Hedgehog (Hh) ligands activate an evolutionarily conserved signaling pathway that provides instructional cues during tissue morphogenesis and, if misregulated, can contribute to developmental disorders and cancer. Fully bioactive Hh is posttranslationally modified by a cholesteryl moiety at the C terminus (Porter et al., 1996) and a palmitoyl group at the N terminus (Pepinsky et al., 1998). Both lipids firmly tether Hh to the plasma membrane of the producing cell to effectively prevent unregulated ligand release. Signaling at distant cells therefore requires regulated Hh removal from the membrane, a process that is facilitated by vertebrate and invertebrate Dispatched 1 (Disp, also known as Disp1) orthologs: genetic studies in flies and mice have revealed that Disp is specifically required in Hh ligand-producing cells and that Disp inactivation reduces ligand release and compromises Hh pathway activity in vivo (Burke et al., 1999; Kawakami et al., 2002; Ma et al., 2002; Nakano et al., 2004). Yet, the mechanistics of Disp-dependent Hh release remained unclear. Long-lasting questions about the Hh pathway are therefore (1) how Disp drives dual-lipidated Hh release from the plasma membrane, (2) whether Disp acts directly or indirectly in the process, and (3) to what carrier – if any – Hh is transferred.

What makes these questions particularly interesting is that the Hh release protein Disp on Hh-producing cells is structurally related to the Hh receptor Patched 1 (Ptc, also known as Ptch1) on Hh-receiving cells (Hall et al., 2019). Both proteins contain 12 transmembrane helices and two extracellular domains and belong to the resistance–nodulation–division (RND) family of transmembrane efflux pumps. In addition, both proteins contain a conserved domain known as the sterol-sensing domain (SSD) that is involved in different aspects of homeostasis of free or esterified cellular cholesterol in other SSD proteins (Hall et al., 2019). These striking structural resemblances between Ptc and Disp and the conserved SSD constitute further unexplained features of the Hh pathway, because they imply that similar – possibly cholesterol related – mechanisms control the opposite functions of Hh release from producing cells and Hh perception at receiving cells.

In this study, to characterize Disp-dependent release of the vertebrate Hh family member Sonic hedgehog (Shh) from the plasma membrane, we produced murine Shh in Bosc23 cells, a derivative of HEK293 cells that endogenously express Disp (Jakobs et al., 2014). Notably, in our in vitro system, we made sure that Shh biosynthesis faithfully undergoes all required posttranslational modifications to generate the dual-lipidated, fully bioactive plasma membrane-associated morphogen. The first posttranslational modification consists of the removal of the Shh signal sequence during translocation into the endoplasmic reticulum. The resulting 45 kDa precursor proteins consist of an N-terminal signaling domain that starts with a cysteine (C25 in mouse Shh) and a C-terminal autoprocessing/cholesterol transferase domain. For the second modification, the autoprocessing/cholesterol transferase domain covalently attaches cholesterol to the C terminus of the N-terminal signaling domain and simultaneously splits the 45 kDa precursor protein at the cholesteroylation site (Bumcrot et al., 1995) to ensure complete C-terminal cholesteroylation of all Shh signaling domains. In contrast, the third essential posttranslational Hh modification – N-lipidation of signaling domains – requires a separate enzymatic activity encoded by the Hh palmitoyltransferase Hhat, the lack or insufficient expression of which results in the secretion of non-palmitoylated inactive Shh (Chamoun et al., 2001). Because HEK293 and Bosc23 cells lack sufficient endogenous Hhat activity (Jakobs et al., 2014), throughout this work, we expressed the 45 kDa Shh precursor together with human Hhat from one bicistronic mRNA (Jakobs et al., 2014). We then compared dual-lipidated Shh release from the plasma membrane of Disp-expressing and Disp-deficient Bosc23 cells using SDS–PAGE and immunoblotting. We found that Disp regulates proteolytic Shh processing from both lipidated membrane anchors (another posttranslational modification called shedding), because Shh shedding from Disp-deficient cells was strongly and specifically reduced when compared to Disp-expressing control cells. [3H]-cholesterol efflux assays further revealed that Disp-deficient cells are impaired in their ability to secrete [3H]-cholesterol into the culture medium, and cholesterol quantification assays showed that the amounts of free membrane cholesterol in these cells are significantly increased. These findings suggest that the primary function of Disp is to control the amount or spatial distribution of membrane cholesterol at the cell surface, and that cholesterol-dependent physical properties of the plasma membrane may in turn control Shh shedding. We support this possibility by demonstrating restored Shh shedding from Disp-deficient Bosc23 cells upon pharmacological cholesterol depletion or overexpression of the putative cholesterol pump Ptc (Zhang et al., 2018). These data link the known structural conservation between Disp and Ptc with a shared membrane cholesterol-related mechanism that is essential for both Hh perception in target cells and – as shown in this study – Hh relay from producing cells.

In the first part of our study, we established essential in vitro conditions for the release of physiologically relevant Hh from Disp-expressing cells into serum-depleted medium (Creanga et al., 2012; Jakobs et al., 2014; Tukachinsky et al., 2012). First, we expressed Shh together with Hhat to minimize the production of non-palmitoylated or only partially palmitoylated overexpressed Shh, as described previously (Jakobs et al., 2014). Second, it is known that solubilization of the dual-lipidated vertebrate Hh family member Shh from the plasma membrane requires a synergistic factor called Scube2 [signal peptide, cubulin (CUB) and epidermal growth factor (EGF)-like domain-containing protein 2; Hall et al., 2019]. Scube2 activity in Shh release critically depends on its C-terminal CUB domain (Creanga et al., 2012; Tukachinsky et al., 2012), which derives its name from the complement subcomponents C1r and C1s, sea urchin protein with EGF-like domains (UEGF) and bone morphogenetic protein 1 (BMP1). CUB domains contribute to protease activities in these proteins (Gaboriaud et al., 2011), possibly by binding to and inducing structural changes in the substrate to boost turnover (Bourhis et al., 2013; Jakobs et al., 2017). Alternatively, Scube2 has been implicated in the transfer of dual-lipidated Shh from Shh-expressing cells to distant receiving cells (Tukachinsky et al., 2012; Wierbowski et al., 2020). To distinguish between these possibilities, we produced dual-lipidated Shh in Disp-expressing Bosc23 cells in the presence or absence of Scube2, and analyzed cellular and solubilized proteins using SDS–PAGE and immunoblotting. Our analyses confirmed previously published reports (Jakobs et al., 2014, 2016) that Scube2 strongly enhances Shh solubilization and revealed increased electrophoretic mobility of most released Shh over that of the corresponding dual-lipidated cellular material (Fig. 1A). Such increased electrophoretic Shh mobility can be best explained by the proteolytic removal of both lipids together with the associated terminal peptides during release for three main reasons. First, the removal of Shh lipids alone, for example by chemical saponification, decreases electrophoretic Shh mobility instead of increasing it (Porter et al., 1996) (schematic in Fig. 1B). This rules out Shh release by hypothetical esterase equivalents. Second, reverse-phase high-performance liquid chromatography (HPLC) directly confirmed lipid loss during Shh release, because, in the presence of Scube2, a cholesterol-modified but non-palmitoylated cellular C25AShh variant (N-terminal cysteine replacement by a serine or an alanine blocks Shh palmitoylation; Hardy and Resh, 2012) converted into less hydrophobic soluble C25AShh (Fig. 1C). Third, Scube2 enhanced the conversion of hemagglutinin (HA)-tagged dual-lipidated ShhHA (in this construct, the HA tag was inserted adjacent to the cholesteroylated glycine 198) into truncated proteins that lacked the tag (Fig. 1D). These findings strongly suggest near-complete Shh delipidation during sheddase-mediated release, which is in line with previous in vitro (Jakobs et al., 2014, 2017) and in vivo (Palm et al., 2013; Schürmann et al., 2018) observations.

Fig. 1.

Scube2 enhances proteolytic Shh processing. (A–D) Schematics of expressed Shh constructs are shown, using PDB:3M1N as a template. P, palmitate; C, cholesterol; CW, Cardin–Weintraub motif representing the N-terminal cleavage site; scissors indicate loss of lipidated terminal peptides during release from Bosc23 cells. CW is shown in red; an inserted HA tag is shown in blue. (A) Increased electrophoretic mobility of soluble Shh (trunc) over the dual-lipidated cellular protein (P+C) results from the loss of both lipidated terminal peptides during release from Bosc23 cells (indicated by scissors), and this process depends on Scube2. Cellular (cells) and released (media) proteins were detected using Shh-specific antibodies (αShh). Residual serum albumin in TCA-precipitated supernatants served as loading control (PonceauS staining, PonS). Bottom: schematic of Shh release. (B) According to previous publications (Pepinsky et al., 1998; Porter et al., 1996), dual-lipidated Hh (cellular lipidated) migrates faster in SDS–PAGE than E. coli-expressed unlipidated Hh (no lipids), although mass spectrometry has determined molecular masses of 20,167 Da for the former form and 19,560 Da for the latter. Increased electrophoretic lipidated Hh mobility, despite higher molecular mass, is caused by SDS association with the large hydrophobic sterol backbone of cholesterol and the C16 hydrocarbon tail of the palmitate. Consistent with this, chemical hydrolysis of the ester bond that attaches cholesterol to Shh decreases the electrophoretic mobility of the delipidated product (Zeng et al., 2001). The observed increase in dual-lipidated Shh electrophoretic mobility during release from human cells (as shown in A) can therefore only result from the additional loss of associated terminal peptides, which more than compensates for this decrease. (C) Reverse-phase HPLC. The elution profile of Bosc23-expressed non-lipidated control C25AShhN (gray line) resembles that of soluble C25AShh (dotted line), but not that of its lipidated cellular precursor (black solid line). Elution profiles are expressed relative to the highest protein amount in a given fraction (set to 100%). fr#, fraction number. (D) Insertion of a C-terminal HA tag supports shedding. Removal of terminal peptides including the 1 kDa C-terminal tag increases the net electrophoretic mobility gain of solubilized proteins (arrowhead) (Jakobs et al., 2014, 2017), as visualized using antibodies against Shh and the HA tag (αHA). Data in A,C,D are representative of at least three independent experiments.

Fig. 1.

Scube2 enhances proteolytic Shh processing. (A–D) Schematics of expressed Shh constructs are shown, using PDB:3M1N as a template. P, palmitate; C, cholesterol; CW, Cardin–Weintraub motif representing the N-terminal cleavage site; scissors indicate loss of lipidated terminal peptides during release from Bosc23 cells. CW is shown in red; an inserted HA tag is shown in blue. (A) Increased electrophoretic mobility of soluble Shh (trunc) over the dual-lipidated cellular protein (P+C) results from the loss of both lipidated terminal peptides during release from Bosc23 cells (indicated by scissors), and this process depends on Scube2. Cellular (cells) and released (media) proteins were detected using Shh-specific antibodies (αShh). Residual serum albumin in TCA-precipitated supernatants served as loading control (PonceauS staining, PonS). Bottom: schematic of Shh release. (B) According to previous publications (Pepinsky et al., 1998; Porter et al., 1996), dual-lipidated Hh (cellular lipidated) migrates faster in SDS–PAGE than E. coli-expressed unlipidated Hh (no lipids), although mass spectrometry has determined molecular masses of 20,167 Da for the former form and 19,560 Da for the latter. Increased electrophoretic lipidated Hh mobility, despite higher molecular mass, is caused by SDS association with the large hydrophobic sterol backbone of cholesterol and the C16 hydrocarbon tail of the palmitate. Consistent with this, chemical hydrolysis of the ester bond that attaches cholesterol to Shh decreases the electrophoretic mobility of the delipidated product (Zeng et al., 2001). The observed increase in dual-lipidated Shh electrophoretic mobility during release from human cells (as shown in A) can therefore only result from the additional loss of associated terminal peptides, which more than compensates for this decrease. (C) Reverse-phase HPLC. The elution profile of Bosc23-expressed non-lipidated control C25AShhN (gray line) resembles that of soluble C25AShh (dotted line), but not that of its lipidated cellular precursor (black solid line). Elution profiles are expressed relative to the highest protein amount in a given fraction (set to 100%). fr#, fraction number. (D) Insertion of a C-terminal HA tag supports shedding. Removal of terminal peptides including the 1 kDa C-terminal tag increases the net electrophoretic mobility gain of solubilized proteins (arrowhead) (Jakobs et al., 2014, 2017), as visualized using antibodies against Shh and the HA tag (αHA). Data in A,C,D are representative of at least three independent experiments.

Based on these findings, and because Scube2-regulated Hh solubilization is known to require Disp (Hall et al., 2019), we expected strongly and specifically impaired Shh shedding from cells made deficient in Disp function. To test this hypothesis, we generated Disp-knockout cells (Disp−/−) using CRISPR/Cas9 genome editing in Bosc23 cells. Sequencing of the targeted genomic loci confirmed deletion of 7 base pairs, leading to a frameshift and a stop codon at amino acid 323 located in the first extracellular loop of the 1524-amino-acid protein (Fig. 2A,A′). We verified that predicted off-target sites were unaffected (Table S1) and confirmed complete Disp protein loss in Disp−/− cells using immunoblotting (Fig. S1A). Transgenic overexpression of murine Disp (Disptg) restored the immunoblot signal, demonstrating effectivity and specificity of the targeting approach (Fig. S1A′). We then analyzed Shh release from the plasma membrane of Disp- expressing and Disp-deficient Bosc23 cells into the supernatant using SDS–PAGE and immunoblotting. Consistent with the findings shown in Fig. 1A, Shh was released from Disp-expressing CRISPR non-targeting control cells (nt Ctrl) in the presence of Scube2, and the electrophoretic mobility of most released Shh was increased over that of the corresponding dual-lipidated cellular material (Fig. 2B, arrow). Notably, Shh shedding from Disp−/− cells was significantly reduced compared to that from nt Ctrl cells (Fig. 2B, arrows, B′). Instead, Disp−/− cells accumulated cellular Shh (Fig. 2B, arrowhead), as has been shown in vivo (Burke et al., 1999). Also consistent with previous observations (Kawakami et al., 2002; Ma et al., 2002), loss of Disp in Bosc23 cells did not affect Shh biosynthesis or autoprocessing of the primary translation product into the 19 kDa cholesteroylated signaling domain (Fig. S1B). Disp loss also did not affect morphogen secretion into serum-depleted medium, because unlipidated control C25SShhN (having the palmitate-accepting cysteine replaced and also lacking the C-terminal autoprocessing/cholesterol transferase domain) was readily released (Fig. 2C, arrows, C′). We conclude that Disp increases shedding of dual-lipidated Shh into bioactive truncated proteins (Fig. S2A,B), but that Disp is not essential for this process per se, because small Shh amounts can also be released from Disp−/− cells (Fig. 2B, arrow). This supports previous observations of non-essential Disp function for Indian Hh signaling in the skeleton (Tsiairis and McMahon, 2008) and in cells or tissues that generate high levels of Hh (Nakano et al., 2004). Notably, in the absence of Scube2 (Fig. 2D,D′), apparently unprocessed Shh (as indicated by similar electrophoretic mobilities of cellular and soluble proteins; Fig. 2D, arrows) is somehow released in a Disp-independent manner, suggesting that the underlying mechanism is physiologically irrelevant.

Fig. 2.

Impaired Shh release from Disp−/− cells. (A) Alignment of targeted disp gene sequences from Disp−/− cells and from non-targeted (nt Ctrl) cells. (A′) Schematic representation of the Disp protein structure. An asterisk indicates the CRISPR/Cas9-generated stop codon introduced at position 323, deleting 11 of 12 TM domains that together represent ∼80% of the protein sequence. L1 and L2 indicate extracellular loops. TM2–TM6 (colored red) constitute the SSD. (B,C) Immunoblots of cellular (c) and released (into the medium, m) Shh (B) and unlipidated control C25SShhN (C) in nt Ctrl and Disp−/− cells in the presence of Scube2. Arrows indicate solubilized Shh and the arrowhead indicates accumulated cellular material in Disp−/− cells. (D) In the absence of Scube2, Shh processing into serum-free medium was abolished in nt Ctrl and Disp−/− cells. Instead, both cell types released similar amounts of unprocessed protein. In B,C,D, anti-β-actin blots (αβ-actin) and Ponceau S staining of residual serum albumin (PonS) serve as loading controls. (B′,C′,D′) Quantifications of relative Shh (B′,D′) and C25SShhN (C′) release from nt Ctrl and Disp−/− cells. Ratios of solubilized versus cellular Shh were determined and expressed relative to Shh release from nt Ctrl cells (black bars). Data are mean±s.d. n=21 in B′, n=8 in C′ and n=5 in D′. ****P<0.0001; ns, not significant (two-tailed unpaired t-test).

Fig. 2.

Impaired Shh release from Disp−/− cells. (A) Alignment of targeted disp gene sequences from Disp−/− cells and from non-targeted (nt Ctrl) cells. (A′) Schematic representation of the Disp protein structure. An asterisk indicates the CRISPR/Cas9-generated stop codon introduced at position 323, deleting 11 of 12 TM domains that together represent ∼80% of the protein sequence. L1 and L2 indicate extracellular loops. TM2–TM6 (colored red) constitute the SSD. (B,C) Immunoblots of cellular (c) and released (into the medium, m) Shh (B) and unlipidated control C25SShhN (C) in nt Ctrl and Disp−/− cells in the presence of Scube2. Arrows indicate solubilized Shh and the arrowhead indicates accumulated cellular material in Disp−/− cells. (D) In the absence of Scube2, Shh processing into serum-free medium was abolished in nt Ctrl and Disp−/− cells. Instead, both cell types released similar amounts of unprocessed protein. In B,C,D, anti-β-actin blots (αβ-actin) and Ponceau S staining of residual serum albumin (PonS) serve as loading controls. (B′,C′,D′) Quantifications of relative Shh (B′,D′) and C25SShhN (C′) release from nt Ctrl and Disp−/− cells. Ratios of solubilized versus cellular Shh were determined and expressed relative to Shh release from nt Ctrl cells (black bars). Data are mean±s.d. n=21 in B′, n=8 in C′ and n=5 in D′. ****P<0.0001; ns, not significant (two-tailed unpaired t-test).

Next, we reversed the Disp−/− phenotype by the overexpression of transgenic V5-tagged Disptg (Stewart et al., 2018). Confocal microscopy of non-permeabilized Disp−/− cells expressing either Shh or Disptg confirmed secretion of both proteins to the cell surface (Fig. 3A,A′). Consistent with their cell-surface localization, co-expressed Disptg restored Shh shedding from Disp−/− cells (Fig. 3B, arrows, B′). We also tested the activity of a murine DispΔL2 variant lacking amino acids 752–972 of the second extracellular loop, located between transmembrane (TM) regions 7 and 8 (Fig. 4A), to determine a possible role of this loop in Shh binding and release (Cannac et al., 2020). This assay did not reveal significantly increased Shh shedding from DispΔL2-expressing Disp−/− cells (Fig. 3B,B′), indicating that the Disp L2 region contributes to the process. Shh shedding from Disptg- and DispΔL2-transfected nt Ctrl cells was also not significantly increased, indicating sufficient endogenous Disp expression in these cells (Fig. 3C,C′).

Fig. 3.

Overexpressed Disptg locates to the cell surface and restores Shh release from Disp−/− cells. (A,A′) Representative confocal planes of Disp−/− cells expressing Shh (A, red) or Disptg (A′, red). Both transgenes were secreted to the cell surface. Nuclei were counterstained with DAPI (blue). Dashed lines indicate the border of cytoplasmic eGFP signals (green). Images are representative of three experiments. Scale bars: 2 µm. (B) Co-expressed transgenic Disptg enhanced processed Shh release from Disp−/− cells (c) into the medium (m). Transgenic DispΔL2, which lacks most of the second extracellular loop, did not release significantly increased amounts of truncated Shh. Empty-vector (EV)-transfected Disp−/− cells served as negative controls. (B′) Quantified relative processed Shh release, as shown in B. Data are mean±s.d. n=10. **P<0.01; ns, not significant; P=0.377 for DispΔL2 (one-way ANOVA with Dunnett's multiple comparison post hoc test). (C) Co-expressed transgenic Disptg or DispΔL2 did not significantly increase Shh release from nt Ctrl cells. (C′) Quantified relative processed Shh release as shown in C. Data are mean±s.d. n=14. ns, not significant (one-way ANOVA with Dunnett's multiple comparison post hoc test). In B and C, arrows indicate solubilized truncated Shh from Disptg- or DispΔL2-expressing cells and the arrowhead indicates solubilized Shh from EV-transfected cells. Anti-β-actin blots (αβ-actin) and Ponceau S staining of residual serum albumin (PonS) serve as loading controls.

Fig. 3.

Overexpressed Disptg locates to the cell surface and restores Shh release from Disp−/− cells. (A,A′) Representative confocal planes of Disp−/− cells expressing Shh (A, red) or Disptg (A′, red). Both transgenes were secreted to the cell surface. Nuclei were counterstained with DAPI (blue). Dashed lines indicate the border of cytoplasmic eGFP signals (green). Images are representative of three experiments. Scale bars: 2 µm. (B) Co-expressed transgenic Disptg enhanced processed Shh release from Disp−/− cells (c) into the medium (m). Transgenic DispΔL2, which lacks most of the second extracellular loop, did not release significantly increased amounts of truncated Shh. Empty-vector (EV)-transfected Disp−/− cells served as negative controls. (B′) Quantified relative processed Shh release, as shown in B. Data are mean±s.d. n=10. **P<0.01; ns, not significant; P=0.377 for DispΔL2 (one-way ANOVA with Dunnett's multiple comparison post hoc test). (C) Co-expressed transgenic Disptg or DispΔL2 did not significantly increase Shh release from nt Ctrl cells. (C′) Quantified relative processed Shh release as shown in C. Data are mean±s.d. n=14. ns, not significant (one-way ANOVA with Dunnett's multiple comparison post hoc test). In B and C, arrows indicate solubilized truncated Shh from Disptg- or DispΔL2-expressing cells and the arrowhead indicates solubilized Shh from EV-transfected cells. Anti-β-actin blots (αβ-actin) and Ponceau S staining of residual serum albumin (PonS) serve as loading controls.

Fig. 4.

Overexpressed Ptctg restores Shh release from Disp−/− cells. (A,A′) Schematic representations of Disp (blue) and Ptc (green). Twelve transmembrane domains (TM1–TM12), two extracellular loops (L1 and L2), and the N- and C-termini are labeled. Conserved SSDs (TM2–TM6) are highlighted in red. DispΔL2 and PtcΔL2 lack most of the second extracellular loops (L2). (B,C) Co-expression of transgenic Ptctg or PtcΔL2 increases Shh shedding from Disp−/− (B) and nt Ctrl (C) cells (c, cellular Shh; m, Shh released into the medium). Arrows indicate solubilized processed Shh from Ptctg- or PtcΔL2-expressing cells, and arrowheads indicate reduced amounts of solubilized Shh from empty vector (EV)-transfected cells. Anti-β-actin blots (αβ-actin) and Ponceau S staining of residual serum albumin (PonS) serve as loading controls. (B′,C′) Quantification of relative processed Shh release as shown in B and C. Data are mean±s.d. n=6 in B′, n=9 in C′. *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Dunnett's multiple comparison post hoc test).

Fig. 4.

Overexpressed Ptctg restores Shh release from Disp−/− cells. (A,A′) Schematic representations of Disp (blue) and Ptc (green). Twelve transmembrane domains (TM1–TM12), two extracellular loops (L1 and L2), and the N- and C-termini are labeled. Conserved SSDs (TM2–TM6) are highlighted in red. DispΔL2 and PtcΔL2 lack most of the second extracellular loops (L2). (B,C) Co-expression of transgenic Ptctg or PtcΔL2 increases Shh shedding from Disp−/− (B) and nt Ctrl (C) cells (c, cellular Shh; m, Shh released into the medium). Arrows indicate solubilized processed Shh from Ptctg- or PtcΔL2-expressing cells, and arrowheads indicate reduced amounts of solubilized Shh from empty vector (EV)-transfected cells. Anti-β-actin blots (αβ-actin) and Ponceau S staining of residual serum albumin (PonS) serve as loading controls. (B′,C′) Quantification of relative processed Shh release as shown in B and C. Data are mean±s.d. n=6 in B′, n=9 in C′. *P<0.05, **P<0.01, ***P<0.001 (one-way ANOVA with Dunnett's multiple comparison post hoc test).

How can control of Shh shedding by Disp be explained? One answer to this question comes from established structural