Hedgehog is required for murine yolk sac angiogenesis

The first site of blood and blood vessel formation in the mouse is the extra-embryonic visceral yolk sac (Boucher and Pedersen, 1996; Palis et al., 1995). Yolk sac morphogenesis is based on complex tissue interactions during and after gastrulation, which are guided by a wide array of signaling molecules. The visceral yolk sac is composed of two cell layers, visceral endoderm and the underlying extra-embryonic mesodermal layer. Blood islands form around 7-7.5 dpc from localized mesodermal masses lying interior to the visceral endoderm (Palis et al., 1995). These masses contain hemangioblasts, the proposed common precursors of both primitive erythrocytes and endothelial cells (Choi et al., 1998; Robb and Elefanty, 1998). Endothelial cells then form an outer layer, surrounding a pocket containing loosely packed, differentiating red blood cells. Mesodermally derived mesothelial cells provide the outer most layer of the blood island, and line the visceral endoderm between blood islands.

Vasculogenesis, the in situ differentiation of mesoderm into endothelial cell-containing tubes, occurs shortly after the onset of blood island formation in the visceral yolk sac. The resulting endothelial tubes then form a three-dimensional network of tubules that constitutes the primary vascular plexus, which then undergoes reorganization, sprouting and remodeling through the process of angiogenesis to form the large vitelline vessels (Flamme et al., 1997; Risau, 1997; Risau and Flamme, 1995). This remodeling is accompanied by the recruitment and differentiation of vascular smooth muscle cells, derived most likely from the mesothelial layer. Hematopoiesis, the differentiation of primitive erythrocytes from mesodermal precursors (hemangioblasts), occurs concomitantly with vasculogenesis. Several genes are involved in the process of vasculogenesis and angiogenesis, including Flk1 (Kdr – Mouse Genome Informatics) (Shalaby et al., 1995) Flt1 (encoding vascular endothelial growth factor receptor 1) (Fong et al., 1995), Cd34 (Cheng et al., 1996), Vegf (Carmeliet et al., 1996; Ferrara et al., 1996), and Tie2 (Tek – Mouse Genome Informatics) (Sato et al., 1995). The importance of these genes in vascular morphogenesis has been well supported by phenotypic analysis of targeted mutations in mice.

Embryonic stem (ES) cell embryoid bodies can form blood island-like structures in vitro and mimic many of the differentiation events that occur in the yolk sac. Pluripotential ES cells proliferate in an undifferentiated state and spontaneously differentiate into embryoid bodies when grown in suspension. Morphologically, embryoid bodies are characterized by an outer layer of extra-embryonic endoderm surrounding a core of cells analogous to the epiblast, the source of all three primary germ layers in the embryo. The embryoid body core then undergoes cavitation, using spatially localized programmed cell death, in a manner analogous to the formation of the amniotic cavity in the embryo (Coucouvanis and Martin, 1995; Coucouvanis and Martin, 1999). At later stages of development, embryoid body differentiation includes the formation of blood island structures containing blood cells and endothelial cells, providing a model to study the processes of vasculogenesis and hematopoiesis (Doetschman et al., 1985; Hirashima et al., 1999; Vittet et al., 1996).

Hedgehog signaling is essential for proper pattern formation and morphogenesis in several species (McMahon, 2000). There are three mouse homologs of the Drosophila hedgehog (Hh) gene: sonic hedgehog (Shh), desert hedgehog (Dhh) and Indian hedgehog (Ihh) (Echelard et al., 1993; Hammerschmidt et al., 1997). All three hedgehog genes in the mouse signal through a highly conserved signaling pathway that includes a receptor patched (Ptch), a twelve pass membrane-bound ligand binding component, and smoothened (Smo; Smoh – Mouse Genome Informatics), a seven pass G-protein-like molecule, that is responsible for transducing the Hh signal (Denef et al., 2000; Murone et al., 1999). Hedgehog-dependent transcriptional activation is mediated by three vertebrate homologs to the Drosophila Cubitus interruptus (Ci) transcription factor: Gli1, Gli2 and Gli3. Downstream target genes include Ptch, whose upregulation serves as an indicator of a hedgehog response, as well as bone morphogenic proteins (BMPs).

Ihh is expressed in the visceral endoderm of the yolk sac beginning at 6.5 dpc (Becker et al., 1997; Grabel et al., 1998) and the inner, mesodermal layer responds to hedgehog signaling by upregulating Ptch expression (Maye et al., 2000). The visceral endoderm has been proposed to be the source of inductive signals necessary for proper yolk sac vascular morphogenesis to occur (Dyer et al., 2001; Farrington et al., 1997; Palis et al., 1995). A recent study supports the hypothesis that Ihh secreted by the visceral endoderm cells promotes the differentiation of posterior epiblast cells into both endothelial and red blood cells (Dyer et al., 2001). Our previous studies using pharmacological inhibitors of Hh signaling in embryoid bodies have also suggested a role for Ihh in yolk sac differentiation (Maye et al., 2000). Consistent with these observations, 50% of Ihh-deficient embryos die at midgestation, perhaps owing to defects in the yolk sac. The other 50% die at birth, most likely because of respiratory failure as a result of a truncated rib cage – one example of the multiple defects in bone morphogenesis observed in the mutant (St-Jacques et al., 1999). A role for hedgehog signaling in blood vessel formation is supported by several observations. These include the hypervascularization of neurectoderm in response to overexpression of Shh (Rowitch et al., 1999), the decreased vascularization of lung tissue in Shh-deficient mice (Pepicelli et al., 1998) and the defective vascularization observed in the zebrafish Shh mutation (Brown et al., 2000).

To determine the role of hedgehog signaling in yolk sac vascular morphogenesis, we now examine the differentiation of Ihh- and Smo-deficient ES cell embryoid bodies and embryo yolk sacs. We find that the absence of hedgehog signaling leads to the inhibition of blood island differentiation in ES embryoid bodies and to defects in vascular remodeling in embryos.

R1 ES cells (courtesy of Patricia Labosky) were electroporated with a linearized Ihh targeting construct designed to replace exon 1 with a neomycin resistant (neo^R) cassette. The construct (St-Jacques et al., 1999) was modified to include a point mutation in the neo^R cassette, making it less resistant to neomyocin (G418, Sigma). This modification facilitates the generation of homozygous transfectants using higher concentrations of G418 (Mortensen et al., 1992). The targeting construct also contains a thymidine kinase gene downstream of the neo^R cassette for selection against random insertion events. Heterozygous cell lines (clones 14 and 37) were recovered after culture of electroporated cells with 200 μl/ml G418 and 2×10^–6 M gancyclovir (FIAU). Heterozygous cell lines were then cultured in the presence of 2 mg/ml G418, resulting in two Ihh^–/– cell lines, clones 14-61 and 14-63I. The identification of null cell lines was confirmed by both PCR and Southern blot analyses. Primers to amplify the targeted construct were: 5′ neo, ATTCGCAGCGCATCGCCTTCTATCGCCTTC; 3′ Ihh, CAGCACCCGGTCTCCTGGCTTTACAGCTGA. PCR conditions were 5 minutes at 95°C, then 10 cycles of 30 seconds at 94°C, 1 minute at 60°C, 4 minutes 30 seconds at 72°C, then 35 cycles of 30 seconds at 94°C, 1 minute at 64°C and 4 minutes 30 seconds at 72°C. Primers to amplify the endogenous Ihh gene have been described previously (Maye et al., 2000). The conditions were 5 minutes at 95°C, then 35 cycles of 30 seconds at 94°C, 1 minute at 59°C, 3 minutes at 72°C, followed by 7 minutes at 72°C. The probe for the Southern blotting is taken from St-Jacques et al. (St-Jacques et al., 1999). The Southern hybridization was performed using Expressyb from Clontech and a digoxigenenin- labeled probe, and was visualized with CSPD (Roche).

Both parental and transfected ES cell lines were maintained on STO or neo-resistant STO fibroblast feeder layers in DMEM (Gibco) and 15% fetal calf serum (Gibco) in the presence of recombinant LIF. Embryoid bodies were formed by removing the stem cells from the feeder layer and culturing them in suspension in medium without LIF (Doetschman et al., 1985). R1 ES cells heterozygous for insertion of a lacZ reporter gene at the Ptch1 locus (courtesy of Matthew Scott) were grown and differentiated as described.

Mice containing a null allele of Ihh were previously generated (St-Jacques et al., 1999). Timed matings between heterozygous animals (identified by PCR as described by St-Jacques et al. (St-Jacques et al., 1999)) were set up to yield litters for harvest at 9.5, 10.5, 11.5 and 12.5 dpc. Part of each embryo was removed for PCR genotyping, using primers and conditions as described by St-Jacques et al. (St-Jacques et al., 1999).

Whole-mount in situ hybridization on embryoid bodies was performed as described by Maye et al. (Maye et al., 2000). Whole-mount immunocytochemistry was performed by blocking freshly fixed (not dehydrated) embryoid bodies all day in 3% powdered milk, 0.1% Triton X-100 in phosphate-buffered saline (PBS) at 4°C., incubating with primary antibody overnight on a rocker at 4°C, followed by five 1 hour washes with blocking solution and incubation with secondary antibody overnight at 4°C. This was followed by five 1 hour washes in blocking solution, followed by a 20 minute incubation in PBS, 0.2% bovine serum albumin (BSA) and 0.1% TritonX-100. Color was developed using DAB (Sigma) plus 0.5% nickel chloride (NiCl). Primary antibodies used were as follows: PECAM1 (Mec13.3 from Pharmingen) 1/500; Flk1 (Vegf-R2, Ly-73 from Pharmingen) 1/50; αSMA (1A4 from Sigma) 1/500; and CD34 (RAM34 from Pharmingen) 1/50. A rabbit anti-rat HRP-conjugated secondary (Sigma) was used with PECAM, Flk1 and CD34 antibodies. A goat anti-mouse HRP conjugate (Transduction Laboratories) was used with αSMA antibodies. Photographs of intact embryoid bodies were taken before dehydrating, clearing and embedding in paraffin. RNA extraction, primers and conditions for RT-PCR were as described by Maye et al. (Maye et al., 2000). Tal1 primers were as described by Robb et al. (Robb et al., 1995).

After removal of part of each embryo for genotyping, each yolk sac and the remainder of the embryo was fixed, either in 3.7% formaldehyde overnight at 4°C or in a zinc fixative (0.05% calcium acetate, 0.5% zinc acetate, 0.5% zinc chloride in 1M Tris pH 7.4) overnight at room temperature. They were then dehydrated into 100% methanol, cleared in toluene, embedded in paraffin and sectioned. After dewaxing in xylene, sections were either stained with Hematoxylin and Eosin, or used for in situ hybridization or immunocytochemistry with the antibodies described above. In situ hybridization was as described by Wickramasinghe et al. (Wickramasinghe et al., 1995). The Bh1 probe was a gift from James Palis. For immunocytochemistry, slides were incubated for 10 minutes in 1% H₂O₂ to quench endogenous peroxidases. After rinsing in PBS, they were blocked in PBS + 0.1% Triton X-100 + 5% appropriate serum for 30 minutes, followed by incubation overnight at 4°C in this blocking buffer containing primary antibody. After three washes in PBS, slides were incubated for 1-2 hours in blocking buffer containing secondary antibody. After three washes in PBS, the signal was visualized with DAB with or without 0.5% NiCl in PBS.

Whole-mount immunostaining was performed on 9.5, 10.5, 11.5 and 12.5 dpc yolk sacs as described at http://cbi.swmed.edu/ryburn/sato/htmprotocols/immunowholemount.htm. Color was developed as described above. Yolk sacs with embryos were then dehydrated and cleared in toluene before paraffin embedding and sectioning.

Slides were photographed using a Nikon E400 light microscope equipped with a Spot camera (Diagnostic Instruments), and processed using Adobe Photoshop. Whole embryos and embryoid bodies were photographed using an Olympus SZX12 dissecting microscope equipped with an Olympus DP11 digital camera.

In order to define the cell type(s) responding to the Ihh signal secreted by the visceral endoderm in ES embryoid bodies, we used an R1 ES cell line that is heterozygous for an insertion of a lacZ reporter gene at the Ptch locus (Goodrich et al., 1997). The insertion places the lacZ reporter gene under the control of the Ptch promoter. As Ptch is consistently a common downstream target of hedgehog signaling (McMahon, 2000) expression of β-galactosidase in these cells provides a measure of hedgehog response. Fig. 1A shows blood island morphology in R1 Ptch-lacZ embryoid bodies at day 13. Ptch-lacZ expression is localized in the mesodermal layer (mesothelial cells), adjacent to the source of Ihh signaling (visceral endoderm stained with eosin) and surrounding the blood islands (arrows, Fig. 1A). The position of the Ptch-expressing cells in a layer surrounding the blood island suggests that they could be endothelial, mesothelial or perhaps vascular smooth muscle cells. We therefore examined the expression of molecular markers for these cell types in both R1 Ptch-lacZ embryoid bodies and in the yolk sac mesodermal layer in Ptch-lacZ heterozygous mice. Platelet-endothelial cell adhesion molecule (PECAM) is an endothelial cell marker, and α smooth muscle actin (α-SMA) is a vascular smooth muscle cell and mesothelial cell marker (Hungerford and Little, 1999). Comparison of Ptch-lacZ expression with Pecam expression demonstrates that PECAM-positive cells lie interior to the zone of Ptch-lacZ-expressing cells (Fig. 1B). By contrast, α-SMA appears to overlap with regions of Ptch-lacZ-expressing cells in the outer most layer of blood islands and in the mesothelial cells that line the endoderm between blood islands (Fig. 1C). In day 10.5 dpc yolk sacs from Ptch-lacZ heterozygotes, β-galactosidase expression is also localized to the outermost layer surrounding blood vessels as well as to the mesothelial cells, between vessels (Fig. 1D). In these yolk sacs, α-SMA expression appears to localize to the same cell types that express β-galactosidase (Fig. 1E). Because upregulation of Ptch is an indicator of hedgehog signaling, these experiments suggest that mesothelial and perhaps vascular smooth muscle cells, but not endothelial cells, are the direct targets of the Ihh signal secreted by the visceral endoderm.

In order to examine the role of Ihh signaling in blood island differentiation, we generated Ihh deficient (Ihh^–/–) ES cell lines using a targeting construct (Fig. 2A) designed to replace exon 1 with a neomyocin resistance (neo^R) cassette that contains a point mutation in the neo^R open reading frame (Mortensen et al., 1992). This modification facilitates the generation of homozygous transfectants using higher concentrations of G418. Cell lines were generated as described in Materials and Methods. Identification of Ihh null cell lines was confirmed using PCR analysis (data not shown) and Southern analysis. Two Ihh heterozygous cell lines (clones 14 and 37, out of 96 screened) and two Ihh null cell lines (clones 14-61 and 14-63I, out of 144 screened) were obtained (Fig. 2B, and data not shown). The two heterozygous cell lines behaved identically in all assays, as did the two homozygous cell lines. Therefore, all studies shown use clone 14 as the heterozygous line and clone 14-61 as the homozygous cell line. As expected, the Ihh-null cell line did not express Ihh mRNA during embryoid body differentiation, based on RT-PCR analysis using a primer that recognizes a segment of the signaling peptide replaced by the (neo^R) cassette (Fig. 2C). Our previous data suggested that the upregulation of Ptch in embryoid body cores is a response to a hedgehog signal emanating from the visceral endoderm layer (Maye et al., 2000). This Ptch response is absent in the embryoid bodies derived from Ihh-deficient ES cells, confirming a role for Ihh signaling (Fig. 2D).

To determine if expression of Shh or Dhh is upregulated as a result of the absence of Ihh expression, we examined the expression of these genes using RT-PCR. Fig. 2C supports our previous report (Maye et al., 2000) and shows that Shh is expressed beginning at day 10 in embryoid bodies derived from Ihh^+/– ES cells. The same pattern of upregulation is also observed in the mutant cell-derived embryoid bodies. Dhh is not expressed by either cell line during embryoid body differentiation (data not shown). These data suggest that the expression of Shh and Dhh are not altered by mutation of the Ihh gene.

If Ihh signaling plays a role in embryoid body differentiation, we would expect that the differentiation of Ihh-deficient cells would be compromised, in comparison with heterozygous cells. The first differentiation event visible in ES cell embryoid bodies is the appearance of an outer endodermal layer surrounding a core of epiblast-like cells (days 1-4). Most ES cell lines generate an outer layer of predominantly visceral endoderm, with some parietal endoderm. Embryoid bodies then begin to form a cavity, by spatially restricted programmed cell death reportedly regulated by BMP2 and BMP4 signaling (Coucouvanis and Martin, 1995) (days 4-7). As cavitation proceeds, the core cells beneath the visceral endoderm differentiate into a columnar epithelial layer analogous to embryonic ectoderm or epiblast, which eventually, as in the embryo, gives rise to mesoderm and mesodermal derivatives (Vittet et al., 1996) (day 7 onwards). By day 7, Ihh-deficient embryoid bodies begin to appear smaller than embryoid bodies derived from heterozygous and wild-type lines (data not shown). On day 10, additional distinctions become apparent, with the Ihh^–/– lines producing embryoid bodies that have largely failed to cavitate compared with the Ihh^+/– lines (Fig. 3A). Cavitation in the mutant embryoid bodies is inhibited by 70% relative to the level observed for the embryoid bodies derived from heterozygous cells (note low power images in Fig. 3A,B). The decreased level of cavitation is accompanied by an apparent increase in extracellular matrix accumulation in the embryoid body interior. Mutant embryoid bodies also appear to have increased levels of a more rounded and loosely associated cell type in their outer layer than embryoid bodies derived from heterozygous cells (see arrows in Fig. 3). Parietal endoderm consists of rounded loosely associated cells that secrete high levels of extracellular matrix components, suggesting that Ihh^–/– embryoid bodies have an increased proportion of parietal endoderm cells. We previously observed an analogous endoderm transition in embryoid bodies generated from wild-type ES cells that were treated with the hedgehog antagonist forskolin (Maye et al., 2000). Despite the apparent inhibition of cavitation observed in the embryoid bodies derived from Ihh-deficient cells, levels of Bmp2 and Bmp4 expression as assayed by RT-PCR, are unaltered relative to embryoid bodies derived from heterozygous cells (data not shown).

By day 13 of development, the majority of wild-type (data not shown) and heterozygous embryoid bodies have matured to form blood islands containing erythrocytes that can be detected by benzidine staining (Fig. 3B). By contrast, Ihh deficient embryoid bodies appear unable to form blood islands (see Fig. 3B). The occasional cell-free pockets that form do not contain blood cells and do not appear to be surrounded by a layer of endothelial cells. To determine if endothelial cells were present in the mutant embryoid bodies despite the absence of apparent blood islands, we examined the expression of PECAM. By day 13, PECAM staining in Ihh^+/– embryoid bodies is robust and present predominantly in the endothelial cells that line the inner layer of the blood island (arrows in Fig. 4). By contrast, the Ihh^–/– embryoid bodies express PECAM at much lower levels and not in the cell layer surrounding the occasional cavity (arrowhead in Fig. 4A). Some diffuse staining is observed in the core region of both heterozygous and mutant embryoid bodies. (Fig. 4A, see Fig. 7). This core staining appeared to be localized to cells undergoing apoptosis, and we therefore used DAPI staining to determine if there were pyknotic nuclei, often observed in dying cells, in these regions. Pyknotic nuclei did coincide with regions of core PECAM staining for both mutant and heterozygous embryoid bodies, suggesting that this staining does not correspond to cells likely to contribute to blood island formation (data not shown). We also examined the expression of Flk1, a VEGF receptor important for vasculogenesis (Shalaby et al., 1995) and expressed by both hemangioblast and endothelial cells. Flk1 is expressed in Ihh^+/– and Ihh^–/– embryoid bodies in a pattern similar to that of PECAM (data not shown).

To further determine the extent of blood island differentiation in the mutant embryoid bodies, RT-PCR analysis was carried out to examine the expression of Tal1, a transcription factor necessary for blood formation (Elefanty et al., 1999; Visvader et al., 1998), Flk1 and Flt1, an angiogenesis marker. Tal1 and Flk1 are expressed in the proposed hemangioblast, and thus their expression levels may reflect the relative abundance of this population of progenitors. Fig. 4B shows that in differentiating embryoid bodies at days 4-13, the levels of Tal1, Flk1 and Flt1 are sharply reduced in the absence of Ihh. These data suggest that in embryoid bodies Ihh is required for the formation of blood islands, including the differentiation of hemangioblast precursors and their yolk sac derivatives, endothelial cells and primitive erythrocytes.

To complement our in vitro studies, we examined the phenotypes of Ihh heterozygous and Ihh mutant embryos and yolk sacs. It has been previously reported that roughly half of Ihh^–/– embryos die between 10.5 and 12.5 days post coitum (dpc), possibly owing to yolk sac defects, the rest die just after birth (St-Jacques et al., 1999). At the end of the midgestation lethality period, at day 12.5, we therefore expect to observe only 12.5% mutant embryos in litters derived from heterozygous matings, whereas at earlier days we expect this percentage to be between 12.5 and 25%. This is consistent with our observation that 16% of the 200 9.5-13.5 dpc embryos examined were homozygous mutants. In whole-mount, 10.5 through 11.5 dpc Ihh^–/– yolk sacs appeared less well vascularized than their heterozygous and wild-type littermates, suggesting a defect in yolk sac circulation (Fig. 5A). By day 10.5 dpc, wild-type and heterozygous yolk sacs are well into vascular remodeling, which produces large vitelline vessels, whereas the Ihh^–/– yolk sacs appear deficient in this process and tend to have fewer, smaller, vessels (see Fig. 5A). This difference is pronounced by days 11.5 and 12.5 dpc, when heterozygous and wild-type yolk sacs appear to have completed vascular remodeling, resulting in an intricate orderly arrangement of large vitelline arteries and veins branching in a fractal pattern to smaller vessels. Day 11.5-12.5 dpc Ihh^–/– yolk sacs clearly contain vessels, but they remain predominantly small and disorganized, suggesting an inability to complete remodeling (angiogenesis), which results in a compromised vascular network (see Fig. 5A, third row). Virtually all mutant embryos examined in whole mount appeared to have compromised vasculature, although the extent of the defect varied from subtle to more pronounced. Fig. 5B shows the pattern of PECAM staining in a whole heterozygous and mutant yolk sac, and supports the observation that the vessels formed in the absence of Ihh signaling are smaller and less well organized (Fig. 5, arrow) and can appear flattened (Fig. 5, arrowhead).

Histological analysis of 9.5 through 12.5 dpc yolk sac sections supports the conclusion that the vessels formed in the Ihh mutant yolk sacs are often smaller and/or flattened in their morphology (day 11.5 dpc embryo is shown in Fig. 6A). These defects are seen as early as 9.5 to 10.5 dpc and become more apparent by days 11.5 and 12.5 dpc, when the distal-most vessels in the mutants may appear flat (collapsed) and/or wider than normal (Fig. 6A).

To address the molecular nature of the vascular defects in Ihh^–/–yolk sacs, we examined the expression of PECAM, Flk1, α-SMA and CD34, an antigen expressed on the surface of hematopoietic precursors, as well as endothelial cells (Wood et al., 1997). At 9.5 dpc, PECAM is expressed robustly in the endothelial cell layer in Ihh^+/– yolk sacs, but to a lesser extent in the mutants (data not shown). This distinction is maintained in both 10.5 and 11.5 dpc yolk sacs (Fig. 6B), and is also observed for Flk1 expression (Fig. 6B). We also examined CD34 expression in day 9.5 to 11.5 yolk sacs. We were able to detect weak expression of CD34 in the 9.5dpc Ihh^+/– yolk sacs; however, none was detected in the Ihh^–/– yolk sacs (data not shown). Stronger expression was detected in the heterozygous yolk sacs at 11.5 dpc, but expression was still absent in the mutant yolk sacs (Fig. 6B). Expression of CD34 was confined to endothelial cells and was never observed on round, hematopoietic-like cells in the heterozygous yolk sacs at the time points we observed.

At 9.5 dpc, α-SMA expression around the blood vessels is faint, in both Ihh^–/– and Ihh^+/– yolk sacs; however, by 10.5 dpc, expression levels around blood vessels increase in heterozygous yolk sacs but remain weak in the Ihh mutant yolk sacs (data not shown). By 11.5 dpc, expression of α-SMA is robust in the Ihh^+/– yolk sacs and encircles the blood vessel whereas in the Ihh^–/– yolk sacs, α-SMA expression appears predominantly restricted to the mesothelial layer (Fig. 6B). These results suggest that Ihh signaling may be necessary for the organization of endothelial cells and vascular smooth muscle cells in the yolk sac.

The interpretation of data obtained from studies examining the role of hedgehog signaling using embryoid bodies deficient for Ihh is complicated by the possibility that other hedgehog genes may be expressed and therefore play roles in embryoid body differentiation. We have recently determined, using RT-PCR analysis, that Dhh mRNA is not detectable in R1 ES embryoid body cultures (data not shown), but our previous studies, and data reported here (Fig. 2C), establish that Shh is expressed after day 11 in embryoid body core cells internal to the endoderm (Maye et al., 2000). To study the role of signaling mediated by all hedgehog genes, we examined the phenotype of Smo-deficient ES-cell derived embryoid bodies. Smo encodes the component of the Hh receptor complex thought to be required for all hedgehog signals, therefore, mutations in this gene effectively eliminate hedgehog signal transduction (Zhang et al., 2001). Derivation of the cell lines will be described elsewhere. The two heterozygous lines analyzed behaved identically to each other, as did the two homozygous cell lines. We therefore report the data obtained for one of the heterozygous lines and one of the homozygous lines.

Smo-deficient embryoid bodies have the same phenotype as the Ihh-deficient cell lines. In contrast to the heterozygous lines, which cavitate and form numerous blood islands, the vast majority of Smo-deficient embryoid bodies fail to cavitate (cavitation is inhibited by 78% compared with heterozygotes) and fail to form blood islands (Fig. 7). Again, Bmp2 and Bmp4 expression levels, based on RT-PCR analysis, are unaffected by the absence of hedgehog signaling in mutant embryoid bodies (data not shown). Smo^+/– embryoid bodies, like Ihh heterozygotes, express robust levels of PECAM in blood island endothelial cells, whereas PECAM staining in Smo null lines does not localize to rare cell-free pockets, and is only found diffusely in core cells (Fig. 7), as in Ihh^+/– embryoid bodies described above.

Smo^–/– embryos die before turning (∼9.5 dpc) (Zhang et al., 2001) and, as described below, display a more severe yolk sac phenotype than Ihh^–/– embryos. To determine the extent of endothelial cell differentiation in Smo^–/– yolk sacs, we analyzed PECAM expression. Whole-mount examination at 9.5 dpc clearly demonstrates that both Smo^+/– and Smo^+/+ yolk sacs form robust vasculature, whereas Smo^–/– yolk sacs fail to undergo angiogenesis and remain arrested in the primary vascular plexus stage (Fig. 8A). In the mutant yolk sac shown, the chicken-wire-like pattern of PECAM staining that characterizes the vascular plexus is restricted to the region most proximal to the placenta and does not extend around the circumference of the yolk sac. Analysis of sectioned yolk sacs reveals a blood cell phenotype not observed in the Ihh^–/– yolk sacs. The tubules of endothelial cells formed in the mutant often contain dense clusters of hematopoietic cells, tightly packed within the tubule (Fig. 8B). This packed phenotype is most frequently observed proximal to the placenta. In situ hybridization analysis for expression of the embryonic globin gene Bh1 (Wilkinson et al., 1987) confirms the identification of these cells as primitive erythrocytes (Fig. 8C). We examined PECAM expression in sections to observe more closely the extent of endothelial cell differentiation. The staining for PECAM was strong in Smo^+/– or Smo^+/+ yolk sacs, highlighting blood islands and blood vessels, and significantly weaker in Smo^–/– yolk sacs (Fig. 8D). These data suggest that the complete absence of hedgehog signaling leads to an inhibition of vascular remodeling, but that differentiation of primitive erythrocytes and some endothelial cells does proceed.

Previous studies have demonstrated the importance of the visceral endoderm layer in yolk sac vasculogenesis (Farrington et al., 1997; Palis et al., 1995) and identified Ihh as a key signal for this process (Dyer et al., 2001; Maye et al., 2000). A recent study using isolated epiblast tissue supports the hypothesis that Ihh secreted by the visceral endoderm cells promotes the differentiation of posterior epiblast cells into blood islands containing endothelial and red blood cells (Dyer et al., 2001). We now demonstrate using a gene targeting approach that hedgehog signaling is essential for normal yolk sac angiogenesis.

Hedgehog signaling molecules play numerous inductive and patterning roles in vertebrate and invertebrate development. In the mouse, the three hedgehog genes, Shh, Ihh and Dhh, have distinct and overlapping functions. In the yolk sac, there is no evidence for expression of Shh, whereas Ihh and Dhh have specific spatio-temporal expression patterns (Becker et al., 1997; Dyer et al., 2001; Farrington et al., 1997; Maye et al., 2000). Ihh is expressed in the visceral endoderm of the yolk sac beginning shortly after gastrulation (∼6.5 dpc) and persisting after 12.5 dpc (Becker et al., 1997; Farrington et al., 1997; Maye et al., 2000). The mesodermal layer responds to hedgehog signaling by upregulating Ptch expression. Others have shown that Dhh is present in the yolk sac mesoderm, and to a lesser degree in the endoderm, from 10.5 dpc to 12.5 dpc (Farrington et al., 1997). These expression patterns coincide temporally with the formation of blood islands and the development of the yolk sac vasculature.

Using a gene targeting approach, we generated Ihh^–/– ES cells and examined their ability, in comparison with Ihh^+/– ES cells, to differentiate as embryoid bodies. We show that loss of Ihh signaling from the visceral endoderm results in an overall decrease in the ability of embryoid bodies to cavitate compared with heterozygous lines. Ihh-deficient embryoid bodies are also severely compromised in their ability to form blood islands and produce primitive erythrocytes. The absence of blood islands may be attributable to the absence of ectoderm cells capable of generating the appropriate mesodermal precursors. Targeted mutagenesis of other genes has been shown to alter blood island differentiation in ES embryoid bodies. VE-cadherin- (a transmembrane protein necessary for endothelial cell-cell interactions) and VEGF-deficient ES cells produce embryoid bodies that contain endothelial and blood cells, but are unable to form a mature endothelial-cell derived vascular network (Ferrara et al., 1996; Vittet et al., 1997).

Ihh is not the only hedgehog gene expressed during embryoid body differentiation. We have shown that although Dhh is not detectable in embryoid bodies, Shh message accumulates in the embryoid body core at about 11 days (Maye et al., 2000). This pattern of Shh expression is also observed in the Ihh-deficient embryoid bodies (Fig. 2C), indicating these cells have hedgehog signaling activity caused by to Shh. Our observation that the phenotype of Smo^–/– ES cell-derived embryoid bodies is virtually identical to Ihh^–/– embryoid bodies suggests that Shh signaling does not contribute significantly to ES embryoid body cavitation or their differentiation into blood islands, perhaps owing to its delayed time of accumulation relative to Ihh. These results indicate the importance of Ihh signaling in ES cell embryoid body blood island development; no hematopoiesis or vasculogenesis can be detected in the absence of functional Ihh.

It has been previously suggested that the mid-gestational lethality observed in Ihh^–/– mice was attributable to defects in yolk sac morphogenesis (St-Jacques et al., 1999). We now present evidence in support of this hypothesis. Although it appears that Ihh^–/– yolk sacs can initiate vasculogenesis and hematopoiesis, they are deficient in their ability to undergo vascular remodeling. Mutant yolk sacs, at 10.5 to 11.5 dpc, are paler than those of heterozygous and wild-type littermates, most probably owing to reduced circulation of erythrocytes in mutants. Blood vessels are present, but they appear smaller and less organized than in the heterozygous yolk sacs. In addition, many of the blood vessels in mutant yolk sacs have a flattened morphology. Similar defects in yolk sac vascularization have also been observed in connexin 45-deficient mice (Kruger et al., 2000) and in yolk sacs deficient in endoglin, a TGF-β binding protein (Arthur et al., 2000; Li et al., 1999).

As defects in vascular morphogenesis could potentially be due to endothelial cell defects and/or vascular smooth muscle cell defects, we examined PECAM and α-SMA staining patterns. We have consistently observed reduced levels of PECAM staining in Ihh^–/– yolk sacs compared with heterozygous and wild-type littermates. This observation is consistent with fewer endothelial cells being present, owing either to limited proliferation or differentiation. Several studies suggest that Shh can act as a mitogen (Britto et al., 2000), and evidence suggests that Ihh plays this role for chondrocytes (Karp et al., 2000). Reduced numbers of endothelial cells may contribute to the observed decrease in the number and diameter of blood vessels, as well as affect endothelial-dependent recruitment of mesenchymal cells to form the vascular smooth muscle cell layer. Reminiscent of the phenotype in endoglin-deficient yolk sacs in which vascular smooth muscle cells fail to encompass blood vessels (Li et al., 1999), we see a decrease in α-SMA staining around blood vessels in Ihh^–/– yolk sacs. The loss of vascular integrity may also explain the collapsed vessels we see upon histological analysis of Ihh^–/– yolk sacs. This type of defect may be attributed to an abnormal vascular smooth muscle cell layer, which can lead to a loss of vascular stability (Folkman and D’Amore, 1996). Flattened vessels may also result from a state of vascular regression caused by poor circulation in the yolk sac (Risau, 1997) and/or defects in the differentiation or recruitment of vascular smooth muscle cells (Darland and D’Amore, 1999). Previous studies support a role for hedgehog signaling in smooth muscle differentiation. In Shh-deficient lungs, the smooth muscle layer around the bronchi is absent (Pepicelli et al., 1998) and in Ihh/Shh double mutants, there is a reduced smooth muscle layer in the gut (Ramalho-Santos et al., 2000).

While this phenotypic analysis can explain the midgestation lethality of 50% of the Ihh^–/– mutant embryos, why do the other 50% survive until birth? Mutations in other genes that result in abnormal yolk sac morphogenesis, such as the gene encoding factor V (Cui et al., 1996) and the gene encoding the tissue factor pathway inhibitor (Huang et al., 1997), also display midgestation lethality affecting only around 50% of the embryos. This observation may reflect the fading importance of the yolk sac after day 10 or so, when the placenta takes over its functions (Cross et al., 1994). If suboptimal vascularization has been sufficient to support development through this early period, the embryos will survive, despite compromised yolk sac function. In the case of the Ihh-deficient embryos, Dhh expression during 10.5 to 12.5 dpc, may provide the necessary cues to promote yolk sac morphogenesis through this critical period, at least in some of the embryos (Dyer et al., 2001). Given the observation that Dhh is expressed in the yolk sac, it was of interest to examine the yolk sac phenotype of Smo-deficient embryos. If Dhh plays a role in yolk sac morphogenesis, we would expect Smo^–/– yolk sacs, which lack response to all hedgehog proteins, to arrest earlier and exhibit a more severe phenotype than Ihh^–/– yolk sacs. Based upon PECAM staining of day 9 yolk sacs, the developing vasculature is arrested in the primary vascular plexus stage, blood vessels do not form and angiogenesis fails to occur.

Histological analysis of Smo^–/– yolk sacs shows that the existing endothelial tubes are tightly packed with a large number of Bh1-expressing hematopoietic cells. The presence of excess blood cells may be attributable to a decrease in endothelial cell proliferation or differentiation, permitting more hemangioblast precursors to choose the hematopoietic pathway. Alternatively, the build up of erythrocytes may be a physical one, owing to the absence of a vasculature that can promote circulation. The decreased levels of PECAM staining observed in the mutant yolk sacs suggests that endothelial cell differentiation is compromised. Similar to Ihh^–/– yolk sacs, Smo^–/– yolk sacs exhibit reduced levels of α-SMA, consistent with a defect in the vascular smooth muscle cell layer. It has been demonstrated that signaling between the endothelial cell layer and the mesenchymal cells is essential for vascular stability and remodeling to occur properly, and that defects in one cell type may affect the function of the other (Conway et al., 2001; Folkman and D’Amore, 1996; Hirschi et al., 1998).

The differentiating yolk sac in vivo can be influenced by a number of cell types not present in the ES embryoid body system, including maternal tissue and extra-embryonic ectoderm. ES cell embryoid bodies deficient for Ihh or Smo exhibit more severe phenotypes than do Ihh^–/– or Smo^–/– yolk sacs. The mutant embryoid bodies are unable to generate endothelial or blood cells or to form blood islands, whereas the Ihh^–/– and Smo^–/– yolk sacs contain endothelial cells and erythrocytes and can form endothelial cell tubes and, in the case of Ihh mutants, undergo suboptimal remodeling. One possibility is that the ability of the Ihh^–/– mutant yolk sacs to progress further than the Smo^–/– yolk sacs is due to the presence of Dhh activity. It has previously been demonstrated that Dhh is expressed in the yolk sac beginning at day 10. There is no Dhh expressed, however, in ES embryoid body cultures, which may explain why both Ihh- and Smo-deficient cell lines apparently have identical defects. The observation that the complete absence of hedgehog signaling in Smo^–/– ES cell embryoid body culture generates a more severe phenotype than the complete absence of hedgehog signaling in the embryo (no endothelial or blood cells differentiate in vitro) suggests that additional signals are present in vivo that are not produced in culture. These signals could be maternal or embryonic in origin and appear to be essential for initiating hematopoiesis and vasculogenesis. The Smo-deficient ES cells provide a useful model for identifying these signals.

We conclude that hedgehog signaling promotes the differentiation of endothelial cells and their remodeling to form blood vessels, but that the initial stages of differentiation can take place in the absence of this signal. Additionally, proper recruitment or differentiation of vascular smooth muscle cells is dependent on hedgehog signaling. Our observation that primitive erythrocytes are present in Smo^–/– yolk sacs suggests that this cell type is able to differentiate in the absence of hedgehog signaling, despite evidence that these cells can be induced by the addition of hedgehog peptides to epiblast cultures (Dyer et al., 2001). In fact, these blood cells may differentiate from the hemangioblast precursor at the expense of endothelial cells in the absence of hedgehog signal. A role for hedgehog signaling in angiogenesis is supported by the recent observation that Shh promotes angiogenesis by inducing angiogenic cytokines, including the VEGFs and angiopoietins (Pola et al., 2001). Our observation that Ihh^–/– embryoid bodies are unable to upregulate the VEGF receptors Flk1 and Flt1 suggests that hedgehog signaling may be essential for upregulating the receptors for the VEGF cytokines as well. These observations suggest that hedgehog peptides may be useful therapeutic agents for promoting blood vessel sprouting.

We thank Ann C. Burke, Stephen Devoto and Michael Weir for critical reading of the manuscript and valuable suggestions; Patricia Labosky for R1 ES cells; Matthew Scott for Ptch1-lacZ ES cells; and James Palis for the BH-1 in situ probe. This work was supported by NSF grant IBN 9604367 and American Heart Association grant 0150284N to L. G.