Mouse dispatched mutants fail to distribute hedgehog proteins and are defective in hedgehog signaling

Hedgehog (Hh) signaling plays a key role in inductive interactions in many tissues during vertebrate development (reviewed byChuang and Kornberg, 2000;Ingham and McMahon, 2001). This process involves both short- and long-range signaling from a localized Hh source. Molecular studies of the Hh pathway have shown that Hh signaling is capable of exerting long-range signaling activity over a distance of tens of cell diameters and inducing distinct cell fates in a dose-dependent manner. For example, sonic hedgehog (Shh) expression in the notochord and floor plate patterns the ventral neural tube as well as the sclerotome of the somites (Chiang et al., 1996;Echelard et al., 1993;Fan and Tessier-Lavigne, 1994;Marti et al., 1995a;Roelink et al., 1995). Similarly, Shh expression in the zone of polarizing activity (ZPA)specifies digit identity along the anteroposterior axis of the developing limbs (Chiang et al., 1996;Lewis et al., 2001;Riddle et al., 1993;Yang et al., 1997).

Extensively studies on Hh signaling in both invertebrates and vertebrates have led to a prevailing model of Hh reception. Hh signal is transduced through hedgehog binding to patched 1 (Ptch), a multipass transmembrane protein (reviewed by Ingham and McMahon,2001; Kalderon,2000). Genetic and molecular studies suggest that Ptch inhibits the signaling activity of smoothened (Smo), a seven transmembrane protein that shares sequence similarity with G-protein-coupled receptors (reviewed byIngham and McMahon, 2001;Kalderon, 2000). Though the molecular mechanism remains to be elucidated, Hh binding to Ptch appears to relieve the Ptch-mediated repression of Smo. As a consequence, activated Smo can initiate the signaling cascade, turning on transcription of key Hh targets.

An attractive model to account for the activity of Hh is the generation of a Hh protein concentration gradient, which provides positional information in the morphogenetic field. The mechanism by which Hh protein moves across tens of cell diameters is not obvious because of the fact that Hh protein is membrane anchored through lipid modification. The Hh protein precursor undergoes autoproteolysis to generate an N-terminal signaling fragment (Shh-N)(Bumcrot et al., 1995;Lee et al., 1994), followed by two types of post-translational modification. A cholesterol molecule is covalently attached to the C terminus of Shh-N(Porter et al., 1996a;Porter et al., 1996b) and a palmitoyl group is added to the N-terminus of Shh-N(Pepinsky et al., 1998) (the resulting lipid-modified form of Shh-N will be denoted as Shh-Np). The role of lipid modification in Hh signaling is not completely understood, but in vitro studies have shown that Shh-Np becomes membrane anchored as a consequence of lipid modification. It is conceivable that an important step in Hh signaling is to release the membrane-anchored Hh from the Hh-producing cells to allow for subsequent `movement' through the morphogenetic field. Interestingly,movement of lipid-modified Hh in Drosophila depends on the activity of tout velou (ttv) in Hh-receiving cells(Bellaiche et al., 1998).ttv encodes a glycosaminoglycan transferase, suggesting TTV generates a proteoglycan that may mediate the transfer of Hh protein between cells. The role of ttv vertebrate homologs, the Ext genes(Stickens et al., 1996), in Hh signaling has not yet been established.

Some insight into the process of Hh release from Hh-producing cells came from the identification of the dispatched (disp) gene inDrosophila that is predicted to encode a twelvepass transmembrane protein and is required in Hh-producing cells to transport lipid-modified Hh protein (Burke et al., 1999).Drosophila mutants in the disp gene display phenotypes reminiscent of hh mutants as Hh protein, instead of moving out of Hh-producing cells, accumulates to a higher level in these cells(Burke et al., 1999). Disp exhibits sequence similarity to an emerging family of multipass membrane proteins, including Ptch, all of which contain a characteristic sterol-sensing domain (SSD) (reviewed by Kuwabara and Labouesse, 2002). These observations suggest that Hh movement is closely linked to lipid modification and likely employs novel cellular mechanisms in releasing and transporting a membrane-anchored cell surface protein. However, the biochemical mechanisms by which Disp facilitates Hh movement remain unknown. To address the issue of Hh movement in vertebrates,we report the identification of the mammalian dispatched gene and studies aimed to understand its role in Hh protein transport during vertebrate embryonic development.

Standard molecular biology techniques were performed as described(Sambrook and Russell,2001).

A mouse EST clone (IMAGE 1430982) containing sequence similarity to theDrosophila disp gene was used to screen a mouse embryonic cDNA library and several partial Disp cDNAs were obtained. The 5′end of the Disp cDNA was obtained by RT-PCR. A full-lengthDisp cDNA (4721 bp) was acquired by ligating together restriction fragments of partial cDNAs. The GenBank Accession Number for mouseDisp cDNA is AY150577.

Mouse Disp cDNA was used to screen a mouse 129/SvJ genomic library. To construct a positive/negative targeting vector for removing exon 8 of the Disp gene (the resulting allele is designated DispΔE8), a 2.7 kb fragment containing sequences upstream of intron 7 was used as the 5′ region of homology (Fig. 3A). A 3.5 kb fragment containing sequences downstream of exon 8 was used as the 3′ region of homology and was inserted upstream of the MC1-tk-pA cassette (see Fig. 3A). A PGK-neo-pA cassette was inserted between the 5′ and 3′ homology regions and replaces the seventh intron and eighth exon of the Disp gene (Fig. 3A). E14Tg2A.4 (E14) feeder-independent ES cells(Nichols et al., 1990) were electroporated with a SalI-linearized targeting vector and selected in G418 and FIAU as described (Joyner,2000). Heterozygous E14 ES cells were injected into blastocysts of C57BL/6 strain mice to generate germline chimeras. Chimeric males were mated with C57BL/6, 129/Sv, 129/Ola or Swiss-Webster females (to maintain theDisp mutant allele in different genetic backgrounds) and heterozygous animals were identified by Southern blotting of tail-tip DNA(Fig. 3B).

Histological analysis, whole-mount in situ hybridization using digoxigenin-labeled probes and section in situ hybridization using³³P-labeled riboprobes were performed as described(Wilkinson and Nieto, 1993). The mouse Disp in situ probe encompasses the last kb of theDisp cDNA.

We collected wild-type, DispΔ E8^+/- andDispΔ E8^-/- embryos at 9.5 dpc (genotypes confirmed by Southern blotting) for western blotting to detect the processing event of Shh. In addition, we transfected COS7 cells, using Lipofectamine Plus reagent(Invitrogen), with expression constructs, which encode either the full-length Shh or the N-terminal fragment of Shh (Shh-N) without post-translational modifications. Transfected cells were harvested 2 days after transfection. To control for the specificity of Shh antibodies, we also collected Shhmutant embryos at a stage similar to that of DispΔE8^-/- embryos. COS7 cells or 9.5 dpc mouse embryos were lysed in buffer containing 50 mM Tris (pH 8.0), 150 mM NaCl, 2 mM EDTA, 1% Triton X-100 (v/v), 0.5 mM DTT, 1 mM PMSF and 2 μg/ml each of aprotinin, leupeptin and pepstatin A. Insoluble materials were sedimented by centrifugation at 20,000 g for 30 minutes at 4°C. The supernatants were transferred into a fresh tube and the samples were separated on 15% SDS-PAGE and transferred onto PVDF membranes for immunoblotting(Harlow and Lane, 1999). The membranes were blocked in 10% w/v fat-free milk powder in phosphate buffered saline (PBS) containing 0.1% Tween 20 overnight and incubated with primary antibody against Shh for 2 hours. The membranes were incubated with secondary antibodies followed by chemiluminescent detection according to manufacturer's instructions (ECL, Amersham Pharmacia Biotech).

We followed a protocol kindly provided by Dr Gritli-Linde(Gritli-Linde et al., 2001)with some minor modifications. The embryos were fixed overnight in Sainte Marie's fixative (95% ethanol, 1% acetic acid) at 4°C. After washing the embryos three times, 30 minutes each, in 95% ethanol, we proceeded to paraffin embedding and sectioning at 5 μM. Tissues were dewaxed in xylene twice, 5 minutes each and rehydrated to water by taking through 100% ethanol twice, 5 minutes each, 95% ethanol twice, 5 minutes each and PBS once for 5 minutes. The endogenous peroxidases were blocked by incubating the slides in 3%H₂O₂ in methanol for 10 minutes in the dark at room temperature. The slides were rinsed in PBS three times, 5 minutes each, and the nonspecific staining was blocked by incubating the slides in PBS with 5%sheep serum, 0.2% BSA and 0.1% Triton X-100 for 40 minutes at room temperature. Slides were incubated overnight at 4°C with the primary antibody (anti-Shh) diluted 1:500 in PBS with 0.2% BSA and 0.1% Triton X-100 in a humidified chamber. The signal was amplified using a Tyramide signal amplification kit (TSA Biotin kit NEL700 or 700A from PerkinElmer). We followed a modified version of the manufacturer's protocol outlined below. The slides were rinsed three times, 5 minutes each, in TNT (0.1M Tris, pH 7.5,0.15 M NaCl, 0.025% Tween 20). Slides were then incubated with goat anti-rabbit biotinylated secondary antibody at 5 μg/ml (Vector laboratories) in TNT solution containing 2% w/v fat-free milk powder for 45 minutes at room temperature in a humidified chamber. Slides were rinsed three times, 5 minutes each, in TNT solution. The slides were incubated for 30 minutes in TNB buffer (0.1 M Tris, pH 7.5, 0.15 M NaCl, 0.5% blocking reagent)in the dark at room temperature in a humidified chamber. The slides were incubated for 30 minutes with SA-HRP diluted 1:100 in TNB buffer at room temperature in the dark in a humidified chamber. The slides were rinsed three times, 5 minutes each, in TNT solution and they were incubated in the Biotinyl-tyramide amplification reagent diluted to a working concentration of 1:50 for exactly 9 minutes in the dark. The slides were rinsed three times, 5 minutes each, in TNT solution and incubated with SA-HRP diluted 1:100 in TNB buffer for 30 minutes at room temperature in the dark and in a humidified chamber. Next, The slides were rinsed three times, 5 minutes each and once for 2 minutes, in TNT solution and developed in DAB solution (Vector laboratories)for 3-20 minutes. Finally, the slides were rinsed in PBS for 5 minutes and counterstained using 0.5% Toluidine Blue with 10 mM sodium acetate (pH 4.6).

We performed sequence analysis of the mouse genome and identified two genes with significant sequence similarity to the Drosophila disp gene(Fig. 1). One of these genes shares greater sequence similarity with disp and is likely the mouse ortholog of disp (which will be referred to as Dispthroughout this report), while the other (referred to asDisp-related) is probably a more distant member of the family. We isolated full-length Disp cDNA clones by screening a mouse embryonic cDNA library using a Disp EST clone (IMAGE 1430982) as a probe in hybridization, as well as amplifying fragments of Disp cDNAs by polymerase chain reaction (PCR). The complete Disp cDNA (4721bp)encodes a predicted protein of 1521 amino acids with a relative molecular mass of 170,047 (Fig. 1A). BothDisp and Disp-related encode proteins with twelve predicted membrane-spanning domains as well as stretches of sequences similar to a conserved domain known as the sterolsensing domain (SSD)(Fig. 1A). Proteins containing the SSD include several classes of proteins that are involved in different aspects of cholesterol homeostasis or cholesterol-linked signaling(Fig. 1B) (reviewed byKuwabara and Labouesse, 2002). Notably, Ptch, the Hh receptor, also contains an SSD. The function of the SSD is not well understood but it has been suggested that SSD plays a role in vesicular trafficking/cargo transport in relation to sterol and/or lipoprotein concentration.

As a first step towards understanding the potential role that Dispplays in Hh signaling, we examined the temporal and spatial expression patterns of Disp in mouse embryos collected from 7.5 days post coitum(dpc) to 18.5 dpc. Shh is first detected at late streak stages of gastrulation (∼7.75 dpc) in the midline mesoderm arising from the node(Echelard et al., 1993)(Fig. 2A). Weak Ihhexpression is also detected in the posterior part of the node at 7.75-8.0 dpc(Zhang et al., 2001). Genetic analysis demonstrated that Shh and Ihh play partially redundant roles in Hh signaling in the mouse node(Zhang et al., 2001). At 7.75 dpc, Disp is barely detectable by whole-mount in situ hybridization(Fig. 2B) but a ∼4.7 kbDisp transcript could be detected at this stage on a Northern blot(data not shown). By late-headfold stage just prior to somite formation,Shh expression is detected in the node and head process(Echelard et al., 1993)(Fig. 2C). At this stage,Disp is only very weakly expressed in cells immediately adjacent to the midline mesoderm (arrowheads in Fig. 2D) as well as at junctions between neural and surface ectoderm(arrows in Fig. 2D). Subsequently, Shh expression is detected in several signaling centers, including the notochord, floor plate and ZPA of the limb and in several endoderm derivatives (Echelard et al., 1993). As somites form (∼8 dpc) and the embryonic axis extends caudally, the notochord, which represents the caudal extension of the head process, also expresses Shh. Disp is initially weakly activated in the notochord and its expression is upregulated by 9.5 dpc(Fig. 2I,M). By 8.5 dpc, whenShh is induced in the floor plate at the ventral midline,Disp expression is only very faintly expressed in the floor plate at this stage as well as at later stages (Fig. 2I,M and data not shown). At ∼9.5 dpc Shh is activated in the ZPA of the forelimb(Echelard et al., 1993) andDisp is broadly expressed throughout the limb mesenchyme as well in the apical ectodermal ridge (AER) (Fig. 2H-J). Expression levels of Shh in the ZPA increase from 9.5 to 10.5 dpc (Echelard et al.,1993). At 10.5 dpc, Disp expression in both fore- and hindlimb is still broad, but its expression is downregulated both in ZPA and surrounding regions (arrow in Fig. 2L). Expression of Disp is also detected in the somite and branchial arches (Fig. 2H,M,L).

At later stages of development, Ihh expression is detected in developing chondrocytes. Ihh expression is first detected at 12.5 dpc in chondrocytes in the center of cartilage condensation of long bones(Bitgood and McMahon, 1995;St-Jacques et al., 1999). At 13.5 dpc, Ihh expression is downregulated in the more mature central cells that are undergoing hypertrophy(Bitgood and McMahon, 1995;St-Jacques et al., 1999). At this stage, the expression domain of Disp largely overlaps with that of Ihh (data not shown). In addition, a strong Dispexpression domain was detected in the articular chondrocytes facing the joint cavity (data not shown). At later stages, Ihh expression is restricted to the prehypertrophic chondrocytes between the zones of proliferating and hypertrophic chondrocytes(Bitgood and McMahon, 1995;St-Jacques et al., 1999)(Fig. 2O); Dispexpression remains associated with Ihh expression in the prehypertrophic condrocytes in addition to maintaining its strong expression in the articular chondrocytes (Fig. 2P). Taken together, these findings suggest a potential role ofDisp in Hh signaling since its expression domains overlap with those of both Shh and Ihh during early mouse embryogenesis.

To better understand the role Disp plays in Hh signaling, we generated a null allele of Disp using gene targeting in mice. TheDisp gene is located on mouse chromosome 1 and the genomic locus consists of eight exons. The eighth exon encodes the last 1193 amino acids of Disp protein, which include all twelve predicted transmembrane domains(Fig. 3A). We targeted the eighth exon to generate a null allele of the Disp gene (designatedDispΔ E8) (Fig. 3B). The gross morphology of homozygous DispΔ E8mutant embryos at 9.5 dpc (Fig. 4B) is remarkably similar to embryos deficient in the Smogene (Zhang et al., 2001)(Fig. 4C). Furthermore, similar to Smo mutants, homozygous DispΔ E8 mutants do not survive beyond 9.5 dpc. By contrast, DispΔ E8 heterozygous embryos cannot be distinguished from their wild-type littermates (data not shown). DispΔ E8 mutants exhibit cyclopia and holoprosencephaly. In addition, DispΔ E8 mutants fail to complete embryonic turning (Fig. 4B). The embryonic lethality observed in homozygousDispΔ E8 mutants is most probably due to defective heart development. DispΔ E8 mutants fail to undergo normal rightward looping of the heart, which remains as a linear tube and is surrounded by a bloated pericardial sac (Fig. 4B,F). All of these phenotypes have been reported in embryos defective in Hh signaling (Chiang et al.,1996; Zhang et al.,2001).

To confirm that the observed defects in DispΔ E8 mutants are due to defective Hh signaling, we examined the expression of the Hh targets,Ptch, Hip1 and Gli1(Chuang and McMahon, 1999;Goodrich et al., 1996;Marigo et al., 1996;Platt et al., 1997). In situ hybridization was used to monitor their expression in DispΔ E8mutants in wholemounts and sections. Expression of Hip1 is known to be completely dependent on Hh signaling(Chuang and McMahon, 1999)while Ptch expression is initially Hh independent but is strongly upregulated upon Hh signal transduction(Goodrich et al., 1996). InDispΔ E8 mutants at 9.5 dpc, Shh is expressed in the notochord, the ZPA, the gut endoderm and the branchial arches(Fig. 5B,J), but expression ofHip1 (Fig. 5F) andGli1 (Fig. 5H) is completely absent in DispΔ E8 mutants. Expression ofPtch is greatly reduced and only weak expression is detected in the sclerotome of the somite, the ventral neural tube and the distal posterior margin of the forelimb (Fig. 5D,L), which may reflect Hh-independent expression ofPtch. Taken together, these findings indicate that Disp is required for Hh signaling during mouse embryogenesis.

To better understand the molecular mechanisms that underlie the defects observed in DispΔ E8 mutant embryos, we performed a detailed histological and marker analysis. Our analysis focused on LR axis determination, the axial structures, the ventral neural tube, the somite and the limb, as the role Hh signaling plays in patterning these structures has been well characterized (Chiang et al.,1996; Lewis et al.,2001; Marti et al.,1995b; Riddle et al.,1993; Roelink et al.,1995; Zhang et al.,2001). In addition, formation of these structures involves both short- and long-range Hh signaling.

Disp mutants are first distinguishable at the six- to seven-somite stages (∼8.5 dpc) by the abnormal morphology of the forebrain, indicative of loss of ventral midline fate, and by a delay in cardiac morphogenesis (data not shown). The failure to complete embryonic turning and the absence of heart looping in DispΔ E8 mutants suggested that LR axis development may be affected, as has been previously reported in Smo mutants(Zhang et al., 2001).Pitx2, which encodes a bicoid-related homeobox protein, is expressed in the left lateral plate mesoderm (LPM) from two- to three-somite (∼8 dpc) to 10 somite (8.5 dpc) stages in wild-type embryos(Piedra et al., 1998;Ryan et al., 1998;Yoshioka et al., 1998).Pitx2 expression is greatly reduced in the left LPM in two- to six-somite DispΔ E8 mutants, whereas expression ofPitx2 in the head mesenchyme and yolk sac is unaltered (data not shown). These results suggest that defective Shh and Ihhsignaling in the node affects the establishment of LR asymmetry(Zhang et al., 2001) inDispΔ E8 mutants.

Analysis of Shh mutant mice suggests that Shh is required for the maintenance but not the formation of the notochord(Chiang et al., 1996). IfDisp is required for Hh signaling, phenotypes resembling the axial defects in Shh mutants should be observed in DispΔ E8mutants. Consistent with this hypothesis, expression of brachyury [which is required for differentiation of the notochord and is normally expressed in the primitive streak, the node and developing notochord(Herrmann and Kispert, 1994)]becomes discontinuous in the rostral region of DispΔ E8 mutant embryos (arrow in Fig. 6B). Though the origin of the floor plate is not completely understood, the floor plate and notochord share similar expression profiles (including Shhand Hnf3b) and there is good evidence to suggest that expression of Shh in the notochord acts short-range to induce floor plate(Le Douarin and Halpern, 2000;Placzek et al., 2000). InDispΔ E8 mutant embryos, Shh(Fig. 5B,J) andHnf3β (Fig. 6D,F)are not detected in the ventral midline of the neural tube, suggesting that the floor plate fails to form. These results indicate that Disp is required for Shh signaling in the axial midline.

Shh signaling from both the notochord and the floor plate plays a key role in patterning the ventral neural tube in a dosedependent manner(Chiang et al., 1996;Roelink et al., 1995). To examine whether dorsoventral patterning of the neural tube is affected inDispΔ E8 mutants, we probed the expression of molecular markers that define different dorsoventral positions in the early neural tube(Briscoe and Ericson, 1999;Briscoe and Ericson, 2001). In the neural tube, Pax3 expression is normally restricted to the dorsal half (alar plate) of the spinal cord from the tail to the diencephalons(Fig. 6G) and Pax6 is only weakly expressed in the alar plate and more strongly throughout the ventral half (basal plate) of the neural tube, except at the ventral midline(Fig. 6I). In DispΔE8 mutants at 9.5 dpc, Pax3 expression in the spinal cord extends ventrally (Fig. 6H),whereas Pax6 expression level is quite low (to a level characteristic of normal alar plate expression) (Fig. 6J). Wnt1 (data not shown) and Wnt3a(Fig. 6P) are expressed in the roof plate in DispΔ E8 mutants. These results indicate that the ventral neural fate is not properly specified in the absence of Disp. Consistent with this conclusion, expression of a set of homeodomain proteins in neuroprogenitor cells (such as Dbx1, Dbx2, Nkx6.1 andNkx2.2) was not detected in DispΔ E8 mutants (compareFig. 6K with 6L and data not shown). Expression of these homeodomain genes is induced or repressed in response to graded Shh signaling (reviewed byBriscoe and Ericson, 1999;Briscoe and Ericson, 2001). Recent studies suggest that the resulting overlapping expression domains of these genes specify different neuronal types, including interneurons and motoneurons, at distinct positions of the ventral neural tube. Loss of the homeodomain code resulted in absence of islet 1 expression, a marker for motoneurons, in DispΔ E8 mutants(Fig. 6N), as well as loss ofEn1, which is expressed in V1 interneurons (data not shown).

Many studies have shown that Shh signaling in the floor plate and notochord induces expression of sclerotomal marker Pax1 and suppresses the dorsal dermomyotomal marker Pax3(Chiang et al., 1996;Fan et al., 1995;Fan and Tessier-Lavigne,1994). In DispΔ E8 mutants at 9.5 dpc,Pax1 expression is not induced in the somite, suggesting that sclerotomal differentiation does not occur(Fig. 6R). By contrast,Pax3 expression in the somite is expanded ventrally(Fig. 6H,T). We then asked whether dermomyotomal development is affected in the absence of Disp. In wild-type embryos, the first myogenic bHLH gene to be expressed isMyf5 at 8 dpc (Summerbell et al.,2000), followed by the activation of myogenin at 8.5 dpc(Tajbakhsh et al., 1997).Myod1 expression is detected about 2 days later at 9.75 dpc(Tajbakhsh et al., 1997). InDispΔ E8 mutants at 9.5 dpc, Myf5 was detected at low levels in the dermomyotome (Fig. 6V). Myogenin and Myod1 expressions are not detected at these stages (Fig. 6X and data not shown). These results suggest that dermomyotomal development is initiated but does not proceed in DispΔ E8 mutants.

Shh signaling from the ZPA specifies digit identity along the anteroposterior (AP) axis of the limb(Chiang et al., 1996;Lewis et al., 2001;Riddle et al., 1993;Yang et al., 1997). As described above, though Shh expression in the ZPA appears to be normal in the forelimb buds of DispΔ E8 mutants at 9.5 dpc(Fig. 5B), Hh targets are either not induced (Hip1 and Gli1)(Fig. 5F,H) or the expression levels are greatly reduced (Ptch)(Fig. 5D,L), suggesting that proper AP patterning is disrupted. Consistent with this, Hand2(dHand) expression, which normally shows broader,Shh-dependent expression over almost half of the AP axis at this stage (Charite et al., 2000)(indicated by the bracket in Fig. 6Y), is truncated in DispΔ E8 mutants (arrow inFig. 6Z). Interestingly,expression of Hoxd13, the most posteriorly restricted Hoxd family member that is regulated by Shh signaling(Zakany and Duboule, 1999)(Fig. 6AA), is only slightly reduced in DispΔ E8 mutants(Fig. 6BB). Shhsignaling is known to induce Fgf4 expression in the apical ectodermal ridge (AER), which regulates proximodistal (PD) outgrowth of the limb bud(reviewed by Martin, 1998)(Fig. 6CC). Fgf4 also functions to maintain Shh expression in the ZPA. In DispΔE8 mutants at 9.5 dpc, Fgf4 expression is not detected in the AER (Fig. 6DD). This could be due to retarded growth of the mutants as well as defective Hh signaling to induce Fgf4 expression. By contrast, Fgf8 expression in the AER of DispΔ E8 mutants cannot be distinguished from that of wild-type embryos (reviewed by Martin,1998) (Fig. 6EE,FF). Dorsoventral (DV) patterning of the limb appears to occur normally in DispΔ E8 mutants(Parr and McMahon, 1995) (data not shown). Together, these findings indicate an absolute requirement ofDisp in multiple aspects of Hh signaling.

Studies in Drosophila suggest that disp is involved in facilitating the movement of the cholesterol-modified form of Hh and does not affect Hh synthesis or processing (Burke et al., 1999). As Shh expression appears to be normal inDispΔ E8 mutants, we asked whether processing of Shh to generate a cholesterol-modified N-terminal fragment of Shh also occurs normally in DispΔ E8 mutants. On western blots, Shh antibodies recognized the unprocessed (upper arrow inFig. 7) as well as the processed form of Shh (Shh-Np) (lower arrow inFig. 7) in wild-type andDispΔ E8^+/- embryos. Shh antibodies also recognized Shh-N, which migrates slower than Shh-Np on an SDS-PAGE. By contrast, neither the unprocessed form of Shh nor the processed Shh-Np or Shh-N could be detected in lysate from Shh mutant embryos. In lysates fromDispΔ E8^-/- embryos, a band running at the same position as Shh-Np was detected by Shh antibodies, suggesting that Shh processing occurs in the absence of Disp. In addition, the ratio of processed to unprocessed (a very small amount) (data not shown) form of Shh inDispΔ E8^-/- embryos could not be distinguished from that of their wild-type littermates. These results suggest that Shh processing occurs normally in the absence of Disp.

To investigate whether the phenotype observed in DispΔ E8mutants is due to defective Hh movement, we examined the distribution of Shh protein in wild-type and DispΔ E8^-/- embryos. Using the procedure described by Gritli-Linde et al., we found that in wild-type mouse embryos at 9.5 dpc, Shh immunoreactivity is strong in the notochord and extends outwards in a graded fashion (arrows inFig. 8A), upwards towards the ventral neural tube along the extracellular matrix (arrowheads inFig. 8A) as previously shown(Gritli-Linde et al., 2001)and downwards towards the branchial pouch (data not shown). Similar patterns of Shh immunoreactivity extending from the notochord were observed on embryo sections where the floor plate has not yet been induced(Gritli-Linde et al., 2001). In DispΔ E8 mutant embryos at this stage, Shh immunoreactivity is confined to the notochord and no immunoreactivity is detected outside the notochord (Fig. 8B). By contrast, in Smo mutant embryos, Shh immunoreactivity is detected in the notochord and extends in a graded fashion though at a lower level than that in wild type (data not shown). Taken together, these results indicate that while DispΔ E8 and Smo mutants share similar phenotypes, the underlying molecular defects are different. Hh transport appears to be normal in Smo mutants but Hh protein is not capable of transducing its signal in Hh-responding cells. By contrast, in the absence ofDisp, processed Hh protein fails to be transported out of Hh-producing cells and Hh-responding cells never receive the Hh signal.

We cloned the mouse dispatched gene and showed that it encodes a putative multipasss membrane protein with an SSD domain. Our phenotypic analysis ofDisp mutant mice demonstrated that Disp null mice phenocopySmo null mice (Zhang et al.,2001), suggesting that Disp is essential for Hh signaling. This conclusion was further supported by a detailed molecular analysis of Disp knockout mice that exhibit defects characteristic of loss of Hh signaling. We also provide evidence to indicate that Dispis not required for Hh protein synthesis or processing but rather is involved in moving Hh protein out of its sites of synthesis. In summary, our results are consistent with studies of Drosophila disp, indicating a conserved mechanism of facilitating Hh protein movement that is essential for proper Hh signaling.

Disp exhibits a dynamic expression pattern during mouse embryogenesis. It is possible that regulation of Disp expression involves Hh signaling. Expression of Disp in midline axial structures is relatively weak, although analysis of Disp mutants strongly suggests that Disp plays an essential role in midline Hh signaling. In this case, it is not known whether Disp is required continuously for proper signaling of Hh protein as initial expression levels ofDisp are low. In addition, Disp expression in the limb becomes downregulated in locations where Shh is upregulated. It is possible that Disp is not continuously required or a low level ofDisp expression is sufficient for Hh transport. It is interesting to note that in many structures Disp is expressed at a lower level in regions of Hh expression and at a higher level adjacent to regions of Hh signaling. One possibility is that Disp could be involved in a feedback mechanism to modulate Hh signaling. Alternatively, expression ofDisp outside Hh expression domains may imply a potential role in processes not mediated by Hh signaling.

Our mutant analysis revealed the essential role Disp plays in Hh signaling, including Shh and Ihh signaling. As the phenotypes observed in Disp mutants and Smo mutants are identical in our analysis, it is most likely that no Hh signal is transduced in the absence of Disp, despite the prominent expression of Hh protein. Hh signaling involves both short- and long-range signaling, and it is somewhat surprising that in Disp mutants even short-range signaling is defective. For example, induction of floor plate does not occur inDisp mutants, and this process requires direct cell-cell contact of ventral midline cells with the notochord and not long-range movement of Hh protein (Le Douarin and Halpern,2000; Placzek et al.,2000). It is possible that the Hh protein is not presented to the cell surface in the absence of Disp, although the Hh protein is properly processed in the secretory pathway of Hh-producing cells. Alternatively, Disp may be required directly in short-range signaling once the Hh protein is localized on the cell surface of Hh-producing cells. For example, Disp may be involved in partitioning Shh into membrane microdomains essential for Hh binding to Ptch or Disp may direct membrane to membrane transfer of Shh between Hh-producing and Hh-responding cells.

As Disp mutants do not survive beyond 9.5 dpc, it has not been possible to assess the role Disp plays in Ihh signaling in the developing chondrocytes and gut endoderm (Bitgood and McMahon, 1995; Ramalho-Santos et al., 2000; St-Jacques et al., 1999) as well as Dhh signaling in the developing testis and peripheral nerves (Bitgood et al.,1996; Parmantier et al.,1999). It is also possible that Disp has Hh-independent functions,because expression of Disp is detected in locations where none of the known Hh proteins is expressed. Answers to these issues will require further genetic and molecular studies.

Although the issue of lipid modification and its role in Hh movement in Hh-responding cells is not yet completely resolved, the crucial step of moving Hh protein out of Hh-producing cells appears to be evolutionarily conserved. Molecular analysis of Drosophila disp revealed its essential role in facilitating movement of the lipid-modified form of Hh protein in Hh-producing cells (Burke et al., 1999). Our studies demonstrate that the mouse ortholog of Dispatched also plays a similar role in Hh transduction. Because Disp-deficient mice phenocopy Smo mutants (Zhang et al., 2001), it is likely that Disp is involved in transporting all three mammalian hedgehog proteins. These results suggest that the molecular mechanism by which lipid-modified Hh is released from Hh-producing cells is conserved. However, it is not known whether Disp is dedicated to facilitate the movement of lipid-modified Hh proteins or it also plays a role in transporting other lipid-modified proteins. The function ofDisp-related is not known, but the fact that its restricted expression domain does not overlap with Hh expression (T'N. K. and P.-T. C.,unpublished) suggests that Disp-related is unlikely to be involved in the same process as Disp.

Generation of an active Hh signal is a highly regulated process. It involves autoproteolytic cleavage, lipid modification and regulated transport. Our studies show that Disp is not required for Hh protein synthesis or processing but rather is involved in moving Hh protein from its sites of synthesis. Mosaic analysis in Drosophila suggests that Disp is only required in Hh-producing cells but not in Hh-receiving cells to facilitate Hh movement, despite ubiquitous expression of disp mRNA(Burke et al., 1999). It is not known whether Disp also functions exclusively in Hh-producing cells for vertebrate Hh signaling. Compared with disp, mouse Dispexhibits a relatively restricted expression domain, although Disp protein distribution has not been determined. How Disp functions to facilitate Hh movement is also not known. Disp contains 12 predicted membrane-spanning domains but its subcellular localization remains to be determined. It is possible that Disp resides in the ER/Golgi to mediate the transport of Hh protein in the secretory pathway. Proteins with SSDs have been implicated in vesicular transport (Kuwabara and Labouesse, 2002) and Disp may be involved in a similar process to direct the movement of Hh-containing vesicles to the plasma membrane. Alternatively, Disp may function on the plasma membrane to promote the release of Hh protein from Hh-producing cells. Interestingly, the topology of Disp bears similarity to that of ion channels or transporters. Cellular and biochemical studies will be required to uncover the molecular mechanisms by which Disp facilitates transport of the lipid-modified form of Hh protein in Hh-producing cells.

We thank Dr Andy McMahon (Harvard University) for providing Shhand Smo mutant mice and all those who supplied probes. We thank Chris Wilson for help with sequence analysis, members of the Chuang laboratory for helpful discussion, and Chris Wilson, Tony Gerber, Didier Stainier, Shaun Coughlin and Tom Kornberg for critical reading of the manuscript. Work in the Chuang laboratory was supported by the Sandler Family Supporting Foundation,the HHMI Biomedical Research Support Program, March of Dimes Birth Defects Foundation and an NIH grant (HL67822).