Intermuscular tendons are essential for the development of vertebrate stomach

The gastrointestinal (GI) tract is a remarkably complex, three-dimensional,specialized system derived from a simple tubal structure. The GI tract is composed of three germ layers: mesoderm (which forms the smooth muscle layer),endoderm (which forms the epithelial lining) and ectoderm [which includes the enteric nervous system (ENS)]. The primitive gut is initially a straight tube. As they develop, each region of the gut is characterized by a unique morphology discernible by gross and microscopic examination. These tissues show regional specific differentiation along the anteroposterior axis. This regionalization is maintained throughout life and is essential for the normal function of the adult gut. Candidate factors involved in gut development include genes that were first identified in Drosophila. These comprise homeotic (Hox and Nkx) and secreted factors [bone morphogenetic protein (BMP) and Hedgehog] (de Santa Barbara et al., 2002).

The motility of the GI tract is ensured by the correct coordination of the visceral smooth muscle cells (SMC) and the autonomous ENS. The ENS originates from neural crest cells that migrate from the dorsal region of the neural tube and colonize the whole gut to establish its innervation(Wallace and Burns, 2005). SMCs derive from the splanchnopleural mesoderm that will form the undifferentiated visceral mesenchyme(Roberts, 2000). Few have investigated the molecular mechanisms involved in the differentiation of the visceral mesenchyme into SMCs. SMCs are present in both vascular and digestive systems; however, the digestive tract is the most abundant contributor of SMCs in humans (Gabella, 2002). In the chick, the gizzard (muscular stomach or antrum) has a thick layer of smooth muscle that facilitates mechanical digestion, whereas the proventriculus (glandular stomach or fundus), which develops anteriorly to the gizzard, has only a very thin smooth muscle layer(Roberts, 2000). The intestines show modest development of smooth muscle layers. In the avian stomach, SMC differentiation is observed from embryonic day (E) 9. Thus, the chick stomach offers an ideal model in which to elucidate the molecular mechanisms that control visceral SMC differentiation. In the GI tract, Bmp4, a ligand that belongs to the transforming growth factor β(TGFβ) superfamily, is expressed in the mesenchyme of the whole chick gut with the exception of the gizzard (Roberts et al., 1998). When overexpressed, Bmp4 causes a reduction in the thickness of the smooth muscle layer of the stomach,demonstrating a regulatory role in gut muscle growth(Roberts et al., 1998). Conversely, the homeotic gene Bapx1 is expressed only in the chick gizzard mesenchyme and acts as a repressor of Bmp4, therefore modulating gizzard smooth muscle development(Nielsen et al., 2001). In addition, we have investigated the function of the BMP pathway during visceral SMC differentiation and found that aberrant modulation of BMP activity altered this process (de Santa Barbara et al.,2005).

In order to identify factors that trigger and control the differentiation of visceral SMC, we carried out a microarray screen to isolate candidate genes. We identified Scleraxis, a member of the basic-helix-loop-helix (bHLH) family of transcription factors, which is expressed in tendon cells of the stomach adjacent to the visceral SMC. We then used the avian retroviral system to specifically misexpress or inactivate Scleraxis in the stomach mesenchyme, and showed that Scleraxis expression defines the intermuscular tendon domains that are established in close association with the visceral SMC.

Fertilized White Leghorn eggs were obtained from Haas Farm, France. Tissues were staged according to Hamburger and Hamilton for early embryogenesis stages and by embryonic day (E) for gastrointestinal tract analysis(Hamburger and Hamilton,1951).

The Fgf8 (Brent and Tabin,2004), sFgfR2b (Mandler and Neubuser, 2004) and GFP(Moniot et al., 2004) viral constructs have been previously described. ShScleraxis associated with the mouse U6 promoter and full-length avian Scleraxis cDNA were cloned into the shuttle vector Slax and then subcloned into the RCAS(A)vector. Full-length avian Scleraxis cDNA was cloned in frame in the Slax-Engrailed vector and then subcloned into the RCAS(A) vector. All vectors were transfected into avian DF-1 cell lines, and viruses harvested and titered using standard techniques. To target the presumptive stomach mesenchyme,misexpression experiments were performed on stage 9-10 embryos, as previously described (Moniot et al.,2004).

Gizzards from stage 25 (referred to as E5 gizzards) were harvested in PBS solution. After collagenase treatment (Sigma) at room temperature for 12 minutes, we isolated the mesenchymal layer using fine forceps (Simon-Assman and Kédinger, 2000). Individual mesenchymal cells were plated on dishes and kept in culture for 24 (E5+1D) and 72 hours (E5+3D) in DMEM, 10% fetal bovine serum in the absence or presence of the different avian retroviruses.

In situ hybridization experiments on whole tissues/embryos and paraffin sections were carried out as previously described(Moniot et al., 2004). Different chick templates for antisense riboprobes were obtained by PCR amplification using specific primer sets (details of primers are available on request). The following plasmids were used: αSMA, Type I Collagen, Fgf7, Fgf10, Fgfr1(Edom-Vovard et al., 2001), Fgfr2 (Brent and Tabin,2004), Fjx (Yamaguchi et al., 2006), Scleraxis (gift from D. Duprez), Sox10 (Moniot et al.,2004) and Tenomodulin. Immunohistochemistry was performed on paraffin sections using polyclonal antibodies against αSMA(Sigma) and Phospho-Histone H3 Ser10 (Upstate), and monoclonal antibodies against Type I Collagen, GAG and Decorin [Developmental Studies Hybridoma Bank(DSHB)].

In cellulo in situ hybridization was performed as previously described(Gregoire et al., 2006). Methyl violet staining was used as described(Bi et al., 2007). Immunofluorescence was performed using monoclonal antibodies directed against Tenascin (DSHB) and Type I Collagen (DSHB), and polyclonal antibodies against Caldesmon (Sigma), Desmin (Sigma) and Sox9(Moniot et al., 2004). The Alexa 488 anti-mouse and Alexa 555 anti-rabbit secondary antibodies(Invitrogen) were used, and nuclei were stained with DAPI (Molecular Probes). Cells were mounted in FluorSave reagent (Calbiochem).

For microarray experiments and quantitative RT-PCR amplification, RNAs were extracted with the RNeasy Kit (Qiagen). Biotinylated complementary RNAs were hybridized to the Affymetrix GeneChip Chicken Genome Arrays using standard manufacturer's protocols (Affymetrix, IRB, CHU Montpellier, France). Fluorescence intensities were quantified and analyzed using the GCOS software(Affymetrix; see Table S1 in the supplementary material). Scleraxisexpression was quantified by quantitative RT-PCR amplification using LightCycler technology (Roche Diagnostics; 95°C for 10 seconds, 60°C for 5 seconds, 72°C for 10 seconds). PCR primers are available on request. mRNA values were determined by LightCycler analysis software (version 3.1),according to the standard curves. Data were represented as the relative mean level of Scleraxis expression relative to 18S standard expression.

For electron microscopy, tissues were immersed in a solution of 3.5%glutaraldehyde in phosphate buffer (0.1 M, pH 7.4) overnight at 4°C. They were then washed and post-fixed in 1% osmic acid plus 0.8% potassium ferrocyanide in the dark at room temperature for 2 hours. After washing,tissues were dehydrated in graded ethanol solutions and embedded in EmBed 812 DER 736. Thin sections (85 nm; Leica-Reichert Ultracut E) were collected. Sections were counterstained with uranyl acetate and lead citrate, and observed using a Hitachi 7100 transmission electron microscope at the CRIC facility (C. Cazevieille, Montpellier, France).

Images of whole-mount tissues and paraffin sections were collected with a Nikon DXM1200 camera connected to a Nikon Multizoom AZ100 microscope.

In order to identify factors that play a role at the onset of differentiation of the visceral SMC, we used microarrays to compare the gene expression profiles of undifferentiated E6 and differentiated E9 chick gizzards (Fig. 1A; see also Table S1 in the supplementary material). We observed a higher expression of well-known smooth muscle markers, such as Caldesmon, Desmin, Smooth Muscle Myosin, Smooth Muscle Actin, and of more recently discovered ones, such as Smoothelin and SRF in E9 gizzards(Wallace and Burns, 2005; Niessen et al., 2005; Mericskay et al., 2007). We also identified a new cluster composed of genes (i.e. Scleraxis, Decorin,Tenomodulin and Type XII Collagen) that are associated with the development and differentiation of tendon tissues(Tozer and Duprez, 2005).

In this study, we focused on Scleraxis, a member of the bHLH family of transcription factors previously reported as an early marker of tendons and tenocytes (Edom-Vovard et al.,2001; Schweitzer et al.,2001). We next examined the expression profile of Scleraxis during GI tract development and found it expressed in two specific subdomains of the stomach mesenchyme from E6 to E9, and at the boundaries of these two domains after SMC differentiation (E12; Fig. 1B). At E9, Scleraxis was expressed also in the caecum, a structure that separates the colon from the small intestine (see Fig. S1 in the supplementary material).

The mesenchyme of the avian embryonic stomach is composed of visceral SMC(that expresses αSMA, a SMC-specific factor) and intercalated ENS cells (positive for Sox10) to allow autonomous contraction. However, the expression of Scleraxis and αSMA or Sox10 was mutually exclusive (Fig. 1C), suggesting that Scleraxis is expressed neither in visceral SMC nor in ENS cells. Analyses by light microscopy of the embryonic gizzard showed a homogenous visceral mesenchymal structure with migrating ENS cells in the outer layer at E6 (Fig. 1D). Conversely, at E12, the gizzard consisted of two well-differentiated smooth muscle areas both associated with two connective structures (Fig. 1D), which, at E9, appeared to be close to the differentiating SMC(Fig. 1E). These connective tissues were positive for Scleraxis, Tenomodulin (another gene identified by our microarray screen) and other tendon markers, such as Type I Collagen (Ros et al.,1995), Four-jointed(Yamaguchi et al., 2006) and Decorin, all of which showed an expression pattern overlapping with that of Scleraxis (Fig. 1C,E)in these two connective tissues. Conversely, Scleraxis and αSMA were detected in two specific and mutually exclusive domains(Fig. 1E). Furthermore, we found that, in the mouse, Scleraxis expression was limited to two small domains in the antrum that correspond to the avian gizzard(Fig. 1F).

An early study highlighted the intimate relation between muscle bundles and these two connective structures in the adult avian gizzard(Watzka, 1932); two centra tendinea were observed in the adult gizzard and were characterized as being rich in collagen fibrils (McLelland,1979; Gabella,1985). Attachment of two skeletal muscles through a central structure was also described in the diaphragm. This was considered to be another category of tendon and was named an intermuscular tendon(Ackerman and Greer, 2007; Murchison et al., 2007). Taken together, these data indicate that, in the stomach, Scleraxisexhibits a restricted expression pattern that defines two tendons closely associated to the two visceral smooth muscles(Fig. 1F). We define them as intermuscular tendons.

Previous studies have shown that the fibroblast growth factor (FGF)signaling pathway is required for the induction of Scleraxisexpression in somitic tendon progenitors and for its maintenance in chick limb tendons (Brent and Tabin, 2004; Edom-Vovard et al., 2002). In order to identify the FGF ligand(s) and receptor(s) that could be responsible for the initiation of Scleraxis expression in E6 stomach, we quantified by RT-PCR their levels in E6 gizzards, and detected robust expression of eight FGF ligands and three FGF receptors(Fig. 2A). Whole-mount in situ hybridization revealed that Fgf7, Fgf10, Fgfr1 and Fgfr2were expressed in the gizzard mesenchyme from E6(Fig. 2B; see also Fig. S2 in the supplementary material). The other detected FGF ligands were expressed mainly in the epithelial layer (our unpublished data). We then investigated whether the FGF signaling pathway could regulate Scleraxisexpression. With this aim, we perturbed the FGF pathway by misexpressing an inhibitory form of the mouse FGFR2 IIb receptor (referred to as sFgfR2b) that is secreted and preferentially sequesters the FGF ligands 1, 3, 7 and 10(Mandler and Neubuser, 2004),and then monitored Scleraxis expression by in situ hybridization and quantitative RT-PCR (Fig. 2C,D). In the gizzard, infection with RCAS-sFgfR2b retroviruses led to a downregulation of Scleraxis expression and a size reduction of the two specific domains (Fig. 2C), whereas expression of RCAS-GFP, a negative control, had no effect on Scleraxis (compare Fig. 2C with Fig. 1B). Efficient retroviral infection of the mesenchyme was confirmed by immunohistochemistry using antibodies against the avian retroviral GAG protein(α3C2), or by in situ hybridization using a riboprobe against the avian retroviral envelop gene (env; data not shown). We monitored the Scleraxis mRNA level by quantitative RT-PCR and observed a strong decrease in RCAS-sFgfR2b in the stomach; this finding was supported also by the observed reduction of Fgf10 mRNA in this condition(Fig. 2D). Conversely, ectopic activation of the FGF signaling pathway along the GI tract following RCAS-Fgf8 misexpression caused multiple morphological defects, mainly in the proventriculus, the gizzard and the caecum(Fig. 2E; see also Fig. S3 in the supplementary material). RCAS-Fgf8-infected stomachs showed ectopic expression of Scleraxis associated with other tendon markers,such as Type I Collagen and Tenomodulin(Fig. 2E; Figs S4 and S5 in the supplementary material). Moreover, in RCAS-Fgf8-infected stomachs, we observed reduced expression of αSMA in the areas of ectopic expression of the tendon markers (Fig. 2E;Fig. S3 in the supplementary material). To analyze the effect of FGF activation on cell proliferation, we used antibodies directed against phosphorylated Histone 3B (PH3), a standard cell cycle marker of the G2/M transition (Fig. 2F). We did not observe significant differences in the number of proliferative cells in the gizzard SMCs following infection of RCAS-Fgf8 or RCAS-GFP,indicating that FGF activation does not induce a global upregulation of the mitotic potential. These data suggest that the FGF signaling pathway controls positively the establishment of tendons in the developing stomach, and recruits mesenchymal cells towards a tendon fate.

Scleraxis belongs to the bHLH family of transcription factors and is one of the rare transcription factors expressed at the onset of tendon development(Edom-Vovard et al., 2001; Schweitzer et al., 2001). In order to investigate more directly the role of Scleraxis in stomach development, we used a loss-of-function technique that involved delivering ShScleraxis to the chick stomach in vivo(Fig. 3A,B)(Harpavat and Cepko, 2006). First, we tested the efficiency of RCAS-ShScleraxis in primary cells derived from stomach mesenchyme and observed a robust reduction of Scleraxis mRNA (Fig. 3C) without any side effect on the identity of the infected cells(data not shown). In embryonic stomachs, RCAS-ShScleraxis led to a strong downregulation of Scleraxis expression(Fig. 3D) and to some minor defects, such as a smaller gizzard and a straight proventriculus(Fig. 3D). Strong retroviral infection was correlated with decreased expression of Scleraxis, Type I Collagen and Tenomodulin(Fig. 3E), suggesting a reduction in size of the tendon domains in the developing stomach. Conversely,we observed an increase of the territory labeled by αSMA, a SMC marker(Fig. 3E).

We next used a gain-of-function approach and ectopically misexpressed full-length Scleraxis along the GI tract(Fig. 4). Although we misexpressed Scleraxis in different regions of the digestive system,we observed effects only in the stomach(Fig. 4). Scleraxismisexpression induced a gross phenotype characterized by an aberrantly dilated gizzard associated with a curved proventriculus(Fig. 4A). RCAS-Scleraxis expressing cells were highly concentrated and organized around territories of endogenous expression(Fig. 4A), whereas the muscle domains were strongly reduced, as revealed by an αSMA riboprobe(Fig. 4A). Although RCAS-Scleraxis expression was correlated with αSMA inhibition,we did not observe an ectopic induction of Type I Collagen, as we had in the case of RCAS-Fgf8 misexpression (compare Fig. 4B with Fig. 2E).

Scleraxis is considered to be a gene activator and recently Type I Collagen has been identified as a target in tendon fibroblasts(Lejard et al., 2007). Therefore, we thought that by converting Scleraxis into a transcriptional repressor, we could block the expression of target genes and counteract its activator function. With this aim, we fused the Engrailed repressor domain in frame to Scleraxis and misexpressed RCAS-Scleraxis-Engrailed in the stomach to repress endogenous Scleraxis function. In embryonic stomachs, RCAS-Scleraxis-Engrailed led to some minor defects,comparable to those observed in RCAS-ShScleraxis stomachs (data not shown). Type I Collagen expression was strongly inhibited in the infected tendon domain and this inhibition was associated with defects in the tendon structure (Fig. 4C). Recently, others genes known to be expressed specifically in the tendons, such as Tenomodulin and Type XIV Collagen, were demonstrated to be targeted by Scleraxis in the limb(Shukunami et al., 2006; Murchison et al., 2007). In contrast to the limb tendon, we found that Type XIV Collagen was not expressed in the intermuscular tendon but only in gastric ENS cells (see Fig. S6 in the supplementary material). Using quantitative RT-PCR to analyze the expression levels of Tenomodulin, we observed that Scleraxismisexpression in the gizzard upregulated Tenomodulin, whereas ectopic expression of Scleraxis in the proventriculus did not(Fig. 4D). Moreover, Scleraxis-Engrailed expression in the gizzard strongly repressed Tenomodulin expression (Fig. 4D). These findings demonstrate that, during the development of the intermuscular tendon, Scleraxis is upstream of Type I Collagenand Tenomodulin, and strongly suggest that Type I Collagenand Tenomodulin are in vivo targets of Scleraxis in the stomach. However, Scleraxis alone cannot induce ectopic expression of Type I Collagen, suggesting the necessity of an additional interacting partner(s).

We then analyzed by electron microscopy the ultrastructure of visceral SMC after infection with different retroviral constructs(Fig. 5). In control gizzards(Fig. 5A), tendon cells showed predominantly rough endoplasmic reticulum and mitochondria, and were surrounded by collagen fibrils; visceral SMC, by contrast, presented mainly intracellular actin formation and thick cytoplasm. In addition, tendon cells were elongated, whereas SMC were rounded. In RCAS-Fgf8 gizzards, SMC were elongated with numerous filopodia and no intracellular actin formation,but numerous organelles and ectopic collagen fibrils were observed in the intercellular region. These ectopic fibrils were not observed in RCAS-Scleraxis SMC. However, ectopic expression of Scleraxisin the gizzard SMC inhibited the formation of intracellular actin and was associated with the presence of numerous organelles(Fig. 5A). Tendon cells, in which sFgfR2b, ShScleraxis and Scleraxis-Engrailed were misexpressed, showed an absence or a strong reduction of collagen fibrils in the intercellular region (Fig. 5B). Moreover, in each of the three conditions ectopic cells with thick cytoplasm, which are not observed in normal tendon cells, were present(compare Fig. 5B with 5A).

In summary, our data indicate that tendon structures are important for the regulated development of the stomach, and that their differentiation is closely coordinated with that of visceral smooth muscles.

We showed that tendon and smooth muscle domains are closely associated in the stomach and that activation of the FGF pathway in the anterior and posterior parts of the stomach mesenchyme induces ectopic tendon domains(Fig. 2), suggesting that modulation of this pathway in the visceral mesenchyme is sufficient to pattern the tendon. To evaluate the capacity of undifferentiated visceral mesenchyme to give rise to tendon cells, we set up primary cell cultures derived from stage 25 gizzards before the onset of Scleraxis expression in the stomach. Throughout the study, we refer to stage 25 gizzard mesenchyme as E5 gizzard mesenchyme. With this aim, we adapted a reliable technique that allowed us to enzymatically separate the gizzard mesenchyme from its endodermal counterpart (Fig. 6A) (Simon-Assman and Kédinger, 2000). Next, E5 mesenchymal cells were cultured for one (E5+1D) and three days (E5+3D). At E5+1D, we observed that some isolated, scattered cells were positive for Scleraxis expression; at E5+3D, we observed the presence of several colonies that were all positive for Scleraxis expression(Fig. 6B). All Scleraxis-positive colonies were also visualized by staining with Methyl Violet, a specific histological tendon dye(Fig. 6C)(Bi et al., 2007). We then analyzed these colonies and found that they strongly expressed Type I collagen and Tenascin (tendon markers) (Tozer and Duprez, 2005), faintly expressed Desmin (mesenchymal derived cell marker) but did not express Caldesmon (SMC marker) or Sox9 (chondrocyte marker) (Lefebvre et al.,1997) (Fig. 6D). Conversely, we observed robust expression of Caldesmon and Desmin in the cells adjacent to the colonies, suggesting that visceral SMC organized around these structures (Fig. 6D).

We have shown that modulation of the FGF pathway and deregulation of Scleraxis expression are associated with reciprocal changes in visceral SMC and tendon cells of the stomach. We therefore assessed the colony-forming efficiency of E5 gizzard mesenchymal primary cultures in which we deregulated Scleraxis expression or the FGF pathway. The colony-forming efficiency was evaluated by Methyl Violet staining(Fig. 6E). We observed that RCAS-Scleraxis and particularly RCAS-Fgf8 misexpression increased the number of colonies (Fig. 6E). RCAS-ShScleraxis did not affect the number of colonies, whereas RCAS-Scleraxis-Engrailed strongly inhibited their formation. However, RCAS-ShScleraxis colonies were smaller, with few cells compared with the control colonies (data not shown). We also analyzed the proliferation rate in these different cultures by immunohistochemistry using anti-PH3 antibodies (data not shown) and by quantitative RT-PCR to evaluate the level of Pcna RNAs(Fig. 6F) and did not observe,like in vivo, strong changes that could be responsible for the differences in colony formation.

These data suggest that E5 gizzard undifferentiated mesenchyme can differentiate into at least two distinct cell types: visceral SMC and tendon cells. These data also indicate that the FGF pathway and Scleraxis are essential for the formation and differentiation of intermuscular tendons.

In this work, we show that embryonic stomach is characterized by the presence of two tendon domains (intermuscular tendons) close to the gastric visceral smooth muscle structures. Moreover, we propose that coordinated development of the tendon and gastric smooth muscle structures is necessary to ensure the correct development of the stomach. This notion is supported by the findings that: (1) the inhibition of tendon formation induced stomach malformations; and (2) the increase of tendon territories happened at the expense of those of visceral SMCs and impaired their differentiation in vivo. Moreover, we show that undifferentiated stomach mesenchyme can give rise to both SMC and tendon cells, suggesting that the development of the intermuscular tendon cells progresses an autonomous way from the stomach mesenchyme, through the FGF pathway and under Scleraxis control.

The majority of smooth muscle forms continuous structures devoid of attachments. However, a few smooth muscles are attached to rigid structures(i.e. the anococcygeus, the rectococcygeus and the costouterus muscles)(Gabella, 1985). The rectococcygeus muscle, for instance, is attached to the anterior surface of the second and third coccygeal vertebrae, and is inserted onto the posterior surface of the rectum through anchoring tendons. In this study, we describe two tendons that connect two smooth muscle bundles in the stomach. The connection of two muscle bundles through a connective structure, which is defined as an intermuscular tendon, has been previously described; for example, in the diaphragm (Ackerman and Greer, 2007). The localization of two intermuscular tendons in the stomach suggests that their physiological role is to diminish the high tension generated by the periodic contraction of the powerful stomach smooth muscle structures.

In this study, we found that some specific tendon markers (i.e. Scleraxis, Tenomodulin, Four jointed, Type I Collagen and Decorin) are associated with the development and differentiation of these structures. However, other limb tendon-specific markers are not expressed. Indeed, a putative target gene of Scleraxis, Type XIV Collagen is not expressed in the intermuscular tendons but is expressed in the gastric ENS cells. These molecular differences between intermuscular tendons and force-transmitting or anchoring tendons suggest that the generation of tendon structure or the organization of the tissue attachment in these tendon subgroups requires molecular modulation.

At E6, Scleraxis expression defines two domains that give rise to the tendon structures of the stomach (Fig. 7). These two domains are induced by the mesenchymal FGF signaling pathway. We also demonstrate that inactivation of Scleraxisexpression extinguishes the tendon domains, whereas ectopic expression blocks the differentiation of the visceral mesenchyme into SMC. Tendons are defined as the connective tissues that connect/attach muscle to bone (musculoskeletal system) and muscle to muscle (diaphragm system)(Tozer and Duprez, 2005; Ackerman and Greer, 2007). Tendon cells can have different embryological origins; for instance, body wall tendons derive from one specific somitic compartment, the syndetome(Brent et al., 2003), and craniofacial tendons from cranial neural crest cells(Crane and Trainor, 2006). No embryonic origin was demonstrated for the intermuscular tendons in the diaphragm (Ackerman and Greer,2007). The digestive tract mesenchyme derives from the splanchnopleural mesoderm and is also colonized by neural crest-derived(Le Douarin and Teillet, 1973; Barlow et al., 2008) and mesothelium-derived vascular SMC (Wilm et al., 2005) cells. Chick/quail chimera experiments demonstrated that colonization by neural crest-derived cells of the stomach gives rise to ENS cells (Le Douarin and Teillet,1973) and not to the tendon domains we identified in this work. An early study of the muscle-tendon structures in the avian gizzard noted the intimate relationship between muscle bundles and tendons in this organ, and suggested that both derive from the same cell type(Watzka, 1932). Our finding that mesenchymal FGF signaling triggers Scleraxis expression and,consequently, the determination of the two tendon domains in vivo and in primary cell culture suggests that, upon FGF activation, selected stomach mesenchymal cells are primed to differentiate into tendon cells and to express Scleraxis.

As previously reported (Brent and Tabin,2004; Edom-Vovard et al.,2002), we found that the mesenchymal FGF signaling pathway positively regulates Scleraxis expression and induces differentiation of the two tendon structures and the expression of differentiation markers,such as Type I Collagen and Tenomodulin, in different parts of the digestive tract. Conversely, Scleraxis is essential for the development of the tendons and Tenomodulin expression. However, Scleraxis misexpression alone cannot induce ectopic expression of Type I Collagen, whereas the transcriptional repressor Scleraxis-Engrailed inhibits Type I Collagen expression in vivo. Scleraxis is a bHLH transcription factor that can interact with the bHLH factor E47 to regulate the expression of Type I Collagen in tendon fibroblasts (Lejard et al.,2007). All these data suggest that Scleraxis requires additive transcription factors or interacting partners to control the differentiation process initiated by activation of the FGF pathway. We can also imagine that Scleraxis proteins present in the musculoskeletal or in the intermuscular tendons might interact with different tissue-specific partner(s) to reinforce these tissue specificities. In addition, other signaling pathways activated by FGF can cooperate with Scleraxis to induce tendon differentiation. Different studies showed that the FGF-WNT regulatory network controls mesenchyme development. Recently, mesenchymal FGF signaling has been reported to regulateβ-catenin-mediated WNT signaling in lung mesenchyme(Yin et al., 2008). The canonical Frizzled receptor Fz8 is expressed in the stomach(Theodosiou and Tabin, 2003)and its expression pattern overlaps with that of Scleraxis, which suggests that Scleraxis and the WNT canonical pathways are induced after activation of the FGF pathway. In future studies we will try to identify the WNT ligand that interacts with Fz8 and that might enable the cooperation with Scleraxis.

In summary, we show that the vertebrate stomach harbors two tendon domains located in the antrum/muscular stomach, which are closely associated with the visceral smooth muscle structures. The possible physiological function of these domains might be to ensure elasticity of the stomach during contraction of the two massive visceral muscles.