The serosal mesothelium is a major source of smooth muscle cells of the gut vasculature

Our understanding of the molecular mechanisms regulating blood vessel formation has advanced greatly in recent years(Carmeliet et al., 1996; Cleaver and Melton, 2003; Ema and Rossant, 2003; Jain, 2003; Lammert et al., 2003). Still,many cellular mechanisms remain unresolved concerning the generation of blood vessels in the vertebrate embryo. One important question concerns the origin of cells that make up vessels. During blood vessel development in the embryonic trunk, body wall and limbs, as well as in extraembryonic tissues,endothelial tubes are thought to induce locally derived mesodermal mesenchyme to differentiate into smooth muscle and pericytes of the vessel wall(Cleaver and Krieg, 1999; Gerhardt and Betsholtz, 2003; Hirschi and Majesky, 2004; Majesky, 2003; Saint-Jeannet et al., 1992). This mechanism is generally accepted as the major form of blood vessel development in vertebrate embryos.

Studies of blood vessel development to coelomic organs, such as those encased in the pericardial and peritoneal cavities, are limited and have focused primarily on the heart. These studies have shown that the mesothelial covering of the embryonic heart [the proepicardium (PE) and its derivative the epicardium] is a major source of cells to the coronary system(Dettman et al., 1998; Manner, 1993; Manner et al., 2001; Mikawa and Fischman, 1992; Reese et al., 2002). Although debate remains about whether all coronary vasculogenic cells are derived from the PE/epicardium (Cox et al.,2000; Drake et al.,1997; Munoz-Chapuli et al.,2002), it is well established that cells of this mesothelium undergo EMT, migration and subsequent differentiation into coronary vessels(Dettman et al., 1998; Mikawa and Gourdie, 1996; Perez-Pomares et al., 2002; Vrancken Peeters et al.,1999). This form of blood vessel development is thought to be a unique mechanism, as its progenitors are derived from an epithelial mesothelium that subsequently produces vasculogenic mesenchyme(Dettman et al., 1998; Gittenberger-de Groot et al.,1998; Wada et al.,2003b).

The origin of vasculogenic cells to the alimentary canal, which is encased in the peritoneal coelom, is unknown. The structure of the gut is conserved among vertebrates and consists of the epithelial mucosa, submucosa, muscularis externa and serosa (serosal mesothelium and underlying mesenchyme)(Netter, 1997; Roberts et al., 1996). The gut is formed by simple embryonic structures: endoderm gives rise to the epithelial mucosa while the other layers are thought to arise from the lateral splanchnic mesoderm (Kiefer,2003). Additionally, neural crest cells migrate into the gut and differentiate into neurons of the enteric plexus(Young et al., 2000; Young and Newgreen, 2001). Still, the origin of the major vessels to the gut is not understood.

Vascular systems of the heart and gut have several striking similarities. Most significantly, the major vessels to heart and gut run on the surface of the organ and are intimately associated with their mesothelial covering(Netter, 1997). These characteristics have led us to test whether a conserved developmental mechanism, similar to coronary vasculogenesis, accounts for blood vessel formation in the gut. In the current study, we provide molecular genetic and experimental data demonstrating that the serosal mesothelium is the major source of vasculogenic cells of the gut. Our data show that the gut is initially devoid of its mesothelial covering and its surface blood vessels. Soon after formation of the tubular gut, non-resident cells migrate to and over the gut to form the serosal mesothelium. Subsequently, a sub-population of these cells undergoes epithelial-mesenchymal transition (EMT), migrates within the gut and gives rise to vascular smooth muscle cells that populate all major vessels of the gut. Our data indicate that the formation of coelomic mesothelium, whether it be epicardial or serosal, is coupled to vasculogenesis, and suggest that elements of a common developmental mechanism regulate the generation of blood vessels to the heart and gut.

The strategy we used to construct the WT280Cre YAC was similar to that described for the WT280LZ YAC (Moore et al., 1998), except that a nuclear localization signal (NLS)-tagged Cre recombinase open reading frame (ORF) was cloned into the SacII site in the 5′ UTR of the human WT1 first exon instead of theβ-Galactosidase ORF. The resultant WT280Cre YAC was isolated and microinjected into fertilized mouse oocytes by following published protocols(Schedl et al., 1993). Transgenic founders were identified by assaying tail DNA with Cre-specific PCR primers, and one such founder (denoted AG11) was found to direct mesothelial expression in embryos derived from matings with Rosa26R reporter mice(Soriano, 1999). An analysis of genomic DNA with three PCR primer pairs targeted to different regions of WT280Cre revealed that AG11 mice retained the short (SVA) arm of the YAC and the Cre gene, but had lost the long (LVA) arm of the YAC (see Moore et al., 1998). Embryos were used from timed matings between WT280Cre transgenic males (referred to as Wt1-Cre animals) and Rosa26R homozygous females, or ICR wild-type mice, where the day of plug is E0.5.

Embryos were either embedded in OCT and snap frozen directly after dissection, or fixed in 4% paraformaldehyde (PFA), protected in 30% sucrose overnight, and then OCT embedded and snap frozen. Frozen sections were generated at 7 μm on a Leica Cryostat. PFA-fixed tissue was used for the detection of β-galactosidase (β-Gal) protein. Immunohistochemistry was performed according to standard protocols(Bader et al., 1982; Reese et al., 1999; Wada et al., 2003a). The following primary antibodies were used: polyclonal Wt1 (C-19) at a dilution of 1:200 to 1:500 (sc-192, Santa Cruz); monoclonal Wt1 (6F-H2) at 1:50 (M3561,Dako); polyclonal β-Gal at 1:5000 (55976, Cappel/ICN); polyclonal cytokeratin at 1:500 to 1:2000 (Z0622, Dako); monoclonal α-SMA (1A4) at 1:100 to 1:200 (A2547, Sigma); monoclonal Pecam at 1:50 (550274, Pharmingen). Mouse monoclonal antibodies to be incubated on PFA-fixed tissue were directly labeled using the Zenon labeling kit (Molecular Probes), according to manufacturer's instructions. Secondary antibodies were Alexa fluorophore-coupled (Molecular Probes) and were used at a dilution of 1:2000.

Whole-mount lacZ staining was performed according to standard protocols (Hogan et al.,1994). Staining was usually allowed to continue overnight,especially in younger embryonic stages. In larger embryos and mice, the gut and other tissues of interested were either dissected out, or the body wall was opened to fully expose the intestinal organs. For histological analysis, lacZ-stained specimens were dehydrated in Isopropanol and subsequently embedded in paraffin. Serial sections (7 μm) were collected and counterstained with Eosin. Hematoxylin and Eosin staining of sections from unstained tissues and embryos was performed following standard protocols.

E12.5 embryos were freed from extraembryonic tissues, but remained attached to the placenta. CCFSE[5-(and-6)-carboxy-2′,7′-dichlorofluorescein diacetate,succinimidyl ester `mixed isomers', Molecular Probes] was diluted to 24 μM in sterile PBS, and, through a small opening in the ventral body wall covering the herniated intestines, injected with a mouth pipet into the cavity surrounding the guts. After CCFSE application, embryos were incubated for 1 hour at 37°C and 5% CO₂ in DMEM/10% FCS under sterile conditions. Subsequently, embryonic guts were isolated and cultured in 4-well dishes (Nunc) in Optimem (Life Technologies), supplemented with 1 mM L-Glutamine (Life Technologies) and a Penicillin/Streptomycin antibiotic mixture at 37°C and 5% CO₂(Natarajan et al., 1999). Embryonic gut explants were fed every 2 days. At least two embryonic guts were fixed at the same time point, day 0 to day 3, with fresh 4% PFA for 30 minutes on ice, washed in PBS, protected overnight in 30% sucrose, OCT embedded and snap frozen. Control explants without CCFSE treatment showed no difference in tissue integrity and viability (not shown).

Images of sections labeled with CCFSE and β-Gal, or β-Gal and Pecam/SMA were used for quantification. Images were viewed in Photoshop, where the color channels for blue (DAPI), red and green were separated for the counting of cell populations. Percentages were collected for each vessel, and an average percentage calculated for the number of vessels counted.

In order to establish when serosal mesothelium is formed over the gut, and to follow its development, we employed immunohistochemical and histological analyses at selected stages of development. Both the Wilms tumor protein (Wt1)and cytokeratin have been reported as being expression markers for the epicardial mesothelium of the heart (Chan et al., 1988; Foley-Comer et al., 2002; Perez-Pomares et al., 1998; Watt et al.,2004), and also for the serosal mesothelium(Armstrong et al., 1993; Carmona et al., 2001; Foley-Comer et al., 2002; Moore et al., 1998). As seen in Fig. 1A,B, the epicardium of the heart was clearly identified by these two proteins. This pattern of Wt1 and cytokeratin expression served as a control for subsequent immunochemical and molecular genetic analyses of mesothelial formation in the gut.

At early stages of mouse embryogenesis [embryonic day (E) 8.5], the gut consists of two epithelial layers: endoderm and lateral splanchnic mesoderm. At E9.5-E10.5, the mesoderm of the gut has thickened into several cell layers by cell proliferation. Antibodies for Wt1 detected strong reactivity in cells of the nephrogenic mesoderm at E9.5 (Fig. 1C), as previously reported(Armstrong et al., 1993; Carmona et al., 2001; Moore et al., 1998). Still, at E9.5, the simple squamous serosal mesothelium that also expresses Wt1 protein was completely absent from the mesenteries and gut tube(Fig. 1C)(Armstrong et al., 1993; Carmona et al., 2001; Moore et al., 1998). At E10.5,Wt1-positive mesothelial cells were first detectable at the proximal base of the dorsal mesentery, whereas the gut tube was still devoid of a mesothelium,as evidenced by the lack of Wt1 and cytokeratin surface staining(Fig. 1D,E). The absence of a mesothelium at this stage was also confirmed by histology, as cells on the embryonic gut surface were irregularly shaped and arranged, and did not have a squamous phenotype (see Fig. S1 in the supplementary material). Residual cytokeratin staining in splanchnic lateral mesoderm of the mesentery(Fig. 1E) was observed at E10.5, as it loses its epithelial nature. One day later (E11.5), Wt1-,cytokeratin-positive cells were present on the gut surface and the peritoneal wall along the entire anteroposterior axis(Fig. 1F,G). Staining of these two markers was confined to the mesothelial lining of the gut and to the mesentery, indicating that, by E11.5, the entire gut was enclosed by mesothelium (Fig. 1F,G). Thus,serosal mesothelial cells are the only cell type to express Wt1 in the gut at this stage, as has been previously reported(Armstrong et al., 1993; Carmona et al., 2001; Moore et al., 1998). Histological analysis at this stage revealed that cells on the gut surface were arranged in a regular manner, although not yet with a squamous phenotype(see Fig. S1 in the supplementary material). By E13.5, however, these cells comprise a thin layer of simple squamous epithelium that is typical for the serosal mesothelium (see Fig. S1). Thus, Wt1- and cytokeratin-positive serosal mesothelial cells appear in a proximal to distal manner from E10.5 onwards; by E11.5 they have completely enclosed the embryonic gut, and by E13.5 they have fully differentiated into a simple squamous epithelium. Genetic lineage marking (see below) confirmed this pattern of differentiation. Once established, the serosal mesothelium was present over the gut surface throughout prenatal and postnatal life (see Fig. S1).

The major blood vessels to the gut run on its surface in association with the serosal mesothelium (Netter,1997). If formation of these vessels is coupled to formation of the mesothelium, they should not be present before its arrival. At early stages of gut development (E8.5-E12.5), an endothelial plexus associated with intestinal endoderm was revealed using Pecam (platelet/endothelial cell adhesion molecule)-specific antibodies(Bogen et al., 1992). This endothelial gut plexus extends into the dorsal mesentery beneath the splanchnic lateral mesoderm. Importantly, at these stages there are no blood vessels detectable on the surface of the gut, as visualized by the absence of Pecam staining in this area (Fig. 2A). Beginning at E13.5, endothelial cells form tubes that extend from the gut plexus to the surface of the gut(Fig. 2B). Although these endothelial tubes penetrate the mesodermal wall of the gut, they are devoid of vascular smooth muscle, as demonstrated by the absence of anti-α-smooth muscle actin (SMA) staining (Fig. 2C). Note that anti-SMA staining visualized visceral smooth muscle in the gut wall within the same sections(Fig. 2C). At E16.5 and later stages, SMA is expressed in vascular smooth muscle cells that have been assembled by endothelial vessels in the mesentery and on the surface of the gut (Fig. 2D,E); the earliest SMA expression in vascular smooth muscle cells of the mesentery and gut was detected at E16.0, uniformly along the anteroposterior axis (B.W. and D.M.B.,unpublished). These data demonstrate, first, that there are no surface blood vessels before the arrival of the serosal mesothelium, and, second, that the endothelial tubes first present on the gut surface extend from the endothelial plexus associated with the endoderm.

To determine the fate of mesothelial cells, we have employed a Wt1-Cre recombinase/Rosa26R system (see Materials and methods) to irreversibly mark Wt1-positive mesothelial cells and their descendants. To trace the lineage of mesothelial cells, whole-mount and immunohistochemical analyses at selected times in development, as well as in neonates and adults, were performed. Whole-mount analysis of E11-E12 embryos revealed no lacZ staining over the herniated intestines, whereas a strong reaction was present along the dorsal body wall and at the root of the mesentery(Fig. 3A). Immunohistochemistry for β-Gal protein on sections confirmed the lacZ-staining pattern at this stage (not shown). From E12.5 forwards then, lacZ-positive cells extend from the base of the mesentery,progressively over the intestines in proximal to distal manner(Fig. 3B,C). Again,immunohistochemistry confirmed the presence of robust β-Gal expression on the surface of the gut, and its colocalization with Wt1 protein(Fig. 3E). Note that at E12.5,β-Gal expression is not continuous over the most distal portions of the gut tube (Fig. 3E), whereas Wt1 protein is expressed in all mesothelial cells(Fig. 3E′,E′′). Thus, the pattern of lacZ/β-Gal staining recapitulated the endogenous pattern of Wt1 expression in the developing serosal mesothelium, although it was delayed by at most one day (compare with Fig. 1D-G) (Armstrong et al.,1993; Moore et al.,1998). At E18.5, lacZ expression was found covering the entire surface of the gut (Fig. 3D).

Previous studies have suggested that Wt1 expression is linked to cells undergoing EMT (Armstrong et al.,1993; Carmona et al.,2001; Pritchard-Jones et al.,1990). To determine whether Wt1-expressing mesothelial cells and their descendants undergo EMT, we performed co-immunostaining for β-Gal and Wt1. In case of EMT, the progeny of Wt1-expressing serosal mesothelial cells were anticipated to react with the β-Gal antibody, but not the Wt1 antibody, in the submesothelial space. As expected, at E13.5, the gut mesothelium was positive for both β-Gal and Wt1(Fig. 4A-C). Moreover, the submesothelial mesenchyme contained numerous cells positive for β-Gal but negative for Wt1. It should be noted that a very limited number of cells in the subserosal space was found to be positive for both Wt1 and β-Gal expression (Fig. 4A-C). Similar results were obtained at newborn stage, suggesting that at least a small portion of serosal mesothelial cells continuously undergo EMT(Fig. 4D-F).

The presence of β-Gal-positive cells in the submesothelial space of the gut suggests that these cells are the progeny of mesothelial cells that have undergone EMT. To experimentally confirm that surface serosal cells undergo EMT, we combined an explant culture system of the embryonic gut with genetic and in vivo labeling of serosal mesothelial cells(Natarajan et al., 1999). Embryonic guts from crosses between Wt1-Cre and Rosa26R were exposed to the lipophilic dye CCFSE, which has been well established to mark only surface cells (Morabito et al., 2001; Perez-Pomares et al., 2002; Perez-Pomares et al., 2004). CCFSE was placed locally on the herniated gut by injecting under the ventral body wall of E12.5 embryos, and, after one hour of incubation, the embryonic gut tubes, including the mesentery, were dissected out for subsequent culture in CCFSE-free medium. This approach should initially mark only surface serosal mesothelial cells (in localized areas), while cells undergoing EMT would be subsequently labeled in the delaminated underlying mesenchyme. In order to quantitatively delineate this process, we counted the number of surface and submesothelial CCFSE-labeled cells that were positive and negative forβ-Gal protein (Table 1). As expected, at the time of CCFSE application, only mesothelial cells in labeled spots on the surface of the gut were positive for CCFSE(Fig. 4G and Table 1; arrowheads define the area of CCFSE labeling). Note that the overwhelming majority (96.1%) of these cells were also β-Gal positive (Table 1). No sub-serosal mesenchymal cells were labeled, indicating the effectiveness of surface labeling and the lack of immediate EMT. After 24 to 48 hours of culture, CCFSE-marked cells in labeled patches were still found in the serosal mesothelium, but were also present in significant numbers in the submesothelial mesenchyme of labeled areas(Fig. 4H,J). Indeed, when CCFSE-labeled patches were analyzed, on average 25.5% of all CCFSE-labeled cells were found in the submesothelial space(Table 1). Of this group of CCFSE-labeled submesothelial cells, on average 74.4% were also positive for the β-Gal marker (Fig. 4J,K and Table 1). Two important findings arise from these results. First, CCFSE-labeled cells in the submesothelium reveal that a subset of mesothelial cells undergoes EMT in this in vitro system (Fig. 4H,J). This result was similar to our observations with serosal EMT in vivo (Fig. 4A-F). Second, a large majority (74.4%) of CCFSE-labeled cells in the submesothelial space are co-labeled for β-Gal (Fig. 4J,K and Table 1),confirming that these marked cells are descendants of serosal mesothelial cells. By contrast, Wt1 expression was largely confined to the mesothelial surface of CCFSE-labeled cultured guts, whereas CCFSE-positive cells were readily apparent in the connective tissue space(Fig. 4H,I), reiterating the in vivo situation (Fig. 4A-C). This indicates that Wt1 expression is downregulated as cells become mesenchymal, as has been previously reported(Carmona et al., 2001; Moore et al., 1999), whereasβ-Gal protein continues to be expressed. As expected, we also foundβ-Gal-positive cells that do not carry CCFSE, owing to the locally restricted uptake of the compound. In addition, we find a limited number of CCFSE-marked cells on the serosal surface (11.4%) and in the submesothelial space (25.6%) that are not reactive for β-Gal(Fig. 4J; Table 1). The implications of this finding are discussed below (see Discussion). Our analysis of CCFSE-labeled domains at random positions along the anteroposterior axis of the embryonic gut did not reveal any qualitative or quantitative differences in EMT. This indicates that EMT occurs in the same fashion independently from the position along the anteroposterior axis of the gut tube. Taken together,these data indicate that a subset of serosal mesothelial cells, like epicardial mesothelial cells, undergoes EMT.

We next determined the fate of mesenchymal cells derived from the serosal mesothelium using the Wt1-Cre/Rosa26R system. Beginning at E16, and at all later stages, lacZ-stained cells were readily observed in association with differentiating blood vessels running in the mesentery and gut(Fig. 5 and data not shown). In the adult, lacZ-positive cells marked all blood vessels of the mesentery and their branches that dive into the gut(Fig. 5A,B). Although lacZ staining was strong in cells of the arteries, weaker staining was found in the accompanying veins (Fig. 5B). Close inspection of lacZ-positive cells in blood vessels revealed the characteristic perpendicular arrangement of smooth muscle in arteries (Fig. 5C)(Cleaver and Krieg, 1999). Note that lacZ-positive cells were also seen in non-vasculogenic regions in all organs of the gut (Fig. 5B).

Wt1 protein is also expressed in the epicardium(Fig. 1A,B). Thus, as a control, we analyzed the hearts of newborn mice that carry both the Wt1-Cre and Rosa26R alleles for descendants of the Wt1-Cre expressing cells. lacZ expression revealed that progeny of the epicardial mesothelium contribute to coronary vessels (Fig. 5D), as has been previously described(Dettman et al., 1998; Mikawa and Fischman, 1992; Mikawa and Gourdie, 1996; Perez-Pomares et al., 2002; Vrancken Peeters et al.,1999).

To determine the identity of the lacZ-expressing cells of developing blood vessels in the gut, we performed immunohistochemistry forβ-Gal, coupled with markers for vascular smooth muscle and endothelium. Immunostaining for β-Gal also serves to corroborate the lacZ-staining pattern, as recent reports have suggested that lacZ analysis of β-Gal activity in tissues does not always fully recapitulate the β-Gal expression pattern(Couffinhal et al., 1997; Mahony et al., 2002). Using standard fluorescence microscopy, β-Gal protein was found in smooth muscle cells that were co-stained for SMA and desmin adjacent to the vessel lumen (Fig. 6A-D, see also Fig. S2 in the supplementary material). However, colocalization of Pecam- withβ-Gal-antibody was not observed to any significant degree(Fig. 6E-H). Of particular interest are three additional findings. First, Pecam-positive endothelial cells of the vessels are oriented perpendicularly with respect to the vascular smooth muscle cells (Fig. 6F-H), which in these images are labeled for β-Gal protein. Second, it should be noted that β-Gal protein is not expressed in the longitudinal and circumferential visceral smooth muscle layers, as seen in Fig. 6A-D. This indicates that the visceral smooth muscle is of different origin than the vascular smooth muscle of the gut. Third, we find cells that express β-Gal protein but are not associated with the vascular system of the gut(Fig. 6), suggesting that the serosal mesothelium contributes to other cell populations in the intestinal tract.

To further delineate the colocalization of β-Gal protein with endothelial and smooth muscle markers, we followed two approaches. First, we performed confocal microscopy on sections from adult intestines. Colocalization of β-Gal protein with SMA was readily apparent(Fig. 7A-C, see also Fig. S2 in the supplementary material), whereas overlap of β-Gal with the endothelial cell marker Pecam in blood vessels was not detected to any significant degree (Fig. 7D-F). Next, in order to quantify these results, we determined the percentage ofβ-Gal-positive cells in sections of SMA- or Pecam-stained blood vessels of the gut (Table 2). Our quantification revealed that, on average, 77.7% of SMA-positive cells of the gut vasculature co-label with β-Gal (from a total of 392 cells counted). By contrast, on average, 6.8% of Pecam-marked cells in the gut are also positive for β-Gal (from a total of 259 cells counted; Table 2). This result indicates that in the gut, a vast majority of vascular smooth muscle cells are descendants of the serosal mesothelium. The presence of β-Gal-negative vascular smooth muscle may indicate an additional source of this cell type.

Confocal analysis of sections from the hearts revealed that most vascular smooth muscle cells were also labeled by the β-Gal marker(Fig. 7G-I), whereas Pecam-labeled endothelium was mostly negative for the genetic lineage tracer(Fig. 7J-L). Quantitative analysis of sections from the heart resulted in a similar, although slightly different, outcome, as, on average, 92.1% of SMA-labeled cells of the coronary vasculature were positive for β-Gal, while, on average, 14% of Pecam-positive cells co-labeled with β-Gal(Table 2).

Once again, confocal experiments revealed that all major arteries and veins were populated by β-Gal-positive cells. Taken together, the results indicate that the smooth muscle cells of blood vessels in the gut are the progeny of serosal mesothelium.

We present data that for the first time link the presence of the serosal mesothelium, the outer lining of the intestinal tube, with the formation of blood vessels to the gut. Our findings show clearly that the Wt1 protein is expressed in the developing gut only in the surface serosal mesothelium, as has been previously observed by others(Armstrong et al., 1993; Carmona et al., 2001; Moore et al., 1998), but not in other cells within the intestinal wall. Based on the expression of Wt1 in the gut exclusively in the serosal mesothelium, we lineage trace the fate of mesothelial cells using a two-component genetic system consisting of a transgenic Wt1-Cre recombinase mouse line and the Rosa26R reporter mouse. We show that serosal mesothelial cells undergo EMT, as in our labeling study the majority of CCFSE-labeled, delaminated cells express β-Gal protein. Furthermore, our data reveal that the majority, but not all, of vascular smooth muscle cells of the gut express β-Gal, which indicates that they are derivatives of the Wt1-expressing serosal mesothelium. A similar mechanism has been reported for the formation of the coronary vessels from the epicardium during heart development(Dettman et al., 1998; Mikawa and Gourdie, 1996; Perez-Pomares et al., 2002; Vrancken Peeters et al.,1999).

On the basis of these data, we propose the following mechanism concerning blood vessel development to the gut (Fig. 8). At early embryonic stages (E9.5-E10.5), the gut tube consists of endoderm and splanchnic mesoderm without mesothelial covering. A vascular plexus associated with the endoderm is present before arrival of the mesothelium, whereas no blood vessels are found on the surface of the gut. From E11, the serosal mesothelium can be visualized on the surface of the gut and the mesenteries, as well as covering the peritoneal cavity. At E12.5,mesothelial cells on the surface of the gut tube undergo EMT and are seen in the submesothelial space of the gut. From E13.5, endothelial tubes extending from the vascular plexus are seen at the surface of the gut. From E16.5, a subset of the descendants of mesothelial cells has differentiated into vascular smooth muscle cells that are found in all arteries and veins of the mesentery and gut. Therefore, we show that progeny of the serosal mesothelium differentiate predominantly into vascular smooth muscle and other non-vascular cells of the gut.

Both epicardial and serosal mesothelia share expression of marker proteins,such as Wt1, cytokeratins and Bves(Armstrong et al., 1993; Carmona et al., 2001; Foley-Comer et al., 2002; Moore et al., 1998; Osler and Bader, 2004). Here,we have repeated earlier experiments that showed that both Wt1 and cytokeratin expression are indicators of the presence of the serosal mesothelium. At stages before E10.5, the surface of the embryonic gut lacks Wt1 and cytokeratin expression, and a simple squamous epithelium cannot be detected. Soon afterwards, robust expression of Wt1 and cytokeratin is detectable in a single layer on the surface of the embryonic gut, indicating the presence of the mesothelium. Futhermore, throughout embryogenesis and adulthood, Wt1 protein is found in the serosal mesothelium. Our analysis did not provide evidence that this protein is present in other cell populations in the embryonic and postnatal gut. However, we cannot exclude the possibility that Wt1 is transiently expressed in non-mesothelial cells in the gut. Still, given that both previously published reports and the current study demonstrate restricted expression of Wt1 to the serosal mesothelium, we conclude that Wt1 is a suitable marker for this epithelial structure.

The intriguing similarities between the epicardium and the serosal mesothelium, and recent findings that the epicardium is the source of the coronary vessels, led us to analyze the fate of the serosal mesothelium. We used a two-component genetic system in the mouse consisting of a Cre recombinase driven by the Wt1 promoter and the Rosa26R reporter mouse line,and have shown that serosal mesothelial cells are specifically and faithfully marked by the reporter (i.e. β-Gal expression). Also, we have combined this genetic system with the use of CCFSE as a surface marker(Morabito et al., 2001; Perez-Pomares et al., 2002; Perez-Pomares et al., 2004)and have revealed that mesothelial cells undergo EMT, as a large majority of CCFSE-positive submesothelial cells co-label for the β-Gal protein. Finally, our quantitative data indicate that the majority of vascular smooth muscle cells in the mesentery and the gut are β-Gal-positive and thus, we postulate, are descendants of the serosal mesothelium. Although the origin of the small, but significant, number of non-β-Gal-labeled vascular smooth muscle cells is unclear, we have no way to determine their origin at present. Possibly these non-labeled vascular smooth muscle cells originate directly from the splanchnic lateral plate mesoderm that overlies the endodermal epithelium. Alternatively, a small number of serosal mesothelial cells may not express the β-Gal protein, or may express at levels below detection. Overall, our current data indicate a faithful and specific expression of this protein in the serosal mesothelium and its descendants. However, the current methods, which have been previously used in related studies(Cai et al., 2003; de Lange et al., 2004; Jiang et al., 2002; Kawaguchi et al., 2002),follow populations of Wt1-expressing cells and are not intended to address matters of clonal differentiation. Those analyses must await methods to mark individual progenitors.

Earlier studies had demonstrated that endothelial cells are produced from the PE in heart development (Mikawa and Gourdie, 1996; Perez-Pomares et al., 2002). Although we detect endothelial cells in small, but consistent numbers in our lineage studies in the gut and heart, the majority of endothelial cells in these vessels are not marked by our genetic labeling system. The original lineage studies of Mikawa and Gourdie(Mikawa and Gourdie, 1996)suggest that angioblasts are a minor population within the PE of the heart. A similar situation could be true for the serosal mesothelium. One explanation for this result is that angioblasts are associated with coelomic mesothelia but are not truly part of the epithelium. In this case, angioblasts may never express mesothelial markers such as Wt1 and, thus, may be largely undetected in our experimental model. Several authors have suggested that migratory angioblasts originating from the liver primordium are intermingled with the PE, and that the PE is composed of both epithelial and mesenchymal cells(Nahirney et al., 2003; Perez-Pomares et al., 1997; Perez-Pomares et al., 1998). These cells would be marked by direct retroviral or vital dye labeling, as employed in previous PE experiments(Mikawa and Gourdie, 1996; Perez-Pomares et al., 2002). Alternatively, if angioblasts are truly part of the advancing mesothelium, it is possible that they never express Wt1 or do not express Wt1-Cre at sufficient levels to be detected in our system.

Previous studies and our current data suggest a recurring relationship between the formation of coelomic mesothelia, either epicardial or serosal,and the delivery of vasculogenic cells to developing organs(Dettman et al., 1998; Kaufman and Bard, 1999; Manasek, 1969; Manner, 1993; Mikawa and Gourdie, 1996; Perez-Pomares et al., 2002; Perez-Pomares et al., 1998; Viragh and Challice, 1981; Vrancken Peeters et al.,1999). Both the heart and the alimentary canal are initially devoid of mesothelia (Manasek,1969; Manner,1993; Meier,1980), but are housed within a common coelom. This coelom is later subdivided into pericardial and peritoneal cavities by the downward growth of the septum transversum (Kaufman and Bard,1999). Mesothelial precursors of the epicardium and pericardial coelom arise in association with the septum transversum as it bisects the common coelom (Viragh and Challice,1981). We can only speculate that a similar population of mesothelial precursors is delivered to the developing gut and peritoneal cavity, as its origins are unclear. The mesothelial precursors of the pericardial cavity have been shown to arise from the dorsal aspect of the coelom (Nahirney et al.,2003), and appear to migrate over the developing organs. Interestingly, our data suggest that the serosal mesothelium also forms dorsally in the peritoneal cavity, at the mesentery close to the urogenital ridges. Nevertheless, we would like to stress that, at present, we have no data providing evidence towards this mechanism. In the heart and the gut, EMT from the mesothelium produces cells that migrate to developing vessels, where described molecular mechanisms of vasculogenesis may regulate cell differentiation (Carmeliet et al.,1996; Cleaver and Melton,2003; Hellstrom et al.,1999; Jain, 2003; Lammert et al., 2003). In addition, it is important to note that both epicardial and serosal mesothelia produce non-vasculogenic progeny that reside within the heart and the gut(Figs 5, 6)(Mikawa and Fischman, 1992; Munoz-Chapuli et al., 2002; Perez-Pomares et al., 1997; Perez-Pomares et al., 1998). Thus, although potential variation may exist, it appears that the vertebrate embryo employs elements of a common or conserved program for the generation of vessels to coelomic organs rather than any wholly variant mechanism in their development.

Coupling vascular development to the formation of mesothelia varies from blood vessel formation in other regions of the embryo. In the limbs, body wall and extraembryonic tissues, vasculogenic mesenchyme is thought to arise from locally derived mesenchyme (Cleaver and Krieg, 1999; Gerhardt and Betsholtz, 2003; Hirschi and Majesky, 2004; Majesky,2003; Saint-Jeannet et al.,1992). The present data, along with previous studies on coronary development, suggest that vasculogenic mesenchyme is `delivered' to organs within the coelom via its encapsulating epithelium later in development. Although known molecular signaling mechanisms are likely to regulate the angioblast/mesenchyme interaction, we propose that coupling the production of vasculogenic cells to the formation of coelomic mesothelium constitutes a distinct yet conserved cellular mechanism in blood vessel development.

Surgeons have long used the omentum, the serosal mesothelium and its connective tissue companion, to repair injured blood vessels and intestines with good success (Bertram et al.,1999; Matoba et al.,1996; Roa et al.,1999; Sterpetti et al.,1992). Although the cellular basis of this reparative function is unknown, it is interesting to speculate that serosal mesothelium may serve as a source of diverse cell types in injury repair. A previous study from our group has shown that mesothelial cell lines can produce vasculogenic cells after stimulation with specific growth factors, suggesting a retention of embryonic potential (Wada et al.,2003a). Thus, it is possible that the serosa may provide a natural source of divergent cells to be used in the repair of damaged adult structures. Finally, a relatively uncommon developmental syndrome, called`Apple Peel Bowel', producing intestinal atresia has been reported(Federici et al., 2003; Pumberger et al., 2002; Waldhausen and Sawin, 1997). Interestingly, this atresia or intestinal wasting is associated with regional loss of the serosa and its associated blood vessels, and is suggestive of a mechanistic relationship between the generation of coelomic mesothelia and vascular development.

We acknowledge R. Pierre Hunt for help with mouse maintenance and embryo preparations; and Farideh Bowles, Theresa Tholkes and R. P. Hunt for the preparation of frozen and paraffin sections and H&E staining. We thank Randy Strich and Kiki Broccoli for help with yeast protocols and YAC DNA isolation; Dorene Davis and Laura Scorr for technical assistance; and Sean Hua(FCCC transgenic mouse facility) and Dr Dawn Kilkenny for help with confocal image capturing(Vanderbilt Cell Imaging Shared Resource). This work was supported by NIH Grants HL67105 and DK58404 (D.M.B.), and HL55373, HD39946 and HL5281 (J.B.E.B.). Additional support came from the Stahlman Foundation of Vanderbilt University (D.M.B.) and an Appropriation from the Commonwealth of Pennsylvania (J.B.E.B.).