The vertebrate intestinal epithelium undergoes continued cell renewal throughout adult life. This process is highly organized, both spatially and temporally, with a defined sequence of proliferation, migration, and differentiation. Proliferating cells are restricted to crypts that are deeply embedded in the submucosal mesenchyme. As cells begin to differentiate, they migrate towards the lumen and are eventually shed, either from the tips of the intestinal villi or from the surface of the colonic epithelium (Gordon and Hermiston,1994; Sancho et al.,2004; van de Wetering et al.,2002). Intestinal growth and differentiation are regulated by reciprocal interactions between the epithelium and the underlying stroma, mediated by extracellular signaling molecules belonging to a number of conserved protein families. Among these is the transforming growth factor beta (Tgf-β) family that includes both Tgf-βs, potent inhibitors of epithelial cell growth, and bone morphogenetic proteins (Bmps), important regulators of embryonic growth and differentiation (Mishina,2003). Mutations in several genes of the Tgf-β signaling pathway have been identified in human gastrointestinal cancers (Grady and Markowitz,2002; Waite and Eng,2003). In addition, germline mutations in SMAD4, a downstream mediator for both Tgf-βs and Bmps, are found in about one third of individuals with familial Juvenile Polyposis (JP), an autosomal dominant condition presenting with hamartomatous polyps comprised of both stromal and epithelial tissue (Howe et al.,1998; Woodford-Richens et al.,2000; Waite and Eng,2003). Germline mutations that predispose to JP have also been detected in the BMP RECEPTOR 1A gene, suggesting that not only Tgf-β signaling but also Bmp signaling plays an important role in intestinal growth control (Howe et al.,2001). In support of this idea, a number of mouse models have been described in which disruption of the Bmp/Tgf-β signaling pathways results in abnormal development and morphogenesis of the gastrointestinal tract/small intestine (Takaku et al.,1999; He et al.,2004). In one recent model, the activity of Bmp ligands was inhibited by transgenic misexpression of the antagonist, noggin. This secreted, extracellular protein was expressed in the epithelium of the developing and adult intestine under the control of regulatory elements of the villin gene (Haramis et al.,2004). By 4 weeks after birth, transgenic mice show many ectopic crypt-like structures containing proliferating epithelial cells. These are found both in the intervillus regions and penetrating into the villi, perpendicular to the normal crypt-villus axis. They undergo further enlargement and fission to give rise to numerous dilated cysts surrounded by hyperplastic stroma with inflammatory cells, pathological features similar to the hallmarks of Juvenile Polyposis.
Here, we describe transgenic mouse models based on the misexpression of noggin under the control of two versions of the well-characterized promoter of the rat liver fatty acid binding protein gene (Fabp1) (Simon et al.,1993,1997). One version (Fabp14× at −132) drives expression in the large intestine, the region where human gastrointestinal cancers most frequently arise. The second version (Fabp1−596 to +21) drives expression in the epithelium of the proximal small intestine from late fetal stages. This transgenic line eventually acquires a phenotype similar to that described by Haramis et al. (2004). However, phenotypic changes are seen earlier. This enables us to follow the onset and progression of abnormalities in villus morphogenesis and gene expression that ultimately lead to a polyposis-like phenotype.
Expression of Bmpr1a and Bmp2/4 in the Adult Intestine and Colon
Bmpr1a mRNA is highly expressed in the villus epithelium of the adult mouse small intestine, but at much lower levels in both the crypt epithelium and the stroma (Fig. 1C; Haramis et al.,2004). Bmpr1a is likewise preferentially expressed in the distal and surface epithelium of the large intestine compared with the stroma (Fig. 1D). Transcripts for Bmp2 are present in epithelial cells at the villus tips and in the surface epithelium of the large intestine (Fig. 1E,F). Analysis of β-galactosidase activity in Bmp4lacZ heterozygotes (Fig. 1A,B) confirms previous observations that Bmp4 is expressed in the small intestine in a subset of stromal cells along the entire villus axis (Haramis et al.,2004) and in mesenchymal cells surrounding the crypts (He et al.,2004). Mesenchymal expression of Bmp4lacZ is likewise seen in the adult large intestine, including cells surrounding the glands.
Effect of Misexpressing Noggin on Villus and Crypt Morphogenesis in the Proximal Small Intestine
We first generated transgenic mice that express noggin under the control of the rat liver fatty acid binding protein promoter that drives expression in the epithelium of the proximal small intestine. We created seven Fabp1−596 to +21:Noggin transgenic lines that gave similar phenotypes. One line was characterized in detail, and the development of polyps followed from birth to 14 weeks. Region-specific epithelial expression of the transgene in adult mice was confirmed by in situ hybridization, and phenotypic abnormalities were seen only where the transgene is active (Supplemental Fig. 1, which can be viewed online at http://www.interscience.wiley.com/jpages/1058-8388/suppmat).
Gross enlargement of the intestine of Fabp1−596 to +21:Noggin transgenic mice could be detected by about 3 weeks of age. Earlier changes were followed by examination of whole mount preparations of the intestine and in histological sections. As shown in Figure 2A–D, the lining of the transgenic small intestine is already abnormal at birth and 1 week of age. Whole mount preparations show a reduced number of enlarged or irregularly shaped villi interspersed with regions containing either rudimentary villi or none at all. Extensive regions lacking villi are also present in older and adult mice (Fig. 2E,F). Taken together, these findings suggest a requirement for Bmp signaling in villus morphogenesis. The formation of villi is normally preceded by the condensation of mesenchymal cells adjacent to the epithelium at the presumptive sites of new villus formation. Previous studies have shown that Bmp2 and 4 are highly expressed in the condensed mesenchyme and may inhibit proliferation in the adjacent future villus tip (Karlsson et al.,2000). To distinguish between a possible role of Bmp signaling in villus growth or villus formation, we examined sections of proximal small intestine of newborn wildtype and transgenic mice for histological signs of early villus formation and proliferation. In both wildtype and transgenic newborn intestine, proliferating cells are absent from the villus epithelium, as judged by Ki67 immunostaining (Fig. 3A,B). In wildtype intestine, groups of condensed mesenchymal cells are observed underlying evaginating epithelial buds; epithelial cells in the bud are not proliferating, while adjacent regions contain many Ki67-positive cells (Fig. 3A,C). In transgenic newborn intestine, there are extensive regions of epithelium that do not show any signs of villus bud evagination or underlying mesenchymal condensations (Fig. 3D). This suggests that Bmp signaling plays a role during early stages of villus morphogenesis.
Postnatally, the size of the proximal small intestine of the transgenic animals increases dramatically (Fig. 4). At two weeks, a significant feature of the abnormal phenotype is an increase in stromal cells, both within the enlarged villi and also surrounding the developing crypts (Fig. 4A,B). By 4 weeks, prominent expansion of the epithelium is observed, with epithelial cells extending into the villi and forming large mucus-filled cysts (Fig. 4C,D). By 14 weeks, a large part of the hyperplastic tissue is of epithelial origin and villi are no longer recognizable (Fig. 4I,J). Despite this major disorganization, most of the epithelial cells retain a normal columnar shape and only some of the larger cysts are lined by a flattened non-columnar epithelium (Fig. 4I, inset).
Proliferation was analyzed in tissue sections by Ki67 immunostaining. At 2 weeks, shortly after crypt formation, no significant difference is seen in the number and location of proliferating cells in crypts located in the intervillus epithelium of transgenic intestine, consistent with the lack of prominent epithelial hyperplasia at this stage (Fig. 4E,F). At 8 weeks, proliferating epithelial cells are confined to crypts, some of which are now considerably enlarged (Fig. 4G,H). Many crypts are in the process of fission and are no longer restricted to a basal location. Some crypts are found within the villi, and appear to originate from both the cystic and luminal epithelium (Fig. 4H). At 14 weeks, proliferating epithelial cells are present throughout the entire tissue. Most of the crypts are much larger than normal and some cysts contain many proliferating cells (Fig. 4J).
Transgenic mice do not survive past 6 months of age most likely because of the extensive overgrowth of the small intestine. This precludes long-term studies of the development of intestinal cancer (Haramis et al.,2004), which is seen in a subset of human patients with JP (Wirtzfeld et al.,2001). In spite of the abnormal proliferation, normal differentiation of goblet cells, enterocytes and neuroendocrine cells is observed in the absence of Bmp signaling. However, Paneth cells were only found in crypts located along the muscularis propria (data not shown).
Inhibition of Bmp signaling also leads to abnormalities in the large intestine Since most human intestinal cancers arise in the colon or rectum, it is important to determine if Bmps also have a growth regulatory function in the large intestine. To test this hypothesis, we generated 2 transgenic mouse lines that express noggin under the control of a variant of the rat Fabp14× at −132 promoter that drives expression in the large intestine (Simon et al.,1997). Both lines show abnormalities of the large intestine, with epithelial and mesenchymal hyperplasia and the formation of polyps lined by ectopic proliferating glands (Fig. 5A–F).
Abnormal Gene Expression in Transgenic Small Intestine
We and others have observed that Bmpr1a is expressed at only low levels in the intestinal stroma of normal mice (Fig. 1; Haramis et al.,2004). Moreover, levels of phosphoSmad1/5 and 8, an intracellular marker of active Bmp signaling, are also very low in the intestinal stroma of wild type mice compared with the epithelium (Haramis et al.,2004). This raises the possibility that the stromal hyperplasia, which is a prominent feature in both our mouse models and Juvenile Polyposis, is a secondary consequence of reduced Bmp signaling in the epithelium. We therefore analyzed in the small intestine of Fabp1−596 to +21 transgenic mice the expression of known regulators of intestinal stromal proliferation that could potentially mediate this effect. Pdgf-A and its receptor Pdgfr-α are positive regulators of stromal proliferation in the mouse (Karlsson et al.,2000) and PDGFR-α mutations have been identified in gastrointestinal stromal tumors (GIST) in humans (Heinrich et al.,2003). Pdgf-A is normally expressed throughout the villus epithelium and its expression is elevated in the transgenic mice (Fig. 6A,D). While Pdgfr-α positive cells are normally dispersed in the subepithelial layer of the small intestine, a continuous layer of Pdgfr-α positive cells lines the columnar epithelium in the transgenic tissue (Fig. 6C,F). There is a concomitant increase in subepithelial Pdgfr-α mRNA expression (Fig. 6B,E). Sonic hedgehog (Shh) and Indian hedgehog (Ihh) are two mitogenic signaling molecules expressed in intestinal epithelium known to play important roles in gut development (Ramalho-Santos et al.,2000; Madison et al.2004). Shh expression in the normal mouse intestine is low, and a significant increase is not observed upon loss of Bmp signaling (Fig. 6G,J). Ihh is highly expressed at the base of normal villi and at reduced levels in the more distal villus epithelium. In transgenic mice, Ihh expression remains high throughout the villus epithelium (Fig. 6H,K). As a consequence, we observe a significant increase in the stromal expression of its receptor and downstream target, Patched, indicating the activation of the Hedgehog pathway (Fig. 5I,L).
The normal development of the proximal intestine involves the formation of numerous villi consisting of non-proliferating epithelium surrounding a core of mesenchyme cells. The villi are interspersed by crypts that contain proliferating cells and stem cells, and penetrate deep into the submucosal stroma. The formation of both villi and crypts involves a tightly orchestrated series of interactions between the epithelial and mesenchymal layers. Our results confirm previous reports (Haramis et al.,2004) that intestinal morphogenesis is severely disrupted when Bmp signaling is inhibited by expression of the antagonist, noggin, in the epithelial layer. Moreover, this disruption leads to a pathological condition that ultimately resembles that seen in humans with Juvenile Polyposis. Taken together, our transgenic models provide the opportunity to examine the sequence of cellular and genetic changes associated with inhibition of Bmp signalling that can lead to a pathologic phenotype.
One of the significant findings of this study is that early formation of villi appears to be inhibited in the proximal intestine of Fabp1−596 to +21:noggin transgenic mice, so that fewer villi are formed and those that do form are grossly enlarged or misshapen. Moreover, this phenotype appears earlier than in transgenic mice carrying the noggin gene under the control of a villin promoter (Haramis et al.,2004; Madison et al.,2004). The precise reason for this difference is not clear, since the Fabp1 promoter is likely to be active later than the villin promoter, but may relate to the level of noggin expressed by the different constructs. In the newborn Fabp1−596 to +21: noggin transgenic mouse, there are extensive regions lacking the small epithelial buds that are the progenitors of the normal mature villi. In contrast, the localized proliferation of epithelial cells and the formation of primary crypts are not disrupted. The precise mechanism by which inhibition of Bmp signaling disrupts villus formation is not clear. However, it is known that Bmp2 and 4 are highly expressed in the condensed mesenchyme underlying sites of future villus formation and may inhibit crypt formation in the overlying epithelium (Karlsson et al.,2000). Recent evidence that Bmp signaling mediates part of the function of epimorphin in forming the crypt-villus axis further supports a morphogenetic role for Bmps in the small intestine (Fritsch et al.,2002).
One feature of the abnormal phenotype generated by transgenic expression of noggin is hyperplasia of the epithelium and the presence of numerous ectopic crypts. Many of these penetrate into the core of the enlarged villi, where they give rise to mucous-filled cysts. It has been argued (Haramis et al.,2004) that some of the numerous crypts are formed by the process of fission, in which a founder crypt gives rise to daughter crypts. Our own findings support this idea. It is also likely that the continued absence of Bmp signaling within the epithelium allows the formation of additional, ectopic, crypts along the villus axis. This suggests that Bmp signaling normally suppresses crypt formation in the villus and restricts the site of crypt formation to the intervillus regions. Again, the mechanisms underlying this suppression are not clear. One possibility is that Bmp and Wnt signaling have an antagonistic function in the intestine similar to that seen in hair follicle induction in the skin, where Wnt signaling promotes and Bmp signaling inhibits hair follicle formation (Gat et al.,1998; Botchkarev et al.,1999).
Another characteristic feature of the transgenic Fabp1:noggin small and large intestine, and of the human Juvenile Polyposis phenotype, is hyperplasia of the stroma. Changes in the stromal environment may indirectly contribute to changes in the number and size of the crypts as well as allow their progressive invasion of villus tissue. We have presented evidence here that the mesenchymal hyperplasia may be due, in part, to increased production by the epithelium of signaling factors that are mitogenic for the underlying mesenchyme. One candidate factor is indian hedgehog, and this idea is supported by the upregulation of the downstream hedgehog target and receptor, patched, in the transgenic mesenchyme. It is also possible that increased Pdgf signaling contributes to overproliferation of the stroma. Further studies are obviously needed to address the relative role of epithelial and stromal factors in the complex sequence of molecular changes that lead to polyp formation in both transgenic mice and humans.
Generation of Fabpl-Noggin Transgenic Mice
The Fabp1-Noggin transgenes were constructed by cloning the coding sequence of Xenopus Noggin from pNoggin5′ (Smith and Harland,1992) into pLPNDon containing the Fabp1−596 to +21 promoter of the rat liver fatty acid binding protein gene (Fabp1) for expression in the small intestine and into pJ51 containing the Fabp14× at −132 promoter for large intestinal expression (Simon et al.,1993,1997). The transgene was released from the vector backbone and injected into one cell (B6D2)F2 embryos. The transgenic lines were maintained as hemizygotes on a mixed B6D2 background. The mouse experiments were performed in accordance with the guidelines of the Vanderbilt Institutional Animal Care and Use Committee.
Paraffin sections (5 μM) of paraformaldehyde-fixed tissues were stained using standard immunohistochemistry procedures. The following primary antibodies and dilutions were used: rabbit-anti-Ki67 (Novocastra), 1:1,000; rabbit-anti-Pdgfreceptor-α (Santa Cruz Biotechnology, CA), 1:200. A biotinylated goat-anti-rabbit secondary antibody was used at a dilution of 1:200. The signal was amplified by using streptavidin-horseradish peroxidase conjugate (Vectastain Elite ABC Kit; Vector, Burlingame, CA) and visualized with 3,3′-diaminobenzidine substrate. The sections were dehydrated and counterstained with eosin.
In Situ Hybridization
Intestines were fixed overnight in 4% paraformaldehyde, embedded in paraffin and sectioned at 5 μM. 35S labeled in situ probes were generated from the following templates using the transcription kit from Roche Molecular Biochemicals (Indianapolis, IN): pNogginD5′ (Smith and Harland,1992); pmBmp-2a (Bmp2) (Lyons et al.,1989); pHh-16.1 (Shh) (gift of Dr. A. McMahon, Harvard University); pM2-3 (patched) (Platt et al.,1997).
Intestines were frozen in OCT and sectioned at 10 μm. The sections were fixed for 5 min in 0.5% glutaraldehyde fixative (1.25 mM EGTA, 2 mM MgCl2 in PBS) and washed twice in PBS before incubation for 24 hr at 37°C in X-Gal staining solution containing 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 2 mM MgCl2, 1 mg/ml X-Gal, and 0.02% NP-40 in PBS. After staining, the tissues were washed twice in PBS and fixed for 20 min in 4% paraformaldehyde and counterstained with eosin.
This work was initiated in the laboratory of Brigid Hogan, whom we thank for support and encouragement. We thank Dr. Anna-Pavlina Haramis for sharing results before publication and Drs. S.K. Dey and Mark Frey for helpful discussion. The Fabpl promoter was a kind gift of Dr. Jeffrey Gordon. We acknowledge Katie Couric for her vigorous encouragement.