Hedgehog signaling and gastrointestinal cancer

Abstract

Hedgehog (Hh) signaling is critical for embryonic development and in differentiation, proliferation, and maintenance of multiple adult tissues. De-regulation of the Hh pathway is associated with birth defects and cancer. In the gastrointestinal tract, Hh ligands Sonic (Shh) and Indian (Ihh), as well as the receptor Patched (Ptch1), and transcription factors of Glioblastoma family (Gli) are all expressed during development. In the adult, Shh expression is restricted to the stomach and colon, while Ihh expression occurs throughout the luminal gastrointestinal tract, its expression being highest in the proximal duodenum. Several studies have demonstrated a requirement for Hh signaling during gastrointestinal tract development. However to date, the specific role of the Hh pathway in the adult stomach and intestine is not completely understood. The current review will place into context the implications of recent published data related to the biochemistry and cell biology of Hh signaling on the luminal gastrointestinal tract during development, normal physiology and subsequently carcinogenesis.

Hedgehog (Hh) signaling is critical during embryonic development. It is involved in the patterning of the neural tube, lung, skin, axial skeleton, and gastrointestinal tract [1–3] (reviewed in [4] and [5]). In multiple adult tissues, Hh signaling remains active and contributes to differentiation, proliferation, and maintenance. By contrast, deregulation of the Hh pathway is associated with birth defects and cancer (reviewed in [6], [7–9]). Specifically, Hh signaling is constitutively active in medulloblastoma, basal cell carcinoma, small cell lung cancer, breast, pancreas, and gastric cancer. As a result, phase 2 clinical trials for Hh signaling inhibitor GDC-0449 are underway for basal cell [10], medulloblastoma [11], ovarian [12], pancreatic, and metastatic colon cancers (Curis & Genentech).

Although a number of reviews on the Hh pathway and cancer have been published, most have focused on extra-gastrointestinal cancers. In addition, new information related to the biochemistry and cell biology of Hh signaling is rapidly accumulating. Therefore the current review will attempt to place into context the implications of this new information for the luminal gastrointestinal tract, during development, normal physiology and subsequently carcinogenesis.

1. Ligands, Receptors, Signal Transduction

There are three members of the mammalian Hh family: Sonic (Shh), Indian (Ihh), and Desert (Dhh) [13,14]. All the ligands bind with similar affinity to a 12-transmembrane receptor called Patched (Ptch1) that restricts Hh signaling by inhibiting a 7-transmembrane receptor called Smoothened (Smo) [15]. The cytoplasmic tail of Smo serves as a nidus for the accumulation of the Glioblastoma transcription factor protein family (Gli) held in a complex with regulatory proteins, e.g. Costal 2 (Cos), Fused (Fu) and Suppressor of Fused (Sufu) [16]. Upon binding of ligand to Ptch1, the inhibition on Smo is removed and processed forms of Gli migrate to the nucleus to bind Hh target genes ([17–19], reviewed in [5,6,20]). Of the three Hh ligands, Shh is most related to its Drosophila homolog and has been studied most intensively [21]. Thus, we will focus on Shh except where noted.

Shh ligand is expressed as a ~45 kilodalton (kDa) precursor that is cleaved autocatalytically, yielding a 19 kDa amino terminal form, and a residual 26 kDa carboxy-terminal form that acts as a cholesterol transferase. In the stomach, Shh processing depends on the acid-activated protease pepsin [22]. However, whether protease-cleaved Shh is lipid modified and occurs in other tissues has not yet been determined. The final Shh protein is cholesterol modified at a C-terminal cysteine and palmitoylated at its amino terminus [23]. These post-translational hydrophobic modifications facilitate long-range signaling despite their intrinsically poor solubility [24,25]. Although poor solubility and long-range signaling appear counterintuitive, Grobe and coworkers, shed light on this concept by demonstrating that Shh forms stable high molecular weight oligomers with high biological activity [26,27]. Additionally, the lipid modified-Hh interacts with cell-secreted proteoglyglans [28,29], heparan [27], and chondroitin sulfates [30], the latter two have been associated with proper Shh and Ihh signalining respectively. Under-sulfated proteoglycans or Shh mutated in the proteoglycan interaction domain impair Hh signaling of various developmental stages during mouse embryogenesis, resulting in decreased proliferation of chondrocytes and overall growth alterations, but not patterning. In Drosophila and mouse embryos, proteoglycans and heparan sulfate facilitate the distribution on tissues to reach target cells, and on the internalization of the Hh-Ptch1 complex by the cell [31,32]. However the role of proteoglycans has only be examined during neural central tube development. There are no functional studies examining the role of these molecules in adult Hh signaling.

ADAM17 and metalloproteases appear to be involved in Hh release from the producing cell, along with Dispatched, a 12-transmembrane protein with a sterol-sensing domain. To test its function, Dispatched null mice were examined and showed a phenotype similar to mice lacking Smo, i.e., embryonic lethality [33–37]. The authors concluded that Dispatched is necessary for the release of cholesterol-modified Hh, and that this modification regulates Hh signaling in the developing nervous system [37], bone [38], and limb bud [39]. To our knowledge, there are no studies addressing the role of Dispatched in adult tissues. We [22,40] and others [41–43] have shown in vitro that recombinant Shh exhibits regulatory activity, albeit at a lower potency compared to tissue generated Shh presumably because it is not lipid-modified as endogenous Shh [22]. Indeed, Robbins and coworkers have found some Shh mutations in holoprosencephaly patients, which are inactive or exhibit low-activity function as dominant-negative ligands [44], suggesting that variable Shh activity contributes to different phenotypes.

Hh ligands bind to Ptch1, which itself contains a cholesterol-binding domain. Vertebrates have Ptch1 [45] and Ptch2 [46] variants. Ptch1 is a transcriptional target of Hh signaling and acts by repressing Smo [15]. The specificity of Ptch1 is somewhat controversial since its function depends on the cell type, and protein domain involved [47,48]. Hematopoietic cell-specific deletion of Ptch1 does not result in constitutively active Hh signaling [49], as Hh signaling has recently been shown to be dispensable for hematopoiesis [50,51], whereas in skin, brain and colon, mutations in Ptch1 result in unregulated Hh signaling causing basal cell carcinoma of the skin [52,53], medulloblastoma of the brain [54,55], and Gorlin syndrome [56,57], an autosomal dominant disorder characterized by the early appearance of basal cell carcinomas and cancers in other organs, e.g., brain, colon and ovary.

In the presence of Hh ligands, the repression of Smo by Ptch1 is relieved [58,59], activated Smo accumulates in primary cilia and Hh signaling is initiated [60]. In addition to Ptch1, other membrane proteins have been identified that can bind to Hh ligands to modulate signaling. Hip (Hh interacting protein) is the best studied [61,62]. Recently Gas1 (growth arrest specific gene 1) [63–66], Cdo (CAM related/down regulated by oncogene) [67] and BOC (brother of Cdo) [68] have been reported to modulate Hh signaling during development [69], including the gastrointestinal tract [70,71]. However their expression and possible function in adult tissues has not been established, though expression of Cdo and BOC messenger RNA has been demonstrated in normal stomach and gastric cancer samples by RT-PCR [72].

Once Smo is released from the inhibiting influence of Ptch1, it translocates to primary cilia and mediates the activation of Gli proteins [73]. In the absence of Hh ligands, Gli proteins are sequestered in the cytoplasm in a complex containing the kinesin-related protein Kif7, the mammalian homolog of Drosophila Cos2 [74,75]; the serine/ threonine kinase Fu, and Sufu [76–78]. As a complex, Kif7 and Fu direct the proteosomal degradation of the full length Gli protein to its repressor form [79,80]. Gli proteins contain a C-terminal zinc finger DNA binding domain and a N-terminal repressor domain (R) [81], which translocates to the nucleus to suppress the expression of Hh targets [18].

The mechanism of Ptch1 repression on Smo is not completely understood, though involvement of 7-dehydrocholesterol (provitamin D3) secretion has been suggested [15]. Also, a recent report has shown that Tow (target of Wingless) protein, is a modulator of Hh signaling that acts downstream of Ptch1 but upstream of Smo. Tow is a Drosophila homolog of a previously uncharacterized protein family that controls the availability of Smo by regulating the degradation of lipophorin particles [82]. Homologs of Tow have only been identified in insects [83] however; proteins with similar functions may exist in vertebrates. Tow is suppressed by wingless, and its over-expression causes planar cell polarity defects [83]. Thus at least in insects, Hh signaling appears to be suppressed by the Wnt pathway.

Ligand-mediated de-repression of Smo stabilizes full length Gli that transactivates some Hh targets, e.g., Ptch1, Gli1, N-myc, and Wnt5a. Other Hh target genes require further cleavage of Gli to produce a N-terminally-derived activator (Gli-A) before their translocation to the nucleus. Gli2 and Gli3, but not Gli1, are further processed from full-length forms via phosphorylation at different residues primarily by PKA, which results in the release of the activator forms [19]. Therefore Gli 2 and 3 proteins are capable of target gene activation or repression.

The genes targeted by Hh signal transduction depend on whether the promoters are occupied by the Gli activator or repressor form. Gli1 and Ptch1 are the classic genes activated by Hh signaling, e.g., Gli2-A [84]. Moreover, Gli2 appears to be the key transcriptional effector of Hh signaling in skin and other organs, including the stomach [85–90]. Studies in epidermal cells have grouped Hh target genes in to 3 classes: Class I are those genes in which expression is controlled by the Gli-R form (i.e. Pax3); Class II promoters are influenced by both activator and repressor forms, and their expression might be regulated in a signal inducible manner (i.e. Ptch1, Cyclin D1, N-myc). Class III genes are only induced by Gli-A forms (i.e. Gli1, Wnt5a) [91,92].

Primary cilia are necessary to process full-length Gli to Gli-R in the absence of Hh ligand [93,94]. By contrast, Sufu is required for maximal Hh signaling, modulating Hh-target genes in mouse embryonic fibroblasts and during neural tube development [95–97]. Recent studies have demonstrated that Gli processing not only depend on Smo activation, but is also actively regulated by the negative Hh regulator Sufu. Kise et al. showed that Sufu functions as a recruiter for GSK3b, the kinase participating in Gli3 processing in the presence of Hh [80]. Furthermore, Chen et al. demonstrated that Sufu antagonizes the activity of Speckle-type POZ/BTB protein (Spop), the mammalian homolog of the Drosophila Hh-induced BTB domain protein Hib, that acts as a Gli-degrading factor [98]. Therefore Sufu prevents the processing of the full-length Gli2 and Gli3 proteins so that they are available for processing in a cilia-independent manner [99,100]. Further studies on the cytoplasmic proteins involved in Hh pathway signaling will provide some understanding of the signal transduction regulations, particularly in adult tissues.

2. Hedgehog signaling in cancer

Hh signaling is not only important during development, but it also has implications in the maintenance of adult tissues. Its deregulation can lead to the development of several cancers, which are now the focus of Phase II clinical trials. Rubin and de Sauvage proposed three basic models for Hh pathway activity in cancer. Type I cancers, e.g., basal cell carcinomas, are ligand-independent because they involve constitutive activation of either the receptor or downstream signaling molecules. Type II cancers (e.g., pancreatic cancers) are ligand-dependent, implying that they involve autocrine or juxtacrine signaling mechanisms. Autocrine or juxtacrine means that the Hh-producing and responding cells are either the same cell (autocrine) or originate from adjacent tumor cells (juxtacrine). Type IIIa cancers are also ligand-dependent, but exhibit paracrine signaling, i.e., tumor cells produce Hh that activates downstream targets within the mesenchyme. Hh-activated mesenchymal cells produce proliferative signals that feedback to the tumor to promote growth and survival [101]. Recently, these authors included a Type IIIb mechanism called “reverse paracrine” signaling, where Hh ligands are generated from mesenchymal cells, and received by responsive tumor cells in the epithelium [9]. Interestingly, none of the models proposed consider the possibility of non-canonical signaling.

Non-canonical Hh signaling has been described in various studies. Some studies have analyzed Ptch1 interaction with cyclin B1. Ptch1 and cyclin B1 have been shown to associate with cyclin dependent kinase-1 (Cdk1) at the cell membrane in vitro (293T cells) [102] and in vivo (human fetal kidney) [103], resulting in decreased proliferation, and reduction of the mitotic index by the sequestration of the cyclin B1/Cdk1 complex. These functions are antagonized by Shh, suggesting that Ptch1 participates in a G2/M phase checkpoint independent of other downstream Hh pathway components. Another report suggests that activation of Ptch1 modulates apoptosis through activation of caspase 3 by cleavage of the C-terminal cytoplasmic tail of Ptch1 [104,105]. Similarly, Shh blocks Ptch-mediated caspase cleavage. Additionally, treatment of mouse fibroblasts with soluble Shh induces reorganization of the actin cytoskeleton, and the formation of lamellipodia. These changes are observed during approximately the first 10 minutes of treatment before changes in Gli1 are observed, suggesting that Shh can induce cell motility separate from the classic canonical signaling through Gli1 [106]. On the other hand, TGFβ signaling activates Gli2 independent of Hh ligands, which subsequently activates other gene targets, such as follistatin, an inhibitor of stromal activins [107,108]. Nevertheless, the existence of non-canonical signaling remains controversial primarily because the evidence for this pathway depends completely upon the absence of Gli1-mediated signaling assessed using a Gli1 reporter assay.

3. Primary Cilia

Primary cilia have been defined as indispensable for Hh signaling in mammalian cells [60,109]. Ptch1 blocks Smo localization to the cilia in the absence of Shh. In the presence of the ligand, Smo migrates to the primary cilia and then activates Gli proteins. In the skin, Shh is expressed in hair follicles [110], while Ptch1 and Glis are expressed in dermal fibroblasts surrounding Shh secreting cells [90]. Primary cilia are present in murine skin and hair follicles throughout morphogenesis, and during hair follicle cycling in postnatal life on both epithelial and mesenchymal cells [111]. Purkinje cells in the brain and granule precursor neurons express primary cilia too [112]. In these tissues, the requirement of primary cilia in Hh signaling has been demonstrated. In the pancreas and prostate not all cells express primary cilia. However, there is synergy between Hh signaling and cilia positive cells in these two organs [113–115].

Interestingly, some cancer types, and non-canonical Hh signals do not require primary cilia. In basal cell carcinoma, the Hh pathway is constitutively active in the absence of ligand due to mutations in Ptch1 or Smo [52,116,117]. Both medulloblastoma, and basal cell carcinoma development can be associated with constitutively active Smo or Gli2. In the case of Smo, primary cilia are necessary for pathway activation. By contrast, cilia ablation was required for these tumors to grow when constitutively active Gli2 was present [118,119].

We have observed the presence of primary cilia expressed in both the gastric epithelium and mesenchyme (unpublished observations). The presence of primary cilia in the gastric epithelium supports the idea that both epithelial-epithelial, and epithelial-mesenchymal Hh signaling might occur. Thus direction of Hh signaling in the gastric mucosa should be viewed with some caution based on the presence of primary cilia. Indeed, non-canonical signaling remains a possible mechanism.

4. Hedgehog in the Gastrointestinal Tract

4.1. The Stomach

Within the luminal GI tract, Hh signaling in the adult stomach has generated the greatest number of studies due to the original observation by Ramalho-Santos et al. who reported that there are metaplastic changes in day old Shh null mice. Subsequently, a study comparing the phenotype of Shh null to Gli 1, 2 or 3 null mice reported hyperplastic rather than metaplastic changes since immunofluorescent analysis of the P0 stomachs were positive for H⁺,K⁺ATPase [120]. Interestingly, the Gli3 null mice mimicked the Shh null phenotype in the stomach [120]. Until E15.5, Shh is expressed at high levels in the squamous forestomach and anterior glandular stomach (Fig. 1). Shh expression tapers down upon reaching the hindstomach (antrum), where Ihh is highly expressed. Gli1 and Ptch1 expression mirrors the expression of Shh [71]. Subsequently at E18.5, Shh and Ihh are expressed in the gastric mucosa and decrease along the intestine [3,121]. By this time, Ptch1 and Gli1 expression is restricted to the mesenchyme while in the embryo; they are also present in the epithelium [122] (Fig. 1). The specific role of Shh during gastric maturation is difficult to define because Shh^−/− mice die in utero or within 24 h after birth when the gastric mucosa is not completely differentiated and functional.

Fig. 1.

Expression of hedgehog pathway molecules in the gastrointestinal (GI) tract. During development, Sonic hedgehog (Shh) is highly expressed in the stomach and its expression decreases along the intestine. Shh is expressed in the colon glands. Indian hedgehog (Ihh) expression peaks at the pyloric junction and is constant in the small intestine and colon. There are no studies addressing the pattern of Desert hedgehog (Dhh) expression in the developing GI tract. The Hh receptor Patched (Ptch1) and the Glioblastoma (Gli) transcriptional factor family proteins are expressed in the mesenchyme, mirroring Shh expression. In the adult GI tract, Shh expression is limited to the stomach and colon, with the proximal stomach being the main source of the morphogen. Ihh is expressed in the small intestine, with highest levels in the proximal duodenum. Ihh is also expressed in the colonic glands. Dhh expression has been reported in the gastric mucosa, but has not been localized in the small intestine. Ptch1, Gli proteins, and Smo are expressed primarily in the mesenchyme of the adult GI tract.

In the normal adult gastric corpus, Shh protein expression is greatest in parietal cells (Table 1, Fig. 1). Recently using Shh-LacZ reporter mice, we have shown that Shh mRNA is also expressed in zymogenic, surface pit, and mucous neck cells, supporting previous reports by other groups [40,72,123] (Table 1). In addition Shh mRNA and protein expression and Hh signal transduction is observed in the antrum [40]. This finding was surprising, as prior studies indicated no Shh expression in the antrum [121].

Table 1.

Expression of Hedgehog related molecules in the stomach and gastric cancer.

Molecule	Normal Stomach	Intestinal Metaplasia	Cancer		Methods	Reference
			Intestinal	Diffuse
Shh	Neck / Gland Epithelium	Low or absent	High	Low or undetectable	a, b	[72]
					a, b	[193]
					b	[154]
					a, b, c	[123]
					a	[125]
Ihh	Pit Epithelium	ND	High	Low	a, b	[72]
					a, b, c	[123]
Dhh	Gland Epithelium	Low or absent	Low	High	a, b	[72]
					a, b, c	[123]
					a	[125]
Ptch1	Pit Mesenchyme	D	High	High	a, b	[72]
					a, b, c	[123]
					a, d	[125]
Smo	Pit / Gland Mesenchyme	D	High	High	a, b	[72]
					a, b, c	[123]
Gli1	Pit / Gland Mesenchyme	Low or absent	High	High	a, b	[72]
					a, b, c	[123]
Gli2	Pit Mesenchyme	Low or absent	High	High	a, b	[72]
					a, b, c	[123]
Hip	D	NR	Low	NR	a, c	[193]
BOC	Pit	ND	High	High	a, b	[72]

D: Detected, ND: Not detected, NR: Not reported, a: RT-PCR, b: Immunohistochemistry, c: Immunofluorescence, d: LacZ reporter.

There is controversy regarding how Hh signaling occurs in the stomach. A major source of the controversy resides in the lack of immunohistochemistry for Ptch1 and Gli to demonstrate that the mRNA levels reflect protein expression. Most studies indicate that the gastric epithelium secretes Shh and activates Hh receptors in the mesenchyme. Both in situ hybridization, and Lac-Z reporter mice for Ptch1 and Gli proteins show robust expression in the gastric mesenchyme supporting this argument [71,124]. Berman et al. showed that there is Ptch-LacZ expression in gastric parietal cells in the gastric epithelium [125]. However other groups have not reproduced their results. They also showed RNA expression of Ptch1, Gli, Shh and Ihh in resected human tumor samples [125]. In addition, Fukaya et al. demonstrated the expression of the three Hh ligands and downstream signaling molecules by both immunohistochemistry and RT-PCR analysis of tissues from human and mouse gastric epithelium [123] (Table 1). Since this study was performed using laser micro dissection, contamination of the epithelial sample by mesenchymal cells cannot be excluded.

Since Hh signaling is essential for the development of viable embryos, defining its role in adult tissues including the stomach requires the use of cell specific knockouts. Unfortunately, there are very few reliable tissue-specific promoters that will drive Cre recombinase in the stomach. Zavros et al. recently used the H⁺,K⁺ATPase-β subunit promoter, which faithfully drives Cre in gastric parietal cells to generate a conditional Shh knockout [126]. These mice develop hypergastrinemia, an expanded mucous pit cell compartment and parietal cells unresponsive to histamine stimulation. Interestingly, D cells (somatostatin secreting cells), zymogenic and mucous neck cells populations decreased.

Prior studies by Todisco and coworkers revealed that Shh stimulates H⁺,K⁺ATPase gene expression [127]. Presumably, chemoreceptors on D cells detect the increase in pH (hypochlorhydria), triggering a regulatory feedback loop that results in a decrease in somatostatin and hypergastrinemia [128]. In point of fact, we previously showed that the loss somatostatin in null mice is the most potent inducer of gastrin gene expression [129]. Since bona fide chemoreceptors have never been identified, the D cell regulation by Shh functioning as a proton-sensing molecule might also explain the acid-mediated feedback regulation. Confirmation of this model awaits demonstration that D cells directly respond to Hh signaling. Another important observation is that loss of Shh did not cause parietal cell atrophy. Thus one might conclude that Shh regulates target genes that modulate parietal cell function, but not survival. The unresponsiveness of parietal cells to histamine, and the changes observed during tubulovesicle formation in the Shh-depleted model [126] may be associated with ezrin, a membrane cytoskeleton linker protein that mobilizes H⁺,K⁺ATPase in the gastric parietal cells by a mechanism dependent on PKA activation [130,131]. Although it has been shown that gastrin regulates ezrin expression by parietal cells in gastrin deficient mice [132], the effects of hypergastrinemia on ezrin expression and localization have not been yet studied.

We have demonstrated that the Shh 45 kDa form processing to the 19 kDa form in the stomach requires cleavage by pepsin A, which itself requires acidic conditions to become active [22]. Acid is released into the gastric lumen by parietal cells, making it available for pepsin A activation. Thus in the stomach, we assume that Shh is released into the lumen to be fully processed and active. However, Shh processing within an acidic endocytic vesicle cannot be excluded. A recent study by Robbins et al. has indicated that the unprocessed form of Shh has biological activity [133]. Indeed, we found that the predominant form of Shh in the transformed, hypochlorhydric stomach is the unprocessed 45 kDa protein [22]. In addition, some human gastric cell lines are responsive to Shh in vitro (unpublished data), raising the possibility that Hh signaling can occur in through autocrine or non-canonical signaling.

Shh activation is acid-dependent in the stomach, therefore it has been suggested that Shh is released at the apical surface of parietal cells where acid is secreted [134]. Zymogenic, surface pit and mucous neck cells also express Shh, however there is no data available on the mechanism of Shh release from these cells. Recently, Etheridge et al. have shown that Disp1 is required to produce and release Shh basolaterally. Ptch1 in the adjacent cells binds to Shh, indicating that both Ptch1 and Disp1 control the trafficking of Shh away from its site of synthesis [135]. If Hh signaling is paracrine with responsive cells located exclusively in the mesenchyme, then the mechanism by which secreted Shh is received by target cells might involve Disp1 and will require further clarification. Megalin has been proposed as the intracellular transporter for Shh in epithelial cells during development and in the efferent duct epithelial cells of adult rat [136]. Megalin is involved in forming vesicles that deliver Shh to the basal membrane of the producing cell [137]. However there is no evidence that Shh-producing cells in the GI tract express megalin, raising the possibility that ezrin could fill that role in the stomach.

The eventual loss of zymogenic and mucous neck cell populations in the Shh conditional null mouse suggests that Shh also affects the trans-differentiation of mucous neck to zymogenic cells. Since the helix-loop-helix transcription factor Mist1 mediates the transition of this cell lineage in the gastric mucosa [138,139], Shh might modulate the function of this transcription factor and subsequently the appearance of these gastric cell types. Additional evidence supporting a role for Shh in cell differentiation has emerged from studies of aspirin-induced gastric ulceration in rats, which showed that Shh was not required for cell proliferation, but rather for differentiation of parietal, and mucous neck cells. By contrast, blocking Hh signaling with cyclopamine, a Smo inhibitor increased the proliferation of gastric progenitor cells without expansion of the parietal and mucous neck cell compartments [59,140]. Other studies have shown that cyclopamine treatment reduces the expression of H⁺,K⁺ATPase in canine parietal cells [127] and mice hypergastrinemic stomach [140]. Collectively, these results suggest that Shh is necessary for progenitor cell differentiation into various gastric cell lineages rather than proliferation and for parietal cell function. Specifically, loss of Shh expands the progenitor cell compartment at the expense of terminally differentiated lineages [124,126].

Depleting Shh in the mouse parietal cells also induced expression of Ihh, Gli1, TGF, Wnt, SNAIL, and cyclin-D1, while E-cadherin was decreased [126]. Modulations in SNAIL and E-cadherin expression have been associated with carcinogenesis, particularly in the intestine and skin, where Hh activation induces SNAIL expression increasing its metastatic potential [141,142]. Therefore one might consider that Shh regulation of these genes, along with the associated changes in proliferation and apoptosis, might explain mucosal transformation during gastric carcinogenesis. Nevertheless, induction of Ihh did not compensate for the loss of Shh in parietal cells. This result raises the possibility that the three Hh proteins possess distinct functions in the gastric mucosa, similar to their distinct roles in the hair follicle and the sebaceous epithelial cell precursors of the skin [143,144].

H. pylori infection is the main cause of gastric inflammation, and has a strong association with intestinal type gastric cancer development in humans. Helicobacter-infection models in rodents have provided valuable information on the early mechanisms involved in gastric carcinogenesis and the participation of Shh in them. During H. pylori infection, acid secretion by parietal cells is compromised and Shh levels are reduced [145,146]. We have recently shown that IL-1 also reduces Shh gene expression in parietal cells, H⁺,K⁺ATPase expression and that gastric acid secretion subsequently blocks the release of intracellular calcium [40]. Therefore, changes in gastric acidity can result in the development of gastric atrophy, which in turn creates a permissive environment for the expression of intestine-specific transcription factors, e.g., CDX2, which correlates with the development metaplasia [147,148].

4.2. Shh in Gastric Cancer

Gastric cancers are classified histologically into two major types: intestinal and diffuse [149]. The intestinal type is thought to develop from a series of mucosal changes, triggered by chronic inflammation (gastritis), then followed by parietal cell atrophy (atrophic gastritis), gastric mucous cell hyperplasia, intestinal metaplasia, dysplasia, and cancer [150]. On the other hand, diffuse gastric cancer is thought to arise from deregulation of stem or precursor cells [151,152]. In addition, the familial form of diffuse gastric cancer is due to mutations in E-cadherin [153].

The two types of gastric cancer show different Hh ligand expression profiles, perhaps due to their cell origin (See Table 1). The analysis of human metaplastic and gastric cancer samples showed that the three Hh ligands, Ptch1, Smo, and the three Gli transcription factors were only weakly expressed or absent. In intestinal type gastric cancer, Ihh and Shh expression were detected. However, the expression of Smo, Gli 1, and 2 was infrequent but low when detectable [72].

Ihh, Shh, and Dhh expression was variable in diffuse gastric cancers, but in contrast to the intestinal-type cancer, diffuse cancers expressed high levels of Ptch1, Smo, and Gli proteins (Table 1). Some diffuse cancer cells showed histological expression of Ihh, while Shh expression was not detected in epithelial cells of the cancer, but instead were expressed in the adjacent fibroblast cells [123,154]. The latter observation is suggestive of the Type IIIb signaling described by de Sauvage and colleagues, in which the ligand is produced in the supporting stromal cells [9,101]. Thus, some diffuse gastric cancers belong to the Type IIIb group [125], while others might express receptor mutations, or non-canonical pathway activation, as suggested by the lack of response to cyclopamine-mediated inhibition of Hh signaling [140,155], although these tumor types are rare. Gastric metaplasia and intestinal cancer cannot be classified by these criteria because they appear to involve the initial loss, then subsequent gain in Hh signaling once the epithelium is irreversibly transformed (Table 1). Therefore, we suggest adding a Type IV signaling pathway to the classification for cancers in which the normal adult epithelium requires Hh signaling. With neoplastic transformation, Hh signaling is lost, with or without gain of function mutations in the Hh signaling molecules, e.g., Ptch1, Smo or the SuFu-Gli complex [140,155].

In summary, the normal gastric mucosa expresses Shh, Ihh and Dhh (Table 1, Fig. 1). Shh regulates parietal, zymogenic, and mucous neck cell differentiation, and proliferation, while Ihh appears to modulate the surface pit cells without recapitulating the function of Shh. The role of Dhh in the stomach is not yet known. During H. pylori infection, Shh expression reduced by IL-1β results in the lack of acid secretion and hypergastrinemia [40]. Even when gastrin stimulates Shh expression, acid-dependent processing of the Shh precursor is compromised, and parietal cells fail to become properly functional [22], possibly due to changes in ezrin. Hypochlorhydria, hypergastrinemia, and low levels of Shh result in parietal cell atrophy, and mucous cell expansion, with impaired zymogenic cell differentiation, along with the expression of intestinal markers, such as Cdx2 and Muc2, along with the expansion of cells with low Hh responsiveness (low levels of Smo and Gli). Diffuse tumor carcinogenesis is associated with hyperplasia of stem or precursor cells, but it is not yet clear whether Hh signaling contributes to its pathogenesis. Although in some of these tumors, the signaling shifts from its normal gastric epithelial to a mesenchymal origin (Table 1). In addition, mutations in the TGFβ growth factor receptor, activins, or mutations in Ptch1, Smo, and other Hh pathway mediators such as Sufu or hedgehog interacting protein (Hip), can result in constitutively active growth signals.

4.3. The Small Intestine and Colon

During gut development (reviewed in [156]), Hh signaling plays an important role in the left-right axis specification, radial patterning by regulating villus formation, smooth muscle genesis, and development of the enteric nervous system. At E11.5 and E12.5, Shh and Ihh are expressed throughout the endoderm of the developing intestine. Shh expression is down regulated in the small intestine at E14.5, with the highest expression remaining in the duodenum. Ihh expression remains unchanged in the developing intestine and colon. At E18.5 their expression is restricted to the intervillus epithelium in the small intestine. In the colon, Shh is expressed at the base of the crypts, whereas Ihh is expressed in surface colonocytes. Ptch1 and Gli1 are exclusively expressed in the intestinal mesenchyme [71,157,158]. Shh but not Ihh signaling is necessary for normal anorectal development [3] (Fig. 1).

The adult small intestinal epithelium is established after the crypts are completely formed by the third postnatal week. The level of Shh mRNA in the adult small intestine and colon are very low compared to those in the stomach, whereas Ihh expression appears to be more robust, and present mainly at the crypt-villus junction [159]. Ptch1 is expressed at low levels and only in the mesenchyme. The distal colon expresses Ihh, but not Shh in epithelial colonocytes, and Ptch1 expression is also restricted to the mesenchyme underlying the Ihh positive epithelium [157,158].

Both in vivo and in vitro studies show that Hh signaling is necessary for enterocyte differentiation, such as brush border staining for villin, and carbonic anhydrase IV [160]. Since Ptch1 expression is exclusively mesenchymal, enterocyte differentiation must be either under paracrine or non-canonical Hh regulation. In vitro, Hh signaling is necessary and sufficient for induction of differentiation markers in the human colon adenocarcinoma HT-29 cell line [161], although these cell lines express Ptch1 and Gli1 [162], suggesting autocrine regulation, which differs from what is currently understood regarding Hh-mediated enterocyte differentiation.

Ulcerative colitis (UC), and Crohn’s disease are the prevalent inflammatory bowel disorders (IBD) with high morbidity. Lees et al. analyzed samples of human colonic inflammation and found that the levels of Gli1, Ptch1, Hip, and Ihh were lower in UC patients. They also identified a germ line variation in Gli1 that resulted in reduced Hh signaling. In vivo analysis of mice heterozygous for Gli1 demonstrated increased acute inflammation when the mice were treated with dextran sulfate (DSS). These authors found that Hh signaling also regulates myeloid cells, including dendritic cells in the colon, during both homeostasis and under inflammatory conditions [163].

The association of Hh signaling with cancers of the intestine or colon is controversial. Mouse models with deleted or silenced Hh signaling molecules that have been analyzed thus far do not develop gastrointestinal cancers, except in the pancreas (reviewed in [122], and [164]). Some reports show that expression of Hh signaling components is increased in hyperplastic polyps, adenomas and adenocarcinomas of the human colon. These Hh components are also expressed in many colon adenoma or adenocarcinoma cell lines [165,166]. In contrast, it has been reported that Ihh antagonizes Wnt signaling in a colon cancer cell line and that loss of Ihh contributes to intestinal tumor formation [160]. Transgenic expression of the Hh ligand-inhibitor Hip in the small intestine also enhances activation of Wnt signaling. In addition, treatment with cyclopamine does not alter the expression of Ptch1 and Gli1 in some colon cell lines [158]. Therefore, Hh participation in colon cancer development might occur by enhancing or inhibiting the Wnt pathway [167].

The distal colon looses Ihh expression early in the process of the adenoma to carcinoma sequence of colorectal cancer. Ihh expression loss is caused by mutations in adenomatous polyposis coli (APC) or other genes that result in uncontrolled Wnt pathway activity [160,167,168]. Recently, Smo was found to contribute to intestinal tumorigenesis by increasing Wnt signaling, in the APC mutant mice treated with cyclopamine, and this role was confirmed by further specific deletion of Smo in the intestine of these mice [167]. Indeed, Smo expression is high in adenomatous polyps with APC mutations compared to normal intestine where it is hardly detectable. Smo can stimulate intestinal tumor cell proliferation by increasing-catenin in the intestinal tumor cells, altering the β-catenin/Tcf dependent Wnt signaling, resulting in the expression of Wnt targets, Gli independent or non-canonical Hh signaling [169]. Collectively, these data suggest that Hh signaling is important for normal enterocyte differentiation, and its role in intestinal or colon cancer development is not direct, involving both mesenchymal paracrine signaling [170] and Hh pathway molecules acting in Wnt signaling [171,172].

4.4. Esophagus, pancreas and liver

Hh signaling in the esophagus during development mirrors the observed expression in the stomach. The esophagus consists of three to four layers of squamous epithelium. During development at E17, some ciliated cells are also observed, however the ciliated cells disappear 1 week after birth [173]. Human and mice esophageal epithelia differ in that the mouse esophagus becomes keratinized one month after birth, while the human esophagus is not keratinized [174]. In the human adult esophagus, Shh expression is restricted to the basal cell layer, while Dhh is expressed in differentiated cells and Ihh is absent. Gli1 expression was localized in the nuclei of the basal and parabasal cells, and keratinocytes differentiation is blocked by cyclopamine, suggesting Hh signaling is necessary for epithelium differentiation but absent in differentiated cells [175]. Hh signaling is activated in most esophageal squamous cells carcinomas, along with several markers of epithelial-mesenchymal transformation (EMT) that are regulated by Gli1 [175], suggesting that esophageal carcinomas arise from aberrant Hh signaling cross-talk between the mesenchyme and epithelium. However the information available on these carcinomas and Hh signaling is scarce.

The pancreas represents an interesting and controversial organ in terms of Hh signaling. In contrast to the other organs mentioned in this review, Shh expression is absent in the embryonic pancreatic bud [176,177]. Ihh, Dhh and Ptch1 are expressed in the pancreas from embryonic day 13.5 onward, and their expression remains constant in adult islet and duct cells [178,179]. Activated Hh signaling in the developing pancreas results in loss of endocrine cells [177] while models where Hh signaling is lost, result in an expansion of the pancreas [178,180]. Recently, Hebrok and colleagues demonstrated that Hh signaling is active in the pancreatic epithelium during development, and Smo mutant mice are glucose intolerant and produce less insulin [181]. Therefore Hh signaling in the developing pancreas is required for endocrine function. Hh signaling in pancreatic cancer has been widely studied. Teglund et al. recently published a comprehensive review of Hh signaling and pancreatic cancer, among other forms of cancer [182]. In summary, most of the studies report the expression of both Shh and Ihh in pancreatic cancer. Apparently, Shh ligand is expressed transiently in the injured pancreas, but remains “on” in the transforming pancreas (intraepithelial neoplasia PanIN) suggesting aberrant Hh pathway activation in the pancreas contributing to cancer development. However Hh signaling alterations are not sufficient cause cancer and mutation of oncogenes, i.e. Kras, are necessary for pancreatic cancer development.

There is little information about Hh signaling in the liver. Hepatic stellate cells and liver epithelial progenitor cells express and respond to Hh ligands [183–185]. Activation of Hh signaling has been reported in hepatocellular carcinoma [186,187] and Hh signaling in the liver appears to be related to EMT involving the transformation of quiescent stellate cells to myofibroblastic stellate cells, via the reduction of Hip expression, as evidenced in a model of induced cirrhosis [188]. Thus the development of hepatocellular carcinoma involves cross-talk between the Wnt and TGFβ pathways and has not been linked exclusively to alterations in Hh signaling.

4.5. Future directions

Hh signaling is extremely complex. Recent studies have highlighted the importance of different pathway members, namely Smo, Sufu, Kif7, Disp1, and other participants, such as primary cilia, proteoglycans, and heparan sulfate, in Hh signaling regulation. Nevertheless, we still lack a complete understanding of the mechanisms involving Hh signaling.

The roles of Ihh and Dhh in the gastrointestinal mucosa, and mesenchyme have yet to be resolved. There are reports of Ihh and Dhh expression in the gastrointestinal epithelium, but one should exercise caution when analyzing these often-contradictory studies. Most of the antibodies commercially available for Hh ligands cross-react with the other family members, and reports using antibodies against Ptch1 and Gli proteins are not consistent, nor are those antibodies commercially available. Arguably, the best tool available to analyze Hh mRNA expression in vivo is in-situ hybridization. This technology has proven very powerful, in particular to establish the localization of Hh ligand and signaling molecules, e.g., Shh in the stomach, in contrast to Ihh expression in the distal intestine and colon. Nevertheless following the corresponding proteins has been problematic using antibodies. Transgenic reporter mice have also provided important insight; however, complete null mice are not suitable for adult tissue analysis due to the critical role of Hh signaling in development. Conditional and inducible deletion models are available for analysis in the intestine due to the existence of well-characterized tissue specific promoters to express Cre recombinase. However, stomach-specific promoters are limited, making development of deletion models a challenge for the field.

Finally, the duration of Hh signaling might be tissue specific. In the skin, Shh is transiently expressed in periods of hair active growth when Hh signaling is intermittent [189]. During development, Hh signaling is active in the intestinal tract, but in the adult, intestinal Hh signaling is due to Ihh rather than Shh. Shh expression is constant in the stomach, as are Ihh and Dhh. Depletion of Shh expression results in an increase in Ihh expression without apparent functional overlap [126]. The turnover time for each cell type in the stomach is different, ranging from 4 days for mucous cells to close to 60 days for parietal cells [190–192]. Shh is important in proliferation and the differentiation of gastric cell types but, to our knowledge, there are no studies on the expression of different Hh ligands over time in the adult gastric mucosa. Hh proteins may exert differential roles related to specific cell turnover cycles, but that concept has yet to be explored.