Mouse Disp1 is required in sonic hedgehog-expressing cells for paracrine activity of the cholesterol-modified ligand

The hedgehog (Hh) family of secreted proteins control tissue growth and patterning in many key developmental processes in both vertebrates and invertebrates. In mammals, there are three hedgehog genes(Echelard et al., 1993): sonic hedgehog (Shh), Indian hedgehog (Ihh) and desert hedgehog(Dhh). All three can act as potent mitogens, survival factors and inducers of distinct cell types in a dose-dependent manner (reviewed by McMahon et al., 2003). Shh is essential for early patterning of the ventral CNS, and somite and anterior-posterior organization of the limb. Shh also plays a redundant role with Ihh in establishment of left-right asymmetry(Zhang et al., 2001). Later in development, Shh is involved in the growth and morphogenesis of several organs, including hair, tooth, lung, gut and kidney, whereas Ihh coordinates growth and differentiation of the endochondral skeleton, Dhh is required for development of the peripheral nerve,testicular organization and spermatogenesis.

As extracellular signals, Hh proteins are synthesized in discrete subsets of cells in many organs and act in short and long-range signaling processes. The best characterized mammalian Hh-target field is the developing ventral neural tube, where progenitors differentiate into several different cell types in response to a morphogen gradient of Shh, issuing from two point sources:the notochord and floorplate (Briscoe and Ericson, 1999; Briscoe and Ericson, 2001; Jessell,2000; McMahon et al.,2003). The notochord is a ventral rod of mesoderm that underlies the neural tube while the floorplate is a population of support cells at the ventral midline of the neural tube that is induced by notochordal Shh. Distinct ventral neural progenitors are induced at specific positions with respect to the source, and apparent concentration, of Hh ligand.

All Hh proteins are synthesized as full-length precursors that undergo an autocatalytic cleavage reaction. This removes the C-terminal catalytic domain and attaches a cholesterol molecule to the C terminus of the N-terminal signaling fragment (N-Shhp; `p' stands for `processed' in the N-Shh signaling moiety) (Porter et al.,1996a). The hydrophobic cholesterol moiety is thought to bind Hh to the cell membrane. The hydrophobicity of the Hh molecule is further increased by the addition of a palmitoyl group to a conserved cysteine residue that is exposed at the N terminus after signal peptide cleavage(Pepinsky et al., 1998). These modifications regulate Hh activity, oligomerization, range of action, potency and might in many cases, shape a signaling gradient(Burke et al., 1999; Chamoun et al., 2001; Chen et al., 2004; Chen and Struhl, 1996; Kohtz et al., 2001; Lee et al., 2001; Lewis et al., 2001; Zeng et al., 2001). The requirement for such modifications may depend on the context in which Hh proteins act (Chen and Struhl,1996; Lewis et al.,2001). As cholesterol modification is unique to Hh-ligands, the role of cholesterol has attracted considerable attention. Forms of Hh ligand have been engineered that lack the autocatalytic cleavage site and C-terminal cleavage domain, and therefore are not cholesterol-modified [Hh-N in Drosophila (Chen and Struhl,1996) and NShh in mammals(Lewis et al., 2001)]. In both instances, palmitoylation appears to be largely independent of cholesterol addition (Chen et al., 2004; Gallet et al., 2003).

Despite dual lipid modifications that normally retain protein on the cell membrane, Drosophila Hh and mouse Shh and Ihh protein can be detected at significant distances from their expression domains(Chen and Struhl, 1996; Gallet et al., 2003; Gritli-Linde et al., 2001; Lewis et al., 2001; Porter et al., 1996a). This long-distance action is an active process involving a transmembrane protein Dispatched (Disp). Disp was first identified in Drosophila studies,where its activity is required specifically within Hh-producing cells for movement of cholesterol modified ligand into Hh target fields(Burke et al., 1999). By contrast, Hh-N signaling is Disp independent(Burke et al., 1999).

The mouse has two Disp homologs, Disp1 and Disp2, but only Disp1 is able to rescue a Drosophila disp mutant phenotype(Ma et al., 2002). Genetic studies in zebrafish suggest that Disp2 is not involved in Hh signaling (Nakano et al.,2004), and mouse studies support this view (H.T. and A.P.M.,unpublished). Three mutant alleles of mouse Disp1 have been described: Disp1^Δ8, Disp1^C829F and Disp1^Δ2(Caspary et al., 2002; Kawakami et al., 2002; Ma et al., 2002; Tian et al., 2004). Analysis of these alleles lead to similar general conclusion, that Disp1 is involved in Hh signaling during early embryonic development. Among these mutant alleles,the Disp1^Δ8 deletion allele(Kawakami et al., 2002; Ma et al., 2002) and Disp1^C829F missense allele(Caspary et al., 2002) are likely to represent null alleles of Disp1. Homozygous Disp1^Δ8 and Disp1^C829Fmutants do not survive beyond E9.5, and exhibit gross morphological features that are similar to Smoothened (Smo) mutants in which all Hh signaling activity is abolished (Zhang et al., 2001). In Disp1^C829F/C829F and Disp1^Δ8/Δ8 embryos, Hh signaling is retained but only in midline cells of the notochord that both produce Shh and respond to Shh signals (Ma et al., 2002). Disp1^Δ2, by contrast, encodes a hypomorphic allele(Tian et al., 2004). Homozygous Disp1^Δ2 mutants die at birth with facial midline patterning defects, characteristic of attenuated Shh signaling (Wallis and Muenke,1999). This hypomorphic allele has permitted us to demonstrate the genetic interaction of Disp1 with specific components of the Hh signaling pathway. By combining a Disp1^Δ2hypomorphic allele with Disp1^C829F, Shh^null and patched 1 (Ptch1^null)alleles, we created a set of graded facial midline and neural tube phenotypes demonstrating that Disp1 gene dose regulates the level of Shh signaling activity in vivo. Furthermore, rescue of the Disp1^Δ2/Δ2 mutant upon removing one copy of Ptch1 suggested that Disp1 functions exclusively in the Hh pathway (Tian et al.,2004).

Although these phenotypic analyses demonstrated a conserved requirement for Disp1 in the Hh signaling pathway and highlighted the importance of Disp1 dose for normal levels of Shh signaling, they did not address the specific cellular or molecular limits to Disp1 action in mammalian Hh signaling. To determine whether Disp1 activity is required in Shh-producing cells or in paracrine signaling in the target field, we used a Shh^Cre knock-in allele to remove Disp1 exon 2 specifically from Shh-producing cells. Our results indicate that Disp1 activity in these cells is essential for Hh signaling to the target field. Furthermore, using an allele that produces N-Shh, we demonstrate that Disp1 activity is required only for the paracrine action of the cholesterol modified form of Shh.

Disp1^C829F/+ mice were kindly provided by Kathryn Anderson (Caspary et al.,2002). The Sox2Cre and Shhⁿ(Shh^-) alleles have been described previously(Hayashi et al., 2002; St-Jacques et al., 1998). The construction of Shh^GFP-Cre (here shortened to Shh^Cre) is described by Harfe et al.(Harfe et al., 2004). N-Shh^C alleles will be reported in detail elsewhere (J.J. and A.P.M., unpublished). In short, Cre-mediated recombination leads to the generation of an allele functionally identical to the N-Shh allele reported by Lewis et al. (Lewis et al.,2001). Mutants were studied on a mixed genetic background,principally 129SV, C57BL6/J and SW.

To remove exon 2 of Disp1, a targeting vector was engineered in which exon 2 was flanked by loxP sites. Exon 2 encodes the amino terminal cytoplasmic domain and first transmembrane domain of Disp1(Tian et al., 2004). After homologous recombination at the Disp1 locus in AV3 ES cells, a heterozygous ES cell line was injected into blastocysts of the C57BL6/J strain to generate chimeras. These were bred with Swiss Webster mice to obtain Disp1^Δ2C/+ offspring. Chimeric males were also bred with β-actin-Cre females, Cre activity in the preimplantation embryo allow the recovery of Disp1^Δ2/+ heterozygous offspring.

Embryos were fixed in 4% paraformaldehyde at 4°C overnight. Whole-mount and section in situ hybridization using digoxigenin-labeled RNA probes was performed as described previously(Schaeren-Wiemers and Gerfin-Moser,1993; Wilkinson,1992).

Immunofluorescence on embryonic sections was performed as described for sections (Yamada et al.,1991). Antibodies and dilutions were as follows: rabbit αFoxa2, 1:8000 (Ruiz i Altaba et al.,1995); αNkx6.1, 1:3000(Cai et al., 2000);αNkx2.2, 1:4000 (Briscoe et al.,1999); αOlig2, 1:5000(Takebayashi et al., 2000);mouse αNkx2.2, 1:50 (Ericson et al.,1997a); αPax6, 1:20(Ericson et al., 1997b);αPax7, 1:20 (Ericson et al.,1996); and αMNR2, 1:20 (DHSB).

For skeletal preparations, 18.5 dpc embryos were processed as described previously (Karp et al.,2000).

To remove Disp1 activity specifically from the Shhexpression domain using the Cre/loxP system, we first generated a Disp1-conditional allele(Disp1^Δ2C) by flanking the first coding exon with DNA recognition sites (loxP) for Cre recombinase (Cre) (see Tian et al., 2004). Mice homozygous for this allele(Disp1^Δ2C/Δ2C) were viable and fertile with no discernible phenotype, indicating that Disp1^Δ2C has a wild-type function (data not shown). Conversion of Disp1^Δ2C into a Disp1^Δ2 allele in the presence of Cre recombinase is highly efficient, as demonstrated by our previous study in which a ubiquitous Disp1^Δ2 allele was generated by crossing a β-actin Cre transgene onto a Disp1^Δ2C/+ background(Tian et al., 2004).

To knock down Disp1 activity specifically within Shh-producing cells, we used a Cre knock-in allele in which sequence encoding a GFP-Cre fusion protein was inserted into the Shh locus(Shh^GFP-Cre, herein Shh^Cre)(Harfe et al., 2004). This allele no longer expresses a normal Shh transcript, and is a null allele for Shh activity. When combined with a Cre reporter line R26R(Soriano, 1999), Shh^Cre gave rise to lacZ expression that was almost identical to Shh mRNA distribution at E10.5(Fig. 1A,B), indicating that the GFP-Cre fusion protein is functionally active within all Shh expression domains. To further confirm the functional specificity of GFPCre fusion protein, Shh^Cre was crossed with a Shh conditional allele(Shh^Δ2C)(Lewis et al., 2001) to remove Shh in the Shh-producing cell.

The specification of ventral progenitor domains in the presumptive spinal cord gives the most detailed read-out of Shh signaling. The ventral half of the mouse neural tube is occupied by five ventricular progenitor populations,from ventral to dorsal, pV3, pMN, pV2, pV1 and pV0, all of which require a direct Hh signaling input for their development(Briscoe et al., 2001; Wijgerde et al., 2002). These cells move laterally and differentiate into V3 interneuron, motoneurons, and V2, V1 and V0 interneurons, respectively(Briscoe and Ericson, 2001; Jessell, 2000; McMahon et al., 2003). Induction of distinct cell types depends on Shh signaling from the notochord and floorplate. A key aspect of these inductive events is that from pV0 to the floorplate (which is also induced by Shh signaling), individual cell identities require a progressively higher concentration of Shh for their induction (floorplate>pV3>pMN>pV2>pV1>pV0)(Ericson et al., 1997a; Roelink et al., 1995). The Shh^Cre/C mutants displayed a cyclopic head (compare Fig. 1D with 1E) and ventral neural tube patterning defects (compare Fig. 1G,H,J,K,M,N) very similar to those of Shh mutants,indicating that Shh^Cre was effective in removing Shh activity from all Shh-producing cells (Fig. 1).

Appropriate crosses were set up to generate mice that were either Disp1^Δ2/Δ2C, Shh^Cre/+ or Disp1^C829F/Δ2C, Shh^Cre/+. In both cases,pups died within 1 day of birth. In the former, Shh^Creactivity should generate Shh-expressing cells that are homozygous for the Disp1^Δ2/Δ2 allele. In the latter, the only Disp1 activity comes from a single Disp1^Δ2hypomorphic allele. In both examples, Disp1 activity is decreased on a background where Shh dose is lowered by the presence of the Shh^Cre null allele. Ordinarily, reducing Shh levels by removing one allele of Shh produces no discernible phenotype(Chiang et al., 1996; St-Jacques et al., 1998) but reduction of Shh dose enhances the neural phenotype in Disp1hypomorphic combinations (Tian et al.,2004) Importantly, as the production of Cre recombinase is linked to Shh expression, some level of Shh signaling must take place prior to Cre-mediated modification of Disp1 alleles. This is evident in Shh^Cre/C embryos where a few Nkx6.1-producing cells are observed (Fig. 1N), while no Nkx6.1 are produced in Shh^-/- embryos(Fig. 1L,M).

In both genotypes (Disp1^Δ2/Δ2C,Shh^Cre/+ or Disp1^C829F/Δ2C,Shh^Cre/+), pups displayed facial midline defects similar to Disp1 hypomorphic mutants (Disp1^Δ2/Δ2,Shh^+/- and Disp1^C829F/Δ2,Shh^+/-), in which Disp1 exon 2 was deleted in the entire embryo, the severity of the phenotype increased as Disp1 activity decreased (Fig. 2A-E). Importantly, no phenotype was observed in Disp1^Δ2/Δ2C, Shh^+/- embryos (data not shown). The midline facial defects were clearly visible at E10.5 by comparing the ventral disposition of the medial and lateral nasal process and proximal mandibular and maxillary arches that are highlighted by the expression of Fgf8 (Fig. 2F-J). In Disp1^Δ2/Δ2C,Shh^Cre/+ embryos, the two nasal pits were positioned closer to the midline, indicating a loss of facial midline structure. A more severe phenotype was seen in Disp1^C829F/Δ2C,Shh^Cre/+ embryos as the two nasal pits start to fuse where Shh-dependent medial nasal cells were also affected(Fig. 2J). The severity of the midline loss in conditional mutants was comparable with Disp1hypomorphic mutants Disp1^Δ2/Δ2,Shh^+/- and Disp1^C829F/Δ2;Shh^+/- (compare Fig. 2G and I with 2H and J). Skeleton preparations of newborn pups revealed that the premaxilla was missing from Disp1^Δ2/Δ2C, Shh^Cre/+ and Disp1^C829F/Δ2C, Shh^Cre/+ mutant embryos,just as in Disp1^C829F/Δ2, Shh^+/- and Disp1^Δ2/Δ2, Shh^+/- embryos(Fig. 2K-O). Although the parietal bone was also lost in Disp1^C829F/Δ2,Shh^+/- embryos, we only observed a delay in ossification of parietal bones in Disp1^C829F/Δ2C,Shh^Cre/+ mutant (compare Fig. 2N,O). As expected from the analysis of Disp1^Δ2/Δ2, Shh^+/-and Disp1^Δ2/C829F, Shh^+/- mutants,expression of Ptch1, one of the principle transcriptional targets of Hh signaling, was greatly downregulated in the frontal nasal process (FNP) of the conditional mutants at E9.5 (Fig. 2P-T, arrows). This confirms that the loss of midline structures of the frontal nasal process in the conditional mutants Disp1^Δ2/Δ2C, Shh^Cre/+ and Disp1^C829F/Δ2C, Shh^Cre/+ was due to the attenuation of Shh signaling in this region. Furthermore, Shhexpression within the ventral forebrain, which is itself a target of a mesendodermal derived Shh signal, was lost in all allelic combinations(Fig. 2U-Y). This most probably explains the initial defect in FNP development in the Disp1 mutant background.

As previously reported (Tian et al.,2004), in the Disp1^Δ2/Δ2,Shh^+/- mutant, Shh signaling is greatly compromised in the ventral neural tube: the floorplate was absent (Foxa2^-, and Shh^-) and the ventral mid-line was occupied by greatly reduced numbers of the ventral-most neural progenitor, pV3 (Nkx2.2⁺) (10%of wild-type numbers; wild type, 60±6; Disp1^Δ2/Δ2, Shh^+/-, 7±3, n=3, P<0.01) (Fig. 3A,B,F,G). The next ventral-most progenitor cells, pMN(Olig2⁺), were reduced to 15% of the wild type number (wild type,71±6; Disp1^Δ2/Δ2, Shh^+/-,10±3, n=3, P<0.01 Fig. 3B,G). The domain occupied by Nkx6.1⁺ cells, which demarcate pV3, pMN and pV2 progenitors, was also significantly reduced in the mutant (wild type, 175±15; Disp1^Δ2/Δ2, Shh^+/-, 60±5, n=3, P<0.01) (Fig. 3C,H). The loss of these ventral cell identities was accompanied by a ventral upregulation of Pax6 and Pax7, two factors whose expression is generally repressed by Shh signaling in the ventral neural tube(Fig. 3C,D,H,I). The severe reduction of the pMN neural progenitor population translated to a significant decrease in MNR2⁺ motoneuron precursors that were also abnormally positioned at the ventral mid-line (wild type, 145±15, Disp1^Δ2/Δ2, Shh^+/-, 80±10, n=3, P<0.01 Fig. 3E,J).

When Disp1^Δ2/Δ2C, Shh^Cre/+ mutants were examined at E10.5, floorplate induction appears to have initiated as marked by sporadic Foxa2⁺cells in the ventral midline (Fig. 3K). However, Shh, which is activated later than Foxa2 and requires Foxa2 activity (Epstein et al.,1999), failed to be induced in the floorplate(Fig. 3K). Furthermore, this apparent floorplate co-expressed Foxa2 and Nkx2.2⁺(Fig. 3B,L; data not shown). Ordinarily, these are only transiently co-expressed but are rapidly restricted to their respective floorplate and pV3 progenitor domains(Fig. 3A,B; data not shown). The number of Nkx2.2⁺ pV3 cells was also reduced in the mutant(wild type, 60±6, Disp1^Δ2/Δ2C,Shh^Cre/+, 25±4, n=3, P<0.01 Fig. 3B,L). Thus, the ventral midline cells represent some intermediate state between pV3 and floorplate identity. Other ventral progenitors marked by Olig2⁺(Disp1^Δ2/Δ2C, Shh^Cre/+,35±5, Fig. 3L) and Nkx6.1⁺ (Disp1^Δ2/Δ2C,Shh^Cre/+, 125±8, Fig. 3M) were reduced to similar levels to those observed in Disp1^Δ2/Δ2homozygous mutants (Tian et al.,2004). As with the face, the overall phenotype observed in conditional Disp1^Δ2/Δ2C, Shh^Cre/+mutant was weaker than that in Disp1^Δ2/Δ2,Shh^+/- mutants, but comparable with that of Disp1^Δ2/Δ2 mutants. This phenotype was significantly enhanced when Disp1 levels were further reduced in Shh-expressing cells in Disp1^Δ2C/C829F, Shh^Cre/+ mutants, where after recombination all Disp1 activity derives from a single hypomorphic Disp1^Δ2allele. In these embryos, there was a complete failure of floorplate development (Fig. 3U); pV3 and pMN cells move to the midline and are reduced to less than 10% of the wild-type numbers, a similar reduction was also observed for Nkx6.1⁺ and MNR2⁺ cells(Fig. 3V-X). Although the severity of the phenotype was enhanced, it was still slightly less severe than that of Disp1^Δ2/C829F, Shh^+/- mutants(Fig. 3P-T), where pV3 Nkx2.2⁺ progenitors were completely absent (compare Fig. 3S with 3X).

Next, we addressed the specific requirement for Disp1 for cholesterol modified Shh ligand. We have previously reported on an N-Shh allele generated by inserting a stop-codon into the endogenous Shh gene at the position where normal cleavage and cholesterol addition occurs(Lewis et al., 2001). The protein produced is identical to that of the normal Shh signal and is expected to undergo N-terminal palmitoylation (Chen et al., 2004). Unfortunately, this allele is dominant lethal,highlighting the importance of cholesterol modification to normal Hh regulation (Lewis et al.,2001). To overcome this dominant lethality and to enable us to address N-Shh activity in a Disp1^C829F/C829F mutant background that lacks all Disp1 activity, we created a conditional N-Shh(N-Shh^C) allele (J.J. and A.P.M., unpublished). Mice heterozygous for this allele are viable and fertile permitting genetic intercrosses with the Disp1^C829F allele. Cre-mediated recombination leads to exclusive production of N-Shh from the endogenous Shh locus, an outcome that is essentially identical to the original non-conditional N-Shh allele. To initiate recombination throughout the embryo, we used a Sox2Cre transgene to induce recombination in the entire embryo (Hayashi et al.,2002). As the resulting N-Shh allele is under identical cis-regulatory control to the wild-type Shh allele, N-Shhexpression was restricted to Shh-expressing cells (data not shown). Unlike Disp1^C829F/C829F mutants, which do not survive beyond 9.5 dpc and show gross defects in neural, somite, cardiac, vascular, facial and limb development (Fig. 4A,B), Disp1^C829F/C829F, NShh^C/Shhⁿ,Sox2Cre embryos were alive at E10.5 but die at or around birth. Morphologically, these embryos were indistinguishable from the N-Shh^C/Shhⁿ, Sox2Cre mutants(Fig. 4C,D), which also employ N-Shh as the only available Shh ligand. At this stage, both Disp1^C829F/C829F, N-Shh^C/Shhⁿ,Sox2Cre and N-Shh^C/Shhⁿ, Sox2Cre mutant embryos are very similar to the wild type, except for midline defects in the frontal nasal process. Thus, Disp1 does not appear to be required for N-Shh activity.

To confirm that the rescue of Disp1^C829F/C829F mutant by NShh was due to active Hh signaling, we examined the expression of the Hh target Ptch1. In the wild-type limb bud, Ptch1 has a graded expression, occupying most of the posterior half of the distal limb bud mesenchyme, including the zone of polarizing activity (ZPA), the domain where Shh is produced in the limb (Fig. 5A,D). In Disp1^Δ8/Δ8 mutants at 9.5 dpc, Shh was expressed in the posterior mesenchyme as in the wild type. However, Ptch1 expression was greatly reduced and restricted to the distal posterior margin of the forelimb, the Shh expression domain (Kawakami et al.,2002). When Disp1^C829F/C829F mutants expressing N-Shh were examined, Shh expression was slightly downregulated in the posterior mesenchyme compared with the wild type, but to a similar degree to Shh expression in N-Shh^C/Shhⁿ, Sox2Cre mutant embryos(Fig. 5B,C). The Ptch1expression domain was also reduced and restricted to the posterior mesenchyme in both of these embryos (Fig. 5E,F), as in N-Shh/Shhⁿ limb buds(Lewis et al., 2001). High levels of Ptch1 expressions were observed(Fig. 5D-F), with no gradient of expression across the AP axis was evident(Fig. 5E,F). However, when the two limb buds from the same embryos were stained for Shh and Ptch1 expression separately and compared, the Ptch1expression domain clearly extended anterior to that of Shh in both Disp1^C829F/C829F, N-Shh^C/Shhⁿ,Sox2Cre and N-Shh^C/Shhⁿ, Sox2Cre embryos(compare the outlined area in Fig. 5B,C with those in E,F, respectively). This result indicates that Hh signaling expanded beyond the Shh expression domain in Disp1^C829F/C829F mutant when non-cholesterol modified N-Shh serves as the ligand. However, recent cell fate studies that demonstrate an anterior expansion of former Shh-expressing cells complicate the interpretation as the observed upregulation of Ptch1 in non-Shh-expressing cells could reflect `historical' signaling by cells which a few hours earlier were localized within the Shh-expressing ZPA (Harfe et al.,2004).

In the spinal cord, the movement of cells that expressed Shh do not complicate this analysis. All ventral neural progenitor types can be specified in the absence of a floorplate by notochordal Shh signaling(Matise et al., 1998; Ding et al., 1998). Furthermore, cell fate studies using the Shh^Cre allele indicate that Shh-expressing cells at the midline only contribute to the floorplate (data not shown). Consistent with previous reports(Caspary et al., 2002; Kawakami et al., 2002; Ma et al., 2002), a complete loss of Disp1 activity abolishes floorplate (Foxa2^-,Shh^-), pV3 (Nkx2.2^-) and pMN (Olig2^-)induction (Fig. 6A,B,G,H). When N-Shh was introduced into a Disp1^C829F/C829F mutant background, we observed floorplate (Foxa2⁺, Shh is not observed,see Discussion), pV3 (Nkx2.2⁺) and pMN (Olig2⁺)populations at correct positions and in comparable numbers to N-Shh^C/Shhⁿ, Sox2Cre embryos(Fig. 6C,D,I,J). N-Shh also rescues the Disp1 hypomorphic mutant phenotype (compare Fig. 6E,K with Fig. 6C,F,L). When N-Shh was introduced into Disp1^Δ2/C829F mutant background(Fig. 6F,L), the phenotype was indistinguishable from that of the NShh^C/Shhⁿ, Sox2Cre control (Fig. 6C,I).

The graded expression of Hh receptor Ptch1 in the ventral neural tube at E9.5 was absent in Disp1^C829F/C829F mutant(Fig. 6M,N)(Caspary et al., 2002) and greatly reduced in Disp1^Δ2/C829F mutants(Fig. 6Q). By contrast,signaling by NShh from the notochord and floorplate positively regulates Ptch1 expression in the ventral neural tube in Disp1^C829F/C829F or Disp1^Δ2/C829F backgrounds(Fig. 6P,R) as in a wild-type Disp1 background (Fig. 6O). Thus, the rescue of Disp1 mutants by N-Shh is achieved by restoring Shh signaling to the ventral neural tube. It is noteworthy that in all samples where we detect Foxa2, but not N-Shh, protein in putative floorplate cells, Shh expression in the floorplate was observed by in situ hybridization analysis (Fig. 6S-X) (see Discussion).

We have taken advantage of distinct alleles of Disp1 to examine the requirement for Disp1 in Hh signaling in the mouse. Previous reports have demonstrated that Disp1 function is essential for Shh and Ihh signaling(Caspary et al., 2002; Kawakami et al., 2002; Ma et al., 2002). Furthermore,a comparative analysis of autocrine and paracrine functions suggested that Disp1 is not required for autocrine signaling where Hh-producing cells are themselves targets of Hh signaling (e.g. Shh in the notochord), but Disp1 is essential for paracrine signaling to responsive cell populations adjacent to the sources of ligand (Caspary et al.,2002; Ma et al.,2002). These data suggest that Disp1 may function in release of active ligand from Hh-producing cells to adjacent target populations as is the case for Disp in Drosophila (Burke et al., 1999).

In this study we have used a variety of genetic strategies to examine Disp1 action. In the previous studies, the precise cellular requirements for Disp1 were unclear. Here, we specifically attenuated Disp1 activity in Shh-expressing cells; strikingly, the phenotypes closely resemble those observed when Disp1 activity is reduced in the entire embryo(Tian et al., 2004). That these phenotypes are not identical is most likely a technical limitation of our approach. In this, we used a conditional hypomorphic allele of Disp1 where essential sequence encoded within exon 2 was flanked by loxP sites and combined this allele with a Shh^Cre allele to `knock down' Disp1 levels exclusively within Shh-expressing cells. In this genotype, Cre-dependent recombination at the Disp1^Δ2C locus can occur only after Shh transcription is initiated; hence, it is likely some Shh signaling occurs while there is a sufficient level of Disp1 to lead to normal signaling(embryos heterozygous for Disp1^null/+ alleles are phenotypically wild-type). Alternatively, as this approach is only expected to modify Ihh signaling where Shh and Ihh expression overlap,wild-type Ihh activity from non-Shh expressing cells may contribute to the weaker phenotype in the conditional allele. The presence of a few Nkx6.1⁺ cells in Shh^Cre/C embryos is consistent with the former explanation.

With these provisos in mind, it is striking that the phenotypes of attenuated Disp1 activity in the Shh expression domain and in the whole embryo are so similar. Clearly, most if not all, paracrine Shh signaling within the face, neural tube and limb is dependent on Disp1 function in the ligand-producing cell, supporting a model in which Disp1 acts in signal production and not target cell response. How Disp1 acts at the molecular and cellular level is not clear. Studies in Drosophila report the accumulation of Hh ligand in Disp1 mutant cells in the posterior compartment of the imaginal disc (Burke et al., 1999), suggesting that Disp1 may regulate the release of bulk ligand from Hh-producing cells. Further work implicates Disp in apical trafficking (Gallet et al.,2003) within Hh-producing epithelia, suggesting that the defective release and ligand accumulation in Hh-producing cells may be secondary to altered membrane trafficking. However, we failed to observe any obvious accumulation of Shh in Shh-producing cells in Disp1-null mutant embryos (data now shown). We also failed to observe any differences in the release of bulk Shh protein into the medium when Shh was expressed in Disp1^C829F/C829F or wild-type fibroblasts(Tian et al., 2004). Whether this represents a difference between polarized epithelia and fibroblasts is unclear. Interestingly, reports of a highly active multimeric complex of lipid modified Hh ligands (Zeng et al.,2001) raises the possibility that Disp1 may function not in general release of ligand, but rather in the formation and/or release of an active fraction that is composed of Hh oligomers. Clearly, there is enhanced bioactivity in media conditioned by Shh-expressing fibroblast when Disp1 is active in these cells (Ma et al.,2002).

Cholesterol-modification of Hh ligands plays a key role in Hh signaling. In Drosophila, removal of cholesterol from Hh increases the range of ligand action, at least in part through a disruption of normal Ptch-dependent feedback control that normally sequesters ligand(Chen and Struhl, 1996). Whereas Ptch1 mediated sequestration of N-Shh is also defective in the mouse,cholesterol modification of N-Shhp is essential for long-range action in the limb (Lewis et al., 2001). Thus, although there appears to be species or context-dependent differences in the role of cholesterol in Hh signaling, in both flies and mice cholesterol-modification of Hh ligands is essential for normal signaling within a multicellular target field. In Drosophila, Disp is required only for signaling by cholesterol-modified forms of Hh(Burke et al., 1999). Our work in the mouse indicates a similar requirement for Disp1. Whereas, Disp1-null embryos arrest at E9.5 dpc with multiple defects,including an absence of ventral cell identities in the neural tube, expression of a single allele of N-Shh is sufficient to rescue many of these deficiencies. In the resulting neural tube, all Shh-dependent ventral cell identities are represented leading to a phenotype identical to that of N-Shh/Shhⁿ embryos on a wild-type Disp1background.

Interestingly, when we examined N-Shh localization in the floorplate and notochord in N-Shh/Shhⁿ embryos, we were unable to detect any immunoreactivity (in contrast to embryos carrying a single wild-type alleles of Shh). Thus, N-Shh appears to be rapidly lost from Shh secreting cells, whereas NShhp is retained and accumulates within the cell, a finding supported by earlier cell culture analyses(Bumcrot et al., 1995; Porter et al., 1996b). These results highlight the challenge faced in moving a cholesterol-tethered Shh ligand from the initial Shh-producing cell into the target field and the vital role Disp1 plays in this process. Recent studies have demonstrated that N-Shh fails to generate a soluble multimeric protein complex, lending further support to a link between Disp1 and oligomeric forms of Shh ligand(Chen et al., 2004; Zeng et al., 2001). That Disp1 shares a sterol sensing domain with Ptch1 and several proteins that regulate cholesterol biosynthesis or trafficking(Carstea et al., 1997; Hua et al., 1996) suggests cholesterol sensing by Disp1 plays some role in regulation of N-Shhp export.

Finally, given that Hh ligands undergo a second lipid modification, an N-terminal palmitoylation, how does this relate to Disp1 function? Early studies first noted that when NShh was highly expressed in tissue cultures, a reduced fraction of secreted ligand was palmitoylated compared with cells expressing wild-type N-Shhp (Pepinsky et al., 1998). However, more recent studies suggest that palmitoylation is largely independent of cholesterol addition(Chen et al., 2004; Gallet et al., 2003). Furthermore, loss of palmitoylation, but not cholesterol, results in a dramatic reduction in Hh and Shh activity(Chamoun et al., 2001; Chen et al., 2004; Lee et al., 2001). However,N-Shh retains biological activity in the neural tube (data herein) and limb(Lewis et al., 2001). Thus, it is likely that the N-Shh allele we have generated gives rise to an N-terminal palmitoylated ligand. If so, this lipid modification does not appear to be sufficient for retention of Shh ligand in Shh-producing cells from our data. Although both nonpalmitoylated and non-cholesterol tethered Shh(N-Shh) ligands both fail to form oligomers(Chen et al., 2004), N-Shh retains bioactivity while bioactivity is lost in non-palmitoylated Shh. One possible explanation for these results is that the ready secretion of N-Shh ligand may counteract the failure of oligomerization in paracrine signaling in the embryo. By contrast, continued membrane retention and an absence of oligomerization of non-palmitoylated cholesterol-modified Shh ligand may lead to an absence of sufficient active signal within the target field to mediate any paracrine signaling in the mouse embryo.

We thank Drs Kathryn Anderson and Tamara Caspary for allowing us to use Disp1^C829F/+ before the publication of this allele and for helpful discussion; Jill McMahon for ES cell injection; and Diane Faria for histology. For the gift of antibodies, we thank Jen Jensen (Nkx6.1), Ariel Ruiz i Altaba (Foxa2), Hirohide Takebayashi (Olig2) and Tom Jessell (Nkx2.2). B.H. was supported by a NRSA postdoctoral grant from NIH (NRSA AR08642). Work in C.J.T.'s laboratory was supported by a grant from the NIH (HD32443) and work in A.P.M.'s laboratory was supported by a grant from the NIH(NS33642).