Heparan Sulfate-modulated, Metalloprotease-mediated Sonic Hedgehog Release from Producing Cells*

The ectodomains of numerous proteins are released from cells by matrix metalloproteases to yield soluble intercellular regulators. A disintegrin and metalloprotease (ADAM) family members have often been found to be the responsible “sheddases,” ADAM17/tumor necrosis factor-α-converting enzyme being its best characterized member. In this work, we show that ShhNp (lipidated and membrane-tethered Sonic hedgehog) is released from Bosc23 cells by metalloprotease-mediated ectodomain shedding, resulting in a soluble and biologically active morphogen. ShhNp shedding is increased by ADAM17 coexpression and cholesterol depletion of cells with methyl-β-cyclodextrin and is reduced by metalloprotease inhibitors as well as ADAM17 RNA interference. We also show that the amount of shed ShhNp is modulated by extracellular heparan sulfate (HS) and that ShhNp shedding depends on specific HS sulfations. Based on those data, we suggest new roles for metalloproteases, including but not restricted to ADAM17, and for HS-proteoglycans in Hedgehog signaling.

The proteins of the Hedgehog (Hh) 3 family are powerful morphogens that control growth and patterning during development. Establishing the molecular mechanisms that generate the Hh gradient is essential for our understanding of how the Hh signal elicits multiple responses in a temporally and spatially specific manner. The Hh spreading mechanism is especially intriguing, because all Hhs are released from the producing cells despite being synthesized as dually lipidated molecules, whereas lipid-modified peptides normally appear firmly tethered to membranes. Both in vertebrates and in Drosophila melanogaster, the Hhs are synthesized as precursor proteins consisting of the N-terminal signaling domain and a C-terminal cholesterol transferase domain. The precursor first undergoes internal cleavage between residues Gly 198 and Cys 199 (of murine Sonic hedgehog (Shh)) linked to the addition of a cholesteryl moiety to Gly 198 of the N-terminal cleavage product (1). This reaction is catalyzed by the C-terminal cholesterol transferase domain through a nucleophilic substitution resembling intein-mediated protein splicing. Inteins, also called protein introns, are parts of protein sequences that are post-translationally excised, their flanking regions (exteins) being spliced together to yield an additional protein product in a self-catalyzed manner. The second lipid adduct that modifies the Hh proteins is palmitic acid, which attaches to the N-terminal cysteine residue exposed after signal peptide cleavage (2), resulting in the formation of the processed (HhNp) protein.
The molecular mechanisms through which the lipid-modified HhNp morphogen is able to diffuse long distances are currently under intense debate, and its ability to do so has been linked to oligomer formation or co-transport with lipoprotein particles (3,4). HhNp gradient formation also depends on the presence of heparan sulfate proteoglycans (HSPGs).
Heparan sulfate (HS) is produced by most cell types in vertebrates and invertebrates. HS biosynthesis (as well as synthesis of heparin, a highly sulfated form of HS) occurs in the Golgi compartment on proteoglycan core proteins (5). Enzymes called the exostosins (Exts) synthesize the (GlcA1,4GlcNAc1,4) n carbohydrate backbone, which is subsequently modified by N-deacetylase/sulfotransferases, O-sulfotransferases, and a GlcA-C5 epimerase. Many growth factors, chemokines, cytokines, and morphogens bind to HS, and the HSPGs are thought to act as co-receptors for these ligands. Expression of the Drosophila Ext family of proteins, encoded by the genes tout velu (ttv), brother of tout velu, and sister of tout velu, is required for the diffusion of lipid-modified fly HhNp (6 -8) yet does not affect the spread of cholesterol-free fly HhN (6,9,10). The glypican (a type of HSPG attached to the cell membrane by a glycosylphosphatidylinositol anchor) proteins Dally and Dallylike were also found to be required for fly HhNp signaling (11). These data indicate that HSPGs are important for Hh function via an unknown molecular mechanism.
In this work, we investigated the idea that proteolytic ShhNp "ectodomain shedding" may underlie ShhNp release. A disintegrin and metalloprotease (ADAM) family members have been found to be the responsible "sheddases" for the release of mem-brane-associated tumor necrosis factor-␣, tumor necrosis factor receptor, L-selectin, transforming growth factor-␣, intercellular adhesion molecule 1, collagen XVII, CD23, epidermal growth factor, heparin-binding epidermal growth factor-like growth factor, and Notch, among others (summarized in Refs. 12 and 13). In most cases, ectodomain shedding of membrane proteins changes their fate, location, and mode of action and thus their biological activity. However, a substrate will only be released if, at a given time point, it is co-localized with the ADAM sheddase. Release of a number of ADAM substrates like the amyloid precursor protein, CD30, interleukin-6 receptor, L-selectin, and collagen XVII was increased by cholesterol-lowering drugs, such as methyl-␤-cyclodextrin (M␤CD) (14 -18), which suggests that location within membrane microdomains rich in cholesterol, called lipid rafts, may play a regulatory role in the ectodomain shedding of ADAM substrates. Consistent with this, lipid rafts have been described as locations of ectodomain shedding mediated by ADAM17 and ADAM19 (12,19), and ShhNp as well as glypican HSPGs are known to localize to lipid rafts (20). Despite the fact that the nature of these microdomains, including their size, composition, organization, and even their very existence remains highly controversial, those reports still suggested a possible link between ShhNp release, HSPGs, and ADAM-regulated ectodomain shedding.

EXPERIMENTAL PROCEDURES
Cloning and Expression of Recombinant Shh-Shh constructs were generated from murine cDNA (NM 009170) by PCR. Secreted, lipidated ShhNp (nucleotides 1-1314, corresponding to amino acids 1-438) was generated in Bosc23 cells (a human 293T derivative, an embryonic kidney cell line routinely used for the expression of lipidated ShhNp (4)), and secreted, unlipidated ShhN (nucleotides 1-594, corresponding to amino acid 1-198 of murine Shh) was generated in mouse melanoma B16-F1 cells and in Escherichia coli to yield high amounts of protein required for some biochemical tests. PCR products were ligated into pGEM (Promega), sequenced, and subsequently released and religated into pcDNA3.1 (Invitrogen) for the expression of secreted, lipidated 19-kDa ShhNp in Bosc23 cells; into pcDNA3.1/myc-HisC (Invitrogen) for the expression of secreted, C-terminally hexahistidine-tagged 28-kDa ShhN His6 (the large size due to the presence of a c-Myc and intervening cloning sequence) in B16-F1 cells; into pGEX4T-1 (Amersham Biosciences) for expression of N-terminally glutathione S-transferase-tagged 50-kDa GST-ShhN in E. coli; and into pWIZ-SEAP (Gene Therapy Systems) for the expression of secreted, N-terminally alkaline phosphatase-tagged 80-kDa sAP-ShhN in B16-F1 cells. To generate the soluble alkaline phosphatase-Shh fusion proteins, the N-terminal sequence of Shh (amino acids  was produced by PCR (sense primer, 5Ј-agatatcaatgtgggcccggcagggggtttg-3Ј; antisense primer, 5Ј-atctagaagccgccggatttggccgcc-3Ј) and ligated into pWIZ after limited HpaI and subsequent XbaI restriction of the vector and EcoRV and XbaI restriction of the PCR product, resulting in replacement of the alkaline phosphatase (AP) stop codon with the codon for ShhN C25. The biological activity of all recombinant proteins was confirmed by Shh-dependent alkaline phosphatase induction in C3H10T1/2 cells using the method described below. Human ADAM17 cDNA was obtained from the German Resource Center for Genome Research (RZPD) and cloned into pcDNA3.1/myc-HisC. ADAM17 was expressed in a hexahistidine-tagged form in B16-F1 for heparin affinity chromatography and in a nontagged, full-length form in Bosc23 cells for gain-of-function studies. To generate catalytically inactive ADAM17, the active site residue Glu 406 was changed into an alanine (21).
Cell Culture-CHO-K1 cells were cultured in DMEM/F-12 (Invitrogen), and Bosc23 and B16-F1 cells were cultured in DMEM (Invitrogen), all supplemented with 10% fetal calf serum and 100 g/ml penicillin/streptomycin. Cells were transfected with plasmids encoding secreted ShhN, ShhNp, or sAP-ShhN using PolyFect (Qiagen). Cells were grown for 36 h, washed with phosphate-buffered saline, and incubated in DMEM or serum-free DMEM for various time periods in the presence or absence of stimulators or inhibitors of shedding, followed by ultracentrifugation at 210,000 ϫ g for 60 min to remove cell debris, including membrane-tethered ShhNp. The supernatant was then trichloroacetic acid-precipitated. Where indicated, OPTI-MEM was used for serum-free protein expression. Chondrocytes were isolated from the cranial third of 17-day-old chick embryo sterna and cultured in agarose suspension cultures under serum-free conditions in the presence of 100 ng/ml insulin-like growth factor I. N-Isobutyl-N-(4methoxyphenylsulfonyl)-glycyl hydroxamic acid (NNGH) (Biomol) was used at 15 M concentration; tissue inhibitor of metalloproteinase-3 (TIMP-3; Ͼ95% pure) (R&D Systems), methyl-␤-cyclodextrin, and chondroitin sulfate B (Sigma) were used at 100 nM, 300 g/ml, and 100 g/ml respectively. 100 g/ml dextran sulfate sodium salt (Roth), chondroitin sulfate sodium salt (Sigma), and heparin sodium salt (Serva) were also used. Phorbol 12-myristate 13-acetate (PMA) was used at 200 ng/ml. Where appropriate, vehicle was used as a control (DMSO or ethanol).
Protein Purification and Analysis-Proteins were resolved by 15% reducing SDS-PAGE and immunoblotted. Monoclonal anti-tetrahistidine antibody (Qiagen) and anti-ShhN (polyclonal goat IgG; R&D Systems) were used for primary detection. Immunodetection of ADAM17 was conducted using monoclonal anti-human tumor necrosis factor-␣-converting enzyme/ADAM17 ectodomain antibody (R&D Systems). Visualization was performed after incubation with peroxidase-conjugated donkey anti-goat IgG (detecting anti-ShhN) and goat anti-mouse IgG (detecting anti-His 4 and anti-ADAM17) followed by chemiluminescent detection (Pierce). In some cases, results from multiple experiments were pooled and signal quantification was conducted using ImageJ (available on the World Wide Web). Data were plotted as released versus cell bound ShhNp relative to the highest value obtained, which was set to 100%. Radiolabeling was conducted as described in Ref [ 3 H]Palmitic acid-labeled cells and media were additionally subjected to heparin-Sepharose pull-down to improve the signal/noise ratio. After SDS-PAGE, gels were dried and autoradiographed or immunoblotted.
To determine heparin binding properties of ShhN and ADAM17, the supernatant of ShhN-and ADAM17-transfected B16/F1 cells was subjected to heparin affinity chromatography (Ä kta protein purifier) using heparin-Sepharose columns (Amersham Biosciences) at 4°C. Proteins were applied to the columns in the absence of salt, and bound material was eluted with a linear NaCl gradient from 0 to 1 M in 0.1 M sodium acetate buffer (pH 6.0). Fractions were precipitated, and eluted proteins were detected immunohistochemically.
ADAM-RNA Interference (RNAi) Using shRNAs-RNAi was used to specifically knock down ADAM17 (AAI46659) expression in Bosc23 cells. Small interfering RNAs (siRNAs) were expressed from short hairpin RNAs (shRNAs) using the MISSION TM shRNA (Sigma) plasmids TRCN0000052168 to TRCN0000052172. Cells were co-transfected with plasmid encoding ShhNp cDNA and shRNA plasmids or empty control plasmid, and transfected cells were enriched by the addition of 5 g/ml puromycin for 30 h before starting the assay. In addition, to confirm findings obtained by shRNAi, ADAM17 was knocked down in ShhNp-expressing Bosc23 cells by using 5 nmol of On-Target plus anti-human ADAM17 pooled siRNA (Dharmacon) using the siGENOME nontargeting siRNA pool as a control. In all experiments, the amount of ShhNp released into the medium was quantified using ImageJ. Data were plotted as released versus cell-bound ShhNp relative to the value obtained with the empty vector control, which was set to 100%. Statistical analysis was performed in Excel using Student's t test (two-tailed, unpaired). All values shown throughout are ϮS.D.
To assess the effectiveness of ADAM knockdown, semiquantitative RT-PCR analysis of ADAM17 and glypican 1 (Glp1) expression in Bosc23 cells transfected with pooled siRNAs specifically interfering with ADAM17 mRNA was conducted. PCR products were quantified using ImageJ, and their relative intensities were calculated. Semiquantitative RT-PCR revealed a 60% reduction in ADAM17 mRNA levels in Bosc23 cells cotransfected with five pooled shRNAs specific for the protease. As a control, glypican 1 mRNA levels were not affected by ADAM17-specific mRNA knockdown.
Analysis of mRNA Expression-RT-PCR analysis of ADAM expression in Bosc23 cells and chondrocytes was conducted following RNA isolation using TRIzol reagent (Invitrogen). cDNA was generated using the First-Strand cDNA synthesis kit (Fermentas) according to the manufacturer's instructions and subsequently used as a PCR template.
Preparation and Analysis of Tissue HS-Tissues or cultured cells were digested overnight with 2 mg/ml Pronase in 320 mM NaCl, 100 mM sodium acetate (pH 5.5) at 40°C, diluted 1:3 in water, and applied to 2.5-ml DEAE-Sephacel columns. For disaccharide analysis, proteins attached to the glycosaminoglycans were ␤-eliminated at 4°C (0.5 M NaOH, 1 MNaBH 4 ) overnight, neutralized with acetic acid, and applied to a PD-10 (Sephadex G25) column (Amersham Biosciences). Glycosaminoglycans were lyophilized, purified on DEAE, as described above, applied to a PD-10 column, and lyophilized. The samples were then digested using heparin lyases I, II, and III (IBEX), and the resulting disaccharides were separated from undigested chondroitin sulfate using a 3-kDa spin column (Centricon), followed by high pressure liquid chromatography analysis using Carbopac PA1 columns (Dionex), using disaccharide mixtures as standards.
Differentiation of C3H10T1/2 Osteoblast Precursor Cells-40 h post-transfection, ShhNp-and mock-transfected BOSC23 cells were stimulated for 30 min with 300 g/ml methyl-␤cyclodextrin in serum-free DMEM. After M␤CD stimulation, the supernatant was discarded, and fresh DMEM without M␤CD was added for 2 h and subsequently used for C3H10T1/2 cell differentiation to avoid nonspecific effects in the assays. Trichloroacetic acid-precipitated proteins were first immunoblotted to check for protein release before being used in the subsequent assays. Conditioned media were sterile filtered, mixed with DMEM containing 2ϫ fetal calf serum and antibiotics, and applied to C3H10T1/2 cells in 15-mm plates. To some samples, 2.5 M cyclopamine, a specific inhibitor of Shh signaling, and ShhN-neutralizing antibody 5E1 (22) were added. Cells were lysed 5 days after induction (20 mM Hepes, 150 mM NaCl, 0.5% Triton X-100, pH 7.4), and AP activity was measured at 405 nm after the addition of 120 mM p-nitrophenol phosphate (Sigma) in 0.1 M glycine buffer, pH 10.4. Background differentiation in the absence of exogenous Shh was subtracted. Assays were performed in triplicate.

Stimulation and Inhibition of ShhNp Release from Producing
Cells-To evaluate Shh release from producing cells, we expressed recombinant murine full-length Shh cDNA, resulting in autoprocessed ShhNp, and constructs encoding Shh amino acids 1-198, resulting in ShhN (Fig. 1, A-D). Gel filtration analysis of media following 2 days of protein expression showed that ShhN as well as C-terminally hexahistidine-tagged ShhN His6 were monomeric in solution, confirming that both forms were produced in the nonlipidated form (Fig. 1E). In contrast, ShhNp formed large multimers, consistent with previous reports that suggested that ShhNp is produced in a dually lipidated form (23). We next analyzed cell-associated and -released morphogens by immunoblot analysis. 40 h after Bosc23 cells were transfected with full-length Shh cDNA or constructs encoding ShhN, normal medium was replaced with serum-free DMEM. After 2 h, serum-free DMEM was harvested and ultracentrifuged in this and all further experiments to rule out the possibility that the soluble morphogen was only bound to membranous remnants. The supernatant was then trichloroacetic acid-precipitated, and cells were directly lysed in SDS sample buffer. As previously reported (4), we found that ShhN was efficiently secreted into the medium ( Fig. 2A), whereas most of the processed, lipid-modified form remained cell-associated (Fig. 2B). However, treatment of ShhNp-expressing cells with 0.3 mg/ml M␤CD, a cyclic oligosaccharide that extracts and depletes cholesterol from living cells, strongly increased the amount of soluble ShhNp (Fig. 2C). M␤CD-mediated ShhNp release was concentration-dependent and time-dependent (Fig. 2D).
To demonstrate biological activity of M␤CD-released ShhNp, we took advantage of a sensitive cell-based bioassay for Shh, the differentiation of C3H10T1/2 osteoblast precursor cells (24). As shown in Fig. 3, ShhNp-conditioned media induced C3H10T1/2 differentiation into AP-producing osteoblasts, demonstrating biological activity of the M␤CD-released morphogen. In contrast, media obtained from transfected cells without M␤CD treatment did not show biological activity (left bar). To verify that this activity was due to ShhNp, we used the teratogen cyclopamine (25) and the ShhN-neutralizing antibody 5E1 (22) as specific Hh inhibitors during C3H10T1/2 differentiation. As expected, the biological activity of ShhNp conditioned medium was completely inhibited by cyclopamine cotreatment (p Յ 0.001, n ϭ 4) and strongly inhibited by 5E1 cotreatment (p Յ 0.001, n ϭ 4).
Ultracentrifugation of media prior to trichloroacetic acid precipitation was always carried out to derive soluble forms of ShhNp and no unreleased forms bound to membranous remnants or detached cells. Additionally, to demonstrate that soluble ShhNp was not simply M␤CD-extracted from the producing cell, we compared the molecular sizes of cell-bound and released ShhNp. SDS-PAGE analysis revealed increased electrophoretic mobility of M␤CD-released ShhNp if compared with the cell-tethered morphogen (Fig. 4A), suggesting proteolytic processing underlying ShhNp release. Because a ShhN zinc-coordinating site structurally analogous to the active site   yields the unlipidated, 19-kDa soluble morphogen (ShhN). D, sAP-ShhN is N-terminally tagged. E, Western blot analysis of fractions eluted from gel filtration chromatography was employed to confirm the lipidation status of the various constructs, also showing that the hexahistidine tag did not interfere with protein secretion. Bosc23 cells transfected with full-length ShhNp released large multimers (top) (23). In contrast, B16-F1-expressed ShhN was exclusively monomeric (middle), indicating the absence of N-terminal palmitate and C-terminal cholesterol (23). Likewise, hexahistidine-tagged ShhN His6 was also monomeric, proving the absence of both terminal lipids. All ShhN forms were detected using ␣-ShhN antibodies. Molecular sizes and the fractions in which the molecular weight standards eluted are indicated.

FIGURE 2. M␤CD induces ShhNp release from Bosc23 cells. 40 h after
Bosc23 cells were transfected with Shh cDNA or mock-transfected, cells were washed, and serum-free DMEM was added. After 2 h, the medium was harvested and subjected to ultracentrifugation and trichloroacetic acid precipitation, followed by SDS-PAGE and Western blot analysis. A, unlipidated ShhN is efficiently secreted into the medium (m), whereas no signal is detected in mock (empty vector control)-transfected cell lysate (cl) and medium, confirming ␣-ShhN antibody specificity. B, lipidated ShhNp is not released into the medium. Again, no signal is detected in mock-transfected cells and medium. C, 40 h after Bosc23 cells were transfected with full-length Shh cDNA, cells were washed, and serum-free DMEM with or without 300 g/ml M␤CD was added. After 2 h, the medium was harvested and subjected to ultracentrifugation and trichloroacetic acid precipitation, followed by SDS-PAGE and Western blot analysis using ␣-ShhN antibodies. M␤CD induced the release of soluble 19-kDa ShhNp into the medium (top). Cellular ShhNp expression levels are shown in the bottom. nt, Shh-transfected, DMEM only; mock, empty vector control. D, M␤CD-induced ShhNp release is concentration-and time-dependent. ShhNp-expressing Bosc23 cells were treated with increasing M␤CD concentrations ranging from 1 to 900 g/ml DMEM (left), or ShhNp release over time was monitored following stimulation with 300 g/ml M␤CD (right).
of zinc hydrolases, such as thermolysin and carboxypeptidase A, had been described (26), we hypothesized that ShhNp autocleavage may have underlay the observed processing of terminal peptides and release of the morphogen. To test this idea, we conducted site-directed mutagenesis of amino acid Glu 177 predicted to be essential for enzymatic activity (26 -28). However, this did not affect ShhNp E177A release into the medium (Fig.  4B), ruling out autoproteolysis as the underlying release mechanism. We thus tested the ability of extracellular metalloproteases to mediate ShhNp release into the medium. To test this hypothesis, we analyzed levels of soluble ShhNp (Fig. 4C, top) and corresponding ShhNp expression (Fig. 4C, bottom) in the presence of two metalloprotease inhibitors. In M␤CD-treated Bosc23 cells, ShhNp release was indeed reduced by 15 M NNGH, a specific inhibitor directed against metalloproteases, and was almost completely abolished by 100 nM TIMP-3, which is effective against the metalloproteases ADAM10, ADAM17, and ADAM28, among other metalloproteases. We also found increased ShhNp release upon PMA treatment that stimulates ADAM17-mediated shedding (29) (Fig. 4D). Again, PMA-stimulated ShhNp shedding was completely blocked by NNGH cotreatment (average of three experiments is shown). Taken together, these results indicated a role of NNGH-and TIMP-3sensitive proteases in ShhNp release, most likely ADAM 10, ADAM17, and/or ADAM28, in agreement with the presence of truncated ShhNp in the medium and reports describing M␤CD as a stimulator of ADAM-mediated ectodomain shedding (12, 14 -18).
ADAM17 Participates in ShhNp Release-To determine a putative role of ADAM17 function in ShhNp release from Bosc23 cells, we analyzed endogenous ADAM expression by semiquantitative RT-PCR. This resulted in the detection of 500-bp amplification products for ADAM10 and ADAM17 (Fig. 5A). We next assessed ShhNp release upon gain of ADAM17 function. ADAM17/ShhNp cotransfection in the FIGURE 3. ShhNp released by M␤CD treatment is biologically active. C3H10T1/2 osteoblast precursor cells were incubated in ShhNp-conditioned medium obtained after M␤CD treatment of ShhNp-expressing Bosc23 cells or control medium, and the relative amount of Shh-induced AP activity was determined as a readout for C3H10T1/2 differentiation into AP-producing osteoblasts and hence for biological activity of the morphogen. Medium obtained from mock-transfected Bosc23 cells was used as a control and subtracted from the other values. As a second control, ShhNp-transfected cells were incubated without M␤CD to demonstrate specificity of M␤CD-induced release (left bar). M␤CD-released ShhNp induced AP activity in C3H10T1/2 cells that was entirely blocked by specific inhibitors of Shh signaling, the teratogen cyclopamine (CA; 2.5 g/ml, p Յ 0.001, n ϭ 4), or the neutralizing anti-Shh antibody 5E1 (1 mg/ml, p Յ 0.001, n ϭ 4), demonstrating biological activity of the M␤CD-released morphogen and specificity of the assay.  Bottom, nonreleased ShhNp tethered to the corresponding Bosc cells. ADAM17 cotransfection but not cotransfection with an empty control vector results in strong ShhNp release. M␤CD-induced ShhNp release was inhibited by NNGH, demonstrating that this inhibitor is effective against ADAM17 function. C, in contrast to overexpressed wild type ADAM17, expression of the inactive A17 E406A protease (21) did not increase ShhNp release, confirming that proteolytic shedding underlies ShhNp release in this assay. A typical result of three independent assays is shown.
presence or absence of M␤CD induced strong ShhNp shedding that was impaired by NNGH cotreatment (Fig. 5B). Furthermore, site-directed mutagenesis of ADAM17 glutamic acid residue Glu 406 , resulting in the inactivation of proteolytic activity (21), prevented ShhNp release and demonstrated specificity of ADAM17-mediated proteolytic shedding (Fig. 5C).
In order to confirm endogenous ADAM function in ShhNp shedding, siRNAs expressed from shRNAs were used to suppress endogenous ADAM17 expression in ShhNp-transfected Bosc23 cells by RNAi. Following M␤CD treatment, one shRNA construct cotransfected with the ShhNp-encoding construct resulted in a significant reduction of ShhNp release (construct 1, p Յ 0.01, n ϭ 6 independent assays) (Fig. 6A, left). Under normal culture conditions (right), cotransfection with three different shRNA constructs suppressing ADAM17 expression led to a significant reduction of ShhNp release (constructs 3-5, p Յ 0.01, n ϭ 6 independent assays). Interestingly, under normal culture conditions (no M␤CD-induced release), immunoblotting again revealed increased electrophoretic mobility of the soluble morphogen upon ShhNp cotransfection with the shRNA control construct (c) and shRNA construct 2, indicating proteolytic processing of the released morphogen (Fig. 6B). In contrast, nontruncated ShhNp was detected following cotransfection with shRNA constructs 1, 3, 4, and 5. This indicates that ADAM17 knockdown results in both decreased morphogen processing and release. This finding was independently con-firmed using four pooled siRNAs specific for ADAM17. Here, ShhNp release was reduced to 64 Ϯ 8.5% of control levels obtained with a pool of nontargeting siRNAs (p Յ 0.006). However, because Ͼ70% efficiency of ADAM17 mRNA knockdown was determined by immunoblotting using an anti-ADAM17 antibody (Fig. 6C), we concluded that other metalloproteases probably contributed to ShhNp release. Still, based on the ADAM17 overexpression and knockdown experiments, we conclude that ADAM17 participates in ShhNp release, consistent with increased levels of truncated, soluble morphogen upon M␤CD treatment, increased release upon PMA stimulation, and decreased release upon NNGH/TIMP-3 treatment.
Heparan Sulfate Participates in ShhNp Release from Bosc23 Cells-We next hypothesized that HS may affect ShhNp shedding, because fly HhNp associates with cell surface HSPGs (30) and HSPG biosynthesis affects Hh function (31). To test this idea, immunoblotting of released and cell-tethered ShhNp in the presence or absence of various forms of HS or other sulfated carbohydrates in serum-free medium was conducted. Indeed, soluble ShhNp was only detected in the presence of HS and heparin, whereas other sulfated carbohydrates, such as chondroitin sulfate and dextran sulfate, did not affect ShhNp release (Fig. 7A). Moreover, the soluble morphogen was again found to be truncated in the presence of HS isolated from a mouse melanoma cell line (B16-HS) (Fig. 7B, arrow), and ShhNp release upon HS treatment was also inhibited by NNGH cotreatment (Fig. 7C).
We next assayed various ubiquitously expressed forms of HS for their ability to modulate ShhNp release. As shown in Fig. 8A, 80 g/ml HS affected ShhNp release to different degrees, depending on the source of HS. Notably, HS derived from the mouse melanoma cell line B16-F1 (B16 HS) mediated ShhNp release very effectively, whereas identical amounts of HS from adult mouse brain and E15 mouse embryo were less efficient. HS isolated from adult mouse lung, kidney, and skin did not enhance ShhNp levels in the medium, but HS isolated from stage E11.5 embryos did (not shown). These results supported the idea that HS from natural sources affects ShhNp release, suggesting in vivo functions for specific HS sulfations. . siRNAi specific for ADAM17 impairs morphogen release and processing. A, immunoblot analysis of released ShhNp following cotransfection with full-length Shh cDNA and plasmids encoding for siRNAs specifically interfering with ADAM17 mRNA. Five constructs matching different sequences of the ADAM17 target were used. All experiments were repeated six times, and graphs express the ratio of shed versus cell-bound ShhNp after immunoblotting and quantification relative to the ratio obtained with the control (c), which was set to 100%. ShhNp release was reduced up to 35% following M␤CD-induced ShhNp release (left, p Յ 0.01 (construct 1), p Յ 0.06 (construct 2), p Յ 0.07 (construct 3), p Յ 0.08 (construct 4), and p Յ 0.03 (construct 5), n ϭ 6), demonstrating that reducing ADAM17 mRNA levels impaired ShhNp shedding. Under normal culture conditions, ShhNp release was impaired even more strongly (right, p Յ 0.04 (construct 1), p Յ 0.07 (construct 2), p Յ 0.01 (construct 3), p Յ 0.01 (construct 4), and p Յ 0.01 (construct 5), n ϭ 6). **, significant reduction in ShhNp release (p Յ 0.01). B, ShhNp immunoblot analysis after Bosc23 cotransfection with full-length Shh cDNA and the same ADAM17-specific siRNAs shows variably sized ShhNp in the medium but not in the cell lysate. ShhNp shows increased electrophoretic mobility in the medium of control siRNAi-treated and construct 2-treated samples (arrow), indicating processing, but not in cells or media derived from cells transfected with constructs 1, 3, 4, and 5 (arrowhead). C, immunoblot analysis of Bosc23 following transfection with the pooled two most efficient siRNAi constructs specific for ADAM17 (constructs 4 and 5). In contrast to cells transfected with the empty control, ADAM siRNAi results in significant down-regulation of protein expression. Two independent knockdown experiments resulted in a Ͼ70% ADAM17 reduction. To test whether the ability to modulate ShhNp release correlates with HS sulfation, disaccharide analysis after heparin lyase digestion was conducted by anion exchange high pressure liquid chromatography (Fig. 8B). We found that the overall sulfation of B16-HS (0.6 sulfates/disaccharide) was similar to that of E15 mouse embryo HS (0.5 sulfates/disaccharide), adult brain HS (0.66 sulfates/disaccharide), and "inactive" HS isolated from mouse lung, kidney, and skin (32). This suggests that the stimulatory function of HS depends on the patterning of HS sulfation rather than simply on high sulfation levels. Consistent with the idea that specific HS sulfation motifs may affect ShhNp release, disaccharide composition differed most between B16 HS and the less active forms. In B16 HS, the relative amount of D2H0 (UA-2S- [1,4]-GlcN, where UA represents iduronic acid or glucuronic acid as stereochemistry is being lost) was 6 -10 times higher than in the less active forms, and the relative amount of D2A6 (UA-2S- [1,4]-GlcNAc-6S) was strongly reduced. This led us to speculate that free amino groups present on GlcN or 2-O-sulfations may be important in ShhNp shedding. To test this hypothesis, ShhNp and soluble ShhN were expressed in wild type Chinese hamster ovary (CHO-K1) cells and CHO-K1 cells deficient in 2-O-sulfotransferase activity (PgsF17) (33), and relative amounts of soluble ShhN and ShhNp secreted from each cell type were analyzed. HS isolated from mutant pgsF17 cells lacks any 2-O-sulfated iduronic or glucuronic acid residues but contains normal levels of 6-O-sulfated residues and elevated levels of N-sulfated D0S0 (UA- [1,4]-Glc-NS) (34). Moreover, wild type CHO-K1 cells and pgsF17 cells do not produce significant amounts of motifs carrying free amino groups (such as D2H0) (34). Comparison of relative morphogen levels revealed increased ShhNp release by pgsF17 cells if compared with CHO-K1 cells (Fig. 8C), again indicating a role of 2-O-sulfation in ShhNp release. This is consistent with the low relative abundance of D2A6 in B16-derived HS. In addition, we suggest that free amino groups are probably not required for ShhNp release due to their absence in pgsF17derived HS. Thus, elevated levels of D2H0 on B16 HS are probably unrelated to its ability to mediate ShhNp release. Taken together, our results demonstrate that HS differentially regulates the release of truncated (processed) ShhNp and that this function probably depends on specific HS sulfation patterns.
ShhNp Is C-terminally Processed-How does HS influence ADAM-mediated release of ShhNp? Because HS-dependent processes require protein/HS interactions, we first conducted heparin affinity chromatography of recombinant soluble ADAM17 and ShhN. As shown in Fig. 9A, we found strong ShhN binding to heparin (eluting at 0.7-0.8 M salt) but no ADAM17 binding (eluting at Ͻ0.2 M salt). This indicates ADAM insensitivity to HS and indicates that HS more likely acts on Hhs alone. We hypothesized that HS binding may have rendered the morphogen more accessible to cleavage, because other proteins such as antithrombin or fibronectin were described to undergo structural changes upon heparin binding  and because ADAM substrates are thought to be cleaved in a conformation-dependent manner. To test this idea, we expressed C-terminally hexahistidine-tagged 28-kDa ShhN His6 in B16-F1 cells (that express ADAM proteases; not shown) and analyzed morphogen sizes by Western blotting. The addition of 130 g/ml mouse embryo-derived HS to ShhN His6 -expressing cells, but not the addition of chondroitin sulfate or dextran sulfate, resulted in the generation of truncated 19-kDa ShhN doublets (possibly processed twice at the C terminus or at both termini), confirming that morphogen processing is HSdependent (Fig. 9B, top). This result is in agreement with the finding that HS (but not chondroitin sulfate and dextran sulfate) enhanced levels of truncated ShhNp in the medium of ShhNp-transfected cells (Fig. 7A). Antibodies directed against the hexahistidine tag (␣-His 4 ) failed to detect truncated ShhN His6 , confirming C-terminal processing (Fig. 9B, bottom). In agreement with the finding that HS sulfation affects ShhNp release (Fig. 8C), we found that HS purified from Ndst1-deficient embryo littermates increased C-terminal processing about 2-fold (n ϭ 3 independent experiments, p Յ 0.001; not shown). HS derived from Ndst1-deficient embryos shows a 60% reduction in N-sulfation as well as 2-O-sulfation but only slightly reduced levels in 6-O-sulfation, (35), consistent with increased ShhNp release by pgsF17 cells (Fig. 8C). Also, increasing the concentrations of embryonic HS results in elevated C-terminal processing of ShhN His6 (Fig. 9C). We next tested whether HS-induced processing results in comparable sizes of ShhN His6 and Bosc23-released ShhNp. 28-kDa ShhN His6 was expressed in B16-F1 cells in the presence of two different preparations of E18 mouse embryo HS that had previously been tested for efficient ShhNp release and compared with soluble ShhNp and unprocessed ShhN (Fig. 9D). Indeed, immunoblotting revealed identical sizes of proteolytically processed ShhN His6 and released ShhNp (arrow), whereas untagged, soluble ShhN was slightly larger (arrowhead). This confirmed comparable HS-dependent processing of different Hh constructs and confirmed that processing was not restricted to the tagged morphogen.
To directly detect the loss of C-terminal cholesterol from released ShhNp via endogenous metalloproteases and to additionally demonstrate that morphogen processing was not an artifact restricted to the tagged protein, ShhNp-overexpressing cells were grown in the presence of [1␣,2␣-3 H]cholesterol. This resulted in cholesterol labeling of 19-kDa ShhNp following autoprocessing of the 48-kDa precursor by the C-terminal cholesterol transferase domain. ShhNp shedding was M␤CD-induced, and the medium was trichloroacetic acid-precipitated, separated by SDS-PAGE, and analyzed by immunoblotting and autoradiography. As shown in Fig. 10A, anti-ShhN antibodies detected both unprocessed (48-kDa) and processed  Shh proteins in cell lysates, and released ShhNp protein was also detected in the medium. However, cholesterol-labeled 19-kDa ShhNp was only detected in the cell lysate, indicating loss of the cholesterol moiety during release. To verify HhNp shedding from a biologically relevant source, we induced its release from cultured primary chondrocytes isolated from chick sternum (Fig. 10B) that endogenously expressed both HhNp and ADAM17, as assessed by RT-PCR and Western blot-ting (not shown). The addition of 100 ng/ml insulin-like growth factor I to those cells resulted in hypertrophy and the production of soluble, biologically active HhNp. 4 Chondrocytes were cultured in the presence of [1␣,2␣-3 H]cholesterol and insulinlike growth factor I, again resulting in the labeling of cell-tethered 19-kDa HhNp. Again, released HhNp showed loss of the radiolabel. These results show that ADAM-mediated HhNp release and C-terminal processing were not artifacts of protein overexpression or the expression of tagged proteins and that HhNp shedding is a biologically relevant process.
ShhNp May Also Be N-terminally Processed-We hypothesized that, in order to release the dually lipidated morphogen from the cell surface, ADAM-mediated shedding probably results not only in C-terminal but also in N-terminal processing, consistent with the presence of double bands in some ShhNp release assays. To test this idea, we expressed N-terminally alkaline phosphatase-tagged 80-kDa sAP-ShhN, incu-4 R. Dreier, unpublished results. bated the protein with embryo-derived HS as described above, and analyzed morphogen integrity by Western blotting. 100 g/ml E18 HS resulted in specific generation of a 19-kDa ShhN doublet from the 80-kDa sAP-ShhN precursor molecule (Fig.  10C), demonstrating N-terminal cleavage by endogenous proteases. Again, to directly test for the presence or absence of the N-terminal palmitic acid moiety following ShhNp shedding, Shh-transfected Bosc23 cells were grown in the presence of [9,10-3 H]palmitic acid and were M␤CD-treated. As shown in Fig. 10D, both unprocessed (48-kDa) and processed (19-kDa) ShhNp proteins were detected in cell lysates, and high levels of released ShhNp protein were also detected in the medium. However, [ 3 H]palmitic acid signals were not detected in the medium, although numerous [ 3 H]palmitic acid-labeled proteins were present in the cell lysate. Exposure for Ͼ6 months still did not yield a detectable signal in any medium sample, suggesting that soluble ShhNp may have also been nonpalmitoylated.
Last, we attempted N-terminal sequencing of the processed morphogen to determine the cleavage site. However, expression of ShhNp in Bosc23 cells failed to yield sufficient protein for analysis. Therefore, we incubated purified E. coli-expressed, N-terminally glutathione S-transferase-tagged ShhN (GST-ShhN) with conditioned, serum-free medium derived from B16-F1 cells and 100 g/ml E18 HS, resulting in the detection of sufficient 19- 45 . This shows that two N-terminal 9-amino acid and 14-amino acid peptides can be released by proteolytic processing in vitro, in agreement with the detection of double bands in some assays.

DISCUSSION
Hh lipidations are essential for proper morphogen secretion and spreading through the extracellular matrix, despite their tight association with the cell membrane. Thus, a cellular mechanism is needed for the release of the lipidated morphogen from its source. In this work, we describe that metalloproteases release the lipidated morphogen from producing cells, resulting in truncated, soluble, and biologically active ShhNp. ADAM17 overexpression increased ShhNp release that was inhibited by ADAM17 RNAi, demonstrating that ADAM17 (together with other metalloproteases) participates in ShhNp release. Redundancy in enzymes and substrates is common among ADAM family members (36, 37) (e.g. processing of EGFR ligands by ADAM17, ADAM9, and ADAM10 and processing of tumor necrosis factor-␣, amyloid precursor protein, and Notch by ADAM10 and ADAM17) (38 -41). The ability of more than one metalloprotease to mediate ShhNp release is also supported by phenotypes of ADAM17-deficient mice that do not show Shh-related defects (42).
If redundancy in enzymes and substrates is common, how do metalloproteases with proteolytic activity selectively recognize their appropriate substrate, and how is the activity of ShhNp sheddases regulated? We suggest that HS may participate in the regulation of ShhNp shedding, as shown for HS isolated from tumor cells and by increased ShhNp release from 2-O-sulfotransferase-deficient pgsF17 cells. As shown in this work, the stimulatory function of HS on ShhNp release does not simply depend on sulfation levels but may more likely depend on a specific sulfation pattern lacking 2-O-sulfate residues. Increased ShhN His6 C-terminal processing in the presence of Ndst1-deficient embryo-derived HS (showing low levels of 2-O-sulfation) also suggests an inhibitory function of 2-O-sulfate on ShhNp shedding. Based on these findings, we suggest a regulatory role of 2-O-sulfation of cell surface HS in ShhNp release; low 2-O-sulfation of a ShhNp binding HS motif may allow for efficient metalloprotease-mediated processing, whereas 2-O-sulfation of the motif may down-regulate ShhNp release. Although this model is attractive, HS-mediated ShhNp stabilization may also add to elevated ShhNp levels following release from the cell. Still, HS-dependent processing of tagged soluble ShhN His6 indicates that ShhNp release is also regulated by HS-modulated ectodomain shedding. This may help explain the observation that expression of Drosophila HS is required for the diffusion of lipid-modified fly HhNp (6 -8) but not cholesterol-free HhN (6,9,10) that is secreted independent of sheddase function.
In addition to the evidence listed above, release of the ShhNp signaling domain via proteolytic shedding is in agreement with the ShhN crystal structure, showing a globular domain and extended N-and C termini susceptible to proteolytic cleavage (26). The most N-and C-terminal peptides are dispensable for biological activity, because ShhN lacking N-terminal residues 25-33 (28) as well as ShhN lacking amino acids 191-198 5 retained full biological activity. Also, lipid modification of terminal peptides is dispensable, because Ptc binding (43) or Ptc1mediated sequestration (44) does not require morphogen lipidation. Our results are also in full agreement with the established regulatory function of ADAMs in shedding processes (13). Here, we add a novel, important ligand to the growing list of sheddase substrates and suggest that ShhNp shedding may be involved in the regulation of spatiotemporal ShhNp gradient formation.
The findings presented in this work provide a new explanation for the paradoxical situation that membrane-tethered, lipidated ShhNp functions as a soluble morphogen. ShhNp shedding, however, seems at odds with our finding and the finding of others that lipidated ShhNp, but not ShhN, forms multimers in solution (4,23,45). Because the shed morphogen is devoid of its lipid anchors, our results challenge the assumption that ShhNp multimers form via lipid sequestration in the interior of the soluble multimer. We thus suggest an alternative mechanism to explain the formation of ShhNp multimers. Vyas et al. (30) recently demonstrated that fly HhNp forms nanoscale oligomers on the surface of producing cells that further colocalize with HSPGs to form visible clusters. Nanoscale oligomers were described to form on the cell surface based on protein-protein interactions, and lipidation in the absence of substantial regions of the ShhN protein was insufficient for clustering. Instead, HhNp lipidation was suggested to allow for loose HhNp preclustering on the cell surface as a prerequisite for the formation of higher order complexes. Therefore, ShhNp lipidation may also serve as a prerequisite for the formation of oligomeric nanostructures and ultimately HSPG-associated complexes on Bosc23 cells. Because HSPGs can undergo proteolytic shedding together with bound ligands (46), ShhNp⅐HSPG complexes may be shed simultaneously, resulting in the observed labile, high molecular weight structures. This hypothesis is compatible with the described stability of ShhNp multimers in the presence of detergent or 500 mM NaCl (which is insufficient for disruption of ShhN/HS interactions) (Fig. 9A) and loss of ShhNp multimerization upon disruption of the HS-binding Cardin-Weintraub sequence (amino acids 32-39) (45). Nonlipidated ShhN, in contrast, cannot undergo lipid-dependent preclustering, resulting in the direct secretion of the monomeric morphogen. Taken together, we suggest that although the presence of multimeric ShhNp on the cell surface results from initial lipid-dependent nanoscale clustering, soluble clusters may be formed and maintained in a lipid-independent fashion. Work currently conducted in our laboratory aims at the characterization of soluble ShhNp complexes and the identification of additional sheddases involved in ShhNp release.