Evidence for a role of vertebrate Disp1 in long-range Shh signaling

The hedgehog (Hh) family of secreted morphogens is essential for normal development and aberrant activation of the Hh pathway has been implicated in many types of cancer. Mammals have three Hh homologs: sonic hedgehog (Shh), desert hedgehog (Dhh) and indian hedgehog (Ihh) (Echelard et al., 1993). Like other morphogens, Hh proteins are secreted from localized sources in the embryo and can signal over short and long ranges. Patterning of the ventral neural tube is one of the best-studied functions of vertebrate Hh proteins. Shh is secreted initially from the notochord underlying the neural plate and later from the floor plate, the ventral midline of the neural tube, which is crucial for proper dorsoventral patterning of several cell types in the developing spinal cord (Albright et al., 2000). Notochord transplantation and co-cultures of explanted notochord with neural plate tissues have demonstrated that contact is required between the notochord, the source of Shh and neural plate tissue for normal, long-range patterning (Placzek et al., 1993; Placzek et al., 1991), indicating that the establishment of the Shh gradient involves cell-cell contact.

Shh, like other members of the Hh family, undergoes significant processing before release. Shh is generated as a 45 kD precursor, which undergoes autocatalytic cleavage producing 19 kD N-terminus and 26 kD C-terminus fragments (Lee et al., 1994; Porter et al., 1995). Concomitantly, a cholesterol moiety is added to the carboxy terminus of the N-terminus fragment (Porter et al., 1996). In addition, a palmitoyl adduct is added to the amino terminus of the N-terminus fragment by the acyltransferase skinny hedgehog, yielding a mature Shh ligand with dual lipid modifications (Buglino and Resh, 2008; Chamoun et al., 2001; Pepinsky et al., 1998). These modifications probably render the mature Shh protein strictly membrane-associated (Peters et al., 2004).

How an insoluble ligand such as mature Shh signals over a long range is not clear. However, long-range Shh signaling is known to require the action of dispatched 1 (Disp1) a twelve-transmembrane, sterol-sensing domain (SSD)-containing protein resembling the Shh receptor, patched 1 (Ptch1) (Kawakami et al., 2002; Ma et al., 2002). Disp1 and Ptch1 share many characteristics with members of the resistance-nodulation-cell division (RND) family of proton-driven transporters. Not only are the overall structures of Disp1 and Ptch1 similar to that of bacterial RND permeases, but aspartic acid residues in the fourth transmembrane span, which are thought to be important for proton translocation, are also conserved (Goldberg et al., 1999; Tseng et al., 1999). It has previously been shown that bacterial RND permeases function as trimers (Nagano et al., 2005). Similarly, it has been shown that Ptch is also assembled into trimers (Lu et al., 2006).

Unlike the restricted expression patterns of Ptch1 and Shh, expression of Disp1 and disp (the Drosophila Disp1 ortholog) is ubiquitous. Flies lacking disp and mice lacking Disp1 have severe disruptions in Hh signaling (Burke et al., 1999; Caspary et al., 2002; Kawakami et al., 2002; Ma et al., 2002). disp and Disp1 mutants most closely resemble mutants in the Hh signal transduction component smoothened (Smo), exhibiting a near complete loss of Hh signaling. Processing and modification of Hh proteins is normal in the absence of disp, but Hh is largely retained in source cells, suggesting a role for disp in the export of Hh (Burke et al., 1999; Caspary et al., 2002; Kawakami et al., 2002; Ma et al., 2002). Disp is also not required for the Hh response, as expression of the Hh target gene decapentaplegic (Dpp) is still induced in disp ^−/− clones abutting the anterior-posterior compartment boundary by Hh released from the posterior compartment of the Drosophila wing imaginal disc (Burke et al., 1999). However, in vertebrates, Shh signaling has a significantly longer range than in the fly imaginal disc, leaving open the possibility of a role for Disp1 in the formation of the long-range Shh gradient. Because Shh is retained in source cells in the absence of Disp1, it has not been possible to test whether Disp1 functions exclusively in Shh-expressing cells or whether it is also required in Shh-responding cells for the establishment of the long-range signal.

Here, we show that, in polarized epithelial cells, Disp1 mediates the basolateral secretion of Shh, and that Ptch1 on adjacent cells is required for the uptake of Shh. We further show that the range of Shh signaling is shortened in Disp1^−/− tissue, even when Shh is produced in Disp1-expressing cells. Together, these results support a model in which the reiterated Disp1-dependent secretion, coupled to Ptch1-dependent uptake, is responsible for the distribution of Shh through a tissue.

Unless otherwise noted, cells were stained after fixation in 4% paraformaldehyde, placed for 30 minutes on ice and then imaged with an LSM5 Pascal confocal or an AxioObserver Z1 fluorescence microscope (Zeiss).

Disp1^−/− cells were derived from Disp1^+/− cells using established procedures (Mortensen et al., 1991). Disp1^+/− embryonic stem (ES) cells were selected at 5 mg/ml of G418 (Invitrogen) until individual colonies were visible. After expansion, the genotype of each surviving colony was verified by Southern blotting (Kawakami et al., 2002). We identified two Disp1^−/− ES cell lines that could be neuralized using established protocols (Wichterle et al., 2002).

Wild-type ES cells (murine AB1) stably expressing Shh were derived by electroporating ES cells with a plasmid containing an elongation factor 1α [Ef1α (Eif1a — Mouse Genome Informatics)] promoter, the Shh coding sequence followed by an internal ribosome entry site (IRES) and the Zeocin (Zeo) resistance gene (Shh-IRES-Zeo). Electroporated cells were selected with 50 μg/ml Zeo (Invitrogen) for 10 days before individual colonies were picked, expanded and analyzed for Shh expression by western blot and immunofluorescence.

HEK 293T cells in 12-well dishes were transfected with V5-tagged wild-type canine Disp1 or Disp1 mutants. After 24 hours, cells were lysed in 1× NativePAGE sample buffer (Invitrogen) containing protease inhibitors (Roche) and 1% n-dodecyl-B-d-maltoside (Invitrogen) on ice for 30 minutes. Lysates were centrifuged at 13,000 g for 30 minutes, run on NativePAGE Bis-Tris gels (Invitrogen) and transferred to polyvinylidene difluoride (PVDF) membranes (Invitrogen). Membranes were probed with anti-V5 antibody (Invitrogen). Sizes were determined using NativeMark protein standards (Invitrogen).

Shh-expressing wild-type or Disp1^−/− ES cells were grown for 24 hours in Optimem (Invitrogen). Supernatants were collected and concentrated approximately 12-fold using Microcon centrifuge filters (Millipore). LDS sample buffer containing DTT (Invitrogen) was added to a final concentration of 1×. Cells were rinsed once in PBS, then lysed in 1× LDS sample buffer containing DTT. Samples were run on 12% Bis-Tris gels (Invitrogen) and transferred to nitrocellulose. Membranes were probed using a rabbit polyclonal anti-Shh antibody (Cell Signaling).

Canine Disp1 deletions were generated using the QuikChange II XL mutagenesis kit (Stratagene). Madin-Darby canine kidney (MDCK) II cells were maintained in DMEM (Invitrogen) supplemented with 10% FBS (Hyclone) and antibiotics. Three hundred thousand cells were plated onto 12 mm polyester transwell filters (0.4 μm pore size; Corning). After reaching confluence, cells were transfected with V5-tagged canine Disp1 constructs. After another 24 hours in culture, the cells were stained with mouse anti-V5 antibody (1:200; Invitrogen) and rabbit anti-ZO1 antibody (1:100; Invitrogen).

Canine Disp1 miRNA target sequences were as follows: miRNA1 GACTGGTTACGTGGAATAACA; miRNA2 TTGCGGGTGAAAGTTTGTTAA. miRNAs were cloned into pcDNA6.2-GW/EmGFP-miR (Invitrogen). For Shh localization experiments, Shh and either of the two Disp1 miRNAs were co-transfected into confluent monolayers of MDCK II cells. Thirty-six hours after transfection, cells were fixed and stained for Shh and GFP. To visualize Shh on the apical surface only, anti-Shh 5E1 antibody (1:10; DSHB) was added for 30 minutes at 4°C before fixation, permeabilization and staining.

For western blots, HEK 293T cells were co-transfected with Disp1 and miRNA plasmids and allowed to grow for 36 hours. Cells were lysed in a small volume of high-salt RIPA lysis buffer (600 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, 50 mM Tris-Cl, pH 7.6), diluted four-fold, run on an 8% Bis-Tris gel and transferred to nitrocellulose. Blots were probed with anti-V5 and anti-GFP antibodies.

KNRK cells (normal rat kidney cells transformed by Kirsten murine sarcoma virus) were grown on glass cover slips and transfected with Lipofectamine 2000 (Invitrogen). Twenty-four hours after transfection, cells were stained for Disp1 mouse anti-Myc (9E10) and rabbit anti-Shh antibodies.

ES cells were differentiated into neuralized embryoid bodies (EBs) using established procedures (Wichterle et al., 2002) and exposed to Shh N-terminus (ShhN; produced in Baculovirus-infected HiFive cells) at day 2. EBs were grown for an additional 4 days in supplemented DFNB with a medium change after 2 days. On day 6, EBs were stained for Pax7 and Nkx2.2.

EBs were derived as described above from AB1 cells expressing Shh and Disp1^+/− or Disp1^−/− cells. Equal numbers of ShhAB1 and Disp1^+/− or Disp1^−/− EBs were co-cultured for 48 hours (for Pax7 analysis) or 72 hours (for HB9 analysis) in a collagen matrix. EB co-cultures were fixed in 4% paraformaldehyde with 4% sucrose for 40 minutes on ice and stained. Ranges of Pax7 repression and HB9 induction were measured in ImageJ at about 20 clearly interpretable EB interfaces and statistic analysis was performed using Prism (GraphPad Software).

MDCK II cells were plated on transwell filters as described above. After cells reached confluence, they were transfected with Shh and Disp1 constructs and grown for 48 hours. For the final 12 hours of growth, 5E1 anti-Shh monoclonal antibody was added to the basolateral chamber 1:1 with normal growth medium. Cells were fixed and stained as described above. Shh retention was quantified by measuring average signal intensity within a circular area slightly bigger than a cell, and secretion was measured using the signal intensity in a 1.5-cell diameter wide circle around each Shh-expressing cell. Over 50 cells and their surroundings were measured, then analyzed with Prism.

To assess the ability of Disp1 to form trimers, we analyzed lysates of Disp1-transfected cells by blue-native gel electrophoresis. Besides a high molecular weight smear, a discrete band of around 480 kD was detected, the expected size for a Disp1 trimer (Fig. 1A). Similar lysates analyzed on an SDS PAGE gel resolved a single 160 kD band (Fig. 2D). As has previously been shown for Ptch and bacterial RND proteins, we found that Disp1 (and Disp2, not shown) assembles into trimers.

During neural tube development, Shh is secreted from two distinct sources: the notochord and the floor plate. Shh is produced in the notochord and signals over a long range to the basolateral side of overlying neural epithelium. Induction of the floor plate occurs in the neural epithelium closest to the notochord, where it is exposed to the highest levels of Shh. The floor plate also expresses Shh and floor plate-derived Shh could be released to the neural epithelium either apically or basolaterally. To determine at which side of an epithelium Shh is transported, we assessed the localization of both Disp1 and Shh in polarized epithelial Madin-Darby canine kidney (MDCK) cells (Mays et al., 1995). We used antibodies against the tight junction marker ZO1 (Fig. 1C-E, red) to visualize the border between the apical and basolateral domains of the MDCK cells (Nigam et al., 1991).

In MDCK cells transfected with V5-tagged canine Disp1, Disp1 protein accumulates at or near the basolateral surface (Fig. 1C). Similarly, Disp1 is detected on the lateral and basal sides of cells in the chicken neuroepithelium electroporated with Disp1 (see Figure S1A in the supplementary material). The carboxy-terminal domain is required for the localization of Ptch1 (Kawamura et al., 2008) and in analogy we tested the role of the homologous domain in Disp1. Complete deletion of the carboxy-terminal domain (ΔCTD; Fig. 1B) resulted in a molecule that neither localized to the surface (Fig. 1D) nor assembled into trimers (Fig. 1A). In addition to deleting the entire CTD, we also generated a series of six smaller (~60 amino acid) deletions spanning the CTD. Deletion of the first 60 amino acids after transmembrane twelve (Del1) also prevented trimerization and surface localization of Disp1, and was indistinguishable to ΔCTD (Fig. 1A,B and data not shown). However, deleting the second 60 amino acids after transmembrane 12 (Del2) resulted in a protein that still trimerized but was localized primarily to the apical surface of the cell rather than the basolateral surface (Fig. 1A,B,E). The remaining four 60-amino-acid-deletion mutants were indistinguishable from wild-type Disp1 in regard to localization and trimer formation (Fig. 1A,B). Taken together, these results demonstrate that Disp1 trimerizes and is localized to the basolateral side of polarized epithelial cells. Additionally, sequences within the CTD appear to be required for the basolateral localization and trimerization of Disp1.

To test whether Shh is secreted apically or basolaterally, a polarized epithelium of MDCK cells grown on transwell filters was transfected with a plasmid encoding Shh and the distribution of Shh on the apical and basolateral sides was determined by the inclusion of 5E1 anti-Shh monoclonal antibody in the apical or basolateral compartment of the transwell culture chambers. To deplete endogenous Disp1 protein, cells were co-transfected with miRNA constructs against canine Disp1.

Addition of the anti-Shh antibody to the basolateral side of the culture revealed significant amounts of secreted Shh in a gradient stretching a few cell diameters from the cells producing Shh (Fig. 2A), in agreement with the polarized Disp1 localization. To determine whether secretion of Shh on the basolateral side of the cell is mediated by Disp1, we co-transfected Shh with miRNA constructs against Disp1. Under these conditions, less Shh could be detected outside of the transfected cells (Fig. 2B), whereas more Shh was retained by these cells as a consequence of significantly reduced levels of Disp1 (Fig. 2C,D). Shh staining was also present on the apical side of transfected cells (Fig. 2E), and this apically localized Shh was unaffected by co-transfection of Disp1 miRNA constructs (Fig. 2F,G). The Disp1-dependent distribution of Shh away from the source suggests that Disp1 is required to produce a form of Shh that can be released. In polarized cells this export probably occurs at the basolateral side, as we find the apical localization of Shh to be Disp1-independent.

To determine whether the Shh distribution in vivo reflects our observations in MDCK cells, we injected 5E1 hybridoma cells into the neural tube lumina of stage 10 chick embryos and visualized the distribution of the anti-Shh antibody 36 hours later. Under these conditions, a significant accumulation of Shh could be detected basally and in the plane of the epithelium, whereas much smaller amounts were associated with the apical surface of the cells (see Fig. S1B,C in the supplementary material). Moreover, although 5E1 is an efficient Shh-blocking antibody (Ericson et al., 1996), no significant reduction in the Shh response was observed (see Fig. S1C in the supplementary material), a further indication that Shh transport in the neural tube is primarily along the basolateral aspect rather than the apical aspect of neural cells.

As Disp1, like other RND permeases, forms trimers, we tested whether mutant forms of Disp1 act as dominant-negative alleles to interfere with Disp1-mediated secretion of Shh. To test this, we co-transfected Shh with either Disp1, Disp1 lacking the CTD (ΔCTD), or a mutant in which the aspartate residues required for proton translocation were replaced with alanine residues (AAA) (Ma et al., 2002). We measured the amount of secreted Shh and the amount of Shh retained by the transfected cells. Compared with cells transfected with wild-type Disp1 (Fig. 3A), both Disp1ΔCTD (Fig. 3B) and Disp1AAA (Fig. 3C) inhibited the secretion of Shh and caused accumulation of Shh in the transfected cells (Fig. 3D). These findings suggest that both Disp1ΔCTD and Disp1AAA interfere with the function of endogenous Disp1, consistent with the idea that its multimerization is required for normal Disp1 function. The observation that Disp1AAA acts as a dominant-negative allele suggests that the proton channel is required for normal Disp1 function. Therefore, Disp1 is likely to function only in compartments where a transmembrane proton gradient exists, such as in early and late endosomes, trans-Golgi and multi vesicular bodies (MVBs). It seems unlikely that Disp1 would function at the plasma membrane, however, as the transmembrane proton gradient is unlikely to be large enough.

Lipid modifications have a profound effect on the signaling range and efficacy of Shh, as measured by the differential responses to the mutants (Lee et al., 2001; Li et al., 2006). We visualized secreted Shh to assess the effect of the lipid modifications on the range of Shh trafficking and the formation of the Shh gradient. MDCK cells were transfected with Shh mutants lacking either the amino-terminal acylation (C25S), the cholesterol moiety on the carboxy terminus (C*) or both (C25S/C*). Fully lipid-modified Shh could be detected a few cell diameters away from the sites of synthesis (Fig. 4A). Loss of the amino-terminal acyl group resulted in a significant increase in the range over which Shh was distributed (Fig. 4C), although it still appeared to be distributed in a gradient in the extracellular space. By contrast, Shh lacking the cholesterol modification was barely detectable in the extracellular space and not distributed in an obvious gradient (Fig. 4B). A Shh mutant lacking both cholesterol and palmitate anchors was entirely undetectable in our assay (Fig. 4D). These results show that the cholesterol moiety is required for retention of Shh in the extracellular matrix, whereas the acyl modification results in a steeper Shh gradient. Shh without the palmitate anchor still requires Disp1 function for secretion (Fig. 4E,F), indicating that Disp1 function is required for the formation of a secreted form of Shh that is retained in the extracellular space. It has previously been shown that Shh without cholesterol is secreted in a Disp1-independent manner (Ma et al., 2002). Consistent with this, we observed co-segregation of Disp1 and Shh, but not of Disp1 and Shh without a cholesterol moiety, in cells where they are co-expressed (see Fig. S2 in the supplementary material). The results above demonstrate that Disp1 acts on cholesterol-modified Shh and that Shh secreted in a Disp1-dependent manner is retained in the extracellular matrix. We have previously shown that the internalization of Shh is mediated by Ptch1 (Incardona et al., 2002; Incardona et al., 2000), which itself is a putative RND transporter, leaving open the possibility that the long-range transport of Shh involves the activity of both Disp1 and Ptch1.

To assess the roles of Ptch1 and Disp1 in the distribution of Shh, we used co-cultures of genetically different cell lines producing and distributing Shh. We generated embryonic stem (ES) cell lines, either Disp1^+/+ or Disp1^−/−, expressing Shh. Shh expression was not stably maintained in the Disp1^−/− cells, and Shh synthesized in Disp1^−/− cells consistently migrated more slowly on SDS PAGE gels, which suggests that Shh undergoes a Disp1-dependent modification (Fig. 5H). As expected based on previous reports, Shh could be detected in the concentrated supernatant of Disp1^+/+ cells, but not in the supernatant of the Disp1^−/− cells (Ma et al., 2002) (Fig. 5H). By growing Disp1^+/+ (Fig. 5A,C,E,) and Disp1^−/− (Fig. 5B,D,F) Shh cells surrounded by wild-type cells (Fig. 5A,B), Disp1^−/− cells (Fig. 5C,D) or Ptch1^−/− cells (Fig. 5E,F), we assessed the role of these proteins in the distribution of Shh away from its source. Visualization of Shh was performed by the addition of 5E1 to the live cultures for 1 hour followed by fixation and microscopy. Shh-expressing cells were labeled with a cell tracker (Fig. 5, blue). As this procedure was performed in the complete absence of detergent, only extracellular Shh was detectable.

Little Shh could be detected in co-cultures of Disp1^+/+ Shh-expressing cells surrounded by Ptch1^+/+ cells (Fig. 5A,C), but significantly more was found when Shh-expressing cells were surrounded by cells lacking Ptch1 (Fig. 5E), indicating that Ptch1 on surrounding cells has an active role in removing Shh from the sites where it is synthesized. Shh was still detected on the cell surface of Disp1^−/− cells (Fig. 5B,D,F) but the presence or absence of Ptch1 on the surrounding cells had no obvious effect on the accumulation of Shh at the surface of the Shh-expressing Disp1^−/− cells. This suggests that most of the Shh that reaches the cell surface in the absence of Disp1 cannot be bound efficiently by Ptch1 expressed by adjacent cells (Fig. 5G). Together, these results indicate that Ptch1 on adjacent cells preferentially binds Shh that is secreted in a Disp1-dependent manner, indicating that both Ptch1 and Disp1 play a role in the trafficking of Shh away from the sites of synthesis. However, whereas Disp1 is required in the cells synthesizing Shh (Fig. 2B), Ptch1 is necessary on the neighboring cells (that do not express Shh) to bind Shh and presumably move it away. An implication of these findings is that the cells surrounding Shh-producing cells affect the distribution of Shh away from the site of synthesis, and that Shh does not diffuse freely, even after secretion mediated by Disp1.

Although these observations suggest a model for the movement of Shh from the source cell to the adjacent cells, it remains unclear how a long-range signal is established. As Shh is distributed several cell diameters away from its source, and because Disp1 is ubiquitously expressed, we next tested whether Disp1 is also required in subsequent events, mediating re-secretion of Shh to cells farther away from the source.

To test whether Disp1 function is required outside of Shh-expressing cells for long-range signaling, we measured the range of Shh signaling in embryoid bodies (EBs) derived from ES cells with or without Disp1 function. To exclude the possibility that the loss of Disp1 cause changes in the responsiveness to Shh, we exposed neuralized EBs derived from Disp1^−/− and Disp1^+/− cells to various concentrations of soluble Shh N-terminus (ShhN) added to the culture medium. In both cases, the dose-dependent response to ShhN was identical, with repression of Pax7 at 2 nM ShhN and induction of Nkx2.2 at 4 nM ShhN. This demonstrates that there is no inherent defect in the Shh response in Disp1^−/− cells (see Fig. S3 in the supplementary material).

To test whether Disp1 is required for contact-dependent long-range signaling, wild-type EBs expressing Shh were co-cultured for 48 hours with EBs with or without Disp1 function. Like neural plate explants, neuralized EBs widely express Pax7 (Fig. 6A,B). Repression of Pax7 is one of the most sensitive responses to Shh and immunostaining with Pax7 antibody was conducted to assess the range of Shh signaling. In cases where Shh-expressing EBs contacted Disp1^+/− EBs, Pax7 staining was significantly reduced throughout the entire EB, on average 174.3±7.6 μm (Fig. 6D,F). By contrast, when Disp1^−/− EBs are grown in contact with Shh-expressing EBs, the range of Pax7 repression was much shorter, 72.2±3.3 μm, and was observed only in cells nearest to the Shh source (Fig. 6E,G).

Consistent with this observation, we found that wild-type Shh-expressing EBs could induce the expression of the motorneuron-specific protein HB9 at an average distance of 38.3±3.7 μm from the source in Disp1^+/− EBs (Fig. 6H,J), whereas the range of induction EBs lacking Disp1 was 18.2±2.1 μm (Fig. 6I,K). All differences were highly significant (P<0.0001, t-test). We obtained similar results using stage 10 chicken notochords as the localized Shh source (see Fig. S4 in the supplementary material).

Although the Disp1^−/− cells are normally responsive to Shh, they fail to efficiently distribute Shh away from its sites of synthesis, supporting the idea that Disp1 has an important function in both the cells expressing Shh and in the cells involved in the transport of Shh. In addition, Ptch1 plays an equally important role in the distribution of Shh. However, as the Ptch1^−/− cells have a constitutively activated Shh response, we cannot test the signaling range of Shh in Ptch1^−/− tissue.

Our results suggest that Ptch1 and Disp1 act in concert to mediate the transport of Shh through tissues. The similarities between Ptch1 and Disp1, such as their ability to trimerize (Lu et al., 2006) and their putative proton channel (Ma et al., 2002), indicates that their function might be conserved with that of the RND transporters in bacteria. In general, the role of Disp1 is in the secretion of Shh, whereas Ptch1 is involved in the uptake of Shh, and the function of both is necessary for long-range Shh signaling. These observations are consistent with a model in which reiterated secretion (by Disp1) and uptake (by Ptch1) are involved in the long-range transport of Shh. The non-directionality of this process, combined with the incomplete secretion of all internalized Shh, would sufficiently distribute Shh in a gradient away from the source.

The ability of 5E1 antibody to recognize Shh on and around live cells has been crucial in our experiments. This method of staining is more sensitive than 5E1 antibody application after fixation and permeabilization. This suggests that either fixation obscures many epitopes, or that the 5E1/Shh complex is more stable than Shh alone, intracellularly or not. Currently, we cannot discriminate between these options, but nevertheless, the distribution of 5E1 in some significant way reflects the range over which Shh is distributed.

All models of how Shh is distributed are currently complicated by the poor understanding of the form of Shh that moves from cell to cell. Biochemical analysis of the Shh in the supernatant of transiently transfected HEK 293T cells reveals that Shh is in a relatively large (~400-1000 kD) complex (Chen et al., 2004; Zeng et al., 2001) (data not shown). However, the form of Shh that is secreted by Disp1 and internalized in a Ptch1-dependent manner is still unknown. The amount of Shh released by stably expressing ES cells is very small and currently not amenable to further biochemical analysis. Although addition of the polysulfated compound suramin has previously been shown to increase the amount of Shh that accumulates in the supernatant (Yang et al., 1997), we find that this release is Disp1-independent and thus unlikely to represent the normal form of transported Shh (data not shown). Together, our observation that Shh-expressing ES cells release only very small amounts of Shh, combined with our observation that Ptch1 on adjacent cells is required for the transport away from the source, supports the idea that free diffusion of Shh is not responsible for the widespread distribution of Shh, but that in fact Disp1 and Ptch1 are participants in an active transport mechanism where Shh is passed from one cell to the next, mediating the distribution of Shh.

Signaling from the notochord to the basolateral side of the overlying neural plate, combined with our observation that the Disp1-mediated export of Shh is from the basolateral side of epithelial cells, supports the idea that both the reception and transport of Shh in epithelia occurs basolaterally. Consistent with this finding are observations that in the Drosophila wing disc, lipid-modified Hh is found almost exclusively on the basolateral side of cells in the anterior compartment (Callejo et al., 2006). However, in Drosophila, conflicting evidence exists about the effect of Disp on the apicobasal distribution of Hh. Although it was originally reported that loss of Disp had no effect on the localization of Hh within Hh-expressing cells (Burke et al., 1999), a later report found that apically localized large puncta of Hh are lost in the absence of Disp function or with cholesterol modification (Gallet et al., 2003). In support of the latter finding, the putative C. elegans Disp ortholog, CHE-14, is apically localized and is required for apical secretion from epithelial cells (Michaux et al., 2000). Nevertheless, in MDCK cells it appears that apical Shh localization is independent of Disp1 function, and perhaps more importantly, we show here that the basolateral localization of Disp1 requires sequences present in its carboxy terminal domain, demonstrating that the polarized Disp1 function is a regulated process.

The apically localized primary cilium, is thought to have a crucial role in the response to Shh as many components of the Shh response accumulate in it (Corbit et al., 2005; Rohatgi et al., 2007). Our observation that Shh can traffic to all cell surfaces in a Disp1-independent manner leaves open the possibility that Shh, after internalization at the basolateral side, moves to the apical surface and initiates the Shh response at the primary cilium. This is consistent with our observation that Shh expressing Disp1^−/− ES cells can transport Shh to the cell surface and that these cells have a significant autocrine Shh response. Importantly, this mechanism would effectively separate the mechanism by which Shh is transported from cell to cell by Disp1 and Ptch1, along the basolateral side of epithelia, from the Shh response at the apical primary cilium. It is possible that a similar phenomenon is present in Drosophila, where, despite the absence of primary cilia, basolaterally transported Hh activates the Shh response at the apical aspect.

Our observation that Ptch1 on adjacent cells is involved in transporting Shh away from the sites of synthesis also indicates that contact is required for the intercellular transport of Shh. This makes it probable that Shh transport is along the basolateral side of an epithelium as here epithelial cells are in close approximation. The commonly observed induction of Ptch1 expression in response to Shh (Marigo and Tabin, 1996) might not only serve as a negative-feedback loop to control the Shh response, but equally important, might also enhance the long-range transport of Shh.

It is probable that Disp1 and Ptch1 are bona fide members of the RND family of proton-driven transporters. RND permeases export a variety of substrates from the bacterial inner membrane, including heavy metal ions, antibiotics and detergents (Poole, 2002; Tseng et al., 1999). Ptch1 and Disp1 not only share the characteristic topology with other RND members, but also the aspartic acid residues in transmembrane 4, which are essential for proton translocation across the membrane (Goldberg et al., 1999). This conservation strongly suggests that Ptch1 and Disp1 are proton-driven transporters. Members of the RND family are known to function as trimers (Middlemiss and Poole, 2004; Murakami, 2008; Murakami and Yamaguchi, 2003). Here, we find that like other RND permeases, Disp1 also trimerizes and that this is likely to be crucial to its normal functioning.

Prokaryotic RND family members can export cargo from the outer leaflet of the inner membrane to the extracellular space, similar to the proposed role of Disp1 in the release of membrane-tethered Shh. It is therefore probable that the Disp1 trimer releases membrane-bound Shh using the energy of a proton gradient. This implies that Disp1 functions in an acidic compartment. In the secretory pathway, trans-Golgi vesicles have a pH of roughly 6.0 (Demaurex et al., 1998; Seksek et al., 1995), which would be sufficient to power Disp1. However, when secreted, the lipid modifications of Shh dictate that it must be present in multimers or some membrane-associated form, such as exosomes or lipoprotein particles (Neumann et al., 2007; Panakova et al., 2005), and intriguingly, trafficking of lipoprotein particles involves multivesicular bodies (MVBs). MVBs are specialized endosomes that transit from early/sorting endosomes to late endosomes, where their exosomal cargo is either secreted or sent to the lysosome for degredation. Newly synthesized proteins are also targeted to MVBs (Fomina et al., 2003; Lakkaraju and Rodriguez-Boulan, 2008), allowing the integration of both recycled and newly synthesized proteins into exosomes. The production of exosomes depends on a low pH in MVBs (Iero et al., 2008) and is compatible with the activity of both Ptch1 and Disp1.

Based on these results and previously reported findings, we propose the following model. Disp1 is active in MVBs and mediates the loading of Shh onto exosome/lipoprotein-like particles, which are then secreted. These particles are specifically recognized by Ptch1 at the surface of adjacent cells, which traffics them into early/late endosomes (Incardona et al., 2002; Incardona et al., 2000), where the particles are disassembled. Shh can either be degraded, trafficked to the apical surface or trafficked into MVBs, where it would be loaded onto exosomes again for re-secretion. This model accounts for the putative function of Disp1 as a proton-driven transporter and explains the high molecular weight complex that Shh is found in outside of cells, the pH-dependent action of Disp1 and Ptch1 in the intracellular trafficking of Shh and the role of Disp1 in the re-secretion of Shh.

We thank P.-T. Chuang (UCSF) for the Disp1^+/− ES cells and Disp1 probe, P. A. Beachy (Stanford University) for Disp1Myc and mouse Disp1 mutants, M. Scott (Stanford University) for the Ptch1^−/− ES cells, K. Mostov (UCSF) for MDCK II cells, C. Casillas for performing the electroporation experiments, Lina Kwong and Maarten Bijlsma for critically reading the manuscript. This work was supported by NIGMS Grant 1R01HD042307 to H.R. and an NSF graduate research fellowship to A.E.