Hedgehog lipids: Promotors of alternative morphogen release and signaling?

Two posttranslational lipid modifications present on all Hedgehog (Hh) morphogens—an N‐terminal palmitate and a C‐terminal cholesterol—are established and essential regulators of Hh biofunction. Yet, for several decades, the question of exactly how both lipids contribute to Hh signaling remained obscure. Recently, cryogenic electron microscopy revealed different modes by which one or both lipids may contribute directly to Hh binding and signaling to its receptor Patched1 (Ptc). Some of these modes demand that the established release factor Dispatched1 (Disp) extracts dual‐lipidated Hh from the cell surface, and that another known upstream signaling modulator called Scube2 chaperones the dual‐lipidated morphogen to Ptc. By mechanistically and biochemically aligning this concept with established in vivo and recent in vitro findings, this reflection identifies remaining questions in lipidated Hh transport and evaluates additional mechanisms of Disp‐ and Scube2‐regulated release of a second bioactive Hh fraction that has one or both lipids removed.


Physiological relevance of the Hedgehog signaling pathway
The Hedgehog (Hh) signaling pathway is crucial for animal development and plays major roles in stem cell maintenance and adult tissue regeneration. [1][2][3][4][5] Aberrant signal activation is also implicated in many cancers, including basal cell carcinoma and medulloblastoma. [6] Detailed studies on the structures and components of proteins involved in the Hh signaling pathway are therefore important to shed light on animal development and push cancer treatment and regenerative medicine further. Because of the complexity of the Hh pathway, the molecular mechanisms of Hh release and signaling are not fully resolved and are currently intensely discussed. This work aims to contribute to the discussion by linking biochemical data on two alternative modes of Hh release from producing cells-one in which both lipids stay attached to Hh and one in which they are getting removed-with available in vivo data and recent cryo-EM-derived models of Hh relay and signal perception.
Overview of the components of the canonical Hh signaling pathway F I G U R E 1 Overview of the canonical Hh signaling pathway. Hh is synthesized as a 46 kDa precursor. After removal of the signal peptide, its intein-like C-terminal domain (HhC) catalyzes the attachment of a cholesterol moiety to the C-terminus of the 19 kDa signaling domain (1). An acyl transferase (Hhat/Ski) attaches a palmitoyl acid to the Hh N-terminus (2). Both lipid anchors firmly tether the Hh molecule to the cell membrane, which is followed by their glypican-assisted multimerization (3). How these large Hh clusters travel to the distant receptor Patched1 (Ptc) is still controversial (4,5), but it is generally accepted that Disp and Scube2 are important factors for Hh release and/or transport to target cells. In the absence of Hh ligand on target cells, Ptc suppresses the activity of Smoothened (Smo,6), which results in destabilization of the glioma-associated oncogene transcription factor (Gli R , 7). Hh binding to Ptc (either with (8) or without its lipid anchors (9)) leads to internalization and subsequent degradation and simultaneously relieves Smo repression. This stabilizes full-length Gli (Gli FL ) and induces Hh target gene expression (10). The bottom panel shows recently published Ptc-Shh interactions, as determined by cryo-EM. [27][28][29]33,35] Shh is depicted in green, two calcium ions complexed by all Hhs in blue, a zinc ion complexed by most vertebrate Hh family members in black, palmitoyl and cholesteroyl moieties in yellow spacefill, and Ptc1 in red. Arrowheads point toward the N-and C-terminal Shh peptides or the lipid anchors. P: palmitoyl group, C: cholesterol. The semi-transparent structure on the far right consists of superimposed structures 6RMG and 6MG8 that reveal four potential sterol binding sites (S1-S4). These sterols are thought to fill a hydrophobic conduit that stretches from the plasma membrane to the Shh protein-binding interface known as Ptch1 in vertebrates [7] ). In Drosophila melanogaster only one Hh ligand is expressed, and three Hh family members are produced in mammals: Sonic Hedgehog (Shh), Desert Hedgehog, and Indian Hedgehog (Ihh). All three ligands share high sequence homology, are extensively posttranslationally modified, and bind to the same primary Ptc receptor to activate the canonical Hh signaling pathway in receiving cells. [8] The series of posttranslational modifications during Hh biosynthesis starts with the removal of the signal peptide from the 46 kDa Hh precursor in the endoplasmic reticulum ( Figure 1). Hht's intein-like C-terminal domain (ShhC/HhC) then catalyzes an autocleavage reaction and simultaneously covalently attaches a cholesterol moiety to the C-terminus of the 19 kDa N-terminal signaling domain ( Figure 1, step 1). [9] In a separate reaction, hedgehog acyltransferase (Hhat, called Skinny hedgehog (Ski) in the fly) attaches a palmitate to the N-terminal cysteine of Hh/Shh via an amide bond ( Figure 1, step 2). [10] After its secretion to the cell surface, dually lipidated Hh associates firmly with the outer plasma membrane and multimerizes into large clusters within cholesterol-enriched microdomains, [11] using heparan sulfate proteoglycans of the glypican family as scaffolds [12] (Figure 1 . Genetic studies firmly established that two proteins facilitate Hh release from the plasma membrane and its subsequent transport to Ptc: one is the transmembrane protein Dispatched1 (Disp, also known as Disp1 in vertebrates) [13,14] and the other is the soluble glycoprotein signal sequence, cubulin [CUB] domain, epidermal growth factor (EGF)related 2 (Scube2). [15,16] Cell culture experiments that used overexpressed Shh led to two alternative models of Hh release by these two proteins: (i) Disp and Scube2 may synergistically facilitate Hh ectodomain shedding by membrane-associated proteases (called sheddases), which cleave the N-and C-terminal lipidated peptides during release and thereby increase Hh hydrophilicity (Figure 1, step 4), [17][18][19] or (ii) Disp might extract dually lipidated Hh from the membrane and hand it off to Scube2, which in turn may shield both hydrophobic lipid anchors away from the aqueous environment [20,21] (Figure 1, step 5).
The lipidated Hh then becomes unloaded from Scube2 with the help of the coreceptors CDON/BOC and Gas1 to ultimately bind to its receptor Ptc [22] (Figure 1, step 8). In contrast to this suggested model, Hh released by ectodomain shedding will leave the terminal lipids associated with the plasma membrane of the producing cell and therefore does not require Scube2 chaperone function as a prerequisite to reach Ptc ( Figure 1, step 9). Hence, depending on the release and transport model, Hh may bind to Ptc at the cell surface of receiving cells either with its lipid anchors still present (Figure 1, step 8) or without them ( Figure 1, step 9). In functional terms, it is currently thought that ligand-free Ptc (Figure 1, step 6) forms a trimer [23] that undergoes conformational cycling to actively pump cholesterol from the inner leaflet of the plasma membrane to hand it off to extracellular sterol acceptors. [24,25] Consistent with such a highly conserved cholesterol transporter function of Ptc, [26] detection of four collective sterolbinding sites (S1-S4) in Ptc suggest a central hydrophobic conduit for cholesterol transport that courses from the inner leaflet of the plasma membrane to the extracellular domain [24,27,28] (Figure 1 bottom row, far right). Through this conduit, the amount of membrane cholesterol is continuously reduced but increases rapidly upon Hh binding to Ptc. [29] This suggests that Hh binding to Ptc stops its transporter-like function ( Figure 1, steps 8 and 9), which in turn increases cholesterol availability in the plasma membrane and indirectly activates the cholesterolresponsive transmembrane receptor Smo. [30] Smo activation as a consequence of Hh binding to Ptc and subsequent degradation of the Hh/Ptc complex stabilizes Gli to induce target gene expression [31,32] ( Figure 1, step 10).

Cryogenic electron microscopy-derived structures of Hh and its receptor Ptc
As outlined in the previous section, depending on the model of Hh release, Hht's lipid moieties might or might not be involved in, and required for, Ptc receptor binding and signaling. Recently, several groups used cryogenic electron microscopy (cryo-EM) to test these possibilities, with different results. Gong et al. [27] and Qi et al. [28] generated high-resolution cryo-EM structures of Ptc bound to bacterially expressed non-lipidated ShhN ( Figure 1, pdb 6DMY, 6RMG), indicating that Hh lipids are not essential for the interaction. In con-trast, Qi et al. [33,34] resolved high-resolution Ptc structures bound to detergent-solubilized dually lipidated Shh obtained by membrane extraction (R&D 8089-SH/CF) ( Figure 1, pdb 6OEV). In these structures, the N-terminal palmitate modification of Shh inserts into the hydrophobic cholesterol transport conduit formed between the two large extracellular domains (ECDs) of Ptc to block its transporter-like function (dubbed "plug-and-stop" model, the insertion site is close to the S3 position shown in the alignment of 6RMG/6MG8 (Figure 1).
Structural and biochemical studies conducted by Rudolf et al. [35] and Qian et al. [36] suggested that Hh signaling is further promoted by the Cterminal Shh cholesterol moiety that binds the first ECD of Ptc and may also plug the putative conduit for sterol transport, yet at a different site ( Figure 1, pdb 6RVD close to the S4 position shown in the alignment of 6RMG/6MG8). Together, these different structures indicate that several Hh binding options to Ptc-with or without one or both Hh lipidsmight be conceivable.
Indeed, the Shh/Ptc interactions presented by Gong et al. [27] and Qi et al. [28] only involve the so-called pseudo-active groove of Shh (pdb 6DMY) ( Figure 1, step 9), which also represents the binding site of the Shh function-neutralizing monoclonal 5E1 antibody [37] (Figure 2A, pdb: 3MXW) and other Hh antagonists. [38][39][40] Notably, a recently developed conformation-specific anti-Ptc nanobody binds to Ptc at a site overlapping with a site recognized by the pseudo-active groove of Shh and potently activates the Hh pathway, likely by locking Ptc into one of two possible conformational states. [24] This nanobodymediated conformational stabilization may in turn block sterol transport through the conduit. Structural comparison of Ptc in the absence or presence of unlipidated Shh supported a similar mechanism, revealing that ligand binding was associated with marked conformational shifts of both ECDs and stabilized a Ptc loop. This, in turn, may restrict sterol access to, or exit from, nearby S3 or S4 sterol binding sites in the conduit. [27] Another mechanism to inhibit cholesterol transport by Ptc may be through sterical blockade of cholesterol exit from the conduit to putative extracellular sterol acceptors upon Shh protein binding at, or close to the sterol exit site ( Figure 1, Shh protein on top of the S4 site is shown in the superimposition of 6RMG/6MG8). [25] These structure-derived functional modes indicate that Hh lipids may not be essential for Ptc activity regulation in this system and that the pseudo-active groove of Shh might be sufficient to elicit potent signaling by blocking cholesterol transport through Ptc or cholesterol export from Ptc to extracellular acceptors. Such Hh-lipid-independent signaling has recently been confirmed experimentally. [25] Seemingly contrary to these findings, initial cryo-EM results that favored lipidmediated Ptc activity regulation showed that the pseudo-active site points away from Ptc ( Figure 1, pdb 6OEV), leading to the question of how 5E1-sensitive physiological signaling via the pseudo-active groove of Shh may occur. This question was later addressed by three independent groups, who revealed that two Ptc molecules engage one Shh molecule: one Ptc molecule interacts with the terminal lipids and the other with the pseudo-active Shh core interface [34][35][36] (Figure 1, pdb 6RVD). Based on these structures, it is expected that the N-palmitate interaction with Ptc increases Shh signaling over that elicited by the pseudo-active groove alone.

F I G U R E 2
Under serum-depleted conditions, the N-palmitate is indirectly required for proper Shh release but not directly for Ptc activity regulation. (A) Top: Crystal structure of the interface of the fab fragment of the neutralizing monoclonal anti-Shh antibody 5E1 (cyan) and Shh (green). [37] Bottom: The 5E1-responsive pseudo-active groove of Shh is also involved in Ptc binding (only the extracellular loops of Ptc are shown). [27] (B) Immunoblots demonstrate that Scube2-mediated solubilized Shh that was released into the medium (m) was truncated when compared with cellular (c) Shh (green arrowhead). Hhat was co-expressed with Shh in the same cell. In contrast, Shh expression without Hhat resulted in similarly sized (red arrowhead) and truncated soluble Shh (green arrowhead) as did expression of non-palmitoylated C25S Shh and non-cholesteroylated ShhN. Data reproduced from [17]. (C) Crystal lattice interactions suggest that the N-terminal peptide of one Shh (green) blocks the Ptc binding site of the adjacent Shh (orange), and vice versa. N-terminal shedding at the Cardin-Weintraub (CW) site (red) truncates the N-terminal peptide and simultaneously exposes the Ptc interacting site of the globular protein part, which is demonstrated by increased 5E1 antibody binding to this site. IP: 5E1 immunoprecipitation of the interacting protein fraction, PD: heparin-pulldown of total protein in the same supernatants. 5E1 coupled to Protein A beads used for IP served as a loading control (load). Data reproduced from. [46,61] (D) Immunoblots (top) were used for protein quantification and visualization of different electrophoretic mobilities of dually lipidated R&D 8908-SH and Shh released from Scube2-and Disp-expressing cells. 20 ng of recombinant dually lipidated Shh (R&D 8908-SH) induced maximal C3H10T1/2 reporter cell differentiation (light gray bars). Of note, protein-amount-adjusted, non-palmitoylated C25A Shh differentiated the reporter cells with a similar efficiency (white bar), indicating that the palmitoyl moiety is not essential for Ptc activity. Consistently, bioactivity of dually lipidated Shh and non-palmitoylated C25A Shh was completely inhibited by neutralizing 5E1 antibodies (dark gray bars). This supports a predominant signaling function of the Shh pseudo-active groove. (E-G) Reverse-phase HPLC analyses using a C4 column revealed that Shh solubilized by Disp and Scube2 under serum-depleted conditions was less hydrophobic than its cell surface associated precursor [17] or lipidated R&D 8908-SH (red line) (E). As the control, unlipidated Δ25-38 ShhN eluted similarly (fr#) in media and cell lysates (F). [17] Solubilized C25A Shh in the serum-free medium was also less hydrophobic when compared with its corresponding cell lysate (G), supporting that protein solubilization goes hand in hand with cholesterol removal. Data shown in F, G adapted from [17,19]

Hh palmitoylation by Hhat potentiates signaling
Hhat was discovered in a genetic screen 20 years ago [10,41,42] and, as described before, covalently attaches a palmitoyl moiety to the Nterminus of the 19 kDa Hh signaling domain before its secretion to the cell surface ( Figure 1, step 2). This posttranslational modification resulted in a 10-to 30-fold increase of released Shh in its ability to convert C3H10T1/2 reporter cells to an osteoblast lineage. [10] Lossof-function mutations in the hhat gene or its fly orthologue ski generated phenotypes typical for those resulting from loss of hh function in mice and in Drosophila. Similar effects were observed in vitro and in vivo when human or murine Shh were made palmitoylation defi-cient by introducing a C24S or C25S point mutation, respectively. [43,44] Together, these findings led to the straightforward conclusion that the N-terminal palmitate is crucial for enabling Hh to bind to and to regulate Ptc directly [42] -a conclusion now strongly supported by some recent cryo-EM structures. [33][34][35][36] Other recent experiments, however, have indicated that the pathway-activating effect of N-palmitoylation might rather be indirect. Here, both lipid anchors initially tether the Hh molecules to the cell membrane until ectodomain sheddases release them in a regulated manner, leaving the lipidated terminal peptides associated with the cell membrane [18,[45][46][47] (Figure 1, step 4). Consistent with this model, if Shh is co-expressed with Hhat under serum-depleted conditions (0%-1%, the accepted and generally used condition to study Shh solubilization and activity [17,20,21,35] ), most solubilized proteins are truncated when compared with their cellular precursors (Fig. 2b, green arrowhead). [48] In contrast, if Hhat is not co-expressed, Shh will only be incompletely palmitoylated-and Nterminally membrane tethered-by low levels of endogenous Hhat. [17] In contrast to the palmitoylated Shh fraction that had both lipids associated with the plasma membrane prior to release ( Figure 2B, green arrowhead), the non-palmitoylated fraction is only associated via its C-terminal cholesterol moiety and can therefore be solubilized without the need for N-terminal processing (red arrowhead), similar to palmitoylation-deficient C25S Shh and non-cholesteroylated (and only incompletely palmitoylated) ShhN ( Figure 2B). Notably, not only is ectodomain shedding one possible alternative mechanism for Hh solubilization, but it may also bioactivate Hh. This is because the unprocessed N-terminal peptide of C25S Shh blocks the pseudo-active putative Ptc receptor binding site of the adjacent Hh molecule in a cluster, while its removal via shedding relieves this inhibition in vitro [46] and in vivo [45] (Figure 2C, top). As a proof of concept, the artificial, gradual truncation of the N-terminus progressively exposes the 5E1 binding pseudo-active groove of non-palmitoylated C25S Shh ( Figure   2C, bottom), which is in line with increased 5E1 reactivity of truncated Shh found by others. [21] The functional blockade of Hh biofunction by unprocessed N-termini of uncleaved molecules can explain why overexpressed palmitoylation-deficient C85S Hh (the fly equivalent of murine C25S Shh) suppresses Drosophila wing and eye development in a dominant-negative manner [18,45,49] and why gradual artificial truncation of the N-terminal unpalmitoylated peptide attenuates this effect. [45] Moreover, site-directed functional inactivation of the N-terminal Hh processing site strongly suppresses, and sometimes even abolishes Drosophila wing and eye development despite the presence of the N-palmitate, and its additional targeted removal restores their formation. [45,50] These in vivo observations support the possibility that the N-palmitate is not sufficient to activate signaling (contrary to pdb 6OEV shown in Figure 1 [33] ) and that palmitoylated membrane anchors may instead restrict morphogen release until site-specific processing switches membrane-bound Hh into delipidated bioactive forms.
Non-essential roles of the N-palmitate for Hh signaling activation on receiving cells can be further supported by in vitro assays that compare dually lipidated Shh (detergent-extracted from the cell membrane of Shh expressing cells, R&D 8908-SH), as used to generate cryo-EM structures that show lipid involvement, [33,34] and Shh released into serum-depleted medium of cells that express it together with the two known physiological release regulators Scube2 and Disp (Figure 2D). Immunoblots of the dually lipidated R&D 8908-SH show one approximately 19 kDa protein band, whereas solubilized Shh displays a truncated band slightly smaller than 19 kDa ( Figure 2D, right blot).
This reveals that dually lipidated R&D 8908-SH is unlike Shh after its release from Disp-and Scube2-expressing cells. This is also directly confirmed by reverse-phase high-performance liquid chromatography (RP-HPLC) analyses of their different hydrophobicities [17,18] : While R&D 8908-SH and cellular Shh elute in late fractions from the C4 column, consistent with their dual lipidation, Disp/Scube2-solubilized Shh elutes earlier and similar to an unlipidated N-truncated Δ25-38 ShhN control [17] (Figure 2E-F). As observed earlier, immunoblotting revealed that palmitoylation-deficient but cholesteroylated C25A Shh was also released as partially processed ( Figure 2D, upper band) and Ntruncated proteins ( Figure 2D, lower band). Again, RP-HPLC analyses confirmed decreased hydrophobicity of solubilized C25A Shh when compared with its cholesteroylated cellular precursor, suggesting loss of the C-terminal lipid during release ( Figure 2F, G). Direct comparison of protein-amount-adjusted dually lipidated Shh (R&D 8908-SH) with solubilized delipidated C25A Shh in C3H10T1/2 osteoblast differentiation assays revealed that both exhibited similar bioactivity ( Figure 2D, white bar), indicating that the N-palmitate is not essential to induce signaling through Ptc. Supporting the specificity of the assay, C25A Shh biofunction is completely inhibited by 5E1 antibodies that bind the truncated protein at a site that overlaps with the Ptc binding site (Figure 2A, 2C, 2D dark gray bar). [21,37] Strikingly, 5E1 also completely blocked bioactivity of the dually lipidated Shh ( Figure 2D, R&D 8908-SH, dark gray bar). This indicates accessibility of the pseudo-active groove of membrane-extracted Shh and explains its similar bioactivity, if compared with truncated C25A Shh in this assay, by a predominant signaling role of this interface that is not significantly increased by Hh lipids. A similar activity suppression by 5E1 was observed for palmitoylated ShhN co-expressed with Ptc in the same cells. [36] Taken together, these experimental findings are a strong reminder that dually lipidated, membrane-purified Shh does not necessarily represent all solubilized Shh forms, and that strong Shh signaling in vitro does not strictly or always depend on the N-palmitate. Such functional variability may be explained by inhibition of the same Ptc sterol conduit via Hh lipid-mediated 'plug-and-stop' [33,36] or by interactions with the pseudo-active Shh core interface that may either arrest Ptc conformational cycling between two alternative states to inhibit sterol transport [24,27] or physically block sterol exit. [25] Hh protein interactions with Ptc alone may therefore be sufficient to activate signaling through impaired cholesterol transport or export from the S4 site (Figure 1), [25] and Hh lipid-mediated 'plug-and-stop' at the S3 site may or may not contribute to it. The decision between these alternative Ptc binding and signaling modes is potentially made during regulated Hh/Shh release from the plasma membrane of producing cells, which may itself depend on the Hh-producing tissue type or developmental stage. For example, dually lipidated Hh relay from producing cells to Ptc via suggested lipophilic carriers such as exosomes [51] or lipoprotein particles [52] can only occur in the presence of these large carriers and in the absence of a dense extracellular matrix (such as in the developing neural tube or bone) that would inhibit their movement. Alternatively, regulated sheddase expression would be required to convert membrane-associated morphogens into desteroylated, highly bioactive soluble Hh and Shh variants, consistent with in vitro and in vivo observations. [53] In this regard, it is important to note that despite its suggested relevance for Ptc-binding-as derived from structural analyses [35,36] -the C-terminally attached Hh sterol is mostly required for the formation of Hh clusters at the plasma membrane of producing cells prior to their release but does not affect signaling much, [54,55] as has recently been confirmed in vitro and in vivo. [18,25,33] Finally, the F I G U R E 3 Scube2 can act as a Shh shedding accelerator. (A) Left: Scheme of antibodies used: Polyclonal anti-Shh (green) detects full-length Shh and monoclonal anti-CW detects only part of the N-terminal peptide (the Cardin-Weintraub (CW) site, red) that represents one cleavage site. Right: Immunoblots showing quantitatively palmitoylated Shh (cells) and Shh released into the supernatant. Scube2 accelerates Shh release in comparison to empty vector control, and increased incubation times enhance Shh release even in the absence of Scube2. (B) Shh without Hhat co-expression results in only partial palmitoylation and thus in partial truncation during Shh release. Release rates are also less dependent on Scube2. (C) Release and N-terminal processing of palmitoylation-deficient C25S Shh is only moderately affected by Scube2. All blots were stripped and reincubated with anti-CW (red). Data reproduced from. [17] PonceauS staining served as loading controls. Cellular anti-Shh signals serve as expression controls presence or absence of release regulators such as Scube2 may also help decide how and in which form Shh gets released from the plasma membrane of producing cells and tissues.

The Hh release regulator Scube2
Scube2 is a member of the you genes in zebrafish and is crucial for unperturbed Hh signaling in vivo. [15,16,56] Genetic studies revealed that Scube2 acts cell non-autonomously downstream of Hh expression but upstream of Ptc. [15,16] Revealing the mechanisms of Scube2 function in both current models of Hh release, postulating that Scube2 is involved in either Hh removal and transport of dually lipidated Hh to target cells [20][21][22] (Figure 1, step 5) or in Hh release by ectodomain shedding [17,48] (Figure 1, step 4, Figure 2D), will thus help in assessing their physiological relevance. Both models agree that, in vitro, co-expression of Shh/Hhat and Scube2 or the addition of exogenous Scube2 increases Shh solubilization into serum-free medium 10-to 30fold, whereas release of insufficiently or non-palmitoylated C25S Shh is less increased [17,21] (Figure 2B) (Figure 1, steps 4 and 5). These models disagree, however, on whether palmitate enhances the physical interaction between Scube2 and Shh to overcome membrane tethering of Shh, [21] or on whether Shh/Scube2 interactions may represent a transient interaction to indirectly increase Hh shedding. [17,48] One way to distinguish between these possibilities is through the molecular architecture of Scube2. Scube2 consists of nine EGF repeats, a highly glycosylated spacer region, and a CUB domain [15,57] that is required for Shh solubilization. [15,17,20,21,48] The CUB domain is evolutionary conserved and, depending on the model, might be involved either in binding the Shh cholesterol [20] or in regulating accessibility of the sheddase target. [48] To our knowledge, lipid-binding functions of other CUB domains have not yet been described, but protease substrate binding and protease activating roles are well-known for several other CUB domain-containing proteins. [58][59][60] Both models of CUB domain-regulated Shh release can be directly investigated by immunoblotting and RP-HPLC: if the idea holds true that hydrophobic, dually lipidated Shh is shielded from the aqueous environment by binding to Scube2, the soluble Shh protein should be of the same size-and hydrophobicity-as cell-bound Shh. If Shh shedding is more applicable, however, the solubilized Shh should become truncated and less hydrophobic upon Scube2-mediated release. As shown earlier, [17,48] Scube2 increases the release of truncated soluble Shh (if co-expressed with Hhat in serum-depleted media, Figure   2B, Figure 3A), in support of the shedding model, and HPLC analyses confirmed Shh delipidation during Scube2-mediated solubilization [17] ( Figure 2E, G). Autoradiography has also confirmed that truncated soluble Shh lacks palmitate, [46] and re-probing of immunoblots with an alternative anti-Shh antibody that recognizes only the potential cleavage site of the N-terminus (the Cardin-Weintraub (CW)-motif) showed that parts of this site are lost during Shh release [61] (Figure 3). In addition, a Scube2 mutant lacking the CUB domain failed to increase Shh shedding, which is in line with previous in vivo and in vitro results. [48] Notably, the published in vitro or in vivo data interpreted to support lipidated Shh transport by Scube2 does not exclude the possibility of Shh shedding, and several might even support it. First, immunoblots shown in some of these studies revealed fractions of truncated (higher electrophoretic mobility) Shh in solution, [10,14,[20][21][22]35,62] similar to what is shown in Figure 2B. Second, 5E1 predominantly binds this higher electrophoretic mobility form, similar to what is shown in Figure 2C, and this truncated form represented the bioactive Shh fraction. [21] Also note that direct conclusive proof of quantitative lipidated Shh transport by Scube2, for example by RP-HPLC of all relevant protein fractions, is unavailable. We further expect that the detection of soluble Shh in the absence of exogenous Scube2 would indicate that it may not be absolutely required for Hh release and transport and that, under certain circumstances, Scube2 might even be dispensable. Indeed, time-dependent release of processed Shh can be observed in the absence of Scube2 in vitro (Figure 3). This backs the idea that Scube2 acts as a release accelerator [17,21] rather than as an absolute prerequisite for Hh transport. Further support for this idea comes from several in vivo studies that demonstrate that shh overexpression compensates for developmental defects that resulted from F I G U R E 4 Disp is a Shh shedding modulator and is not absolutely required for Shh release into serum-depleted media. (A) Cryo-EM structure of Drosophila melanogaster Disp (violet) bound to non-lipidated C85II HhN (green). [71] Arrowheads point toward Shh N-and C-terminal peptides. (B) Release of truncated Shh from CRISPR/Cas9-generated Disp knockout cells (Disp -/-) was decreased when compared with truncated Shh release from CRISPR non-targeted guide RNA-treated cells (nt Ctrl). The same effect was observed for palmitoylation-deficient C25A Shh release. In contrast, release of lipidation-deficient C25S ShhN was not affected by Disp knockdown, and the soluble protein was not truncated. (C) Impaired release of truncated Shh can be overcome by treatment with the cholesterol-lowering drug methyl-beta-cyclodextrin (MβCD), a well-established shedding enhancer. Note that truncation of solubilized Shh excludes the possibility that MβCD extracts the dual-lipidated protein from the plasma membrane. This experiment indicates similar outcomes of Disp-and MβCD-enhanced Shh release. Data reproduced from. [19] Anti-GAPDH and PonceauS stainings served as loading controls you/scube knockdown, [15,16,63] which is incompatible with an essential shuttling role of Scube2. Moreover, scube2 overexpression does not result in hyperactivated Hh signaling in zebrafish, [15] and a zebrafish you ty97 mutant that expresses truncated Scube2 lacking its CUB domain displayed milder Hh phenotypes than shh, gli2, smo, or disp1 mutants [15,[64][65][66][67][68] and only in a subset of Hh-dependent tissues. [15,68] Consistent with this, mice made deficient in Scube2 function show only minor developmental defects, [68] and no known Scube ortholog is expressed in Drosophila melanogaster, although most other Hh signaling components are conserved. Still, Drosophila Hh function can be replaced with mammalian Shh despite the lack of a soluble Scube2like acceptor in Drosophila. [42] This in vivo finding therefore also supports that Scube2 is not essential for Shh transport. Finally, chimeric proteins that consist of the 19 kDa Hh/Shh signaling domains fused to green fluorescent proteins that carry the C-cholesterol are fully functional in flies [69] and can drive mouse development to term. [70] Such largely unperturbed biofunctions, despite insertion of the 25 kDa GFP tag between 19 kDa Hh/Shh and the terminal C-cholesterol, are difficult to reconcile with unperturbed association with Scube2 and Ptc receptor binding, [22] especially if these interactions critically involve both lipids (as suggested by pdb 6RVD, Figure 1). Taken together, these in vitro and in vivo observations are less compatible with direct and essential Scube2 chaperone functions in lipidated Hh transport and are more compatible with an alternative indirect role as a Hh release modulator.

Disp as a Hh release regulator
Both models of Hh release-in lipidated unprocessed or truncated processed form-postulate important roles of Disp in these processes. Disp is a 12-pass transmembrane protein that shows structural similarities to the primary Hh receptor and cholesterol transporter Ptc [13,71,72] (Figure 4A) and Niemann-Pick disease protein 1 (NPC1). [71,72] Genetic screens revealed that loss-of-function mutations in the disp gene phenocopy hh null phenotypes [13] due to disrupted Hh release from producing cells. [13,14] As a consequence, in disp knockout cell clones, Hh accumulates at the cell surface, [13,73] impairing long-range but not juxtacrine Hh signaling. [73] As outlined before, one way to explain this observation is that Disp may extract lipidated Shh from the cell membrane and hand it off to soluble Scube2, [20,22,74] similarly (yet in inverse direction) to what has been described for the transport and hand over of free (unesterified) cholesterol from NPC2 to NPC1. [75] Such Shh extraction was suggested to be driven by the transmembrane sodium gradient, [74,76] but the mechanism by which Disp may release lipidated Hh clusters and hand them over to Scube2 is still elusive. Notably, in vivo, the proposed hand-off mechanism from Disp to Scube2 [20,22] is strongly challenged: Drosophila melanogaster does not express any Scube orthologue, yet replacing fly Disp with murine Disp, which is claimed to require Scube2 as a Hh acceptor, fully rescued the disp -/phenotype in flies. [62] This important observation supports the idea that Disp function is conserved among species, yet that there is no strict in vivo requirement for Scube2 acting as an acceptor for lipidated Hh suggested to be handed over from Disp. Furthermore, all high-resolution cryo-EM-derived structures of Disp and Hh were obtained using non-lipidated Hh/Shh [71,72,76] (Figure 4A), while the alternative use of dually lipidated Shh (R&D 8909-SH/CF) resulted in a lower resolution structure in which the Shh secondary structure and both lipid moieties could not be clearly resolved. [72] This suggests favored interactions of Disp with the non-lipidated ligand that may possibly represent a stage that directly follows Shh shedding, for example to down-regulate Disp activity in a way similar to that possible for Ptc. [29] The other model of Scube2-modulated Shh shedding requires that the role of Disp in Hh release must be indirect, possibly involving a Disp-modulated messenger molecule that in turn affects shedding.
This messenger molecule might be plasma membrane cholesterol, as indicated by the known structural similarities between Disp and the established cholesterol pumps Ptc [26,29] and NPC1. [75] Further support for a sterol-pumping function of Disp comes from the detection of sterol-like densities in the Disp transmembrane domains, [72,76] again similar to what has been described for Ptc [28,29] and NPC1. [77] If correct, Disp may regulate Shh shedding by modulating the plasma membrane cholesterol content, which is known to affect shedding of many other membrane-associated proteins in experimental in vitro settings. [78][79][80][81][82] Disp-and membrane-cholesterol dependent Hh shedding may possibly be achieved by changing the half-life of cholesterolenriched microdomains that assemble and store membrane-associated lipidated Hh, which in turn may change access of certain sheddases to these sites. [79,80,83] Of note, such a sheddase-modulating role coupled to a cholesterol transporter function of Disp has most recently been supported [19] : first, it was confirmed that Disp knockout (Disp -/-) impaired truncated Shh release in vitro ( Figure 4B), in agreement with previous in vitro [20] and in vivo studies. [13,62,84,85] In further support of previous in vivo findings, disp knockout reduced the release of cholesteroylated but palmitoylation-deficient C25A Shh, whereas the sheddase-independent secretion of cholesteroylationand palmitoylation-deficient C25S ShhN was not affected ( Figure 4B).
Treatment of Disp -/cells with the well-established cholesterolenriched microdomain disruptor and shedding enhancer methyl-betacyclodextrin (MβCD) [81,86] restored Shh processing and release (Figure 4C), suggesting similar modes of Disp-regulated and MβCDenhanced Shh solubilization. [19] Indeed, the study also demonstrated increased amounts of free plasma membrane cholesterol in Disp -/cells and reduced Disp-dependent efflux of [3] H-labeled cholesterol from these cells into the media. Such indirect permissive Disp function via control of membrane cholesterol content, rather than direct extraction of dual-lipidated Shh, is consistent with published observations that (i) Disp is not absolutely required for paracrine Ihh signal-ing in the skeleton, [84] (ii) that shh overexpression can overcome Disp restriction [66] similar to what has been observed for Scube2, [15,16,63] (iii) that murine Disp can functionally replace Drosophila Disp despite the lack of a Scube2 ortholog in flies, [62] and (iv) that cells or tissues that generate high levels of Hh are partially independent of Disp. [13,20,62,85] Taken together, these findings suggest that Disp may export membrane cholesterol to indirectly increase Shh release, but at this stage does not rule out the additional possibility that the cholesteroylated Hh C-terminus may also represent a Disp substrate for export. This possibility needs to be investigated in the future.

Conclusions and prospects
In this work, we discussed that although lipid-modified Shh can bind and signal to Ptc, three important processes acting upstream of Ptc binding may play previously underestimated roles in Hh release. The first important posttranslational process is Hh N-palmitoylation by Hhat. As discussed in this perspective, it is considered essential for direct Ptc binding and signaling, but it can also facilitate sheddasemediated processing of the extended associated N-terminal peptide.
The second and third important processes are the modulating roles of Scube2 and Disp in the proteolytic processing of lipidated terminal Shh peptides. Therefore, although the available structural data provides excellent support for lipid-modulated Ptc binding and signaling, much of the in vivo and in vitro data supports an additional mode of Hh signaling that does not strictly depend on both lipids. We note that the possibility of more than one defined mode to release dually lipidated, monolipidated or delipidated Hhs from different cell types or tissues may provide an excellent strategy to release morphogens with different signaling strengths and activity ranges from just one membrane-associated precursor. The decision for these modes may be influenced by properties of the extracellular matrix, [47,87] the plasma membrane [19] or by the presence or absence of regulating factors such as Scube2. [17] Together with the fact that the pseudo-active Ptc binding site constitutes a hot-spot for mutually exclusive interactions with various extracellular regulators and with the matrix, [40,[88][89][90] and that this site also plays specific roles in non-cell autonomous signaling in some early embryonic tissues, [91] alternative modes of Hh release may be part of how Hh morphogens can function differently as morphogens or as growth factors in different cellular contexts, tissues, and developmental stages. Therefore, detailed characterization of tissue-or celltype specific Hh release modes continues to represent an important topic for future investigation.

MATERIAL AND METHODS
All experimental findings shown in this work, except for comparative studies that include commercial R&D 8908-SH Shh, were reproduced from previously published work using the original methods and protocols, as described below.

Cloning of recombinant proteins
Shh expression constructs were generated from murine cDNA (NM_009170: nucleotides 1-1314, corresponding to amino acids 1-438 and human Hhat cDNA (NM_018194). Both cDNAs were cloned into pIRES (Clontech) for their coupled expression from bicistronic mRNA to achieve near-quantitative Shh palmitoylation.

Protein detection
Cells were seeded into six-well plates and transfected with 0.5 or 1 μg Shh constructs and 0.5 μg Scube2 by using Polyfect (Qiagen).
Cells were grown for 2 days at 37 • C with 5% CO 2 in DMEM containing 10% FCS and penicillin-streptomycin (100 μg/ml). Media were changed to serum-free DMEM for 6 h, harvested, and centrifuged at 300 × g for 10 min to remove debris. Supernatants were incubated with 10% trichloroacetic acid (TCA) for 30 min on ice, followed by centrifugation at 13 000 × g for 20 min to precipitate the proteins. For 5E1 immunoprecipitation, 10 μg monoclonal 5E1 antibodies (DSHB) were coupled to 2.5 mg Protein A-sepharose beads (Sigma) per immunoprecipitation (IP). Five hundred microliters of medium per IP was incubated overnight on a rotator. For pulldown controls, 40 μl of heparin-sepharose beads was added to the other half of Shh-containing media (500 μl medium per pulldown) and incubated overnight on the same rotator. All precipitates were analyzed by reducing SDS-PAGE and immunoblotting by using goat-α-Shh antibodies (R&D Systems, AF464) or rabbit-α-Shh antibodies directed against the N-terminal CW-peptide sequence KRRHPKK (Cell Signaling) followed by incubation with horseradish peroxidase-conjugated secondary antibodies.
Where indicated, dual-lipidated, HEK293-derived human Shh (R&D Systems, 8908-SH) served as a size control and to quantify Bosc23expressed, TCA-precipitated proteins on the same blot. In cholesterol depletion experiments, serum-free medium was supplemented with 800 μg/ml MβCD for 6 h prior to TCA precipitation and subsequent immunoblotting analysis. GAPDH and PonceauS served as loading controls for cells and protein-conditioned supernatants, respectively.

Reverse-phase HPLC
Bosc23 cells were transfected with expression plasmids for dually lipidated Shh, truncated unlipidated Δ25-38 ShhN control protein and cholesteroylated (yet non-palmitoylated) C25A Shh. Two days after transfection, cells were lysed in radioimmunoprecipitation assay buffer containing complete protease inhibitor cocktail (Roche, Basel, Switzerland) on ice and centrifuged at 1000 × g at 4 • C for 10 min. Supernatants were ultracentrifuged at 35 000 rpm for 30 min at 4 • C in a TLA45 Rotor (Beckmann) to remove cellular debris, TCA precipitated, and the washed and dried TCA precipitate dissolved in 100 ml RIPA buffer. RIPA-dissolved proteins from supernatants and soluble wholecell extracts were then acetone precipitated. Protein precipitates were resuspended in 35 μl of (1,1,1,3,3,3) hexafluoro-2-propanol and solubilized with 70 μl of 70% formic acid, followed by sonication. Reversephase HPLC was performed on a C4-300 column (Tosoh, Tokyo, Japan) and an Äkta Basic P900 Protein Purifier. To elute the samples, we used a 0%-70% acetonitrile/water gradient with 0.1% trifluoroacetic acid at a flow rate of 1 ml/min at room temperature for 30 min. Eluted samples were vacuum dried, resolubilized in reducing sample buffer, and analyzed by SDS-PAGE and immunoblotting by using anti-Shh antibodies. 100 ng of dually lipidated, HEK293-derived human Shh (R&D Systems, 8908-SH) were used as a control. Signals were quantified with