Sonic hedgehog processing and release are regulated by glypican heparan sulfate proteoglycans

A major challenge in developmental biology is to understand how cells coordinate their developmental behavior with that of their neighbors. Cells often use signaling molecules from a local source that act at a distance on target cells. The proteins of the Hedgehog (Hh) family are one such group of signaling molecules that control development and contribute to cancer formation and progression in the adult (Theunissen and de Sauvage, 2009; Yauch et al., 2008). Vertebrates produce three structurally and functionally related Hh proteins [Sonic Hh (Shh), Indian Hh (Ihh) and Desert Hh]. Production of the active 19-kDa Shh protein begins with covalent attachment of a cholesterol moiety to the carbonyl moiety of its C-terminal glycine residue. In a second step, Hh acyltransferase (Hhat) attaches a palmitoyl group to the NH₂-terminal cysteine of the Shh signaling domain. The dually lipidated molecule constitutes the active morphogen (Jacob and Lum, 2007). Upon secretion to the cell surface, lipidated Hh proteins multimerize and then transport to cells expressing the Hh receptor Patched (Panáková et al., 2005; Zeng et al., 2001). The paradox is that these membrane-tethered Hh molecules serve as long-range morphogens, suggesting that specific mechanisms are required for their release and transport.

One group of molecules involved in Shh solubilization are a disintegrin and metalloproteinase (ADAM) sheddases (Dierker et al., 2009; Ohlig et al., 2011). Sheddases release numerous membrane-associated molecules from the cell surface (summarized in Tellier et al., 2006; Yang et al., 2006), while the cell retains the membrane-associated remnant peptides. However, because redundancy in enzymes and substrates is common (Kheradmand and Werb, 2002), activities and colocalization of sheddases with their substrate(s) requires tight regulation. One recently described Hh shedding regulator is the glycoprotein Scube2 (signal peptide, cubulin domain, epidermal growth factor-like protein 2) (Jakobs et al., 2014). Another group of molecules that regulate Hh signaling is the glycosylphosphatidylinositol (GPI)-linked glypican (Gpc) heparan sulfate proteoglycans (HSPGs, Gpc1–Gpc6 in vertebrates) (Chatterjee and Mayor, 2001; Mayor and Riezman, 2004; Rietveld et al., 1999), called Division abnormally delayed (Dally) (Nakato et al., 1995) and Dally-like protein (Dlp) (Khare and Baumgartner, 2000) in the fly. Gpc functions often depend on their heparan sulfate glycosaminoglycan chains, which are produced by virtually all cells as part of the cell surface and extracellular matrix (Esko and Lindahl, 2001). The unbranched heparan sulfate chain is synthesized by glycosyltransferases Ext1 and Ext2, or by the Tout-velu (Ttv) and Sister of tout velu orthologs in Drosophila. The nascent heparan sulfate chain consists of repeating glucuronic acid and N-acetylglucosamine (GlcAβ1,4GlcNAcα1,4)_n disaccharide units that undergo modification by one or more of the four N-deacetylase/N-sulfotransferase isozymes [or the single Sulfateless (Sfl) ortholog in the fly (Toyoda et al., 2000)]. The resulting N-sulfoglucosamine (GlcNS)-containing chain is further modified by one 2-O-sulfotransferase, several 3-O- and 6-O-sulfotransferases, and a GlcA C5-epimerase, the latter converting GlcA into iduronic acid (IduA) (Esko and Selleck, 2002; Lindahl et al., 1998). Finally, heparan sulfate sulfation in the extracellular compartment is reduced by the secreted 6-O-endosulfatase 1 (Sulf1) and its homolog Sulf2 in vertebrates, or by the Sulf1 fly ortholog (Lamanna et al., 2006). Together, these activities result in vertebrate heparan sulfate consisting of multiple unmodified NAc (N-acetylated), highly sulfated N-sulfated, and mixed NAc and N-sulfated domains (Carlsson et al., 2008). In Drosophila, heparan sulfate consists of a single extended N-sulfated domain (Kusche-Gullberg et al., 2012). Heparin represents a highly sulfated form of heparan sulfate, mostly consisting of disulfated (26%) and trisulfated (73%) disaccharides and a minor monosulfated or non-sulfated fraction containing GlcA or IduA linked to GlcNAc and GlcNS (1%) (Saad et al., 2005). Heparan sulfate and heparin binding of various morphogens, cytokines, chemokines, growth factors, proteases and protease inhibitors is mediated by clusters of basic amino acids on their surfaces. In vertebrate Hh proteins, heparan sulfate is mainly bound by the Cardin–Weintraub motif BBBxxBB (B represents a basic amino acid, x represents any amino acid) (Cardin and Weintraub, 1989; Chang et al., 2011; Farshi et al., 2011; Rubin et al., 2002), but Cardin–Weintraub residues also constitute the N-terminal Hh sheddase cleavage site (Ohlig et al., 2012). This suggests a possible mode of processing regulation by competitive heparan sulfate binding, consistent with recently described HSPG bioactivities in Hh-producing cells (Ayers et al., 2010; Bilioni et al., 2013; Oustah et al., 2014; Wojcinski et al., 2011). In the present study, we provide evidence that supports this functional link. We found that Gpc HSPGs regulate Shh release from producing cells in vitro and in vivo and that heparan sulfate regulates proteolytic Shh processing in vitro. We further show that purified heparan sulfate directly stimulates Shh processing and that this ability depends on the source of heparan sulfate. This finding indicates that there are heparan-sulfate-dependent Hh ligand or release factor scaffolding or activator functions in the extracellular compartment. It further reveals a previously unknown role of cell surface HSPGs in the precise control of spatiotemporal Hh signaling.

Embryonic null mutants for the genes dally and dlp (Ayers et al., 2010; Gallet et al., 2008; Han et al., 2004b; Takeo et al., 2005; Williams et al., 2010) and mutants for heparan sulfate biosynthetic genes ttv (Bellaiche et al., 1998; The et al., 1999), brother of ttv and sister of ttv (Bornemann et al., 2004; Han et al., 2004a; Takei et al., 2004) have phenotypes similar to hh-null mutants. These defects have been attributed to impaired Hh transport and reception. Moreover, it was recently reported that overexpression of the heparan-sulfate-modifying extracellular enzyme Drosophila sulfatase (Sulf1) in Hh-producing posterior wing disc cells increased Hh patterning activity in vivo and that sulf1 knockdown in these cells impaired Hh function (Wojcinski et al., 2011). This finding implied that the heparan sulfate expressed in Hh-producing cells can also regulate Hh biofunction (Oustah et al., 2014; Wilson and Chuang, 2006). To confirm this function, we downregulated HSPG expression in vivo, using Hh-dependent Drosophila wing development as a read-out. In the fly, wing development is initiated by Hh expression in the posterior wing disc compartment in response to the Engrailed (En) transcription factor (summarized in Crozatier et al., 2004; Hartl and Scott, 2014). Hh proteins are then released from posterior cells to move anteriorly. The position of the longitudinal L3–L4 wing veins results from this anterior Hh movement and can therefore be used to quantify the positional information provided by Hh, i.e. reduced Hh biofunction reduces the L3–L4 intervein region and concomitantly increases the L2–L3 intervein space (Crozatier et al., 2004; Lee et al., 2001; Mullor et al., 1997; Strigini and Cohen, 1997) (schematic in Fig. 1). We used the en-Gal4/UAS system to express various RNA interference (RNAi) constructs to target heparan sulfate biosynthetic enzymes or HSPG core proteins in Hh-producing posterior cells in vivo. Initial knockdown of the heparan sulfate biosynthetic genes sfl and ttv led to strong pleiotropic defects, lack of wings, or larval or pupal lethality that prevented further analysis (data not shown). Drosophila that were partially deficient in dally and dlp expression, however, survived and showed significantly impaired Hh-dependent wing patterning. A quantitative wing analysis revealed that the ratio between the L3–L4 and L2–L3 intervein areas decreased from 0.96±0.005 [en>GFP RNAi control (n=42)] to 0.86±0.007 (en>dally1, −10.23%), 0.9±0.007 (en>dally2, −6.27%), and 0.86±0.008 (en>dally3, −10.2%) (all mean±s.d., n=20, Fig. 1). This result demonstrates that impaired dally expression in all three independent RNAi fly lines (dally1–dally3) affected Hh function. Two independent RNAi lines targeting dlp expression also displayed reduced Hh function (en>dlp1, 0.82±0.008, −15.1% and en>dlp2, 0.9±0.012, −6.13%). Thus, both Drosophila glypicans have overlapping functions in Hh-producing cells. Because these observed phenotypes strongly resemble wing patterning defects seen upon RNAi-mediated sulf1 knockdown (Wojcinski et al., 2011), we hypothesized that heparan sulfate modulates Hh multimerization prior to its release (Vyas et al., 2008) or impairs Hh release from the cell surface.

All Hh proteins are dual-lipidated proteins that firmly tether to the membrane of the cell that produces them. C-terminal Hh lipidation uses a strictly coupled cleavage and cholesterol esterification reaction (Bumcrot et al., 1995; Lee et al., 1994; Liu, 2000; Tabata and Kornberg, 1994) to render all 19-kDa Hh signaling domains insoluble (Liu, 2000; Marti et al., 1995; Porter et al., 1996a,b). The second Hh lipid modification is palmitic acid attached by Hhat to conserved N-terminal Hh cysteine residues (Amanai and Jiang, 2001; Chamoun et al., 2001; Lee and Treisman, 2001; Micchelli et al., 2002; Pepinsky et al., 1998) (Fig. 2A). We and others have shown that proteolytic removal of N- and C-terminal lipidated peptides (called shedding) releases bioactive Shh morphogens from the surface of producing cells in vitro (Damhofer et al., 2015; Dierker et al., 2009; Jakobs et al., 2014; Ohlig et al., 2011). We further showed that because of the lack of the N-terminal Hhat acceptor cysteine (C25S or C25A in mouse Shh) (Hardy and Resh, 2012) non-palmitoylated Hh proteins require only C-terminal shedding for their solubilization, whereas most Shh N-termini are left intact (Jakobs et al., 2014). Here, we confirm this observation by SDS-PAGE and immunoblotting of N-palmitoylated Shh and non-palmitoylated Shh^C25A expressed in Bosc23 cells (Fig. 2A), a HEK293-derived cell line commonly used to analyze Shh production and release. To monitor C-terminal Shh processing, we tagged both proteins with hemagglutinin (HA), resulting in the extended C-terminal membrane anchor N¹⁹⁰SVAAKSG-YPYDVPDYA-G¹⁹⁸ (G198 represents the cholesterol-modified glycine; underlined italicized letters represent the tag) (Jakobs et al., 2014). To confirm N-terminal processing during release, we used anti-Cardin–Weintraub antibodies raised against the Cardin–Weintraub peptide K³³RRHPKK³⁹ located adjacent to the palmitoylated cysteine residue (Ohlig et al., 2012). Polyclonal anti-ShhN antibodies were used to detect both full-length and truncated proteins on the same (stripped) blot (schematic in Fig. 2A). As expected, cellular Shh–HA was converted into soluble morphogens lacking most anti-Cardin–Weintraub and all anti-HA antibody binding (Fig. 2A, asterisk). Cell-surface-associated HA-tagged Shh^C25A was converted into a C-terminally processed, but N-terminally unprocessed soluble form (yellow band), showing that the N-terminal processing of unpalmitoylated morphogens was impaired (Ohlig et al., 2012). Consistent with our previous observation that Shh palmitoylation in vitro is incomplete (Jakobs et al., 2014), Shh processing in CHO-K1 cells was strongly enhanced after stable hhat transfection (CHO-K1^tg cells, Fig. 2B, asterisk). These results demonstrate that both Hh lipidations contribute to Hh association with the cell membrane (Chamoun et al., 2001; Konitsiotis et al., 2014; Ohlig et al., 2011) to ensure that only fully processed Hh is released (Ohlig et al., 2012). This raises the question of how Hh shedding is controlled.

On the basis of previously described Hh colocalization with cell-surface glypicans prior to their release (Ayers et al., 2010; Vyas et al., 2008), we hypothesized that HSPGs might play this role. Consistent with this hypothesis, we observed that heparan sulfate expressed on the surface of 2-O sulfotransferase-deficient CHO-pgsF-17^tg cells (Xu and Esko, 2014) interfered with N-terminal Shh processing in the presence of 600 μg/ml methyl-β-cyclodextrin (MβCD), a cyclic oligosaccharide that activates sheddases and thus forces proteolytic processing of membrane-associated proteins (Fig. 2B) (Murai et al., 2011). The assembly of heparan sulfate in these and other cells is not template driven; rather, it is thought to depend on the organization and specificity of biosynthetic enzymes and the availability of precursors. Thus, 2-O sulfotransferase deficiency not only reduces heparan sulfate 2-O-sulfation, but also strongly enhances N-sulfation and moderately elevates 6-O-sulfation by an unknown compensatory mechanism (Axelsson et al., 2012). These secondary biosynthetic effects lead to complex changes in the heparan sulfate domain structure and overall heparan sulfate sulfation in these cells. N-terminal Shh processing was also impaired in pgsA-745^tg cells, a mutant CHO cell line defective in heparan sulfate and chondroitin sulfate biosynthesis (Fig. 2C). In contrast, comparable Shh^C25A processing from CHO-K1^tg wild-type cells and pgsA-745^tg cells revealed that C-terminal proteolysis was largely heparan sulfate independent, at least in the presence of the tag (Fig. 2D). From this observation, we draw two major conclusions. First, that glycosaminoglycans, probably of the heparan sulfate kind, regulate Shh processing. Second, N-terminal processing is rate limiting in Shh release. The latter conclusion is supported by the detection of increased soluble product amounts from non-palmitoylated Shh^C25A in comparison to palmitoylated protein products (compare relative amounts of soluble and cellular material produced in CHO-K1^tg cells; Fig. 2C,D, left lanes) and the findings of others (Chamoun et al., 2001; Konitsiotis et al., 2014). Finally, we confirmed that Hh palmitoylation is not a strict prerequisite for N-terminal processing, because N-truncated Shh^C25A was detected in CHO-K1^tg but not in heparan sulfate-deficient cells (Fig. 2D, left, asterisk). The possibility that pgsA-745^tg cells lack sheddase expression was ruled out by reverse transcription PCR (RT-PCR), which revealed robust ADAM10 and ADAM12 expression in all cell lines (data not shown). C-terminal Shh processing in these cells also confirmed sheddase expression (Fig. 2D).

Vertebrates express six Gpc family members (Gpc1–Gpc6) that stimulate or suppress Hh function in receiving cells, depending on the Gpc isoform (Capurro et al., 2008; Li et al., 2011). We therefore tested whether Gpc proteins in producing cells can also regulate Shh shedding in an isotype-specific manner. First, we expressed Shh alone in CHO-K1 cells or CHO-K1^tg cells as a control and quantified relative Shh amounts in media. As shown in Fig. 3A (Step 1), most Shh released from Hhat-transfected CHO-K1^tg cells was N-terminally processed, as demonstrated by a size shift and strongly impaired anti-Cardin–Weintraub antibody binding (center column, center row, right lane). In contrast, most solubilized proteins produced in Hhat-untransfected cells remained N-terminally unprocessed (left lanes). Again, both observations confirm that palmitate tethers Hh to the cell membrane and thereby prevents the release of N-terminally unprocessed lipidated protein termini. To quantify the amounts of unreleased surface-associated Shh, we added fresh media containing 600 μg/ml MβCD to the same cells (Fig. 3A, Step 2). MβCD solubilized substantial amounts of previously unreleased protein from the surface of CHO-K1^tg cells in the same (N-truncated) form, whereas only small amounts of remaining unpalmitoylated Shh were released from Hhat-untransfected CHO-K1 cells (bottom). These findings show that Shh shedding at the cell surface is strictly controlled and that this control involves the lipidated N-terminal Shh peptide.

To analyze Gpc functions in this controlled process in detail, we next expressed Shh and Gpc1 from a bicistronic mRNA in CHO-K1^tg cells (Fig. 3B). Under serum-free conditions, we found that Gpc1 strongly impaired Shh release in comparison to the control Shh expressed in its absence (compare Fig. 3A,B, center rows), despite comparable Shh production (Fig. 3A,B, top rows). The addition of MβCD revealed that the unreleased material resided on the cell surface (Fig. 3B, bottom row). Shh release was also strongly impaired in Gpc1 and Shh co-transfected pgsF-17^tg cells, pgsA-745^tg cells, and a CHO cell line deficient in heparan sulfate but not chondroitin sulfate biosynthesis (pgsD-677^tg cells). The observation that not even MβCD released N-terminally processed proteins from these mutant cells supported the hypothesis that the presence of properly modified heparan sulfate is required for Shh solubilization from the cell surface. Fig. 3C extends these findings, showing quantified soluble Shh co-expressed with Gpc1–Gpc6 from CHO-K1^tg, pgsF-17^tg cells and pgsA-745^tg cells relative to the material expressed in the absence of exogenous Gpcs. We observed that Shh solubilization as a consequence of Gpc1–Gpc6 co-expression was variably reduced in all cell lines (Fig. 3C). This finding indicates that most Gpcs suppress Shh shedding under serum-free conditions (Fig. 3C) as well as in the presence of MβCD (Fig. 3D). Finally, we determined that the solubilized morphogens showed similar activities in the Shh-responsive cell line C3H10T1/2 (Nakamura et al., 1997), indicating that Gpc isoforms do not impact upon Shh biofunction in receiving cells (supplementary material Fig. S1).

Shh and Gpc coexpression was also analyzed in Bosc23 cells. We first determined that Gpc2 and Shh, and Gpc6 and Shh expression levels and protein colocalization on the cell surface were comparable (Fig. 4, see yellow punctate staining in D and H) (Vyas et al., 2008). Next, we isolated cell lysates and supernatants from pIRES-Shh-Gpc-transfected Bosc23 cells and analyzed cellular and solubilized Shh by SDS-PAGE and immunoblotting as described above. Fig. 5 shows that the Shh solubilization was mostly unchanged after normalization of protein production rates to the Shh control level. With the exception of Gpc2, our findings confirm that the Gpc core proteins play only minor roles in the regulation of Shh release from Bosc23 cells. Both findings suggest that direct Shh–heparan-sulfate interactions represent the main determinant for Shh release. Alternatively, Gpc knockdown or overexpression might affect Shh release by modulating indirect changes in sheddase expression (Purushothaman et al., 2008).

To distinguish between these two possibilities, we analyzed heparan-sulfate-regulated Shh processing under cell-free conditions. Preliminary results revealed that non-lipidated ShhN^C25S was processed in CHO-K1^tg cells but not in pgsA-745^tg cells (supplementary material Fig. S2), suggesting that heparan-sulfate-regulated Shh processing can also occur independently of Hh membrane association. We confirmed lipid-independent protein processing in detail by comparing cellular and solubilized forms of Shh and C-terminally hexahistidine-tagged ShhN (ShhN–His₆) (Fig. 6A). Shh and ShhN–His₆ were transfected into Bosc23 cells, expressed in [9,10(n)-³H] palmitic acid containing serum-free medium, concentrated by heparin-sepharose pulldown, and blotted onto polyvinylidene difluoride (PVDF) membranes for SDS-PAGE, immunoblotting, and autoradiography analysis of the same blot. As expected, a fraction of soluble 19-kDa Shh generated from the cell-tethered precursor was processed, as indicated by a molecular mass shift and associated loss of the radiolabel (Fig. 6A, left, asterisk). Another soluble protein fraction represented by an upper medium band also lacked the radiolabel, demonstrating that this fraction does not simply represent cellular (lipidated) material that was leaked into the medium or released by some other means (Creanga et al., 2012; Palm et al., 2013). Instead, most of these proteins represent Shh that escaped N-palmitoylation during biosynthesis and were therefore released in an N-terminally unprocessed (but C-terminally processed) form. This result is supported by the observation of processing products obtained from 28-kDa ShhN–His₆. A fraction of this protein was N-terminally processed, as indicated by a small size shift and the absence of the radiolabel (Fig. 6A, right, arrow). Alternatively, ShhN–His₆ was C-terminally processed into 19-kDa ShhN, as indicated by the substantial size shift and the presence of the radiolabel (Fig. 6A, right, arrowhead). Combined N- and C-terminal processing removed both peptides and produced a protein that was indistinguishable from fully processed Shh (asterisks). This finding confirms that neither N-terminal Shh palmitoylation nor C-terminal cholesterylation are essential prerequisites for proteolytic Shh maturation under the conditions used in this assay, allowing us to expand our analysis of ShhN–His₆ processing in vitro.

As stated earlier, we hypothesized that Gpc heparan sulfate chains act as assembly sites for proteases and their substrates, in turn bringing them into close proximity to facilitate substrate turnover. To test this idea, we harvested and centrifuged 19-kDa Shh and 28-kDa ShhN–His₆ expressed into serum-free Dulbecco's modified Eagle's medium (DMEM), and we incubated the soluble proteins with pooled glycosaminoglycans isolated from E12.5–E18 mouse embryos (EHS). EHS was used because of its essential role in Hh signaling in vivo (Grobe et al., 2005; Pallerla et al., 2007; Yasuda et al., 2010). After EHS-conditioned media were incubated for 0–3 h at 37°C to allow for protease–substrate assembly and Shh processing, samples were precipitated with trichloroacetic acid (TCA) and analyzed by SDS-PAGE and immunoblotting as described above. Consistent with our expectations, both proteins were not processed in the absence of exogenous EHS (Fig. 6B, 0 to 3 h), but EHS potentiated Shh and ShhN–His₆ processing by an unidentified Bosc23 protease. The 28-kDa ShhN–His₆ form was converted into 19-kDa proteins lacking C-terminal hexahistidines and the N-terminal Cardin–Weintraub peptides (asterisks). These proteins were comparable in size to processed Shh and bioactive Ihh endogenously secreted from cultured chick chondrocytes (Fig. 6C) (Dierker et al., 2009). This result demonstrates that lipidated and non-lipidated Shh was processed by similar, or the same, proteolytic enzyme(s) (Fig. 6C, asterisk) and that heparan sulfate potentiates processing of both proteins in a direct way.

We next compared biological activities of truncated and unprocessed soluble proteins. Supernatants of ShhN–His₆-expressing Bosc23 cells were incubated with 20 µg/ml heparin, EHS, or mixed heparan sulfate derived from B16-F1 and PC-3 cancer cell lines (CHS) as an approximation of Shh-sensitive pancreatic and skin cancers. We observed that heparin was inactive and that CHS aided the conversion of ShhN–His₆ into an unstable product (Fig. 6D, blot). As shown earlier, EHS enhanced the conversion of 28 kDa ShhN–His₆ into bioactive 19 kDa proteins (Fig. 6D, chondrocyte assay) (Dreier et al., 2008). In this bioassay, supernatants from mock-transfected Bosc23 cells were inactive, as were heparin- and CHS-treated samples. Insulin-like growth factor I (IGF-I) served as a positive control, and EHS alone was inactive. The same was observed for Shh (Fig. 6B,D). We draw two conclusions from these findings. First, that ShhN–His₆ processing increases Shh biofunction (Ohlig et al., 2011), and, second, that ShhN–His₆ processing depends on the source of heparan sulfate (as supported by results shown in supplementary material Fig. S2B). Quantitative disaccharide analysis of heparin-lyase-digested heparan sulfate by liquid chromatography electrospray ionization tandem mass spectrometry (LC-Esi-MS/MS) (Lawrence et al., 2008) revealed that overall EHS sulfation (0.86 sulfates per disaccharide) was increased over that of CHS (0.6 sulfates per disaccharide) (Fig. 6E). Variances in overall charge and disaccharide composition of EHS and CHS are consistent with, and likely the cause of, the observed differences in Shh processing.

Ligand-dependent activation of Hh signaling is associated with tumorigenesis in a subset of epithelial cancers, including colon, pancreatic and ovarian cancer (Theunissen and de Sauvage, 2009), and HSPGs regulate Shh activity and Shh-dependent cancer cell proliferation (Chan et al., 2009). Importantly, Hh does not activate autocrine cancer cell signaling, but instead signals to the stromal compartment (Tian et al., 2009; Yauch et al., 2008), which in turn provides a more favorable environment for tumor growth. RT-PCR analysis of the pancreatic tumor cell line PANC1 confirmed the expression of components required for Shh production and release, but not for Hh reception (supplementary material Fig. S3). This observation is in line with the described shedding of endogenous Shh from the PANC1 cell surface (Damhofer et al., 2015). We confirmed Shh–HA shedding from PANC1 cells by the nonphysiological and physiological release factors MβCD and Scube2, as indicated by enhanced electrophoretic product mobility in comparison to the cell-bound protein precursor (Fig. 7A). Importantly, adding ‘processing-active’ heparan sulfate isolated from CHO-K1 cells and E12.5 mouse embryos to Shh-expressing PANC1 cells also increased Shh–HA processing and solubilization (Fig. 7A). We further observed that 14 µg/ml heparan sulfate isolated from tumor cell lines B16 Bl6, B16 F1, PC3 and HeLa, or heparan sulfate from different mouse embryos, stimulated Shh processing to different extents (Fig. 7B). Heparan sulfate isolated from PC3 and B16 cells, as well as from E12.5 and E14.5 mouse embryos, potentiated Shh shedding from PANC1 cells, whereas HeLa-derived heparan sulfate and heparan sulfate isolated from E14.5 and Sulf-deficient mouse embryos was inactive. Notably, 6-O oversulfated heparan sulfate isolated from Sulf1 and Sulf2 double mutant mouse embryonic feeder cells was also inactive. Heparin and PANC1 heparan sulfate served as negative controls. These findings reveal that heparan sulfate, depending on its source, acts as a new Shh release stimulator in cultured pancreatic cancer cells.

To determine whether specific heparan sulfate sulfations and Shh release are linked, we conducted disaccharide analysis of our heparan sulfate isolates (Fig. 7C,D). We found that overall heparan sulfate sulfation of ‘processing (p)-active’ heparan sulfate isolates varied considerably: 0.8 sulfates per disaccharide for PC3 heparan sulfate, 0.86 sulfates per disaccharide for B16 F1 heparan sulfate, and 0.82 sulfates per disaccharide for B16Bl6 heparan sulfate; 0.73 sulfates per disaccharide for E12.5 heparan sulfate and 0.73 sulfates per disaccharide for E14.5 heparan sulfate. ‘Inert’ HeLa heparan sulfate contained 0.73 sulfates per disaccharide, heparan sulfate from Sulf1- and Sulf2-null mouse embryonic feeder cells contained ∼1 sulfate per disaccharide (Lamanna et al., 2006), and heparin contained ∼2 sulfates per disaccharide. Thus, we found no correlation between overall heparan sulfate sulfation and its ability to stimulate Shh-processing from PANC1 cells. Quantitative disaccharide analysis also failed to reveal any specific ‘signature’ disaccharide in active versus inactive heparan sulfate (Fig. 7C,D).

Tissue patterning and growth in development is a dynamic process that depends on the tight control of spatiotemporal morphogen production, spread and reception. Our work emphasizes the key role of Gpcs in the regulation of Hh release from producing cells and tissues in vitro and in vivo. Gpcs are heparan-sulfate-modified proteins that are bound to the outer surface of the plasma membrane through their GPI anchors. This leads to their association with lipid rafts (Mayor and Riezman, 2004), specialized membrane microdomains that serve as organizing centers for the assembly and trafficking of multiple signaling molecules and their receptors. Notably, because of their N-terminal palmitoylation and C-cholesterylation, Hh proteins also cumulate in lipid rafts (Resh, 2006; Rietveld et al., 1999) and aggregate with Gpcs (Vyas et al., 2008). This suggests that GPI-linked HSPGs on the surface of producing cells act as storage sites for Hh morphogens. Subsequent Hh solubilization depends on the proteolytic processing of lipidated terminal peptides because the extraction of even a single cholesterol molecule from a lipid bilayer requires free energies ranging from 60–90 kJ/mol (Bennett et al., 2009), decisively preventing the discharge of lipidated Hh proteins. Alternative membrane-coupled Hh transport by transcytosis, cytonemes or exosomes is incompatible with the one-way route taken by authentic, dual-lipidated and multimeric Shh to the cell surface but not back into the cell (Gao and Hannoush, 2014).

How is Hh and HSPG assembly linked with subsequent proteolytic Hh release? One attractive idea is that not only do Gpcs serve as platforms for Hh assembly and storage, but they also associate with Hh release factors, such as Hh sheddases (Dierker et al., 2009) or the sheddase activator Scube2 (Jakobs et al., 2014) (Fig. 8). Indeed, we show that sheddase function and specificity can be achieved in two heparan-sulfate-regulated ways. First, extracellular heparan sulfate, depending on its site of production, might activate or attract sheddases to Shh morphogens already bound to these chains. This finding is supported by the observed absence of (N-terminal) Shh processing in heparan sulfate-deficient cells and the potentiation of Shh processing by exogenous heparan sulfate in vitro. Hh sheddase activation by heparan sulfate is further consistent with the known increase of Hh solubilization by Suramine, a polysulfated compound (Etheridge et al., 2010) and with Syndecan 2 (another cell surface HSPG) acting as a docking receptor and activator for MMP7 (Ryu et al., 2009). Moreover, ADAM12 associates with HSPGs through its cysteine-rich domain (Iba et al., 2000), and heparan sulfate binding to ADAM12 transiently activates the enzyme (Sorensen et al., 2008). Finally, heparin activates MMP2 and MMP9 through MT-1 (Butler et al., 1998) and heparan sulfate also activates MMP2 (Koo et al., 2010). Notably, we observed pro-MMP9 (but not pro-MMP2) activation by the same heparan sulfate found to be active in Shh release (U.P., unpublished observations). The soluble protease(s) involved in Shh processing in our assays, however, remain to be identified.

Binding of heparan sulfate chains to the Cardin–Weintraub cleavage site might constitute a second level of processing control (Fig. 8). Consistent with this possibility, heparan sulfate strongly interacts with Shh (Carrasco et al., 2005; Farshi et al., 2011; Zhang et al., 2007) and did not potentiate its processing, and heparin dose-dependently blocked Shh processing and release (supplementary material Fig. S4). Examples for other highly charged, inactive glycosaminoglycans are pgsF-17 heparan sulfate and heparan sulfate isolated from Sulf1- and Sulf2-null mouse embryonic feeder cells, which both show significantly increased heparan sulfate 6-O sulfation (Bai and Esko, 1996; Lamanna et al., 2006). In vivo and in vitro, heparan sulfate 6-O sulfate at the cell surface is removed by the activity of soluble 6-O sulfatases (Sulfs): lack of Sulf activity therefore leads to heparan sulfate oversulfation (enhancing the Cardin–Weintraub interactions with heparan sulfate), whereas Sulf gain-of-function reduces heparan sulfate 6-O sulfation and charge (loosening Hh interactions). Notably, removal of Sulf1 in the posterior Hh-producing Drosophila wing compartment significantly impaired Hh signaling (Zhang et al., 2007), and concomitant Sulf1 overexpression extended the Hh gradient (Wojcinski et al., 2011). In the vertebrate neural tube, Shh-producing cells must also upregulate Sulf1 to reach their full inductive potential, confirming that Sulf1 stimulates Shh activation by decreasing heparan sulfate 6-O sulfation at the ligand source (Oustah et al., 2014). We support these observations and suggest that they can be explained by Sulf-regulated competitive heparan sulfate binding to the Cardin–Weintraub sheddase target site, which in turn impairs Hh processing and release.

Overall, our work, by characterizing Shh solubilization in response to HSPG expression and heparan sulfate modifications, reveals major Gpc functions in the switch between morphogen precursor accumulation at the cell surface and its release in processed form. To a large extent, this direct regulation depends on the charge of heparan sulfate and its domain structure controlling substrate accessibility. This raises the very important, yet poorly investigated question of how cells regulate heparan sulfate charge and structure because, unlike in the case of DNA, RNA and protein biosynthesis, the assembly of heparan sulfate is not template driven. Rather, heparan sulfate assembly seems to rely on ‘ad hoc’ integration of environmental clues to increase or decrease the frequency of binding sites, for example, through the expression of Sulf proteins. This raises the exciting possibility that the inherent, template-independent plasticity of heparan sulfate might also regulate sheddase activities and their substrate specificities in other developmental settings and in pathological conditions such as inflammation, cancer, rheumatoid arthritis and Alzheimer disease.

Fly RNAi strains UAS-sfl-RNAi, UAS-dally-RNAi, UAS-dlp-RNAi, and UAS-ttv-RNAi were obtained from the Bloomington Drosophila Stock Center or from the Vienna Drosophila RNAi Center (Vienna Stock Center: dally3 #v14136, dlp2 #v100268; Bloomington Stock Center: dally1 #33952, dally2 #28747, dlp1 #34091). UAS-GFP-RNAi flies and en-GAL4 flies were kindly provided by Christian Klämbt, University of Münster, Germany. UAS-RNAi fly lines were crossed with en-Gal4 to suppress target gene expression under engrailed control in the Hh-expressing posterior compartment. Flies were maintained at 25°C, and RNAi expression experiments were conducted at 29°C. Right wings of adult F1 flies were mounted in Hoyers medium, photographed, and analyzed by MoticImage. Phenotypic changes in Drosophila wing patterning were quantified by dividing the L3–L4 intervein area by the L2–L3 intervein area (L3–L4/L2–L3).

Shh constructs were generated from murine cDNA (NM 009170) by PCR using primers for the wild-type gene or primers carrying the desired point mutations and tag sequences. Secreted lipidated Shh (nucleotides 1–1314, corresponding to amino acids 1–438) was generated in Bosc23 cells [a human 293 T derivative, an embryonic kidney cell line (Zeng et al., 2001)], PANC1 cells (Damhofer et al., 2015) and CHO cells affected in their ability to synthesize heparan sulfate. Secreted unlipidated ShhN (nucleotides 1–594, corresponding to amino acid 1–198 of murine Shh) was also produced. PCR products were ligated into pGEM (Promega), sequenced and subsequently released and religated into pcDNA3.1 (Invitrogen) for the expression of lipidated 19-kDa Shh, or into pcDNA3.1/myc-HisC (Invitrogen) for the expression of soluble, C-terminally hexahistidine-tagged 28-kDa ShhN–His₆ (the large size is due to the presence of a c-Myc and intervening cloning sequence). Gpcs and Shh were cloned into pIRES for bicistronic expression of Shh (CAP dependent) and Gpc (CAP independent) in the same cells.

Bosc23 cells and PANC1 cells were cultured in DMEM (Gibco BRL) supplemented with 10% fetal calf serum (FCS) and 100 µg/ml penicillin-streptomycin. Cells were transfected by using PolyFect (Qiagen). Cells were grown for 36 h, washed with phosphate-buffered saline (PBS), and incubated in serum-free DMEM for various times in the presence or absence of stimulators or inhibitors of shedding, followed by ultracentrifugation at 210,000 g for 60 min to remove membrane-tethered Shh. Serum-free supernatants were TCA precipitated. Endogenously produced Ihh was isolated from chondrocytes derived from the cranial third of 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 (IGF-I) (Dreier et al., 2008). All procedures and protocols met the guidelines for animal care and experiments in accordance with national and European (86/609/EEC) legislation. Several CHO cells with defects in proteoglycan synthesis (pgs) were included in this study and compared with wild-type CHO K1 cells to analyze the effect of HSPGs on Shh. Lacking functional xylosyltransferase, the CHO cell line pgsA-745 is devoid of glycosaminoglycans (heparan sulfate and chondroitin sulfate). CHO pgsF-17 cells are defective in 2-O-sulfation of iduronic acid and GlcUA, which increases 6-O-sulfation (Bai and Esko, 1996). CHO pgsD-677 cells are defective in heparan sulfate, but not chondroitin sulfate biosynthesis (Lidholt et al., 1992), leading to increased compensatory chondroitin sulfate expression on the cell surface. Where indicated, CHO cell lines stably expressed Hhat to increase N-palmitoylation of recombinant Shh. Comparable Hhat expression was confirmed by RT-PCR. CHO cells were cultured in medium containing 50% DMEM and 50% DMEM/Hams F-12 (Lonza) in the presence of 10% FCS and 100 μg/ml penicillin-streptomycin. Where indicated, 600 μg/ml MβCD was added to serum-free medium. CHO cells were transfected by using PolyFect (Qiagen). At 24 h after transfection, cells were washed once with 1×PBS, and 1 ml of Scube2-conditioned serum-free medium was added to the cells. Media were harvested after another 24 h and processed. Based on the established role of Scube2 glycoproteins (signal sequence, cubulin domain, epidermal growth factor 2) in Hh release from 293 T cells (Tukachinsky et al., 2012), HEK293S cells (Creanga et al., 2012), and Bosc23 cells (Jakobs et al., 2014), Scube2-conditioned medium was used in all Shh release experiments.

Proteins were resolved by 15% reducing SDS-PAGE and western blotted. Antibodies against ShhN (polyclonal goat IgG; R&D Systems), Cardin–Weintraub (R&D Systems), or HA were used for primary detection, and horseradish peroxidase (HRP)-conjugated donkey anti-goat IgG (detecting anti-ShhN), goat anti-rabbit IgG (detecting anti-Cardin–Weintraub), and goat anti-mouse IgG (detecting anti-HA) were used for secondary detection, followed by chemiluminescent analysis (Pierce) and signal quantification by ImageJ (http://rsb.info.nih.gov/ij/). In all Gpc coexpression analyses, data were plotted as released versus cell-bound Shh relative to the value obtained with the control, which was set to 100%. Radiolabeling was conducted as described previously (Ohlig et al., 2011). Briefly, 1 mCi [9,19(n)-³H] palmitic acid was added to Shh-transfected Bosc23 cells in serum-free medium at 24 h after transfection, and cultured for an additional 40 h. [³H]palmitic-acid-labeled cells were lysed, and cells and media were subjected to heparin–Sepharose pulldown to improve the signal-to-noise ratio. After SDS-PAGE and western blotting, PVDF-immobilized proteins were first immunodetected and membranes subsequently autoradiographed.

Shh and HA-tagged Gpc-transfected cells were fixed and incubated at 4°C overnight with antibodies against Shh and the HA-tagged Gpc proteins. They were visualized by using a LSM 510 confocal microscope (Carl Zeiss, Jena, Germany). Thresholded Pearson's correlation coefficients were determined by Volocity.

For RT-PCR analysis in Bosc23 and PANC1 cells, RNA was isolated by using TriZol reagent (Invitrogen) and cDNA was generated with the First-Strand cDNA Synthesis Kit (Fermentas). Primer sequences used for PCR can be provided upon request.

Tissues or cells were digested overnight with 12 mg/ml pronase in 320 mM NaCl and 100 mM sodium acetate (pH 5.5) at 40°C, diluted 1:3 in water, and applied to 2 ml DEAE Sephacel columns. For disaccharide analysis, proteins attached to the glycosaminoglycans were β-eliminated overnight at 4°C (5 M NaOH, 10 M NaBH₄), neutralized with acetic acid, and applied to a PD-10 (Sephadex G25) column (Pharmacia, Uppsala, Sweden). Columns were pre-equilibrated with 25 ml 10% ethanol and a 2.5 ml sample was loaded per column. Glycosaminoglycans were eluted with 10% ethanol, lyophilized, and resuspended in double-distilled water. For compositional disaccharide analysis, 10 µg glycosaminoglycan samples were first digested by using heparin lyases I, II and III (1.5 mU each in 100 µl reactions) (IBEX, Montreal, Canada) at 37°C for 1 h, and the resulting disaccharides were separated from undigested material by using a 3 kDa spin column (Centricon, Bedford, MA). Analysis was then carried out by liquid chromatography–mass spectrometry (LC-MS). Disaccharides were separated on a C18 reverse phase column (0.46×25 cm) (Vydac) with the ion-pairing reagent dibutylamine (Sigma), and eluted species were evaluated with a quantitative mass spectrometric method (Lawrence et al., 2008).

At 40 h post-transfection, supernatants from Shh and mock-transfected Bosc23 cells were harvested and added to C3H10T1/2 cells. Cyclopamine (2.5 µM), a specific inhibitor of Shh signaling, and the ShhN-neutralizing antibody 5E1 (Ericson et al., 1996) were added to confirm specificity of the assay. At 5–6 days after induction, cells were lysed (20 mM Hepes, 150 mM NaCl, 0.5% TritonX-100, pH 7.4) and Shh-induced alkaline phosphatase activity was measured after addition of 120 mM p-nitrophenolphosphate (Sigma) in 0.1 M glycine buffer (pH 10.4). Chondrocytes were isolated from the cranial third of 17-day-old chick embryo sterna and cultured for 14 days in agarose suspension culture under serum-free conditions in the presence (positive control) or absence of 100 ng/ml IGF-I. Shh-induced chondrocyte hypertrophy resulted in a switch from collagen II production to collagen X release.

All statistical analysis was performed in Prism by using a Student's t-test (two-tailed, unpaired, confidence interval 95%). All error estimates are s.d.

We thank Jeff Esko (UC San Diego, USA) for CHO cells; Dorothee Spillmann (Uppsala University, Sweden) for confirming heparan sulfate-mutant CHO-cells identities; Thomas Dierks (University of Bielefeld, Germany) for Sulf-deficient cells and tissues; Ruey-Bing Yang (Academia Sinica, Taiwan) for human Scube2; and Eva Korpos, Jeannine Wegner and Sabine Kupich for supporting this project.