Heparan Sulfate and Transglutaminase Activity Are Required for the Formation of Covalently Cross-linked Hedgehog Oligomers*

Sonic hedgehog (Shh) signaling plays major roles in embryonic development and has also been associated with the progression of certain cancers. Here, Shh family members act directly as long range morphogens, and their ability to do so has been linked to the formation of freely diffusible multimers from the lipidated, cell-tethered monomer (ShhNp). In this work we demonstrate that the multimeric morphogen secreted from endogenous sources, such as mouse embryos and primary chick chondrocytes, consists of oligomeric substructures that are “undisruptable” by boiling, denaturants, and reducing agents. Undisruptable (UD) morphogen oligomers vary in molecular weight and possess elevated biological activity if compared with recombinant Sonic hedgehog (ShhN). However, ShhN can also undergo UD oligomerization via a heparan sulfate (HS)-dependent mechanism in vitro, and HS isolated from different sources differs in its ability to mediate UD oligomer formation. Moreover, site-directed mutagenesis of conserved ShhN glutamine residues abolishes UD oligomerization, and inhibitors directed against transglutaminase (TG) activity strongly decrease the amount of chondrocyte-secreted UD oligomers. These findings reveal an unsuspected ability of the N-terminal hedgehog (Hh) signaling domain to form biologically active, covalently cross-linked oligomers and a novel HS function in this TG-catalyzed process. We suggest that in hypertrophic chondrocytes, HS-assisted, TG-mediated Hh oligomerization modulates signaling via enhanced protein signaling activity.

Hedgehog (Hh) 4 family members are involved in tissue patterning and progenitor cell proliferation by activation of distinct target genes in a concentration-dependent manner. In vertebrates, three closely related Hh morphogens (Sonic hedge-hog (Shh), Indian hedgehog (Ihh), and Desert hedgehog) have been described, and a single ortholog is expressed in Drosophila melanogaster. Production of active morphogen begins with the cleavage of the signal sequence followed by autocatalytic cleavage of the 45-kDa precursor molecule to yield a 19-kDa N-terminal signaling domain. This domain becomes C-terminal-cholesterol-modified during processing and N-terminal-palmitoylated, resulting in a dual-lipidated molecule tightly bound to the surface of producing cells that constitutes the active morphogen (called ShhNp, if derived from the Shh precursor, or IhhNp, if derived from the Ihh precursor) (1). On the (Drosophila) cell surface, lipidated morphogen monomers are organized into suboptical oligomers that are further recruited to preexisting heparan sulfate proteoglycan (HSPG)-rich membrane subdomains to form large, visible multimeric clusters (2). Morphogen release from the cell surface then depends on the expression of Dispatched (3) and ADAM (a disintegrin and metalloprotease) family members. The latter mediate morphogen shedding in an HSregulated way, as has recently been shown for ShhNp in transfected Bosc23 cells, a HEK 293T-derived cell line (4).
In addition to HS-regulated ShhNp shedding, HS is also involved in the formation of the Hh extracellular gradient, which in the fly depends on the HS co-polymerase Tout-velu (Ttv, exostosin in vertebrates) and the glycosylphosphatidylinositol (GPI)-linked HSPGs Dally and Dally-like, corresponding to vertebrate glypicans (5)(6)(7). Exostosins synthesize the HS (GlcA1,4GlcNAc1,4) n carbohydrate backbone, which is 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 (8). In this work we show a new HS function in hedgehog signaling; that is, the HS-dependent formation of oligomeric substructures stabilized by transglutaminase-catalyzed cross-links.
Eight transglutaminases (TGs) are encoded in the human genome, and at least two have been shown to be secreted forms that participate in extracellular matrix remodeling; they are tissue transglutaminase 2 (TG2) and factor XIIIa (fXIIIa). TGs catalyze covalent ⑀-(␥-glutaminyl)lysyl interprotein cross-links in a Ca 2ϩ -dependent manner, resulting in the generation of supermolecular structures with extra rigidity and resistance against proteolytic degradation (for review, see Ref. 9). TG activity is involved in the development of the heart, lung, and the central nervous system in addition to salivary gland devel-opment, blood clotting, and other processes. The two extracellular TGs, TG2 and fXIIIa, undergo up-regulation, which is physiological in growth plate hypertrophic chondrocytes and pathological up-regulation in osteoarthritic cartilage (10). TGmediated cross-linking of in vitro substrates amyloid ␤-A4 (11), ␣-synuclein (12), the microtubule-associated Tau protein (13), and myelin basic protein have also been implicated in the pathogenesis of Alzheimer disease, Parkinson disease, and progressive suprabulbar palsy in which the abnormal accumulation of insoluble proteinaceous aggregates causes progressive neuronal death (14). Here, we identified members of the Hh family as new targets for TG cross-linking activity, adding new substrates to the extensive list of TG-cross-linked extracellular proteins.

EXPERIMENTAL PROCEDURES
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. In some assays ShhN (resulting in a non-cholesterol-modified but biologically active 19-kDa molecule) was expressed instead of ShhNp (resulting in the biologically active, lipidated 19-kDa morphogen that undergoes multimerization upon secretion to the cell surface) to yield sufficient protein for biochemical analysis. PCR products were subcloned into pGEM (Promega), sequenced, and subsequently cloned into pcDNA3.1/myc-HisC (Invitrogen) for expression in Bosc23 and B16-F1 cells and into pFastBac (Invitrogen) for expression in Sf9 cells and into pGEX4T-1 (Amersham Biosciences) for expression in Escherichia coli. A C-terminal His 6 tag was added to some constructs, resulting in the generation of non-lipidated 28-kDa ShhN His6 (the large size is due to the presence of an additional Myc tag and intervening cloning sequence). Secreted, lipidated ShhNp (nucleotides 1-1314, corresponding to amino acids 1-438) was generated in human Bosc23 cells, and secreted, unlipidated ShhN (nucleotides 1-594, corresponding to amino acids 1-198 of murine Shh) was generated in Bosc23 or mouse melanoma B16-F1 cells. C-terminal-truncated proteins were also generated that were C-terminal-fused to a GPI target sequence derived from human CD55 (containing no glutamine residues, accession number NP 001108224) to allow for the generation of high local concentrations of ShhN peptides on the surface of transfected cells.
Cell Culture, Protein Purification, and Analysis-Human Bosc23 and mouse melanoma B16-F1 cells were cultured in DMEM (Invitrogen) supplemented with 10% fetal calf serum (FCS) and 100 g/ml penicillin/streptomycin and were transfected with plasmids encoding the secreted forms of ShhN and ShhNp using PolyFect (Qiagen). Cells were grown for 36 -48 h, and the medium was harvested and ultracentrifuged at 210,000 ϫ g for 60 min to remove proteins bound to membranous remnants. Proteins were then trichloroacetic acid-precipitated or subjected to heparin-Sepharose (Sigma) pulldown followed by three washing steps in phosphate-buffered saline and analyzed by SDS-PAGE. Where indicated, proteins were not eluted from the heparin beads; instead, the beads were mixed with SDS sample buffer, boiled, and briefly centrifuged, and the sample was loaded onto the gel. Chondrocytes were isolated from the cranial third of 17-day-old chick embryo sterna and cultured in agarose suspension cultures in DMEM supplemented with 100 g/ml penicillin/streptomycin and 25 mM ␤-aminopropionitrile, a lysyloxidase inhibitor, under serumfree conditions in the presence or absence of 100 ng/ml insulinlike growth factor I (IGF-I) or 25 ng/ml 3,5,3Ј,5Ј-tetraiodothyronine (thyroxin, T4) for 9 -14 days. This resulted in the secretion of endogenous, lipidated hedgehog; as it is unknown whether IhhNp was exclusively produced or whether ShhNp was also present, the chondrocyte-generated Hh is referred to as HhNp. Sf9 cells from the ovarian tissue of Spodoptera frugiperda (German Collection of Microorganisms and Cell Cultures, DSMZ) were grown in Grace's insect medium (Invitrogen) supplemented with 10% FCS and 10 g/ml gentamycin. For intracellular expression of ShhN, cells were infected using the Bac-to-Bac baculovirus system (Invitrogen). E. coli BL21 cells (Stratagene) were grown in LB medium containing 100 g/ml ampicillin. To induce the formation of multimers, proteins were incubated with heparin sodium salt (100 g/ml, Sigma), chondroitin sulfate sodium salt (100 g/ml, Sigma), dextran sulfate (100 g/ml, Sigma), and heparan sulfate fractions isolated from mouse embryos and coupled to Affi-Gel beads (Bio-Rad).
Recombinant ShhN proteins were analyzed by fast protein liquid chromatography (Ä kta Protein Purifier (GE Healthcare)) using HisTrap columns for the enrichment of Sf9-expressed proteins or a Superdex200 10/300 GL column for gel filtration analysis equilibrated with phosphate-buffered saline at 4°C. Eluted fractions were trichloroacetic acid-precipitated before being subjected to SDS-PAGE. Proteins were analyzed by boiling in SDS sample buffer containing 2% SDS, 100 mM dithiothreitol followed by reducing SDS-PAGE and Western blotting. Monoclonal antibody 5E1 that binds to biologically active ShhN/ShhNp and IhhN/IhhNp was used for the detection of polyvinylidene difluoride-bound hedgehog as well as to block its biological function in differentiation assays (Developmental Studies Hybridoma Bank, Iowa City, IA). We also used ␣-ShhN (polyclonal goat IgG; R&D Systems) that detects biologically active and inactive forms of HhN and HhNp. Tagged proteins were detected by anti-histidine (␣-4xh, Qiagen) and anti-Myc (␣-Myc, Roche Applied Science) monoclonal antibodies. Secondary detection was performed by incubation with peroxidase-conjugated donkey-␣-goat IgG (detecting anti-ShhN) or goat-␣-mouse IgG (detecting 5E1, ␣-4xh and ␣-Myc, all Dianova) followed by chemiluminescent detection (Pierce).
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 of DEAE Sephacel columns. For disaccharide analysis, proteins attached to the glycosaminoglycans were ␤-eliminated overnight at 4°C (0.5 M NaOH, 1 M NaBH 4 ), neutralized with acetic acid, and applied to a PD-10 (Sephadex G25) column (GE Healthcare). Glycosaminoglycans were lyophilized, purified on DEAE as described above, applied to a PD-10 column, and again lyophilized. The samples were then digested using heparin lyases I, II, and III, and the resulting disaccharides were separated from undigested chondroitin sulfate using a 3-kDa spin column (Centricon, Bedford, MA) followed by HPLC analysis using Carbopac PA1 columns (Dionex, Sunnyvale, CA). For the production of HS-beads, HS isolated from mouse embryos at various embryonic (E) stages was Pronase-digested and DEAE-purified as described above and subsequently coupled to AffiGel beads (Bio-Rad) via the HS-associated peptides according to the manufacturer's instructions.
Differentiation of C3H10T1/2 Osteoblast Precursor Cells-C3H10T1/2 cells were grown in DMEM supplemented with FCS and antibiotics as described above. Post-transfection, Shh and mock-transfected Bosc23 cells were cultured in DMEM, 10% FCS for at least 30 h. Conditioned media were then sterilefiltered, mixed 1:1 with Opti-MEM containing 10% FCS and antibiotics, and applied to C3H10T1/2 cells in 15-mm plates. Chondrocyte culture supernatants were sterile-filtered, and FCS was added to a final concentration of 10%, mixed 1:1 with DMEM, 10% FCS containing antibiotics, and subsequently used for differentiation. To some samples 2.5 M cyclopamine was added, which specifically blocks Hh signaling via binding to the Hh signaling molecule Smoothened (15). Additionally, 1 g/ml 5E1 was added to the medium to inhibit Hh signaling via binding of the morphogen, effectively blocking HhN interaction with its receptor Patched (16,17). Cells were lysed 5 days after induction (20 mM Hepes, 150 mM NaCl, 0.5% Triton X-100, pH 7.4), and alkaline phosphatase activity was measured at 405 nm after the addition of 120 mM p-nitrophenol (Sigma) in 0.1 M glycine buffer, pH 10.4. Assays were performed in triplicate. Statistical analysis was performed in Excel using Student's t test (two-tailed, unpaired). All values shown in the text and Figs. 2 and 3 are ϮS.D.
Mass Spectrometric Analysis of ShhN Oligomers-ShhN was expressed in Sf9 cells, and the lysate was incubated with heparin-Sepharose beads. The pulled-down morphogen was then applied to SDS-PAGE, and Coomassie-stained bands corresponding to immunoreactive bands were excised. Tryptic digest and ESI-MS/MS analysis followed by MASCOT data base searches identified ShhN peptides in the oligomeric structure. We conducted 2 independent analyses with identical results, resulting in the identification of 13 ShhN peptides (Probability-based Mowse score 200) and also 7 peptides from mouse Ihh (due to 88% sequence identity between both proteins). Peptides KLTPLAYK (amino acids 39 -46), QFIPN-VAEK (47-55), ELTPNYNPDIIFK (76 -88), ELTPNYNPDIIFK-DEENTGADR (76 -97), LNALAISVMNQWPGVK (107-122), AVDITTSDR (146 -154), AVDITTSDRDR (146 -156), and LAVEAGFDYVYYESK (165-179) were detected in both runs. Notably, due to the high abundance of lysine and arginine residues in ShhN, most of the remaining sequence was fragmented into peptides of less than 5 amino acids, not included in the data base search. Thus, the identified peptides covered Ͼ49% of total protein and Ͼ66% of protein excluding the small peptides.
Probably for the same reason, the oligomer linkage site was not identified.
Computation-The murine ShhN crystal structure (PDB code 1VHH) was displayed using the SwissPdbViewer.

Chondrocytes Secrete Stable Morphogen Oligomers That
Assemble into Higher Order Labile Multimers-Previous studies have demonstrated the presence of ShhNp hexamers that dissociated at high concentrations of salt and detergent (18,19). In contrast to those "labile" multimers formed from the monomeric morphogen, cultured primary chick chondrocytes endogenously secrete stable morphogen oligomers that were resistant to boiling in the presence of 2% SDS, 100 mM dithiothreitol and could, thus, clearly be distinguished from the monomer by standard denaturing SDS-PAGE (Fig. 1, A and B). As it is unknown whether IhhNp was exclusively produced or whether ShhNp was also present, the chondrocyte-generated oligomers are referred to as HhNp oligomers. Furthermore, we termed these oligomers undisruptable (UD) HhNp, and the term UD "oligomers" will be used throughout this paper to distinguish these stable forms from the morphogen multimers described previously (18,19), which do not resist dissociation after treatment with denaturants, including SDS. Immunoblotted UD HhNp oligomers were detected by polyclonal anti-ShhN antibodies as well as by monoclonal, conformation-dependent 5E1 antibodies, the latter being commonly used to detect the biologically active morphogen. The production of UD HhNp oligomers was enhanced by the addition to the medium of 100 ng/ml IGF-I or 25 ng/ml thyroxin (T4). The addition of IGF-I resulted in the generation of the 19-kDa monomer of HhNp as well as three species of UD HhNp-oligomers with apparent molecular masses of about 75-, 120-, or 180-kDa, corresponding to tetrameric, hexameric, or decameric morphogen oligomers, respectively (Fig. 1A). In the presence of T4, the cells generated almost exclusively the 75-kDa species of UD HhNp (Fig. 1B). Under all conditions, however, the 75-kDa HhNp oligomers were the predominant form detected in our assays. Protein aggregation was not a consequence of trichloroacetic acid precipitation or SDS-PAGE because UD oligomers were not detected after trichloroacetic acid precipitation of monomeric ShhNp secreted from transfected human Bosc23 cells. Only ShhNp monomers of 19 kDa were apparent that were also detected upon 5E1 immunoprecipitation (Supplement 1).
Based on previous reports demonstrating the presence of stable ShhNp oligomers in mouse embryos (20), we also analyzed proteins derived from embryonic day (E) 10.5 and E11.5 mouse embryos homogenized in SDS sample buffer or from cultured embryonic cells derived from such embryos. As shown in Fig.  1C, 5E1 immunoblotting revealed the presence of 75-kDa UD oligomers in addition to a 45-Da protein, likely representing the precursor molecule. In contrast, no monomeric morphogen was detected.
We next asked whether UD oligomers additionally assemble into labile higher order multimers comparable to labile ShhNp complexes described by others (18,19). To answer this question, we fractionated chondrocyte-conditioned medium over a Superdex200 gel filtration column. The various fractions elut-ing from the column were analyzed by immunoblotting after reducing 12% SDS-PAGE. Both polyclonal ␣-ShhN and monoclonal 5E1 antibodies detected labile Ͼ600-kDa HhNp multimers that were dissociated by SDS treatment into SDS-resistant 75-kDa oligomers (Fig. 1D, arrow). We also detected a minor fraction of UD HhNp oligomers eluting in fractions around 70 kDa. Very small amounts of 19-kDa HhNp monomers (Fig. 1D, arrowhead) apparently failed to form any higher order oligomer or labile multimer. Thus, gel filtration confirmed the presence of chondrocyte-produced UD oligomers consistent with our previous findings and confirmed that UD complex formation was no SDS-PAGE-induced artifact. These results also show that Hh proteins were endogenously expressed as labile HhNp multimers consisting of UD oligomers, which probably are stabilized by covalent cross-linking.
We next confirmed that chondrocyte-secreted UD oligomers were not residual labile multimers left behind after inadequate denaturation and/or reduction of disulfide-bonds. As shown in Figs. 2, A and B, even prolonged (up to 60 min) boiling in Laemmli buffer or denaturation by boiling for 5 min in sample buffer containing 6 M urea or guanidine hydrochloride failed to disrupt chondrocyte-produced UD HhNp oligomers. To test for the biological activity of these UD oligomers, we took advantage of a sensitive cell-based bioassay, the differentiation of C3H10T1/2 osteoblast precursor cells (21). Conditioned media derived from unstimulated primary chondrocytes after 9 days of culture and 14 days of culture that contained no monomeric HhNp (Fig. 1A, left lane) induced the differentiation of C3H10T1/2 cells into alkaline phosphataseproducing osteoblasts, demonstrating biological activity of the UD oligomers, consistent with their 5E1 reactivity (Fig. 2C). To verify that the C3H10T1/2 differentiation was due to HhNp activity, we used the teratogen cyclopamine (CA), an inhibitor of Hh-dependent signal transduction (15,22), and the neutralizing anti-Shh antibody 5E1 (16). Indeed, UD FIGURE 1. Endogenously secreted HhNp forms SDS-PAGE-resistant oligomers that form higher order, labile multimers. Immunoblot analysis of chondrocyte-secreted HhNp protein is shown. Chondrocytes isolated from chick sternum were cultured in agarose culture in serum-free medium for 9 -14 days. HhNp-containing media were ultracentrifuged, trichloroacetic acid-precipitated, and analyzed by immunoblotting after reducing 12% SDS-PAGE. HhNp proteins were detected with monoclonal 5E1 or polyclonal ␣-ShhN antibodies. A, in medium harvested from unstimulated chondrocytes (w/o) after 14 days in culture, the ␣-Shh antibody 5E1 detected secreted 75-, ϳ120-, and ϳ180-kDa oligomers (arrows). No 19-kDa monomeric HhNp was detectable (arrowhead). The addition of 100 ng/ml IGF-I resulted in the production of 19-kDa HhNp protein and increased generation of ϳ120-kDa oligomers. 5E1 reactivity suggests biological activity of the HhNp oligomers. B, using polyclonal ␣-ShhN antiserum, UD HhNp oligomers of ϳ75 kDa were detected upon stimulation with 25 ng/ml thyroxin (T4) after 14 days in culture. C, immunoblot analysis of embryonic day (E) 10.5 and 11.5 mouse embryo lysates is shown. A 45-kDa morphogen (possibly representing the full-length precursor) as well as 75-kDa oligomers were detected. Morphogen monomers were not detected. D, chondrocytes isolated from chick sternum were cultured in agarose culture in serum-free medium in the presence of T4 for 14 days. Conditioned medium from those cells was then ultracentrifuged and fractionated over a Superdex200 gel filtration column equilibrated in phosphate-buffered saline, and the various fractions were analyzed by immunoblotting after reducing 12% SDS-PAGE to confirm the presence of HhNp multimers. Both polyclonal ␣-ShhN and monoclonal 5E1 antibodies predominantly detected Ͼ600-kDa HhNp multimers that were disrupted into ϳ75-kDa oligomers upon SDS-PAGE. A minor fraction of 75-kDa oligomers did not form any higher order, labile multimers. Only very low levels of monomeric HhNp and no labile HhNp multimers formed from the 19-kDa monomer were detected (arrowhead). FIGURE 2. Chondrocyte-expressed HhNp oligomers are SDS/dithiothreitol-resistant and biologically active. Prolonged boiling in reducing Laemmli buffer for 5-60 min (A) and denaturation with sample buffer containing 6 M urea (U) or 6 M guanidine hydrochloride (G.HCl) after the 5 standard minutes of boiling (B) failed to disrupt ϳ75-kDa UD Hh oligomers (arrow), demonstrating unusual stability of the complexed morphogen. The approximate size of monomeric ShhN is indicated by an arrowhead. C, HhNp oligomers are biologically active. C3H10T1/2 osteoblast precursor cells were incubated for 5 days in chondrocyte-conditioned media in the presence or absence of the teratogen CA that specifically inhibits Hh-driven differentiation into alkaline phosphatase-producing osteoblasts. Conditioned media obtained after 9 and 14 days of chondrocyte culture induced alkaline phosphatase activity that was inhibited by CA co-treatment (2.5 g/ml, n ϭ 3, p Յ 0.001) and 5E1 treatment (1 g/ml, n ϭ 3, p Յ 0.001). Differentiation of C3H10T1/2 cells is expressed as relative alkaline phosphatase activity of lysed cells after a 5-day induction, measured at 405 nm after the addition of 120 mM p-nitrophenol phosphate.
HhNp activity was inhibited about 10-fold by the addition of 2.5 M CA or 1 g/ml 5E1 to chondrocyte-conditioned media (p Յ 0.001, n ϭ 3).
UD HhNp Oligomers Show Enhanced Biological Activity-UD HhNp oligomers must have evolved because of specific advantages over the monomeric form. We, thus, asked whether the biological activity of UD HhNp oligomers was increased if compared with monomeric ShhN, as had been described for the labile multimers formed from ShhNp. To answer this question, we analyzed the supernatants of Bosc23 cells expressing cDNAs encoding hexahistidine-tagged (unlipidated) ShhN His6 and ShhNp (Fig. 3A) as well as the supernatant of HhNp-secreting chondrocytes by gel filtration chromatography followed by immunoblot analysis (Fig. 3B). Consistent with the findings of others (18,19), ShhN His6 only occurred as a monomer in solution. In contrast, ShhNp formed labile multimers consisting of the 19-kDa morphogen and ranging in size from 300 to ϳ10 3 kDa. UD HhNp oligomers formed higher order, labile multimers slightly exceeding 600 kDa in size that consisted of 75-kDa UD oligomers. As shown in Fig. 3C, UD HhNp effectively induced the differentiation of C3H10T1/2 cells, whereas approximately equal or even higher amounts of ShhNp and ShhN (Fig. 3B) were biologically less active. In all cases differentiation of C3H10T1/2 cells was strongly impaired by co-treatment with 2.5 M CA or 1 g/ml 5E1, confirming that UD oligomers rather than undefined factors in the conditioned media were responsible for the enhanced biological activity.
UD Oligomerization Depends on a Highly Conserved N-terminal Glutamine Residue-To determine the mechanism underlying the formation of stable oligomers from morphogen monomers, we investigated ShhN oligomerization after expression in B16-F1 mouse melanoma cells. To allow for high local protein concentrations on the surface of transfected cells and to target the structures into the same membrane microdomains described for the lipidated morphogen (23), individual peptides were C-terminal-linked with the human CD55 GPI recognition sequence. As shown in Fig. 4A, several truncated Shh peptides readily formed high molecular weight oligomers. The shortest peptide still able to form oligomers (most likely tetramers) in this system ranged from amino acids 25 to 52 (after processing of the signal sequence, amino acids 1-24) (Fig. 4B). Comparison of this amino acid sequence with various corresponding peptides of vertebrate and invertebrate Hhs revealed three notable areas of high sequence conservation; that is, a block of conserved residues required for N-terminal acylation (24) ranging from 25 to 30 in the mouse Shh nomenclature, the highly conserved heparin binding Cardin-Weintraub (CW) sequence (amino acids 33-39)(25), and a third block of unknown function, including an absolutely conserved glutamine residue in position 47 (Gln-47). Molecular modeling using the ShhN crystal structure (26) revealed that Gln-47 was situated on the N-terminal, extended domain of the morphogen (Fig. 4C). This location is consistent with Gln-47 serving as a substrate of TG enzymes (27) that recognizes glutamine residues at the surface or, more generally, within terminal extensions protruding from compactly folded protein domains. Thus, we hypothesized this residue to be a TG target, resulting in the formation of ⑀-(␥glutaminyl)lysyl cross-links with an undefined lysine residue present on a second protein. Consistent with this idea, a lysine residue adjacent to Gln-47 also is absolutely conserved and always is preceded by uncharged and aliphatic residues. Embedding of lysine residues into such sequences has been reported to enhance their reactivity as TG substrates (28). Conditioned media from ShhN His6 -or ShhNp-transfected cells (A) or from cultured chondrocytes were ultracentrifuged and fractionated over a Superdex200 gel filtration column equilibrated with phosphate-buffered saline, and the various fractions were analyzed by immunoblotting after reducing 12% SDS-PAGE to detect the monomeric morphogen and UD oligomers as well as labile morphogen multimers. B, conditioned medium from ShhN His6 -transfected Bosc23 cells revealed the exclusive presence of tagged, 28-kDa monomeric morphogen. In contrast, ShhNp formed labile multimers of Ͼ300 kDa. Chondrocyte-secreted HhNp also formed labile, higher order multimers from 75-kDa UD oligomers. C, HhNp UD oligomers show elevated biological activity if compared with monomeric ShhN and ShhNp multimers. 19-kDa ShhNp, 75-kDa UD HhNp, and 28-kDa ShhN His6 were added to C3H10T1/2 osteoblast precursor cells and incubated for 5 days. To verify that the biological activity in the conditioned media was due to Hh activity, two inhibitors of Hh activity were again used: the teratogen CA and the monoclonal, inhibitory antibody 5E1. Conditioned media obtained after 14 days of chondrocyte culture induced strong alkaline phosphatase activity in C3H10T1/2 cells that was effectively inhibited by CA co-treatment (2.5 g/ml, n ϭ 3, p Յ 0.001) and 5E1 co-treatment (1 g/ml, n ϭ 3, p Յ 0.001), demonstrating strongest biological activity of multimeric UD HhNp oligomers. The biological activity of ShhN or ShhNp was significantly lower despite a comparable or higher concentration of the morphogen (shown in B). Differentiation of C3H10T1/2 cells is expressed as relative alkaline phosphatase activity of lysed cells after a 5-day induction after subtraction of background alkaline phosphatase expression, measured at 405 nm after the addition of 120 mM p-nitrophenol phosphate.
However, although Gln-47 was located in Hh proteins within a highly conserved motif consisting of apolar amino acids, which was described to allow for efficient substrate recognition by TG (29), this residue was also followed by a proline residue in position ϩ2. The consensus QXXP (whereas X stands for any amino acid) is a poor TG target, at least in gluten peptides in vitro (30). Thus, we first added rTG2 to Sf9-expressed, mutated ShhN lacking all three glutamine residues (Gln-47, -101, and -117, ShhN 3xQ ) (Supplement 2) after heparin-Sepharose pulldown from crude cell lysates. This resulted in a morphogen that, unlike the wild type ShhN control, could not be covalently linked by recombinant TG2 in vitro (Fig. 4D, arrow). Oligomerization was restored in a Sf9-expressed ShhN mutant lacking only Gln-101 and Gln-117 (ShhN 2xQ ), confirming that Gln-47 is part of the isopeptide linkage (Fig. 4E, arrow). Importantly, we confirmed that mutant forms were biologically active (Supplement 3), demonstrating that glutamine mutagenesis did not result in a misfolded protein with altered characteristics. Taken together, we conclude that Gln-47 is the preferred residue for morphogen oligomerization.
To directly test for TG-mediated isopeptide bond formation in UD oligomers, we treated GPI-linked ShhN  , expressed in B16-F1 cells, with 20 mM cystamine (C), an alternative substrate of several TGs preventing isopeptide cross-link formation. As shown in Fig. 4F, protein aggregation was indeed reduced in the presence of the reagent. Next, to demonstrate whether endogenously produced UD oligomers were also linked via TG activity, we incubated cells derived from E11.5 mouse embryos in the presence of 5 and 10 mM cystamine. As expected, treatment resulted in impaired UD oligomer formation in a concentration-dependent manner (Fig. 4G).
We confirmed these results by specific inhibition of chondrocyte HhNp oligomerization. Here, IGF-I-induced formation of UD HhNp oligomers was also abrogated by 20 mM cystamine (Fig. 5A) and reduced by 1-5 M L-682.777, a specific inhibitor of TG2 and fXIIIa (Fig. 5B) or 50 and 100 M dansylcadaverine (D) (Fig. 5C). In contrast, we found that the addition of 1 mg/ml heparin (Fig. 5D) boosted HhNp oligomerization in IGF-I-stimulated, HhNp-expressing chondrocytes. This was particularly the case for UD HhNp oligomers with molecular masses Ͼ75 kDa (Fig. 5D). These data suggest that in some systems such as the Ca 2ϩ -rich environment in bones, TG and HS secreted from hypertrophic chondrocytes cooperate in the formation of biologically active UD oligomers (Supplement 4).
Heparan Sulfate Is Necessary for the Formation of UD Oligomers from Recombinant, Monomeric ShhN-Because the amount of chondrocyte-secreted UD HhNp was not sufficient for biochemical characterization, in vitro reproduction of HSmodulated UD complex formation was attempted using recombinant murine ShhN (amino acids . ShhN shows 92% amino acid sequence identity and 98% identity and similarity to chick IhhN, implying comparable biochemical properties. Based on the results shown in Fig. 5D and following the   . Comparison of the amino acid sequences of mammalian N-terminal Hh peptides and Drosophila Hh reveals the presence of three highly conserved motifs; that is, a block of conserved residues required for N-terminal acylation (24) ranging from amino acids 25 to 30 (Mouse Shh), the highly conserved heparin binding CW sequence (B, basic residue) (25), and a third block of unknown function including an absolutely conserved glutamine (*) residue. Boxes represent absolutely conserved amino acid residues in at least eight gene products, and gray columns indicate the presence of conserved amino acids. C, shown is a ribbon diagram of ShhN (26). The position of the CW sequence and absolutely conserved residues Lys-46 and Gln-47 is indicated. Both residues are located on the N-terminal-extended peptide. D, ShhN (amino acids ) and a mutant lacking glutamine residues Gln-47, -101, and -117 (ShhN 3xQ ) were expressed in Sf9 cells. The lysate was incubated with heparin beads and 2 g/ml tissue transglutaminase 2 (rTG2), pulled down, and analyzed. The mutant protein bound to heparin but, in contrast to the wild type form, did not undergo TGmediated protein dimerization. E, Sf9 cell lysate after expression of mutant ShhN 2xQ lacking glutamine residues Gln-101 and -117 was incubated with heparin beads in the presence or absence of 2 g/ml rTG2. ShhN 2xQ oligomers were detected in the presence of TG2 but not in its absence. F, the addition of 20 mM cystamine (C), a specific inhibitor directed against transglutaminase activity, strongly inhibits aggregation of ShhN  . G, cystamine co-treatment also strongly impairs UD oligomerization of endogenous morphogen expressed in primary cells isolated from E11.5 mouse embryos (arrow). *, 45-kDa Hh precursor.
reasoning that HS is critical for Hh signaling, we first incubated recombinant monomeric forms of ShhN with the highly sulfated glycosaminoglycan heparin. Heparin coupled to Sepharose beads was used to pull down recombinant 19-kDa ShhN monomers that were first enriched by metal ion affinity chromatography on HisTrap columns. Subsequent SDS-PAGE and immunoblotting resulted in the detection of in vitro generated UD ShhN oligomers (Fig. 6A). Importantly, UD protein oligomers were detected by conformation-dependent 5E1 antibodies, supporting the finding that UD HhNp oligomers were biologically active. At this stage, tagged forms of ShhN allowed us to confirm the specificity of ␣-ShhN and 5E1 antibodies for the oligomeric morphogen. As shown in Supplement 5, tagged oligomers were indeed detected by these antibodies as well as by monoclonal antibodies specifically directed against the Myc tag or hexahistidine tag. To directly confirm the identity of in vitro generated ShhN UD oligomers, two bands (of ϳ70 and ϳ120 kDa, arrows) corresponding to 5E1-and ␣-ShhN-positive bands on Western blots were excised from Coomassie-stained SDS-PA gels (Fig. 6A, arrows) and subjected to trypsin digestion. The tryptic peptides were then analyzed by nano-LC-ESI-MS/MS and compared with the MASCOT data base. We conducted 2 independent analyses with identical results, resulting in the identification of 13 ShhN peptides (Probability-based Mowse score 200) and also 7 peptides from mouse Ihh (due to 88% sequence identity between both proteins). Notably, because of the high abundance of lysine and arginine residues, ShhN trypsin digestion resulted in fragmentation into numer-ous small peptides of less than 5 amino acids that were not analyzed. Thus, the identified peptides covered Ͼ49% of total protein and Ͼ66% of protein excluding all small peptides, confirming that both UD oligomers were ShhN-derived. Importantly, no other proteins were co-identified in the excised bands, ruling out the presence of ShhN heteromultimers.
Recombinant, Hexahistidine-tagged 28-kDa ShhN His6 secreted from B16-F1 cells also formed ϳ60-kDa ShhN His6 olig-    was expressed intracellularly in Sf9 cells (ShhN). ShhN was incubated with heparin beads, and the heparin beads as well as the eluate were analyzed by reducing SDS-PAGE and immunoblotting using polyclonal ␣-Shh antiserum and the monoclonal ␣-Shh antibody 5E1. Results of duplicate experiments are shown. Heparin mediates the formation of SDS/dithiothreitol-resistant, Ͼ60-kDa ShhN oligomers (arrows) from monomeric ShhN (arrowhead). The identity of oligomeric ShhN was confirmed by nano-LC-ESI-MS/MS of excised bands from the Coomassie-stained gel (arrows). Note the strong reactivity of ShhN oligomers with antibody 5E1, confirming the biological activity of the morphogen oligomer. B, HS was isolated from embryos of various developmental stages, covalently coupled to Affi-Gel, and incubated with hexahistidine-tagged, B16-F1-expressed monomeric ShhN His6 . HS-bound ShhN His6 was then pulled down, and beads were directly subjected to SDS-PAGE and immunoblotting. Only HS derived from E11.5 embryos induced protein oligomerization, and HS derived from E13.5 and E14.5 embryos bound the monomeric morphogen. C, shown is an analysis of disaccharide composition of HS shown in B. HS was isolated from E11.5-E17.5 embryos, and samples were digested with heparin lyases. The resulting disaccharides were analyzed by anion exchange HPLC. HS from E11.5 embryos was under-sulfated, resulting in increased levels of non-sulfated (D0A0) and N-sulfated (D0S0) disaccharides and a decrease in 6O-sulfated and 2O-sulfated disaccharides, especially D0A6 and D2A6, if compared with HS derived from E13.5-E17.5 embryos. The relative amount of free amino groups (D2H6 and D2H0), however, was strongly elevated in E11. omers (most likely dimers) upon heparin affinity chromatography, as did E. coli-expressed ShhN (not shown). The various UD oligomers formed from monomeric ShhN expressed in different procaryotic and eucaryotic systems indicated that heparin universally mediated formation of UD oligomers from the monomeric morphogen. However, this highly sulfated form of HS is exclusively produced in connective tissue-type mast cells and, thus, cannot be expected to affect UD oligomerization during developmental processes in vivo. Therefore, to evaluate the function of differentially sulfated embryonic HS, HS isolated from E11.5-E17.5 mouse embryos was coupled to Affi-Gel beads. HS beads were then added to monomeric ShhN His6 secreted from B16-F1 mouse melanoma cells, and the HS-bound protein was pulled down and analyzed by immunoblotting. As shown in Fig. 6B, only E11.5 mouse embryo-derived HS induced UD oligomerization. Additionally, ShhN bound strongly to HS isolated from E11.5-E14.5 mouse embryos but not to HS isolated from later embryonic stages, indicating variations in HS-dependent ShhN binding and multimerization during development. To test whether the ability to up-regulate UD ShhN His6 oligomer formation correlates with the degree of HS sulfation, disaccharide analysis after heparin lyase digestion was conducted by anion exchange HPLC (Fig.  6C). Surprisingly, disaccharide analysis of E11.5 HS showed a very low sulfation level (0.2 sulfates per disaccharide) in comparison to E15.5 HS (0.95 sulfates per disaccharide) or heparin (2.5 sulfates per disaccharide, not shown). Additionally, disaccharide analysis of E11.5 HS revealed high levels of free amino groups (D2H0, UA2S-GlcN) but a very low amount of 6-O sulfation (D0A6, UA-GlcNAc6S and D2A6, UA2S-GlcNAc6S). The somewhat surprising finding of low sulfation on E11.5 HS as well as extensive sulfation present on heparin, both mediating morphogen multimerization, may be explained by the use of purified, overexpressed protein in the latter case. This may have resulted in extensive HS-covering by the morphogen monomer, allowing for TG-mediated covalent linkage of adjacent molecules. In the in vivo situation, however, we suggest that UD oligomer formation may depend on limiting specifically sulfated residues and the patterning of HS sulfation. Possibly, the size variation of HhNp UD oligomers secreted from chondrocytes may even be mediated by specific sulfation patterns present on co-expressed, cell surface-tethered HS.

DISCUSSION
The detection of oligomer formation by SDS-PAGE and/or Western blotting and inhibition of protein cross-linking by bifunctional amines is widely used to identify TG protein substrates. In this work we demonstrate by SDS-PAGE the formation of HhNp oligomers starting from the endogenously produced monomers and confirmed the identity of in vitro generated ShhN oligomers by employing two different Hh-specific antibodies in combination with two monoclonal antibodies directed against Myc and hexahistidine tags as well as LC-ESI-MS/MS analysis of in vitro generated ShhN oligomers. Notably, heparin increased HhNp complex formation in chondrocyte cultures, and ShhN effectively formed oligomers in the presence of specific forms of HS. We also made the observation that the formation of large (120 and 180 kDa) oligomers was efficiently inhibited by a range of TG-specific abortive substrates and an inhibitor; however, the 75-kDa form was less affected. Indeed, the incomplete inhibition of UD oligomerization by a range of TG inhibitors is not compatible with the mode of de novo direct polymerization of proteins in a singlephased reaction, which constitutes one type of TG-mediated protein cross-linking reaction. Examples for this type of reaction would be the formation of the vaginal plug after copulation in rodents (31) and lobster hemolymph clotting (32). Notably, another type of known TG action is only incompletely affected by inhibitors or analogue substrates of TGs (9) and can be described as the enzymatic "spot-welding" of non-covalent (e.g. reversible) assemblies. Here, cross-linking occurs in a twophase system that uses a preformed polymer scaffold, and TG contributes only to the stabilization (maturation) of a preformed reversible polymer structure (9). An example for this second type of TG-mediated reaction is the coagulation of blood. After limited proteolysis of fibrinogen by thrombin, fibrin molecules self-assemble into an array of protofibrils and fibrils, ultimately forming a clot network. The cross-linking enzyme, fXIIIa, then introduces a few strategically located bridges into the preformed, labile polymer to stabilize the clot. Analogous to this process, we suggest that by binding to cellsurface HS chains, membrane-tethered, lipidated HhNp proteins or preformed nanocomplexes (2) may first assemble into labile multimers similar or identical to those described previously (Fig. 7). In vivo, HhNp protein binding to specifically sulfated HS "scaffolds" may bring Gln-47 and an unidentified lysine residue provided by another monomer into close proximity, allowing for subsequent TG-mediated isopeptide linkage. This will result in UD oligomer formation within labile multimers.
However, although our model would also predict the formation and release of UD oligomeric HhNp from Drosophila cells, surface plasmon resonance studies showed negligible binding of the fly morphogen to HS but strong binding to heparin (33). In contrast, ShhN bound not only to heparin but (although to a lesser extent) also to porcine intestinal HS that was employed in the study. These observations may reflect a strong preference of fly HhNp for highly sulfated regions or forms of HSPGs possibly related to sequence variations within the HS binding CW motif (murine ShhN, GKRRHPKK; fly HhN, LGRHRARN; CW consensus, XBBBXXBX, B indicates basic amino acid residues (25)). Alternatively, HS binding of fly HhNp may require specific binding motifs that were likely absent from the porcine HS used in the assay (33) but may be expressed on functional fly HS in vivo. This possibility is supported by the finding that Drosophila HhNp colocalizes with HSPGs in vivo and that fly HhNp/HSPG colocalization was abolished upon deletion of the HS binding CW sequence (2). It is further supported by specific HS requirements for efficient ShhN binding and UD oligomerization in vitro, as shown in this work. The requirement for specific HS motifs may, thus, also explain why ShhN or ShhNp expressed in transfected cell lines were only found as monomers or as labile multimers, in contrast to endogenously expressed HhNp isolated from cultured chondrocytes (19). Possibly, heterologous expression systems lack the specifically sulfated HS isoforms required for UD oligomerization and may also lack TG expression, resulting in the generation of only labile morphogen multimers (representing the preformed polymer) (Fig. 7). In contrast, endogenous production of HhNp in vivo may be coordinated with specific HS synthesis and TG secretion, resulting in the regulated formation of covalently cross-linked and biologically active UD oligomers within the preformed polymers. For these reasons, the formation of UD oligomers found in the in vivo situation was undetectable in previous studies, relying on protein expression in transfected cells. Importantly, TG avidly binds to heparin, but heparin binding only slightly affects TG activity in vitro (34). This suggests that UD oligomerization is not merely a result of HS-dependent TG activation.
What may be the function of UD oligomers in specific developmental settings, and what may be their site of expression? Elevated activity of TG2 and fXIIIa have been described as additional features of hypertrophic growth plate chondrocytes (35,36). Notably, the hypertrophic phenotype can be induced by thyroid hormones in growth plate chondrocytes, and T4 also stimulates TG activity in the extracellular matrix of articular chondrocytes (37). We, thus, suggest that at in the Ca 2ϩ -rich environment of the developing bone, TG-producing hypertrophic chondrocytes may secrete UD oligomers with specific advantages over the labile multimer or the monomeric form. These advantages may include the possibility of allosteric control, higher local concentration of active sites, larger binding surfaces, and the generation of novel active sites at the subunit interfaces. Those features may all strongly affect the formation of the extracellular morphogen gradient and interaction with the hedgehog receptor Patched. In this work we show enhanced biological activity of UD oligomers and possibly also of labile multimers over the monomeric form. This may be due to the clustering of multiple molecules of the hedgehog receptor Patched, in turn effectively reducing its interaction with Smoothened. Alternatively, increased biological activity of UD oligomers may be explained by increased protein stability in the extracellular matrix due to resistance against proteolytic degradation, consistent with other TG extracellular matrix substrates (9).
In addition to being involved in normal developmental processes, as shown in mouse embryos or hypertrophic chondrocytes isolated from chick embryos, our findings may also play a role in pathophysiological conditions, such as osteoarthritis FIGURE 7. Model of HS-assisted, TG-mediated HhNp oligomerization. We suggest that in a first step, monomers located on the surface of the producing cell bind to cell surface HS. Non-permissive HS, e.g. HS lacking sulfated motifs allowing for correct morphogen positioning, may result in the formation of labile multimers upon release into the medium (a). Permissive HS, however, may mediate proper HhNp alignment upon HS binding, bringing Gln-47 and a lysine residue of an adjacent monomer into close proximity (b). This represents the first-step formation of labile oligomers in the absence of TG activity (c). In the presence of TG, HS scaffolding may then facilitate covalent linkage between two adjacent monomers, resulting in the generation of UD oligomers (d). Consistent with this model, a low abundance of sulfated NS domains present on the HS chain may favor morphogen oligomerization by increasing the probability of monomers to interact. This possibly explains the ability of E11.5 mouse embryo HS, which contains relatively fewer sulfated binding domains, to mediate UD oligomerization more efficiently than the more highly sulfated HS forms. Thus, the situation depicted in d may be restricted to cell types or tissues that coordinate HS sulfation during biosynthesis and TG expression with endogenous morphogen secretion. The other situation (a-c) may occur in transfected cells lacking permissive HS sulfation and/or TG expression.
(OA). Pathologic hypertrophic differentiation also occurs in articular chondrocytes in OA in situ and has the potential to promote OA progression through pathological calcification (38). Both fXIIIa and TG2 are molecular markers of chondrocyte differentiation in the growth plate (35,39), and TG2 and fXIIIa expression as well as TG activity are up-regulated in human knee OA cartilage chondrocytes (40). Because hypertrophic chondrocytes express TG-linked HhNp oligomers that strongly induced C3H10T1/2 differentiation into osteoblasts in our assays, the findings presented in this work may result in a better understanding of the pathology of OA. In addition to a putative role of UD oligomerization in OA, the novel mechanism suggested here comprising multimerization of Hh proteins by HS scaffold modulation followed by TG-catalyzed cross-linking may also be involved in initiation and progression of neurodegenerative disease. In these scenarios TGs have been implicated in the abnormal accumulation of insoluble and protease-resistant proteinaceous aggregates (14). In addition, HS is co-distributed with the abnormal prion protein, PrP(Sc), even in very early disease stages in scrapie-infected mice (41). This raises the possibility that HS may also serve as a scaffold for protein aggregation in these pathophysiological conditions.