An Emerging Role of Sonic hedgehog Shedding as a Modulator of Heparan sulfate Interactions

Background: Sonic hedgehog is released from expressing cells by proteolytic cleavage (shedding) of lipidated N-terminal peptides Results: The heparan sulfate (HS)-binding Cardin-Weintraub motif represents one N-terminal protease cleavage site in vitro and in vivo Conclusions: This results in impaired HS-binding of solubilized proteins Significance: Shedding may facilitate Sonic hedgehog diffusion through the HS-rich extracellular matrix

providing first in vivo evidence for Shh shedding and subsequent solubilization of N-terminally truncated proteins.

Introduction:
The proteins of the Hedgehog (Hh) family are powerful morphogens that control growth and patterning at various developmental stages. In vertebrates, the function of Sonic Hedgehog (Shh) -one of the three members of the Hh protein family (Shh, Indian Hh and Desert Hh) -has been thoroughly characterized (reviewed by (1)). Shh is essential for patterning of the ventral neural tube (2), for specifying vertebrate digit identities (3,4) and for the control of axon guidance in the developing nervous system (5). Given these various functions, it is not surprising that down regulation of Hh signaling leads to severe developmental abnormalities. In the adult, Hh pathway activation is involved in maintaining the stem cell niche, including the cancer stem cell niche (6), and Hh activity up regulation contributes to the formation and progression of various cancers (6)(7)(8). Thus, tight control of Hh secretion and spreading is essential and its molecular characterization required to better understand how Hh signals elicit dosedependent responses in temporally and spatially specific manner.
The Hh spreading mechanism is especially intriguing, because all Hh family members are released from producing cells despite being synthesized as dually lipidmodified, insoluble molecules (9). Both in vertebrates and in Drosophila melanogaster, Hh synthesis starts with precursor proteins that undergo a series of post-translational modifications (10). Following cleavage of an N-terminal signal peptide, Hh proteins undergo intein-related processing that involves internal cleavage between residues G198-C199 (mouse Shh nomenclature) (11)(12)(13)(14) and covalent attachment of a cholesteryl adduct to G198 of the Shh signaling domain (11,(15)(16)(17). This modification results in protein multimerization on the cell surface and subsequent release of the protein in multimeric form (18). The second lipid adduct that modifies Hh proteins is palmitic acid, which attaches to the conserved Nterminal cysteine (C25 in murine Shh) via the primary amine exposed after signal peptide cleavage (19). This unusual N-acylation is catalyzed by the product of the skinny hedgehog (Ski) gene (also designated sightless, central missing, raspberry or Hh acyltransferase (Hhat) (20)(21)(22)(23)). Although this hydrophobic modification would be expected to further decrease solubility (and thus morphogen diffusion), palmitoylation is absolutely required for Hh biological activity: Unpalmitoylated multimeric proteins are 10x-30x less active than N-palmitoylated wild type forms (21,24). The questions of how Hh lipidation and activity are linked, and how dually lipid-modified Hhs are released from producing cells and spread through the extracellular compartment are currently under intense debate. So far, Hh release has been explained by lipid-dependent micelle formation, co-transport with lipoprotein particles, or ectodomain solubilization via proteolytic removal of both terminal, lipidated peptides (18,(25)(26)(27)(summarized in Fig. 1).
In the latter scenario, dually lipidated Shh multimers (from hereon called ShhNp, standing for Shh N-terminal processed signaling domains) are released from transfected cells via ADAM (A Disintegrin And Metalloprotease)-mediated ectodomain shedding (18,25). Shedding, however, does not only remove both lipidated peptide anchors, but also activates ShhNp ectodomains in the process. Initially, on the cell surface, ShhNp N-terminal peptides interact with adjacent ShhNp zinccoordination sites in the cluster, thereby blocking Patched (Ptc)-receptor binding to these sites ( Fig. 2A) (18,(28)(29)(30). This renders the surface-tethered molecule inactive. By removing N-terminal acylated peptides, ADAMs expose Ptc-binding zinc-coordination sites and thereby couple ShhNp release with its biological activation. The established role of N-palmitoylation for the biological activity of secreted Hhs (21) is therefore indirect: N-acylation merely serves to anchor inhibitory N-terminal peptides to the cell membrane as a prerequisite for their subsequent sheddase-mediated removal. Thus, any absence of N-palmitoylation (in the respective acyltransferase-deficient mutant, or by site directed mutagenesis of the acceptor cysteine residue into a serine (C25S mutagenesis in mouse ShhNp C25S )) restricts sheddase activity towards the C-terminus and results in the release of inactive protein clusters with their zinc-coordination sites (the Ptc binding sites) still blocked. The indirect role of N-acylation for ShhNp function is strongly supported by restored biological activity of otherwise inactive ShhNp C25S upon N-terminal deletion mutagenesis (artificial truncation of the unpalmitoylated protein to mimic processing) (18). This suggests that Npalmitate does not directly contribute to receptor binding on receiving cells, at least in vitro.
In this work, we further support these findings. First, we show that in addition to Nterminal deletion mutagenesis, insertional mutagenesis as well as forced proteolytic processing restore biological activities of otherwise inactive ShhNp C25S (31). We also demonstrate that the N-terminal Cardin-Weintraub (CW) motif (32,33) is processed during sheddase-mediated ShhNp release in vitro and in vivo. Because this motif contributes to ShhNp/HS interactions (30,34,35), CW-processing strongly reduces HS-binding of truncated proteins. On the basis of these findings, we suggest that sheddase-mediated ShhNp release is coupled to functional morphogen activation and impaired HS binding of solubilized clusters, resulting in their facilitated diffusion into the HS-rich extracellular compartment.

Materials and Methods:
Mice: Endogenous murine (m)ShhNp was isolated from forebrains of 8 week old C57/Bl6 mice (n=6). Brains were separately homogenized in 20mM Tris/HCl pH8.0, 137mM NaCl and 10% glycerol, and subjected to ultracentrifugation. mShhNp in supernatants was then analyzed by heparin and heparan sulfate affinity chromatography (below) in three independent experiments. Additionally, All 6 cleared homogenates were concentrated via heparin-agarose pulldown or 5E1-immunoprecipitation (also described below). Precipitates were washed and binding of Shh-specific antibodies 5E1 and αCW analyzed by immunoblotting. All experimental procedures were done in accordance with Society for Neuroscience and European Union guidelines.
Cloning and expression of recombinant proteins. Shh constructs were generated from murine cDNA (NM_009170) using primers carrying desired point mutations or deletions by PCR. PCR products (nucleotides 1-1314, corresponding to amino acids 1-438) were ligated into pDrive (Qiagen), sequenced, and subsequently released and religated into pcDNA3.1 (Invitrogen) for the expression of secreted, lipidated 19kDa ShhNp in Bosc23 cells or HEK293 cells. PCR products (nucleotides 1-594, corresponding to amino acid 1-198 of murine Shh) were also cloned into pcDNA3.1 for expression of unlipidated 19kDa ShhN in Bosc23 cells. Where indicated, a hemagglutinin (HA) tag was inserted between N-terminal amino acids G32 and K33 by sitedirected mutagenesis (Stratagene), resulting in internally HA-tagged ShhNp. ShhNp C25S (C25 is required for N-terminal palmitoylation during release (18)) was generated by site directed mutagenesis (Stratagene). N-terminally truncated forms and ShhNp 5xA or ShhNp ΔCW (both lacking HS-binding Cardin-Weintraub function) were likewise generated by site-directed mutagenesis and sequenced. Primer sequences can be provided upon request.
Constructs used in this study are presented in Supplement Fig. 1.
Cell culture and protein analysis. Human Bosc23 or HEK293 cells were cultured in DMEM (PAA) with 10% fetal calf serum (FCS) and 100µg/ml penicillin/streptomycin and were transfected using PolyFect (Qiagen). Cells were then kept for 36 hours before the medium was harvested. In most assays, mutated or wild type ShhNp were secreted into serumcontaining or serum-free media without any additional treatment. Proteins secreted into serum-containing media were enriched by heparin-sepharose (Sigma-Aldrich) pulldown over night. Where indicated, Methyl-βcyclodextrin (Sigma-Aldrich) was used at 300µg/ml and 600µg/ml in serum-free DMEM. Proteins secreted into serum-free media were Trichloroacetic acid (TCA)precipitated. All proteins were analyzed by 15% SDS-PAGE, followed by Western blotting using PVDF-membranes. Immunodetection was conducted using polyclonal αShhN (goat IgG; R&D Systems), monoclonal 5E1 (DSHB, University of Iowa, USA) (36) or polyclonal αCW-antibodies directed against the HS-binding Cardin-Weintraub sequence (rabbit IgG, Cell Signaling) for primary protein detection. Visualization was performed after incubation with peroxidase-conjugated donkey-α-goat, α-mouse or α-rabbit IgG (Dianova) followed by chemiluminescent detection (Pierce). For radiolabeling, 1mCi [9,19(n)-3 H] palmitic acid was added to Shh-transfected HEK293 cells cultured in 35mm dishes for 28 hours under serum-free conditions prior to ShhNp release over night. 3H palmitic acid-labeled cells as well as the harvested media were subjected to heparin sepharose pulldown to improve the signal to noise ratio. After SDS-PAGE, gels were immunoblotted and the same blot analyzed by autoradiography.
Chromatography. Proteins were purified by FPLC (Äkta Protein Purifier (Pharmacia)) at 4°C. Gel filtration analysis was performed using a Superdex200 10/300 GL column (Pharmacia) equilibrated with PBS at 4°C. Eluted fractions were TCAprecipitated, resolved by 15% SDS-PAGE and immunoblotted. Signals were quantified using ImageJ. To determine HS-binding of endogenous and recombinant forms of Shh, mouse brain lysates or supernatants of Shhtransfected Bosc23 cells were subjected to HS affinity chromatography (Äkta Protein Purifier). HS columns were generated as follows: Embryonic day (E)18 mouse embryos were digested over night with 2mg/ml pronase in 320mM NaCl, 100mM sodium acetate (pH 5.5) at 40°C, diluted 1:3 in water and applied to 2.5ml DEAE sephacel columns. GAGs were applied to PD-10 (Sephadex G25) columns (Pharmacia). GAGs were lyophilized, again purified on DEAE as described above, applied to PD-10 columns and again lyophilized. Via the peptides attached to the HS chains, samples were coupled to an NHS-activated sepharose column according to the manufacturers protocol (Pharmacia). Proteins were applied to the column in the absence of salt and bound material was eluted with a linear NaCl gradient from 0-1M in 0.1M sodium acetate buffer (pH 6.0). Fractions were TCAprecipitated, resolved by 15% SDS-PAGE and immunoblotted. The same E18 HScolumn was used for all experiments shown in this work.
Shh reporter assays. C3H10T1/2 cells (37) were grown in DMEM supplemented with 10% FCS and antibiotics. 24h after seeding, Shh-conditioned media were mixed 1:1 with DMEM containing 10%FCS and antibiotics, and applied to C3H10T1/2 cells in 15-mm plates. To some assays, 2.5µg/ml 5E1 function blocking antibodies were added (36). Generally, due to variable expression levels, mutant and wild type proteins required adjustment to comparable levels before induction of C3H10T1/2 differentiation. For adjustment, proteins were detected by immunoblotting, and detected bands were quantified by ImageJ. Cells were lysed 5-6 days after induction (20mM Hepes, 150mM NaCl, 0.5% TritonX-100, pH 7.4) and osteoblast-specific AP activity was measured at 405nm after addition of 120mM pnitrophenolphosphate (Sigma) in 0.1M glycine buffer, pH 9.5. Assays were performed in triplicate.
Chondrocytes were isolated from the cranial third of 17-day old chick embryo sterna by over night digestion with collagenase and cultured in agarose suspension cultures under serum-free conditions. Cells were suspended in 0.5% low melting agarose in DMEM and allowed to sediment on the culture dishes which were pre-coated with 1% high melting agarose in water. Cells were grown at densities of 2x10 6 cells/ml in DMEM mixed 1:1 with conditioned, serum-free medium from Shh expressing cells. Media also contained 60µg/ml β-aminopropionitrile fumarate, 25µg/ml sodium ascorbate, 1mM cysteine, 1mM pyruvate, 100units/ml penicillin, and 100µg/ml streptomycin. After 14 days in culture, newly synthesized chondrocyte proteins were metabolically labeled with 1µCi/ml of 14C proline (250Ci/mmol, NEN Life Science products) for 24h. Collagens were isolated after limited digestion with pepsin and were analyzed by SDS-PAGE followed by fluorography (38). 5E1 immunoprecipitation. 10µg monoclonal 5E1 antibodies were coupled to 2.5mg ProteinA-sepharose beads (Sigma) per IP. 500µl-1ml lysate or medium were incubated overnight on a rotator and analyzed by SDS-PAGE and immunoblotting. For pulldown controls, 40µl heparin-sepharose beads were added to 500µl-1ml Shhcontaining media and incubated overnight on a rotator. Experiments were performed in triplicate.
Immunohistochemical detection. sAP-ShhN and mutant forms generated by site-directed mutagenesis were cloned into gWIZ-SEAP (Gene Therapy Systems, San Diego), expressed in Bosc23 cells and secreted into the medium. Conditioned media were normalized for sAP activity and applied to 4% paraformaldehyde (PFA)-fixed frozen embryonic day (E)12 mouse embryo sections over night, followed by three washes with PBS. sAP-ShhN bound to HS was directly visualized by addition of NBT-BCIP (Roche). Control slides were heparinase I-III digested (IBEX, Montreal, Canada) (50mM Hepes, 100mM NaCl, 1mM CaCl 2 , 5µg BSA/ml, pH 7.0) at room temperature over night to confirm Shh/HS specificity. sAP-ShhN was also incubated in the presence of 1M NaCl to prove Shh/HS specificity.
Statistical analysis. All statistical analysis was performed in Prism using Student's t-test (Two-tailed, unpaired, CI 95%). All error estimates are standard deviations of the mean.

Isolation of N-terminally truncated ShhNp from mouse brain
In our model of Shh release, sheddase-mediated processing of inhibitory N-terminal peptides is directly linked to biological activation of solubilized protein clusters.
Unprocessed N-palmitoylated peptides ( Fig. 2A, black ribbon) otherwise interact with zinc-coordination sites of adjacent molecules in the cluster (yellow molecule, zinc is shown as a black sphere). Because Shh zinc-coordination sites represent binding sites for the Shh receptor Ptc (colored in orange) (28,29), unprocessed N-termini prevent receptor-binding and thereby render the protein inactive. In this scenario, palmitate facilitates shedding via membraneproximal positioning of N-terminal inhibitory peptides as a prerequisite for processing, and also prevents solubilization of any incompletely activated (partially processed) morphogen clusters via their firm attachment to the cell membrane.
To confirm these in vitro derived data, we aimed to detect ShhNp processing in vivo. To this end, brain samples from 8-week old C57/Bl6 mice were lysed in detergentfree buffer and centrifuged to deplete membrane-tethered murine (m)ShhNp. We then assessed binding of the soluble protein to physiologically relevant HS (Fig. 2B). To this end, HS derived from embryonic day (E)18 C57/Bl6 mouse embryos was coupled to a HiTrap column and protein binding analyzed by FPLC affinity chromatography. The soluble fraction was also subjected to heparin affinity chromatography. Recombinant ShhN and mutant ShhN ΔCW lacking the HS-binding CW-motif served as positive and negative controls, respectively. Consistent with previous observations, control ShhN eluted from HS at 0.6M NaCl and from heparin at 1M NaCl (30). In contrast, brain-derived 5E1-reactive mShhNp (40) did not bind to HS but eluted from heparin at 0-0.75M NaCl. In this regard, endogenous mShhNp resembled ShhN ΔCW which also binds heparin but not HS (30,35). This suggested that endogenous, mouse brain-derived mShhNp lacks functional CWresidues required for HS-binding of the solubilized morphogen.
This finding prompted us to directly confirm loss of N-terminal CW-residues during release. Murine brain samples were homogenized in detergent-free buffer and subjected to ultracentrifugation. Soluble mShhNp in supernatants was then incubated with agarose-linked heparin in the presence of 0.5M NaCl to reduce unspecific interactions, or immunoprecipitated using ProteinA sepharose-coupled 5E1 antibodies (Supplement Fig. 2A). Like Ptc, monoclonal 5E1 binds the Shh zinc-coordination site ( Fig.  2A, 5E1 epitope in red), residue histidine 180 being bound by both, the receptor and the antibody ( Fig. 2A, pink) (31). 5E1 binding thus competes with Ptc binding (thereby inhibiting Shh signaling) (36), and ShhNp binding of the antibody and the receptor is blocked by unprocessed N-terminal peptides (18). As a consequence, 5E1 binds Nterminally truncated ShhNp but not the unprocessed precursor. For subsequent Shh detection, we employed SDS-PAGE and PVDF-immunoblotting using αCWantibodies raised against the N-terminal peptide immunogen K 33 RRHPKK 39 (the CWmotif). The same blot was then stripped and incubated with 5E1 antibodies. Both antibody specificities were confirmed using recombinant Shh controls (Supplement Fig.  2B, C). Consistent with impaired HS-binding shown in Fig. 2B, 5E1-reactive endogenous proteins were clearly detected in the immunoprecipitated material, but not after heparin-pulldown (Fig. 2C). Notably, 5E1immunoreactive mShhNp on the same (stripped) blot lacked αCW-antibody reactivity, in contrast to recombinant control proteins. Thus, results shown in Figs. 2B and 2C suggest N-terminal processing of solubilized mShhNp in vivo.
As an approximation for the Nterminal mShhNp processing site, we next determined minimum requirements for αCWantibody binding. To this end, we employed artificially truncated forms of recombinant ShhN that lacked variable numbers of Nterminal amino acids, and determined that amino acids R 35 HPKK 39 were required and sufficient for αCW-antibody binding to soluble proteins (Fig. 2D, residues are colored blue in A, and Supplement Fig. 2C). The observed lack of mShhNp αCWantibody reactivity thus suggests processing of endogenous proteins at arginine (R)35 or downstream residues.
Notably, proteolytic CW-processing is consistent with two published findings: 1) Site-directed exchange of all five basic CWresidues for alanines (ShhNp 5xA ) increases morphogen processing (30)

N-terminal ShhNp C25S truncation restores Ptc-binding
Thus, to rule out false positive results due to the possible release of monomeric truncated ShhNp C25S variants, we confirmed their unimpaired multimerization by gel filtration (Fig.  4A). Unimpaired multimerization of all forms allowed us to subsequently conduct biological tests as a read-out for their Ptc-binding capacities. To this end, we took advantage of Shhdependent C3H10T1/2 osteoblast precursor cell differentiation as a sensitive bioassay (37). As shown in Fig. 4B, wild-type control ShhNp strongly induced C3H10T1/2 differentiation into alkaline phosphatase (AP)-producing osteoblasts. In contrast, negative control media obtained from mocktransfected Bosc23 cells as well as full-length ShhNp C25S were inactive, as expected. In agreement with restored 5E1 binding shown before, we observed variable C3H10T1/2 differentiation induced by truncated ShhNp C25S variants: Here, ShhNp C25S;Δ26-33 and ShhNp C25S;Δ26-34 showed strongest relative activities (control ShhNp activity was set to 100%, Fig. 4B). ShhNp C25S;Δ26-34 also induced Gli1-dependent Firefly luciferase expression in Light-2 cells (41) and led to Shhdependent differentiation of primary chondrocytes (38) (Supplement Fig. 3). However, we noticed that induced C3H10T1/2 and Light 2 activities varied considerably, possibly due to the presence of serum factors or unspecific effects of Nterminal serine 25 replacing the palmitate acceptor cysteine. Yet, variable but restored biological activities observed in all three assays indicate preferred processing at, or close to, CW-residues R34/R35 during ShhNp release in vitro.
MβCD-treatment of ShhNp C25Stransfected HEK293 cells also resulted in increased morphogen solubilization (via stimulated processing of membrane-linked Ctermini), as expected. Notably, this treatment also resulted in N-terminal protein processing (Fig. 5A, double asterisks), in contrast to normal culture conditions that always produced soluble intact ShhNp C25S (Fig. 5A,  left). Based on our model of coupled Nterminal processing and activation (Figs. 3C, 4B), we expected increased bioactivity of MβCD-released ShhNp C25S over the unprocessed form (18,21,24,27,(48)(49)(50). As shown in Fig. 5B MβCD-treatment of mocktransfected HEK293 cells did not induce C3H10T1/2 differentiation. We draw two conclusions from this experiment: First, restored ShhNp C25S bioactivity demonstrates that N-palmitate is not directly required for ShhNp signaling on receiving cells, consistent with previous in vitro findings (Fig. 4B). Second, restored bioactivity of Nprocessed ShhNp C25S is consistent with restored 5E1-binding and signaling pathway activation shown in Figs. 3C, 4B and Supplement Fig. 3, confirming coupled ShhNp N-terminal truncation and functional activation.
Notably, HA-tagged ShhNp C25S induced prehypertrophic chondrocyte differentiation into Collagen X-producing hypertrophic chondrocytes, and HA-tagged ShhNp C25S;5xA was also active. These observations confirm Shh activity regulation by N-terminal peptides, and that N-palmitate is not directly required for Shh biofunction on receiving cells. We hypothesized that the inserted 9 amino-acid HA-tag disturbed peptide/protein contacts, resulting in facilitated Ptc-binding of unprocessed clusters. To confirm this possibility, heparin-sepharose pulldown and 5E1-immunoprecipitation of ShhNp C25S and HA-tagged ShhNp C25S were conducted as described above (Fig. 3C), and PVDF-blotted proteins analyzed by αShhN and αHAdirected antibodies. ShhNp C25S;Δ26-38 served as a positive (5E1-binding) control (Fig. 3C). As shown in Fig. 6C (top), HA-tagged and untagged ShhNp C25S bound to heparin. ProteinA-coupled 5E1, however, immunoprecipitated only HA-tagged ShhNp C25S , but not the untagged form. Based on these findings, we conclude that insertional mutagenesis rescues ShhNp C25S biofunction by rearranging inhibitory peptides away from the Ptc-binding zinccoordination site.

Site-directed CW-mutagenesis facilitates N-terminal ShhNp processing
Next, we investigated the possibility that ShhNp activation in vitro depends on proteolytic processing of specific, "permissive" N-terminal target sites. Based on the observed loss of αCW-antibody reactivity during processing, the HS-binding CW-motif may represent one such target. To test this idea, we compared ShhNp and ShhNp 5xA processing and release, the latter construct lacking all five basic CW-residues. This abolished ShhNp 5xA binding to HiTrap column-coupled E18 embryo-derived HS, as assessed by FPLC (Fig. 7A) (30,33). We next aimed to directly confirm ShhNp depalmitoylation during processing and release. To this end, ShhNp and ShhNp 5xA transfected HEK293 cells were grown in the presence of [9,10(n)-3 H] palmitic acid for specific N-terminal radiolabeling. Soluble and cell-bound proteins were harvested and PVDF-immunoblotted; the same blot was then subjected to autoradiography to detect C25-linked 3H palmitic acid. As shown in Fig.  7B, cell-tethered 19kDa proteins (top bands, single asterisk) were labeled and largely unprocessed, consistent with their 3H palmitic acid signals. These bands represented the dually lipidated morphogens. In contrast, increased electrophoretic mobility ShhNp 5xA protein signals lacked 3H palmitic acid (Fig.  7B bottom, double asterisks), thus representing the N-terminally processed proteins (Fig. 3C). Notably, in this assay, ShhNp processing was strongly reduced if compared to ShhNp 5xA . Because CW-residues associate with cell-surface HS prior to release (52), this observation suggests that HSbinding may (down)regulate sheddasemediated processing and release, possibly via blockade of susceptible CW-residues.

HS binding of CW-processed Shh is reduced
Because HS/Shh-interactions are CW-dependent, as shown in this work and elsewhere (30,33,35), and because HSPGs are critically involved in Shh biology (53-58), we aimed to confirm that CW-processing at or beyond residue 35 reduces protein binding to HS (Fig. 2B). This would provide an explanation for the ability of solubilized ShhNp to travel long distance in the HScontaining extracellular compartment (56,(59)(60)(61)(62), effectively escaping HS-mediated protein immobilization and internalization.
To test this possibility, we first compared binding of processed and unprocessed proteins to FPLC-columncoupled E18-derived HS (Fig. 2B) (30). We first compared ionic interactions of ShhNp C25S (the N-terminally unprocessed multimer) with the unprocessed, alkaline phosphatase (sAP) tagged monomeric protein (sAP-ShhN). Both proteins eluted at 0.6M NaCl, demonstrating comparable HS-binding of monomeric and multimeric unprocessed proteins (Fig. 8A). Control sAP did not bind HS, demonstrating specificity of the assay. Next, comparable AP-activities of sAP, sAP-ShhN and CW-truncated sAP-ShhN Δ25-35 were applied to the same E18 HS-column, and sAP-activities in the flowtrough were monitored to confirm saturation of HS binding sites. Here, in notable contrast to strong sAP-ShhN binding to HS (30), sAP-ShhN Δ25-35 did not bind to E18 HS. This confirmed CW-dependent Shh/HS binding in vitro, and that CW-truncation impairs Shh association with physiologically relevant forms of HS, such as embryonic HS.
An embryonic tissue known to be patterned by Shh is the developing neural tube. To directly test this tissue, E12 embryo sections were incubated with equal amounts of the same sAP, sAP-ShhN and sAP-ShhN Δ25-35 proteins used above. As a readout for Shh binding to tissue HS, sAP activity was detected colorimetrically (Fig. 8B) (34). sAP-ShhN yielded a strong signal, and negative controls sAP and sAP-ShhNp 5xA , HS digestion with Heparinase I-III (H-ase) prior to sAP-ShhN incubation or sAP-ShhN incubation in the presence of 1M NaCl demonstrated CW-dependent, specific HS binding of the protein. As expected, sAP-ShhN Δ25-35 failed to bind HS on tissue sections. Based on this and previous results, we conclude that HS/ShhN interactions depend on the CW-motif (30,34,35), and that CW-processing during release reduces HS-binding of solubilized proteins. Conversion of HS-associated cell-surface clusters into soluble unreactive forms may thus underlie unimpaired extracellular ShhNp diffusion.

Discussion:
How morphogen gradients arise has attracted much controversy. So far, micelle formation and lipoprotein-mediated transport, morphogen transport by transcytosis (the sequential endocytosis and exocytosis of bound ligands (63)) and transport via "bucket brigade" (receptor-bound morphogen on one cell moves by being handed over to receptors on adjacent cells (64)) have been suggested. One major argument against extracellular Hh morphogen diffusion, which would represent another possible mechanism, was its apparent incompatibility with dual ShhNp lipidation and hydrophobicity. However, as confirmed in this work, ShhNp shedding represents a mechanism to overcome this problem. Another major argument against morphogen diffusion was that cell surface receptors or non-receptors, such as HSPGs, would strongly impair extracellular movement of "sticky" molecules. Indeed, extracellular glypicans represent a major group of Hhbinding HSPGs that modulate morphogen activity, mobility and stability (53,57,58,(65)(66)(67). Here, negatively charged HS chains represent the main direct interactors between HSPGs and highly conserved CW-motifs present in all Hhs (30,33,35). The resulting Hh/HS interactions are essential for Hh multimerization on producing cells (52), however, similar interactions in the HSPGrich extracellular matrix would be expected to prevent long-range diffusion of soluble proteins (64,68).
In this work, we suggest that CWprocessing during release resolves this problem via conversion of HS-binding surface proteins into unreactive soluble clusters. This mechanism would allow for unimpaired HSPG/ShhNp interactions on the cell surface, yet facilitate protein diffusion upon removal of essential CW-amino acids. Therefore, we postulate that cell-surface shedding serves three functions in Hh biology: it is required for the solubilization of cell surface-linked morphogen clusters, their biological activation during release, and their associated reduction of HS-binding. We are aware of the possibility that N-terminal peptides may get cleaved at different CWamino acids during release, possibly resulting in variable HS-binding of alternatively truncated soluble proteins. In addition to variable HS-binding, N-terminal Shh processing at different sites may also predetermine biological responses of receiving cells (Fig. 4 and Supplement Fig. 3) at the molecular level. This is indicated by alternative transcriptional activities of responding C3H10T1/2 cells after stimulation with differently truncated ShhNp, as shown in Supplement Fig. 4. Thus, alternative Nterminal processing may represent a new regulator for the various biological responses that Shh can elicit. Interestingly, and consistent with this idea, N-terminal proteolysis also specifically regulates biological activities of secreted Wnt morphogens. Wnt family members are essential for embryogenesis, pathogenesis and regeneration, and like Hhs, palmitoylated Wnt proteins initially locate to the cell surface and require solubilization to reach their targets. In this process, Wnts can undergo proteolysis-induced processing of eight N-terminal amino acids, resulting in Wnt oxydation-oligomerization and functional inactivation (69). Therefore, Wnts and Shh represent two examples for the emerging theme of functional morphogen control by proteolytic removal of regulatory N-terminal peptides.
In contrast to proteolysis-induced Wnt inactivation, however, removal of ShhNp N-termini activates the oligomerized protein (40). This is supported by restored ShhNp C25S biofunction and 5E1-binding after deletion mutagenesis, as shown in this work. Furthermore, enhanced N-terminal processing in the presence of MβCD is associated with increased electrophoretic mobility, associated loss of αCW-reactivity and biological activation of truncated ShhNp and ShhNp C25S . Notably, these observations may be refelected by recently demonstrated bioactivity of HEK293-expressed, 5E1-immunoprecipitated lower molecular weight ShhNp, and the lack of such an activity in the remaining higher molecular weight protein (Fig. 6 in (70)). Because we and others showed that higher molecular weight ShhNp is palmitoylated, in contrast to increased electrophoretic mobility ShhNp (the bottom band) which is not (Fig.  7B) (70), we conclude that N-acylation, although essential for ShhNp biological activity per se, is not directly involved in ShhNp signaling on receiving cells. This confutes the third argument against sheddasedependent morphogen diffusion: the assumed direct role of N-palmitate for Hh bioactivity.
Instead, we suggest that N-palmitate ensures quantitative N-terminal ShhNp processing as a prerequisite for complete activation of all solubilized molecules, as incompletely processed clusters will remain tethered to the cell surface. As a consequence, protein concentrations at any position in the responsive field correlate with their biological activities (their Ptc-binding capacities). In contrast, in the absence of quantitative N-terminal processing, biological activities of solubilized clusters would vary, depending on the relative number of processed N-terminal peptides. As a result, protein concentrations at any position in the field would not reliably correlate with Shh signaling activities. This, in turn, would affect morphogen-dependent tissue patterning in unpredictable ways. In the most extreme case, e.g. in acyltransferase-deficient mutants, processing would be restricted to C-terminal peptides in the cluster, resulting in the solubilization of inactive proteins and complete decoupling of protein/activity gradients. This essential "cleavage/activation control" function may explain the unusual attachment of palmitate via an amide bond to the α-amino group of all Hhs, whereas palmitate in most other proteins is thioester (S)-linked (71). S-linked palmitate, however, is susceptible to enzymatic deacylation by palmitoyl-protein thioesterases which would result in a situation comparable to that in acyltransferase-deficient mutants. In contrast, amide-linked palmitate is resistant to thioesterase activity and thus restricts possible modes of Hh release to shedding, ensuring the release of fully activated proteins.
As discussed above, Npalmitoylation may simultanously ensure quantitative CW-processing and inactivation required for unimpaired ShhNp diffusion through the matrix. Endogenous expression of 5E1-reactive but αCW-antibody and HSunreactive mShhNp in the in vivo mouse brain strongly supports this idea.
Hybridoma Bank maintained by the University of Iowa, Iowa City, USA. The excellent technical and organizational assistance of Sabine Kupich is gratefully acknowledged   (18,39). The N-terminal peptide of one molecule (black ribbon) overlaps 5E1-and Ptc-binding sites of the adjacent molecule in the cluster (in yellow). 5E1-binding depends on Shh residues T125, D147, R153, S177, A179 and H180 (shown in red) (31). Ptc-binding residues are shown in orange, residue H180 bound by both, 5E1 and Ptc, in pink. The αCW-antibody-reactive CW-peptide R 35 HPKK 39 is colored blue. Polyclonal αShhN binds to multiple, undefined epitopes. N: N-termini (C25). C: C-terminus B) Ionic heparin-and E18-HS interactions of mouse brain-derived mShhNp and recombinant wildtype (ShhN) and CW-deficient (ShhN ΔCW ) proteins. For HS-affinity chromatography, E18 embryonic HS was coupled to a FPLC column, proteins applied, and bound proteins eluted with a linear (0-1.5M) NaCl gradient. All proteins bound to heparin, in contrast, only ShhN bound to HS. ShhN ΔCW and mShhNp were both detected in the flowthrough (ft), suggesting that the endogenous protein lacked functional HS-binding CW-residues. C) mShhNp and recombinant control proteins were concentrated by heparin-agarose pulldown in the presence of 0.5M NaCl to suppress weak ionic interactions, or 5E1-immunoprecipitation. As shown in B, heparin bound the recombinant protein but not endogenous mShhNp, again indicating loss of basic amino acids. Consistent with this, immunoprecipitated mShhNp was 5E1 reactive but lacked αCW-antibody reactivity in all assays (n=6). One representative experiment is shown. rec c: Recombinant control Shh confirmed specificity of both antibodies on the same (stripped) blot. D) Mapping of the αCW binding site. Truncated proteins (N-terminal amino acids are indicated, C25 and S25 represent full-length forms) were expressed and concentrated by heparin-pulldown, followed by immunoblotting. All truncated proteins were expressed at high levels, as indicated by polyclonal αShhN reactivity. However, αCW-specific antibodies failed to bind all proteins truncated beyond residue R35 on the same (stripped) blot, confirming that endogenous mShhNp was likely processed at or C-terminal of this residue.

Fig. 3: ShhNp C25S N-terminal truncation restores 5E1 binding. A)
Intermolecular interactions observed in the human Shh crystal structure. The 5E1-binding site is shown in red, the Ptc-binding site in orange, residue H180 bound by 5E1 and Ptc in pink. N-terminal amino acids blocking these sites are labelled. B) All mutant proteins lack N-terminal C25, preventing palmitoylation, and also variable numbers of N-terminal amino acids. Red: Ski/Hhat acyltransferase recognition motif. CW: Cardin-Weintraub motif (blue). C) ShhNp C25S and N-terminally truncated forms were expressed in Bosc23 cells and soluble proteins either 5E1-immunoprecipitated or concentrated by heparinsepharose pulldown. Proteins were immunoblotted and detected by αShhN antibodies. Consistent with a previous report, 5E1 failed to immunoprecipitate ShhNp C25S (18). Consecutive N-terminal truncations, however, resulted in restored 5E1 binding, especially of ShhNp C25S;Δ26-37 , ShhNp C25S;Δ26-38 , ShhNp C25S;Δ26-33 and ShhNp C25S;Δ26-34 . Proteolytically processed ShhNp 5xA served as a positive control (30). Right: Corresponding monomeric proteins (ShhN) were N-terminally truncated and analyzed as described above (B, right). Comparable 5E1-binding of truncated and untruncated monomers confirms Shh clustering as a prerequisite for protein inactivation in trans.

Fig. 4: Restored biological activity of N-terminally truncated, palmitoylation-deficient
ShhNp C25S variants. All mutant proteins lacked N-terminal C25, preventing palmitoylation, in addition to consecutive N-terminal peptides (Fig. 3B). Comparable amounts of mutated proteins were employed in all following assays, as determined by immunoblotting. A) Comparable 600kDa and 100-300kDa multimerization was detected by gel filtration of ShhNp C25S and compound mutant proteins (ShhNp C25S;Δ26-31 to ShhNp C25S;Δ26-38 ). Elution profiles are expressed relative to the highest protein level in a given run, which was set to 100%. B) C3H10T1/2 osteoblast precursor cells were incubated with single and compound mutant proteins, and the relative amount of Shh-induced APproduction was determined as a biological readout. Medium obtained from mock-transfected Bosc23 cells and ShhNp C25S -conditioned medium were used as negative controls, and ShhNp conditioned medium as a positive control (set to 100%, other values were expressed relative to ShhNp activity). In this assay, ShhNp C25S;Δ26-33 and ShhNp C25S;Δ26-34 showed activities comparable to that of the wild-type (103%±20% and 99.5%±11% relative to the wild-type control, p=0.9 and 0.94, n=6). All truncated ShhNp C25S double mutants showed significantly increased biological activities if compared to non-truncated ShhNp C25S (p<0.05 in all forms). *** denote significant ShhNp activity if compared to ShhNp C25S (p<0.01). n.s.: not significant (p>0.05).  40 hours after HEK293 cells were transfected with full-length Shh cDNA or cDNA encoding for the palmitoylation-deficient protein, cells were washed and serum-free DMEM with or without 300µg/ml and 600µg/ml methyl-β-cyclodextrin (MβCD) was added. After 4 hours, the medium was harvested, subjected to centrifugation and TCA-precipitation, followed by SDS-PAGE and Western-blot analysis using αShhN-and αCW-antibodies. Cellular protein expression levels (cell lysates) are shown on top. Right: MβCD treatment resulted in forced shedding and increased release of N-terminally truncated ShhNp and ShhNp C25S into supernatants. Reduced αCW-reactivity of processed proteins indicates cleavage at or in close vicinity to CW-residue R35 (bottom blot). Results from the same (stripped) blot are shown. Left: Control expression of N-terminally truncated ShhNp (double asterisks) and unprocessed ShhN C25S (asterisk) in the presence of 10% FCS and without MβCD. B) C3H10T1/2 osteoblast precursor cells were incubated with aliquots of the same ShhNp and ShhNp C25S -conditioned media analyzed above, and relative amounts of Shh-induced AP-activity were determined. Medium from mock-transfected HEK293 cells was used as a control. *** denote significance between biological activities of ShhNp and unprocessed ShhNp C25S expressed without MβCD (p<0.01). In contrast, the observed difference between MβCD-released ShhNp and ShhNp C25S was insignificant (n.s., p>0.05).

Fig. 7: CW-mutagenesis abolishes HS-binding and increases ShhNp processing. A)
Deletion of all five basic CW-residues fully abolishes ShhNp 5xA binding to E18 embryo-derived HS. Averaged elution profiles of 3 independent ShhNp and ShhNp 5xA analyses on the same HS-coupled affinity column are shown. Error estimates are standard deviations of the mean. B) Immunoblot (top) and autoradiograph (bottom) of [9,10(n)-3 H] palmitic acid-labeled, full-length ShhNp and ShhNp 5xA in cell lysates (c) and media (m). Full-length 19kDa ShhNp (asterisk) was detected in lysates and media by Western blotting and 3 H-autoradiography. Unprocessed, radiolabeled ShhNp 5xA was also detected in the cell lysate. In contrast, truncated soluble ShhNp and ShhNp 5xA (double asterisks) derived from the labeled cell-bound forms lacked 3H palmitic acid, indicating N-terminal processing. Note that relative processing of ShhNp 5xA was increased if compared to the wild-type protein. The presence of unprocessed soluble ShhNp in media is due to extensive cell death caused by serumfree culture conditions required for labeling. C) C3H10T1/2 osteoblast precursor cell differentiation is significantly induced by processed ShhNp, but not by unprocessed ShhNp C25S (1.1±0.08 arbitrary units versus 0.3±0.01 arbitrary units, n=6, p=0.0023). In this assay, ShhNp 5xA was inactive despite its N-terminal processing (0.27±0.05 arbitrary units, p=0.57 compared to ShhNp C25S ). ** denote significance, n.s: not significant (p>0.05). D) Size comparison of processed ShhNp 5xA with ShhNp C25S and N-terminally truncated proteins. Comparable size of ShhNp C25S;Δ26-38 demonstrates cleavage at or close to position 37 (arrow). E) Analysis of ShhNp, as described in D. In contrast to ShhNp 5xA , ShhNp is cleaved at or in proximity to position 34/35 (arrow), as confirmed by reduced αCW-antibody-reactivity of processed proteins.

Fig. 8: CW-truncated ShhN does not bind to HS. A)
Ionic E18-HS interactions of mono-and multimeric proteins. HS-binding of unprocessed multimeric ShhNp C25S was comparable to that of alkaline-phosphatase (sAP)-tagged sAP-ShhN, the latter representing the directly secreted, unlipidated monomer. Equal amounts of sAP control protein and CW-truncated sAP-ShhN Δ26-35 , as determined by comparable AP-activity in both samples, were applied to the same E18-HS column, and sAP activities in the flowthrough (ft) confirmed saturation of Shh binding sites on the column. In contrast to sAP-ShhN, no sAP and sAP-ShhN Δ25-35 elution could be observed, demonstrating strongly impaired HS-binding of the N-terminally truncated protein. B) sAP-ShhN ligand binding to horizontal embryonic neural tube (dashed line) sections. PFA-fixed E12 mouse embryos were sectioned and treated with equal amounts of sAP and sAP-ShhN 5xA control proteins, sAP-ShhN Δ25-35 , wild-type sAP-ShhN, and sAP-ShhN in the presence of 1M NaCl or Heparinase (H-ase) I-III to remove cell surface HS. Bound sAP-ShhN was detected by its sAP-activity upon incubation with NBT-BCIP. In contrast, the truncated CW-mutant did not bind to HS.