCCTTGGT-3′ for conventional and real-time PCR,respectively.

Neonatal spleen cells were dislodged with forceps and a single-cell suspension prepared by filtering through a 30-μm strainer. Cells were incubated on ice for 1 hour with 1 μg/ml Ter119, B220, Gr1 (Ly6g - Mouse Genome Informatics), Cd4, Cd8a or Mac1 (Itgam - Mouse Genome Informatics)primary Ab (B-D Pharmingen), followed by washing in PBS and further incubation on ice for 30 minutes with fluorophore-conjugated secondary Ab. Flow cytometry was performed on a Becton Dickinson FACScan and the data were analyzed using FlowJo software (Tree Star, Ashland, OR).

We previously implicated the Barx1-null mutation in homeotic transformation of the rostral gut; however, unexplained embryonic lethality had restricted analysis to embryos at E12.5 or earlier. Crossing the null mutation [deletion of homeobox-encoding exons 2, 3 and part of exon 4(Kim et al., 2005)] to introduce the C57BL/6 genetic background permitted Barx1^-/- mice to survive to birth, although they succumb soon thereafter to respiratory distress that is likely to result from cleft palate; this defect probably reflects significant Barx1 expression in developing branchial arches (see Fig. S1A in the supplementary material)(Tissier-Seta et al., 1995). Barx1^-/- pups appeared in the expected proportion in over 25 litters and our findings did not vary with C57BL/6 contributions between 50% and 98%. Barx1^-/- stomach was found to be greatly reduced in size (Fig. 1A) and escaped leftward rotation, thus presenting as a midline structure (data not shown). The villiform lining of neonatal Barx1^-/- stomach(Fig. 1B) contrasts with the flat mucosa in control mice (Fig. 1D) and carried two distinct surfaces: the distal 1/4 to 1/3 is lined by villi of the intestinal variety, whereas the proximal stomach has a highly atypical mucosa with gastric features. Periodic acid Schiff (PAS),which stains gastric but not intestinal epithelium uniformly, highlighted this difference (Fig. 1C).

To define the nature of the homeotic transformation, we studied gastric lineage markers. In wild-type mice, a stratified squamous epithelium extends from the esophagus to line the rostral dome (fundus) of the stomach(Fig. 1D,E), followed abruptly by the glandular mucosa of the corpus (Fig. 1D,F), which is flat and carries three major cell types in adults(Karam and Leblond, 1992). In mice, Muc5ac, H⁺/K⁺-ATPase and pepsinogen or intrinsic factor (IF; also known as Gif - Mouse Genome Informatics) are specific molecular markers of these respective lineages: foveolar (pit), parietal(oxyntic) and chief (zymogenic) cells. The distal (antral-pyloric) stomach has a similar lining, with modified cell ratios and folds that impart a scalloped appearance (Fig. 1D,G). Neutral mucins in gastric pit cells stain with PAS, whereas intestinal goblet cells are fewer in number and produce acidic mucins with affinity for both PAS and Alcian Blue; intestinal goblet cells also express trefoil factor 3 (Tff3)abundantly (Chinery et al.,1992). Two homeodomain transcription factors permit further distinction between gut segments: Pdx1 expression is scattered in distal stomach and uniform in duodenal epithelium(Offield et al., 1996),whereas Cdx2 is exquisitely specific to the intestine(Silberg et al., 2000). We used these features to characterize Barx1^-/- stomach.

The mucosal lining of Barx1^-/- mid-stomach was found to be thickly folded and the glandular morphology highly disorganized, with branching structures and epithelial nests deep in the mesenchyme(Fig. 1H,I). Although this mucosa was folded, it lacked authentic villi and showed characteristic stomach features: apical PAS staining of pit cells, scattered Pdx1 expression and complete absence of Cdx2 or Alcian Blue staining(Fig. 1J-M). By contrast, the intestinal features of the caudal ∼1/3 of Barx1^-/-stomach mucosa were signified by the classic villus morphology, uniform expression of Cdx2 and Pdx1, and presence of goblet cells that stain with Alcian Blue, PAS and Tff3 antibody (Fig. 1J-M,O). A sharp boundary (dotted lines in Fig. 1C,J-O) always separated the intestinal (distal) and gastric (proximal) types of epithelia; cells expressing gastric intrinsic factor (Fig. 1N) and Cdx2 (Fig. 1M) or Tff3 (Fig. 1O), for example, were never mixed. By contrast, although molecular markers of stomach glands were well represented in the proximal mucosa, there was a loss of the usual boundary between corpus and antral epithelia in the Barx1^-/- mutant. In particular,H⁺/K⁺-ATPase- and Pdx1-expressing cells, which normally show little overlap in distribution, mixed freely in the reduced glandular zone in the Barx1^-/- stomach and Pdx1, normally an antral marker, was expressed in cells abutting the squamous mucosa(Fig. 2A,B).

Proximal Barx1^-/- foregut showed additional anomalies and even more extensive mixing of cell types. First, radial asymmetry was evident, with a squamous mucosa on one surface and a cuboidal epithelium on the other (Fig. 2C). Second,the differentiated squamous epithelium was interspersed with strongly PAS⁺ cells of the foveolar type(Fig. 2D). These cells, which typify stomach glands and are never seen in normal forestomach or esophagus,appeared throughout the tubular portion of the mutant foregut; we also observed ectopic expression of Muc5ac, the mucin responsible for PAS staining in gastric glands (Fig. 2E). Third, many cells lining the Barx1^-/- proximal foregut extended numerous apical cilia, a feature restricted to a few scattered cells in control littermates. Ultrastructural analysis highlighted this finding(Fig. 2F), which is characteristic of prospective squamous epithelia and is likely to indicate incomplete differentiation of Barx1^-/- proximal foregut. Finally, the esophagus was significantly truncated and we never identified a passage lined by a contiguous squamous epithelium; probably as a result of this, the stomach lay mainly in the thorax instead of the abdomen. Thus,judging by its epithelium, the structure that overtly resembles an esophagus appeared to be a highly dysmorphic fundus with a mixed squamo-glandular lining. Sox2, a molecular marker of foregut squamous epithelium(Que et al., 2007), was expressed in proximal but not distal Barx1^-/- foregut(Fig. 2G,H). Cells with smooth-muscle morphology and expression of smooth muscle actin appeared in the correct distribution in the peripheral sub-epithelium, although the muscle layer was discontinuous and less well differentiated than in controls(Fig. 2I,J).

Absence of Barx1 thus results in marked foregut shortening and blurring of gastric squamo-glandular and corpus-antral boundaries(Fig. 2K). We never identified cells with dual stomach-intestine characteristics, but the sharply demarcated gastro-duodenal junction was shifted anteriorly, into the stomach proper. An exterior boundary was always discernible between the stomach and duodenum, but the pyloric sphincter was absent (data not shown). As Barx1expression is exclusively sub-epithelial(Kim et al., 2005; Tissier-Seta et al., 1995),the mucosal anomalies in Barx1^-/- mice must reflect mesenchymal influence over the differentiation of the overlying endoderm. Structural defects such as reduced stomach size, fusiform shape, fundic dysmorphogenesis and pyloric sphincter agenesis, are likely to represent functions intrinsic to the mesenchyme.

We previously proposed that Wnt antagonists are prominent targets of Barx1 regulation in gastric mesenchyme (Kim et al., 2005). The prospective stomach shows a wave of Wnt activity after E9, and we proposed that the usual decline in this activity results from Barx1-regulated production of secreted frizzled-related proteins (Sfrps). Recombinant fetal cell culture results supported this idea, but death of Barx1^-/- embryos precluded direct genetic confirmation. Having overcome fetal lethality, we crossed 129/Sv-C57BL/6 hybrid Barx1^+/- and TOPGAL transgenic (Tg) mice, which carry lacZ cDNA linked to multimerized Wnt-response elements and report faithfully on Wnt signaling (DasGupta and Fuchs, 1999). If the model is correct, proximal stomach endoderm in Barx1^-/-;TOPGAL^Tg embryos should, unlike control TOPGAL^Tg embryos, continue to express β-gal late in gestation. Indeed, between E16.5 (Fig. 3B) and birth, Barx1^-/- embryos carrying one copy of the Wnt-reporter transgene showed prominent β-gal activity throughout the proximal foregut, a region we characterized as an atypical gastric fundus with mixed squamous-glandular epithelium(Fig. 2). Residual β-gal activity in control transgenic stomachs was minimal by E16.5(Fig. 3A,C) and undetectable in E18.5 stomach (data not shown) and at any stage in the developing esophagus. By contrast, the signal in Barx1^-/-;TOPGAL^Tgfundic stomach appeared sooner and stronger than in any other site of embryonic Wnt activity; this signal localized to the endoderm(Fig. 3D). We confirmed lacZ expression by RNA in situ hybridization in E16.5 foregut, where signal was readily detected in mutant (Fig. 3F) but not control TOPGAL^Tg(Fig. 3E) samples.

To monitor Wnt signaling independent of the TOPGAL reporter, we examinedβ-catenin localization. In E18.5 Barx1^-/- foregut,innumerable cells showed unambiguous localization in the nucleus(Fig. 3G,H), whereas the signal in littermate control foregut always appeared at cell-cell junctions(Fig. 3I). We also mated Barx1^+/- mice with another Wnt-reporter strain, Axin2^lacZ. Insertion of lacZ cDNA into the mouse Axin2 locus, a ubiquitous target of canonical Wnt signaling(Jho et al., 2002), accurately marks sites of Wnt activity (Yu et al.,2005). Again, we readily detected prominent β-gal activity in the atypical fundus in E18.5 Barx1^-/-;Axin2^lacZembryos (Fig. 3K), but only weak residual signal in the stomach and none in the esophagus of Barx1^+/-;Axin2^lacZ littermates(Fig. 3J). Together, these data powerfully validate the idea that Barx1 functions in part to attenuate Wnt signaling in developing stomach endoderm.

To test the model independent of Barx1, we evaluated the consequences of forced β-catenin activation, an obligate mediator of canonical Wnt signaling (Clevers, 2006). We crossed mice carrying a floxed, activating β-catenin allele(Harada et al., 1999) with those expressing Cre recombinase under control of the endogenous sonic hedgehog (Shh) gene (Harfe et al., 2004). As Shh is highly expressed in stomach endoderm by E8.5 (Echelard et al.,1993), the progeny from this mating should express unrestrained Wnt activity in epithelial progenitors. Crosses between Shh^Cre/+ and ROSA26R mice(Soriano, 1999) indicated efficient Cre-mediated recombination in E10.5 embryos (data not shown) and we detected significant nuclear β-catenin in E18.5 Shh^Cre/+;Catnb^+/lox(ex3) foregut mucosa(Fig. 4A,B). This experimental model mimics local Wnt activation, although differences in timing and dose might not replicate the Barx1^-/- stomach exactly. Nevertheless, stomach size was reduced in Shh^Cre/+;Catnb^+/lox(ex3) embryos and its glandular epithelium resembled that of the Barx1^-/-gastric corpus (Fig. 1B,H,I):it was thickly folded, branched and contained many epithelial nests deep in the mesenchyme (Fig. 4C,E). As in Barx1^-/- fundus(Fig. 2C,D), the lining of the distal esophagus was radially asymmetric, with well-formed squamous mucosa on one side and cuboidal epithelium containing PAS-staining cells on the other(Fig. 4F,G). Although intestinal villi were absent from Shh^Cre/+;Catnb^+/lox(ex3) stomach, the abnormal gastric epithelium contained many cells expressing Cdx2(Fig. 4H,I), an intestinal epithelial marker (Silberg et al.,2000) that is never found in normal stomach epithelium(Fig. 4J and data not shown). Ectopic β-catenin activation in the developing mouse stomach therefore mimics the heterotopia seen in Barx1-null embryos, albeit with some differences, and confirms that attenuation of endogenous Wnt activity is required for normal stomach epithelial differentiation.

The position, size, morphology and histology of lower abdominal organs are preserved in Barx1^-/- embryos and neonates (data not shown). By contrast, the spleen never appeared in the usual position, apposed to the greater curvature of the stomach, as shown in Fig. 5A for a control neonate;instead, it was markedly hypoplastic and embedded within the dorsal pancreas(Fig. 5B). Associated with this fully penetrant anomaly was failure of the dorsal and ventral pancreatic buds to fuse (Fig. 5B), a defect we attribute to the absence of stomach rotation. Barx1^-/-spleen harbored typical blood cells, including those with the size and features of megakaryocytes (Fig. 5C and data not shown), and insulin(Fig. 5D) and Pdx1 (red box in Fig. 7D) immunostaining confirmed that they reside in the immediate vicinity of the pancreas. Flow cytometric and immunohistochemical analyses revealed normal proportions of all blood lineages (Fig. 5E and data not shown). Thus, Barx1 loss mispositions the spleen and causes marked hypoplasia without compromising blood or lymphocyte colonization per se.

As in control littermates (Fig. 6A), the pre-splenic mesenchyme appeared in E9.5 Barx1^-/- mouse embryos as a cell aggregate within the dorsal mesogastrium, next to the dorsal pancreatic anlage(Fig. 6B). Focal β-gal activity in E10.5 and E11.5 Barx1^-/- embryos that also carry the Tlx1^lacZ knock-in reporter gene(Kanzler and Dear, 2001)confirmed activation of a genetic program for spleen specification (data not shown; Tlx1 is also known as Hox11). Whereas mesothelial invagination normally separates the spleen and dorsal pancreas as they enlarge in the ensuing 2 days(Fig. 6C), Barx1^-/- spleen showed little growth and remained attached to the pancreatic primordium (Fig. 6D). To understand the basis for the unexpected role of Barx1 in spleen development, we re-examined its expression domain. At E9.5 and E10.5,Barx1 expression is reported in a columnar cell layer termed the splanchnic mesodermal plate, which is likely to correspond to the future spleen capsule(Hecksher-Sorensen et al.,2004). We observed that the level of Barx1 mRNA in this structure, which is contiguous with the mesogastrium, was comparable to that in stomach mesenchyme, but Barx1 mRNA was excluded from wild-type spleen anlage at all stages, including and beyond E9.5(Fig. 6E-G and see Fig. S1A in the supplementary material). A specific antiserum helped verify prominent mesothelial expression of Barx1 protein(Fig. 6J,K). Both mRNA and protein staining indicated that mesothelial Barx1 expression is limited to the region surrounding the stomach, spleen and caudal surface of the liver, and does not extend into the mesenteric lining of intestinal loops(Fig. 6H,I and see Fig. S1 in the supplementary material). These data implicate mesothelial Barx1 expression in expansion and morphogenesis of adjacent spleen mesenchyme and segregation of the spleen from the dorsal pancreas.

Studies that combined Barx1 deficiency with a Wnt-reporter permitted us to ask whether Wnt signaling also occurs in the prospective spleen and if Barx1 might influence it. We did not detect β-gal activity at any point in the developing spleen in TOPGAL^Tg or Barx1^-/-;TOPGAL^Tg embryos(Fig. 7A-C and data not shown),which suggests that Barx1 controls spleen development through a different mechanism.

Mice carrying null mutations of the Tlx1(Dear et al., 1995; Roberts et al., 1994), Bapx1 (Nkx3-2 - Mouse Genome Informatics)(Lettice et al., 1999; Tribioli and Lufkin, 1999), Pbx1 (Brendolan et al.,2005), Wt1 (Herzer et al., 1999) or Pod1 (Tcf21 - Mouse Genome Informatics) (Lu et al., 2000)genes show splenic atrophy or asplenia, usually in conjunction with other defects. Mutant embryos typically initiate but fail to sustain spleen development as cells die, fail to proliferate, or change potential. Bapx1,Pbx1 and Tlx1 are early splenic markers and another homeobox gene, Nkx2-5, is also suspected to regulate spleen development(Brendolan et al., 2007). These factors therefore represent good candidates for dependence on Barx1 and their deficiencies might in part mediate aberrant spleen development in its absence. To address this possibility, we assessed expression in Barx1^-/- mice of transcription factor genes implicated in spleen development. Expression of Tlx1, Nkx2-5, Pbx1, Bapx1 and Pod1 mRNAs appeared identical in Barx1^-/- and control embryos (Fig. 7D) andβ-gal staining in the splenic anlage of Barx1^-/-;Tlx1^lacZ embryos was indistinguishable from that in Barx1^+/-;Tlx1^lacZ littermates(data not shown).

Each of the transcripts we tested was expressed in the spleen primordium in wild-type mice, and incidentally also in stomach mesenchyme, but was absent from the mesothelial envelope (Fig. 7D). Only Wt1 showed a distinctive pattern, with prominent mesothelial expression, similar or lower levels in splenic mesenchyme, and absence from stomach tissue(Fig. 8A). Thus, among the genes previously implicated in spleen development, Wt1 is the best candidate for cell-autonomous regulation by mesothelial Barx1. Indeed, Wt1 mRNA is appreciably reduced in Barx1^-/-mesothelium (Fig. 8B), whereas Barx1 mRNA expression is preserved in the embryonic stomach and mesothelium of Wt1 mutants (B.-M.K., J. Alberta, D. Housman and R.A.S., unpublished). Conventional and quantitative RT-PCR confirmed reduced Wt1 mRNA levels in isolated Barx1^-/- spleen(Fig. 8C; residual expression is likely to derive from spleen mesenchyme), and Wt1 immunostaining in Barx1^-/- and control embryos matched results from RNA in situ hybridization. We detected Wt1 in both wild-type and Barx1^-/- spleen anlagen; signals were prominent in wild-type mesothelium and substantially reduced in Barx1-null spleen,particularly in the mesothelium (Fig. 8D). By contrast, Wt1 signals were preserved in embryonic kidney(Fig. 8E), the site of highest native expression.

Patterning of the vertebrate gastrointestinal tract serves as a model for fetal epithelial-mesenchymal interactions. In a quest for regulators of these interactions, we identified Barx1 as a factor that is expressed abundantly,transiently and selectively in the mesenchyme and mesothelium of the developing stomach. We have now characterized its functions in development of the stomach and spleen and its powerful role as a homeotic regulator of abdominal organogenesis. Our results indicate that Barx1 influences development of two adjacent organs by different mechanisms:non-cell-autonomous inhibition of canonical Wnt signaling in stomach endoderm and cell-autonomous disruption of Wt1 gene expression in splenic mesothelium.

Mice lacking Barx1 present a severe, invariant and completely penetrant form of visceral homeosis, with posteriorization of the proximal foregut. The esophagus is considerably shortened or, in the absence of markers that can distinguish mouse esophagus from squamous forestomach, might be missing entirely. Instead of the usual domed morphology, the fundic stomach is tubular, and cuboidal cells expressing neutral mucins and Muc5ac, which are usually confined to the glandular stomach, interrupt its squamous lining. Zones similarly blur in the body of the stomach, where cells normally restricted to the antrum/pylorus mix freely with corpus gland cells. By contrast, the next epithelial boundary is strictly preserved; intestinal villi occupy the entire distal stomach, but stomach and intestinal cells do not overlap in morphology or expression of regional markers. Homeosis in the Barx1^-/- gut thus harbors unique features, with blurring of rostral organ and epithelial boundaries and anterior shifting of intestinal mucosa.

These gastrointestinal abnormalities extend away from the Barx1expression domain both rostrally (esophagus) and caudally (pyloric sphincter),a phenomenon that is reminiscent of homeotic transformations in the limbs and axial skeleton (Capecchi, 1996; Izpisua-Belmonte and Duboule,1992). However, the major anomalies occur precisely in the domain of fetal Barx1 expression, in the gastric fundus and body, and suggest a dual role for Barx1 in stomach mesenchyme. One group of functions, likely to be intrinsic to the mesenchyme, drives sub-epithelial differentiation and generates the correct organ size and shape. Mesenchymal mass is reduced in Barx1^-/- mice but its viability seems intact and smooth muscle appears in the right location; we have not addressed the mechanisms behind the role of Barx1 in stomach morphogenesis. A second group of non-cell-autonomous functions helps specify the overlying endoderm, as we previously inferred in part from findings in recombinant embryonic cell cultures (Kim et al., 2005). Our characterization of Barx1^-/- stomach reinforces this function, extends our understanding and establishes the role of Barx1 in suppressing endodermal Wnt activity. We demonstrate that its absence permits persistent Wnt signaling in stomach endoderm, which is likely to disrupt mucosal specification and differentiation as a direct consequence. However,the scope of stomach and spleen defects in Barx1^-/-embryos, coupled with the lack of canonical Wnt signaling in normal spleen primordium, implies that Wnt inhibition represents only a facet of Barx1 mechanisms, albeit one that is vital in stomach differentiation. Furthermore,we cannot rule out the possibility that Barx1 regulation of spleen morphogenesis also involves Wnt signaling through non-canonical pathways.

Unexpectedly, Barx1^-/- mice have a misplaced and severely hypoplastic spleen of a form not observed in other animal models. Some reptiles (Falkmer, 1985)and mice lacking the pancreas-determining factor Ptf1a(Krapp et al., 1998) show isolated endocrine pancreatic progenitors scattered within the spleen. By contrast, Barx1^-/- mice reveal a novel phenotype in which a discrete spleen is embedded within intact pancreatic parenchyma. Molecular understanding of spleen development is incomplete, but the organ is known to originate as a mesenchymal condensation within dorsal mesogastrium, in close apposition to the dorsal pancreas(Brendolan et al., 2007; Hecksher-Sorensen et al.,2004; Thiel and Downey,1921). Mice with defects in late pancreas development, in which the mesenchyme is unaffected, usually have an intact spleen, whereas loss of pancreas mesenchyme, as observed, for example, in transgenic mice with ectopic Shh expression, is strongly correlated with asplenia(Ahlgren et al., 1996; Apelqvist et al., 1997; Harrison et al., 1999); these observations signify a role for dorsal pancreatic mesenchyme in some aspects of spleen development. However, Barx1 mRNA and protein are conspicuously absent from spleen and pancreas anlagen, but appear at high levels in the epitheliod lining of these organ primordia. Barx1 is thus unique among regulators of spleen development in exerting a pivotal influence exclusively from the mesothelium and its expression pattern suggests that it moderates spleen development indirectly, much as mesenchymal Barx1 helps specify adjacent stomach endoderm. The splanchnic mesodermal plate is a known source of developmental signals, including fibroblast growth factors 9 and 10(Hecksher-Sorensen et al.,2004). A key role for the prospective capsule in spleen development is independently revealed in dominant hemimelia (Dh)mutant mice, which lack this layer and are asplenic(Green, 1967; Hecksher-Sorensen et al.,2004); our findings suggest that some Dh effects might be mediated through Barx1.

All transcription factor genes expressed only in spleen primordium and previously implicated in its maturation are expressed normally in Barx1^-/- spleen. These findings are consistent with the preservation of hematopoietic potential and indicate that Barx1 is dispensable for their expression. Features of the mutant phenotype point instead to functions not previously explored in spleen development. First, normal mesothelium seems to exert a Barx1-dependent trophic effect that enlarges the organ and imparts its characteristic shape. Alternatively, the mutant mesothelium might limit expansive and morphogenetic capacities inherent to the spleen anlage, and we cannot exclude the possibility that the spleen defects in Barx1^-/- mice follow mainly from stomach malrotation and attendant disturbance in configuration of the omental bursa. A second function, separation of the spleen from the dorsal pancreas, is arguably better attributed to cell-autonomous properties of the mesothelium, and it is here that Wt1 loss might be especially pertinent. Unlike other genes implicated in the specification, survival or expansion of the spleen primordium, Wt1 alone is expressed in the mesothelium (in addition to spleen mesenchyme); this overlap with the Barx1 expression domain adds plausibility and significance to the result that mesothelial Wt1expression depends on Barx1. In both Barx1^-/- and Wt1^-/- embryos, the spleen is initially specified in the correct location and ultimately much reduced in size but not absent, and Tlx1 expression is not perturbed. Wt1^-/- spleen primordium is also reported to have a shorter connection to the prospective pancreas (Herzer et al.,1999), although perhaps not as short as we observe in Barx1^-/- mice. Taken together, these observations raise the possibility that Barx1 control over spleen development might be exercised in part through Wt1 gene regulation in the dorsal mesothelium. It is interesting that Wt1 mRNA is reduced in Tlx1^-/-splenic mesenchyme, but not in Tlx1^-/- or Pbx1^-/- mesothelium(Brendolan et al., 2005; Koehler et al., 2000).

Mice with targeted disruption of another homeobox gene, Bapx1,reveal markedly different consequences of failure of the spleen and dorsal pancreas to separate (Asayesh et al.,2006). Bapx1^-/- pancreatic endoderm undergoes metaplastic conversion to intestinal cyst-like structures, a defect attributed to persistent contact with spleen mesenchyme past E13.5, the stage by which the two organs have normally separated. The authors argued that other mouse models of asplenia avoid the same outcome because they do not expose the pancreatic epithelial primordium directly to spleen mesenchyme(Asayesh et al., 2006). As such contact is evident in Barx1^-/- embryos, we suggest that either the metaplastic defect in Bapx1^-/- pancreas is unique to that genotype, or the Barx1^-/- spleen lacks the putative required factors.

Abdominal Barx1 expression is restricted to the stomach wall and mesothelium and we identify significant and distinct developmental functions in each of these locations. Our results also make a persuasive argument for Barx1-mediated inhibition of Wnt signaling in stomach endoderm and against a role for canonical Wnt signaling in spleen development. They hence demonstrate that positional and morphogenetic functions conferred by this homeobox gene occur through distinct mechanisms, even over the short distance that separates the stomach wall from its mesothelium. The pathways we have elucidated thus far - inhibition of canonical Wnt signaling in endoderm and regulation of Wt1 gene expression in mesothelial cells - represent early steps in appreciating the basis for homeobox gene functions in the gastrointestinal tract. Barx1 is likely to regulate additional events that contribute not only to foregut patterning and spleen expansion, but also to control of stomach size and shape and pyloric sphincter formation. Characterization of other such pathways will add to the growing understanding of abdominal organogenesis.

This work was supported in part by grant number R01DK61139 (R.A.S.) and R01NS033642 (A.P.M.) from the National Institutes of Health. J.M. is a fellow of the Charles H. Hood Foundation. We are grateful to Susumu Ito for assistance with electron microscopy; Zhao Chen and SunTaek Kim for help with flow cytometry; Walter Birchmeier for providing Axin2^lacZreporter mice; Cliff Tabin and Mark Taketo for sharing Shh^Cre and Catnb^lox(ex3) mice,respectively; Terry Rabbitts and Licia Selleri for providing Tlx1^lacZ mice; Julia Alberta and David Housman for sharing Wt1-mutant mouse embryos; Christopher Wright and David Alper for generous gifts of Pdx1 and intrinsic factor antisera, respectively; Richard Harvey, Thomas Lufkin, Robert Schwartz and Licia Selleri for plasmids; and Andrea Brendolan, Licia Selleri and Mike Verzi for critical review of the manuscript.