α2β1 Integrin Is Not Recognized by Rhodocytin but Is the Specific, High Affinity Target of Rhodocetin, an RGD-independent Disintegrin and Potent Inhibitor of Cell Adhesion to Collagen*

We have recombinantly expressed a soluble form of human α2β1 integrin that lacks the membrane-anchoring transmembrane domains as well as the cytoplasmic tails of both integrin subunits. This soluble α2β1 integrin binds to its collagen ligands the same way as the wild-type α2β1integrin. Furthermore, like the wild-type form, it can be activated by manganese ions and an integrin-activating antibody. However, it does not bind to rhodocytin, a postulated agonist of α2β1 integrin from the snake venom ofCalloselasma rhodostoma, which elicits platelet aggregation. Taking advantage of the recombinantly expressed, soluble α2β1 integrin, an inhibition assay was established in which samples can be tested for their capability to inhibit binding of soluble α2β1 integrin to immobilized collagen. Thus, by scrutinizing the C. rhodostoma snake venom in this protein-protein interaction assay, we found a component of the snake venom that inhibits the interaction of soluble α2β1 integrin to type I collagen efficiently. N-terminal sequences identified this inhibitor as rhodocetin, a recently published antagonist of collagen-induced platelet aggregation. We could demonstrate that its inhibitory effect bases on its strong and specific binding to α2β1 integrin, proving that rhodocetin is a disintegrin. Standing apart from the growing group of RGD-dependent snake venom disintegrins, rhodocetin interacts with α2β1 integrin in an RGD-independent manner. Furthermore, its native conformation, which is stabilized by disulfide bridges, is indispensibly required for its inhibitory activity. Rhodocetin does not contain any major collagenous structure despite its high affinity to α2β1integrin, which binds to collagenous molecules much more avidly than to noncollagenous ligands, such as laminin. Blocking α2β1 integrin as the major collagen receptor on platelets, rhodocetin is responsible for hampering collagen-induced, α2β1 integrin-mediated platelet activation, leading to hemorrhages and bleeding disorders of the snakebite victim. Moreover, having a widespread tissue distribution, α2β1 integrin also mediates cell adhesion, spreading, and migration. We showed that rhodocetin is able to inhibit α2β1 integrin-mediated adhesion of fibrosarcoma cells to type I collagen completely.

Integrins are cell adhesion molecules that consist of two noncovalently associated subunits, ␣ and ␤ (for review see Refs. 1 and 2). The subfamily of integrins sharing the ␤ 1 subunit are well known receptors for extracellular matrix molecules, such as collagens, laminins, and fibronectin. The subfamily of ␤ 3 subunit containing cytoadhesins comprise the platelet integrin, ␣ IIb ␤ 3 which binds fibrinogen/fibrin (3) and the vitronectin receptor ␣ V ␤ 3 . The latter ones, along with several ␤ 1 integrin, such as the fibronectin receptor ␣ 5 ␤ 1 integrin, recognize a linear arginyl-glycyl-aspartyl sequence within their respective ligands, such as fibrinogen, fibronectin, and vitronectin (4). In contrast, the collagen binding integrins ␣ 1 ␤ 1 and ␣ 2 ␤ 1 recognize arginine and aspartate/glutamate residues of different collagen chains, which are in close proximity to each other within the triple helical collagenous framework of the collagen (5-7), thus forming a completely different spatial structure than the linear RGD peptide.
Integrin-mediated cell adhesion not only anchors the cell mechanically within the extracellular matrix of the tissue but also elicits several cellular responses, such as cell spreading and migration, cell proliferation, and differentiation (for review see Ref. 2). A well studied example of cellular response triggered by integrin-ligand interaction is platelet activation and aggregation (8,9). Thrombocytes abundantly possess the platelet integrin ␣ IIb ␤ 3 on their surface, which unless activated does not bind to fibrinogen/fibrin (3). Ablation of endothelial cells from the blood vessel wall or other injuries of blood vessels make type IV and type I collagen of the basement membrane and the underlying connective tissue, respectively, accessible to platelets. Once getting in contact with collagen, platelets avidly bind to collagen via their collagen receptors (9,10), such as the ␣ 2 ␤ 1 integrin, GPVI, or indirectly via von Willebrand factor, which binds to both collagen and the vWF receptor on the platelet surface. Receptor-mediated adhesion to collagen elicits a cascade of signals within the platelets, which eventually results in secretion of platelet granula, in platelet aggregation and activation of platelet integrin ␣ IIb ␤ 3 which then binds to fibrin with high affinity. Insoluble fibrin, which has been produced by the enzymatic blood clotting cascade and provides a scaffold, which together with platelets form the blood clot as the first and essential step in hemostasis. The key role of the ␣ 2 ␤ 1 as the sole integrin collagen receptor on platelets is drastically manifested in patients with severe bleeding disorders, caused either by a genetic defect or lack of the integrin ␣ 2 subunit (11) or by auto-antibodies against the integrin ␣ 2 subunit (12).
Furthermore, snake, leeches, and ticks have developed natural inhibitors of integrin-ligand interactions, called disintegrins, that target at the integrin-mediated platelet adhesion to fibrinogen/fibrin and collagen (9). By inhibiting blood clotting, their venoms lead to severe bleeding, hemorrhages, or even death of their victims. Besides proteolytic enzymes, disintegrins are mainly responsible for these poisonous effects. Most of the known disintegrins contain a linear RGD sequence placed within a rather flexible loop, which prevents the RGDdependent platelet integrin ␣ IIb ␤ 3 from binding to fibrin (13). However, very little is known about disintegrins that act on the interaction of ␣ 2 ␤ 1 integrin with collagen, the initial step of platelet activation and aggregation. From the venom of the Malayan pit viper (Calloselasma rhodostoma), rhodocytin/aggretin (14,15), and, more recently, rhodocetin (16) have been shown to induce and inhibit, respectively, collagen-elicited platelet activation and aggretion. However, no direct proof was provided that rhodocytin/aggretin and rhodocetin are the agonist and antagonist, respectively, that interact directly and specifically with ␣ 2 ␤ 1 integrin among the different collagen receptors of blood platelets.
Using a recombinantly expressed, soluble human ␣ 2 ␤ 1 integrin, we could rule out that rhodocytin/aggretin binds directly to ␣ 2 ␤ 1 integrin. On the other hand, having established an inhibition assay with the purified soluble ␣ 2 ␤ 1 integrin apart from whole platelets, we could isolate an inhibitor of C. rhodostoma venom that inhibits the binding of soluble ␣ 2 ␤ 1 integrin to collagen on the molecular level. N-terminal sequencing identified this inhibitor to be the lately published rhodocetin (16). We could demonstrate that rhodocetin binds directly to the ␣ 2 ␤ 1 integrin. Rhodocetin efficiently competes with collagen for the ␣ 2 ␤ 1 integrin, even though it does not contain any collagenous triple helix domain, which has been surmised to be a prerequisite for high affinity binding to collagen-binding integrins. Nevertheless, the native conformation that is stabilized by disulfide bridges is essential for binding to ␣ 2 ␤ 1 integrin. In contrast to the majority of snake venom disintegrins, rhodocetin binds to ␣ 2 ␤ 1 integrin in an RGD-independent manner.
Of even more general importance, ␣ 2 ␤ 1 integrin is not only the integrin receptor for collagen on platelets but also abundantly expressed in various tissues (17), suggesting an important role of ␣ 2 ␤ 1 integrin within the organism. Having proved rhodocetin to be a very specific ␣ 2 ␤ 1 -integrin antagonist, we have started to test rhodocetin as a tool in studying ␣ 2 ␤ 1related functions on the cellular level and have demonstrated that rhodocetin can efficiently and entirely inhibit ␣ 2 ␤ 1 integrin-mediated adhesion of HT1080 fibrosarcoma cells to collagen.

EXPERIMENTAL PROCEDURES
Production of the cDNA Constructs of a Recombinant Human Soluble ␣ 2 ␤ 1 Integrin-The transmembrane and cytoplasmic domain of the integrin ␣2 subunit were substituted for a GGSTGGG spacer and the dimerizing motif of the transcription factor Fos. The cloning strategy started from the construct pUC-HygMT-␣ 3 fos, which was described in a previous paper (18). Briefly, the cDNA sequence coding for the ␣ 3 ectodomain within the pUC-HygMT-␣ 3 fos was replaced by the cDNA sequence coding for the ectodomain of the integrin ␣ 2 subunit. To this end, pUC-HygMT-␣ 3 fos was cleaved by SalI and dephosporylated by calf intestine phosphatase. The 7.2-kilobase pair-long vector fragment still contains the Fos-coding sequence of the original construct pUC-HygMT-␣ 3 fos, yet lacking the complete sequence coding for the integrin ␣ 3 ectodomain. The human cDNA coding for the signal sequence and the N-terminal 912 amino acids of the mature ␣ 2 ectodmain were excised from pFneo-␣ 2 construct (19) using SalI and BglII. The cDNA coding for the C-terminal 131 amino acids of the ␣ 2 ectodomain and the first few amino acids of the GGSTGGG spacer, the latter one of which contains the SalI restriction site, were obtained by polymerase chain reaction using the ␣ 2 cDNA of pFneo-␣ 2 as template, and the oligonucleotides ATGCTGAAATTCACTTAACAAGATCTACC with the BglII site underlined and GCCGCCCGTCGACCCTCCTGTTGGTACTTCGGCTTTCTC with the SalI site underlined as forward and reverse primer, respectively. In a triple ligation the SalI-cleaved vector fragment and the cDNA fragments for both the N-and C-terminal part of the ␣ 2 ectodomain were ligated to the pUC-HygMT-␣ 2 fos construct coding for the soluble ␣ 2 ectodomain, which bears at its C terminus the short spacer sequence GGSTGGG and the dimerizing motif of Fos. The pUC-HygMT-␤ 1 jun construct was generated as described in a previous paper (18).
Establishing a Stable, ␣ 2 ␤ 1 Secreting Schneider Cell Clone-Both constructs were transfected in an equimolar ratio into Drosophila Schneider cells, using TransFast TM Transfection Reagents (Promega, Madison, WI) according to the manufacturer's instructions. Transfected cells were selected under 0.1 mg/ml hygromycin B. After two rounds of subcloning by limited dilution and after screening for positive clones by a sandwich ELISA 1 described below, the stable clone ␣ 2 ␤ 1 -G1.2 was established, which after induction of the metallothionine promoters upstream of both integrin ␣ 2 and ␤ 1 ectodomain cDNAs secreted soluble ␣ 2 ␤ 1 integrin into the cell supernatant in concentrations of about 40 g/liter.
To screen hygromycin B-resistant clones for their ability to secrete soluble ␣ 2 ␤ 1 integrin, supernatants of transfectant clones were tested in a sandwich ELISA 4 -5 days after induction by copper sulfate. For the sandwich ELISA, the mouse monoclonal anti-integrin ␣ 2 antibody JA218 (kindly provided by Danny Tuckwell, University of Manchester, UK) (20) was immobilized to the plastic surface of a microtiter plate at 8 g/ml in TBS (50 mM Tris/HCl, pH 7.4, 150 mM NaCl ) with MgCl 2 (TBS/MgCl 2 ). After blockage of nonspecific binding sites on the microtiter plate with 1% (w/v) heat denatured BSA in TBS/MgCl 2 (BSA/TBS/ MgCl 2 ), the cell supernatants were added into the coated wells. The antibody JA218 captured the soluble ␣ 2 ␤ 1 integrin, which was then detected by an rabbit anti-human ␤ 1 integrin-antiserum as primary antibody and goat anti-rabbit IgG-antibodies coupled to alkaline phosphatase (Sigma) as secondary antibody, diluted 1:300 and 1:600, respectively, in BSA/TBS/MgCl 2 . Before each antibody incubation and the final enzymatic detection reaction, wells were washed three times with TBS/MgCl 2 . As substrate of alkaline phosphatase, p-nitrophenylphosphate tablets were used according to the manufacturer's instructions (Sigma). Absorbance was measured at 405 nm using an ELISA-reader (Dynatech, Burlington, MA).
Isolation of Recombinant Human Soluble ␣ 2 ␤ 1 Integrin-In spinner flasks, ␣ 2 ␤ 1 G1.2 cells were grown in Sf900 Medium (Life Technologies, Inc.) containing 0.1 mg/ml hygromycin B and 10% fetal calf serum. Once they had reached a density of about 12 million cells/ml, they were induced by addition of copper sulfate at 0.6 mM. Simultanously, glucose was added to 0.1% (v/w) and glutamine was added to 0.8 mM. Cell supernatant was harvested 5 days after induction and concentrated by ultrafiltration in a YM30 membrane cartridge (Amicon, Witten, Germany). Protease inhibitors aprotinin, leupeptin, and pepstatin were added at 1 g/ml. Mn 2ϩ ions that increase integrin affinity to ligands were added to a final concentration of 1 mM. The concentration of dithiothreitol (DTT) was adjusted to 2 mM, before the concentrated cell supernatant was loaded onto the collagen I column. The collagen I column had been generated by covalently coupling bovine type I collagen to cyanogen bromide-activated Sepharose 4B CL according to the manufacturer's instruction (Amersham Pharmacia Biotech). The loaded collagen I column was washed with TBS containing 2 mM MgCl 2 , 1 mM MnCl 2 , and 2 mM DTT (wash buffer A). After a stringent wash with buffer A with a NaCl concentration of 300 mM, the collagen I column was washed with buffer A, before the soluble ␣ 2 ␤ 1 integrin was eluted with TBS containing 20 mM EDTA. Immediately after elution, MgCl 2 was added to 30 mM, and the eluate fraction was neutralized with 2 M Tris/HCl, pH 8.0. The ␣ 2 ␤ 1 containing eluate fractions were concentrated by ultrafiltration.
Diluted with Mono Q buffer A (20 mM Tris/HCl, pH 8.0, 1 mM MgCl 2 ), the ␣ 2 ␤ 1 containing solution was loaded onto a Mono Q column and eluted with a linear gradient of 0 to 50% Mono Q buffer B (1 M NaCl in Mono Q buffer A) within 60 min. The ␣ 2 ␤ 1 containing eluate fractions were concentrated by centrifugational ultrafiltration using a Centricon 50 tube (Amicon, Witten, Germany). Protein concentration was determined using the bichinonic acid assay according to the manufacturer's instructions (Pierce). Purity was assessed by SDS-polyacrylamide gel electrophoresis (PAGE) and Coomassie staining.
Binding of Soluble ␣ 2 ␤ 1 Integrin to Various Extracellular Matrix Molecules-Bovine type I collagen and chicken type II collagen was kindly provided by Peter Bruckner (University of Mü nster, Germany). Type IV collagen, the type IV collagen fragment CB3[IV], type V collagen, and murine Laminin-1 (Engelbreth-Holm-Swarm-Laminin) were gratefully obtained from Klaus Kü hn, Rupert Timpl, and Albert Ries (Max-Planck-Institute for Biochemistry, Martinsried, Germany). Collagens were plated in 0.1 M acetic acid, except for CB3 [IV], which like laminin was coated in TBS/MgCl 2 onto the microtiter plate. After the wells were blocked with a BSA/TBS/MgCl 2 , the integrin dissolved in the same solution was allowed to bind to the immobilized substratum. MnCl 2 , activating antibody 9EG7 or EDTA were added as indicated. The activating monoclonal anti ␤ 1 integrin antibody 9EG7 (21) was isolated from cell supernatant according to standard protocols. The 9EG7 hybridoma was kindly provided by Dieter Vestweber (University of Mü nster, Mü nster, Germany). After a 2-h incubation at room temperature, nonbound integrin was washed away with HEPES wash buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 1 mM MnCl 2 ) twice. Then collagen-bound ␣ 2 ␤ 1 integrin was covalently cross-linked to the substratum with 2.5% glutaraldehyde solution in HEPES wash buffer for 10 min at room temperature. After washing the plate three times with TBS/MgCl 2 , the amount of bound ␣ 2 ␤ 1 was measured in an ELISA-like procedure with a rabbit anti-human integrin ␤ 1 subunit antiserum as primary antibody and an anti-rabbit IgG-antibody conjugated to alkaline phosphatase as secondary antibody, diluted 1:300 and 1:600, respectively, in BSA/TBS/MgCl 2 . Each antibody incubation, all of which lasted for 1.5 h, was followed by washing the plate with TBS/ MgCl 2 three times. For detection, p-nitrophenylphosphate tablets (Sigma) were used as substrate for the alkaline phosphatase according to the manufacturer's recommendations. The yellow reaction product was measured at 405 nm in an ELISA reader.
Separation of the Snake Venom Proteins from C. rhodostoma-Snake venom lyophilizate from C. rhodostoma (Sigma) was dissolved in TBS, pH 7.4, containing 1 mM EDTA (TBS/EDTA) at a protein concentration of about 200 mg/ml. The proteins were separated by gel filtration on a Superose 6 column HR30/30 (Amersham Pharmacia Biotech) using TBS/EDTA at 0.3 ml/min. Two distinct pools of fractions were able to inhibit the binding of soluble ␣ 2 ␤ 1 integrin to immobilized type I collagen. The fractions containing the Low molecular weight Calloselasma inhibitor (LMW-CI) was diluted in 20 mM MES/NaOH, pH 6.5 (Mono S buffer A) and passed through a Mono S HR5/5 column (Amersham Pharmacia Biotech). The retained proteins were eluted with a linear gradient of 0 -20% Mono S-buffer B (1 M NaCl in Mono S-buffer A) within 60 min. In the third purification step, the LMW-CI containing solution was adjusted to pH 8.5 by diluting into 20 mM Tris/HCl, pH 8.5 (Mono Q-buffer A). The LMW-CI was eluted from the Mono Q column using a linear gradient of 0 -50% Mono Q-buffer B (1 M NaCl in Mono Q-buffer A). The elute fractions containing LMW-CI were concentrated in a Centricon 10 tube by centrifugal ultrafiltration. To reduce contaminating proteins any further, a final gel filtration on a TSK G3000SWXL column (TosoHaas, Stuttgart, Germany) was performed at 0.4 ml/min. N-terminal sequencing by Edman degradation identified LMW-CI to be identical to rhodocetin (16).
Protein concentration was determined by bichinonic acid. Purity of LMW-CI and the apparent molecular masses of its subunits were assessed by SDS-PAGE and Coomassie staining.
Inhibition ELISA: Inhibition of ␣ 2 ␤ 1 Binding to Immobilized Monomeric Type I Collagen-Dissolved in 0.1 M acetic acid at 40 g/ml, type I collagen was coated as monomeric molecule onto the plastic surface of a microtiter plate at 4°C overnight. After washing with TBS/MgCl 2 , nonspecific binding sites on the plastic surface were blocked with BSA/ TBS/MgCl 2 for 2 h at room temperature. Then soluble ␣ 2 ␤ 1 integrin was added as a 6 g/ml solution in BSA/TBS/MgCl 2 either without any inhibitor (positive control of 100% binding), in the presence of a snake venom fraction, or with 10 mM EDTA (nonspecific binding; negative control with 0% binding). To increase the binding signal of ␣ 2 ␤ 1 integrin, both 1 mM MnCl 2 and a 3-fold molar surplus of integrin-activating antibody 9EG7 was added. To prevent any protease activity of the snake venom that could degrade the ␣ 2 ␤ 1 integrin or the collagen substratum, resulting in a likewise decrease of binding signals, the following protease inhibitors were added to final concentrations as follows: 2 g/ml of each aprotinin, leupeptin, and pepstatin, and 2 mM of each 1,10-phenanthroline and phenylmethylsulfonyl fluoride. After having bound to the immobilized collagen ligand in either the presence or the absence of inhibitor for 2 h at room temperature, nonbound ␣ 2 ␤ 1 integrin was washed off the plate with HEPES wash buffer. After chemical fixation, the bound integrin was measured in the ELISA-like procedure described above. As blank value, the binding signal obtained in the presence of EDTA was subtracted from all other values. To calculate relative binding values, the binding signal of ␣ 2 ␤ 1 integrin to type I collagen without any inhibitor was taken as 100%.
Titration of Immobilized Rhodocetin with Soluble ␣ 2 ␤ 1 Integrin-Both native and inactive rhodocetin were coated onto a microtiter plate at 50 g/ml in TBS/MgCl 2 at 4°C overnight. Rhodocetin had been inactivated by heat denaturation at 95°C for 20 min in the presence of 40 mM DTT, followed by blockage of free thiol groups with 120 mM iodacetamide for 10 min at room temperature. After the microtiter plate was blocked with BSA/TBS/MgCl 2 , soluble ␣ 2 ␤ 1 integrin at different concentrations was incubated with the immobilized rhodocetin. Soluble ␣ 2 ␤ 1 integrin was dissolved in BSA/TBS/MgCl 2 containing 1 g/ml of each aprotinin, leupeptin, and pepstatin, as well as 0.5 mM phenylmethylsulfonyl fluoride and 1,10-phenanthroline. After a 2-h incubation at room temperature, wells were washed twice with HEPES wash buffer. Bound ␣ 2 ␤ 1 integrin was fixed, and its amount was determined by ELISA as described above. Nonspecific binding signals measured as ␣ 2 ␤ 1 binding to the blocking agent BSA were subtracted from the binding values for ␣ 2 ␤ 1 binding to native and denatured rhodocetin, respectively. The titration curves were linearized, and a K d value was determined according to the algorithm given by Heyn and Weischet (22).
RGD Peptide Inhibition Assay of ␣ 2 ␤ 1 Binding to Rhodocetin-Inhibition of ␣ 2 ␤ 1 binding to immobilized rhodocetin by RGD peptide was performed similarly to the titration experiments. After the microtiter plate was coated with rhodocetin at 50 g/ml overnight at 4°C and blocked with BSA/TBS/MgCl 2 at room temperature for 2 h, soluble ␣ 2 ␤ 1 at 15 g/ml was added either in the absence or presence of various concentrations of the linear GRGDSP peptide (Bachem, Heidelberg, Germany) for 2 h at room temperature. Then unbound ␣ 2 ␤ 1 integrin was removed by washing with HEPES wash buffer twice. Bound ␣ 2 ␤ 1 integrin was fixed with 2.5% glutaraldehyde in HEPES wash buffer. Its amount was determined by ELISA as described above. The binding signals were corrected for the blank values measured as ␣ 2 ␤ 1 binding to BSA and afterward normalized to the noninhibited binding of ␣ 2 ␤ 1 to rhodocetin in the absence of GRGDSP peptide (positive control, 100% binding).
Circular Dichroism Spectroscopy of Rhodocetin-The buffer of the rhodocetin solution was changed to 20 mM sodium phosphate, pH 7.0, 50 mM NaCl by gel filtration on a TSK G3000SWXL column (TosoHaas, Stuttgart, Germany). The rhodocetin containing eluate fractions were concentrated in a Centricon 10 tube by centrifugal ultrafiltration to reach a concentration of about 0.3 mg/ml. The CD spectrum was recorded from 190 to 260 nm in a 0.1-mm cuvette in a CD spectrophotometer type CD6 (Jobin Yvon, Paris, France). Temperature was controlled by a self-constructed Peltier element cuvette holder. The relative amount of secondary structures (␣ helix, parallel, and anti-parallel ␤ strands, random coil) were calculated with the deconvolution program of CDNN by Gerhard Böhm (23).
Inhibition of Cell Adhesion to Collagen by Rhodocetin-Monomeric bovine type I collagen at a concentration of 0.2 g/ml in 0.1 M acetic acid was immobilized onto a microtiter plate at 4°C overnight. After being washed with TBS/MgCl 2 for three times, the plate was blocked with BSA/TBS/MgCl 2 for 2 h at room temperature. HT1080 fibrosarcoma cells at a density of 500,000 cells/ml in Dulbecco's modified Eagle's medium were plated onto the plate for 35 min in a tissue culture incubator at 37°C in both absence and presence of various concentrations of rhodocetin. Adherent cells were detected by staining with crystal violet (24). Briefly, adherent cells were fixed with 70% (v/v) solution of ethanol for 7 min and stained with a 0.1% (w/v) solution of crystal violet in destilled water. After washing the wells, cell bound dye was extracted with a 0.2% (v/v) Triton X-100 solution, and its amount was measured in an ELISA reader at 560 nm. Experiments with cells were done in triplicates. The adhesion signal of HT1080 cells measured on BSA was considered nonspecific background and subtracted from the adhesion signals of cells on type I collagen. Adhesion signals in the presence of rhodocetin were normalized to the adhesion signal of the noninhibited cell adhesion to type I collagen without any inhibitor.

Production and Isolation of Recombinant Soluble Human
Integrin-A recombinant soluble human ␣ 2 ␤ 1 integrin that consists of the ectodomains of both ␣ 2 and ␤ 1 integrin subunits being noncovalently associated by the dimerizing motif of Fos and Jun, respectively, was secreted by transfected Drosophila Schneider cells. Affinity purification of the cell supernatant on a type I collagen column yielded not only the soluble ␣ 2 ␤ 1 integrin but also a protein of 45 kDa as determined by SDS-PAGE under reducing conditions (Fig. 1, lane 5). Edman degradation of the latter one revealed its N-terminal sequence as STEFSEDLLDEDLDLDIDE and, thus, identified the 45-kDa protein as Drosophila BM40 (GenBank TM accession number AJ1333736). Interestingly, BM40 was abundantly expressed by Schneider cells. Like the soluble ␣ 2 ␤ 1 integrin, it bound to type I collagen column in a divalent cation-dependent manner. About 10 times more BM40 than soluble ␣ 2 ␤ 1 integrin was eluted from the type I collagen column by EDTA. However, binding of BM40 to type I collagen did not interfere with ␣ 2 ␤ 1 integrin binding to its collagen ligand. Being a less acidic protein than BM40, the soluble ␣ 2 ␤ 1 integrin was further purified by anion exchange chromatography on a Mono Q column, from which the soluble ␣ 2 ␤ 1 integrin was eluted at lower ion strength than the highly acidic BM40. Yields of soluble ␣ 2 ␤ 1 integrin ranged from 30 to 40 g/liter of cell supernatant.
Characterization of the Recombinant Soluble ␣ 2 ␤ 1 Integrin-In SDS-PAGE, the soluble ␣ 2 ␤ 1 integrin heterodimer was separated into the Fos zipper containing ␣ 2 -ectodomain, ␣ 2 -Fos, and the Jun zipper containing ␤ 1 -ectodomain, ␤ 1 -Jun, which run at 150 and 95 kDa, respectively, under nonreducing conditions and at 140 and 100 kDa, respectively, after reduction ( Fig. 1, lanes 1 and 4). The identity of the ␣ 2 band was proven by N-terminal sequencing. Edman degradation revealed the sequence YNVGLPEAKI in agreement with the mature human integrin ␣ 2 subunit (19), demonstrating that the human ␣ 2 subunit was correctly processed proteolytically within the insect cells. Like the wild-type form on human cells, the human ␤ 1 -Jun chain expressed by the insect cells was N-terminally blocked and thus inaccessible to Edman degradation. However, it was identified in Western blot by a polyclonal antiserum against the human integrin ␤ 1 subunit (data not shown). Unlike other integrin ␣ subunits, the ␣ 2 -ectodomain is not proteolytically processed into a heavy and light chain. Neither was the human soluble ␣ 2 ␤ 1 integrin cleaved in the heterologous expression system of the Drosophila Schneider cells. Having very similar isoelectric points, the ␣ 2 ␤ 1 integrin and BSA could not efficiently be separated by anion exchange chromatography leading to a slight contamination of BSA in the ␣ 2 ␤ 1 integrin preparation.
The soluble ␣ 2 ␤ 1 integrin was able to bind to collagen types I, II, IV, and V and to laminin-1 (Engelbreth-Holm-Swarm-Laminin) (Fig. 2). The highest binding signals were observed to type I and II collagen, which is in good agreement with results of wild-type ␣ 2 ␤ 1 integrin (25). Like the wild-type form, the soluble ␣ 2 ␤ 1 integrin gave a smaller binding signal on the basal membrane collagen, type IV collagen, and likewise to its triple helical fragment CB3 [IV], which comprise the binding sites for both ␣ 1 ␤ 1 and ␣ 2 ␤ 1 integrin (25). A significantly lower binding signal was measured to type V collagen, which together with type I collagen forms the collagen fibrils of the connective tissue. As a ligand without any collagenous triple helix, laminin-1 was bound by the soluble ␣ 2 ␤ 1 integrin, albeit with a much lower binding signal than the collagenous ligands. The latter finding corroborated studies of wild-type ␣ 2 ␤ 1 integrin binding to laminin-1 (26). Identical to cell membrane-anchored wild-type ␣ 2 ␤ 1 integrin, soluble ␣ 2 ␤ 1 required divalent cations to recognize its ligands. Therefore, EDTA abolished ␣ 2 ␤ 1 binding (Fig. 2). The soluble ␣ 2 ␤ 1 integrin seemed to be regulated by extracellular factors in a manner similar to that of the the wild-type ␣ 2 ␤ 1 integrin on the cell surface, because integrinactivating Mn 2ϩ ions and the activating monoclonal antibody 9EG7 increase the binding signal of soluble ␣ 2 ␤ 1 integrin to its ligands (Fig. 2). Taken together, the soluble ␣ 2 ␤ 1 integrin showed ligand binding properties similar to the membraneanchored wild-type ␣ 2 ␤ 1 integrin. However, no detergent was needed to extract the soluble ␣ 2 ␤ 1 integrin or to keep it in solution. Furthermore, unlike the detergent-extracted wildtype ␣ 2 ␤ 1 integrin, soluble ␣ 2 ␤ 1 integrin remained active even after a longer storage period of several months.
Whole Snake Venom of C. rhodostoma Inhibits Binding of Soluble ␣ 2 ␤ 1 Integrin to Immobilized Type I Collagen-The strong binding signal of soluble ␣ 2 ␤ 1 integrin to immobilized type I collagen (Fig. 2) was diminished and completely inhibited by the crude snake venom of C. rhodostoma in a dose-dependent manner with an IC 50 value of about 50 g/ml (data not shown). Like other snake venoms, C. rhodostoma venom contains several proteases that could be detected by zymogram developed with gelatin. Rhodostoxin (kistomin and major hemorrhagin), a metalloprotease (27,28), and ancrod, a serine protease (29), could be detected in the zymogram among other proteolytic activities (data not shown). To rule out the possibil- ity that any snake venom protease diminishes the ␣ 2 ␤ 1 binding signal to immobilized collagen, protease inhibitors directed against all four classes of proteases, such as aprotinin, leupeptin, phenylmethylsulfonyl fluoride, pepstatin, and 1,10-phenanthroline, were added to the venom protein fraction when applied in the inhibition ELISA to test its capability to inhibit binding of soluble ␣ 2 ␤ 1 to immobilized type I collagen by a nonproteolytic interaction.
Rhodocytin/Aggretin Does Not Inhibit Binding of Soluble ␣ 2 ␤ 1 Integrin to Type I Collagen-Rhodocytin or aggretin are the two names of a 29-kDa protein of C. rhodostoma venom, which induces activation and aggregation of thrombocytes (15,30). It was isolated from the snake venom according to Shin and Morita (31). In SDS-PAGE (Fig. 3, lane 1), the purified rhodocytin/aggretin showed a molecular mass of about 29 kDa under nonreducing conditions. Being a disulfide cross-linked heterodimer, it was cleaved under reducing conditions into two subunits of 19 and 15 kDa (Fig. 3, lane 4). The N-terminal sequences of both subunits, GLEDDFGWSPYDQ[H/(Q)] 2 and DPSGWSSYEG[H/(G)](H)YK, proved their identities as ␣ and ␤ chains, respectively, of rhodocytin/aggretin (14,31). To test the postulated interaction of soluble ␣ 2 ␤ 1 integrin with rhodocytin/aggretin, the latter one was immobilized on a microtiter plate, and the binding of soluble ␣ 2 ␤ 1 was tested. Whereas the soluble ␣ 2 ␤ 1 binds to immobilized monomeric type I collagen in a divalent cation-dependent manner, no binding to immobilized rhodocytin/aggretin was observed (Fig. 4A). A similar result was obtained when wild-type ␣ 2 ␤ 1 integrin, which had been purified from platelets (kindly provided by Albert Ries and Rupert Timpl, Max-Planck-Institute for Biochemistry, Martinsried, Germany), was used (data not shown).
Because immobilization may have caused inactivation of rhodocytin/aggretin, binding of soluble ␣ 2 ␤ 1 to soluble rhodocytin/aggretin was tested by measuring the capability of the snake venom component to inhibit ␣ 2 ␤ 1 integrin binding to immobilized type I collagen. However, rhodocytin/aggretin does not prevent ␣ 2 ␤ 1 integrin from binding to collagen (Fig. 4B). Both the binding test and the inhibition test rule out any direct interaction between rhodocytin/aggretin and soluble ␣ 2 ␤ 1 integrin on the molecular level.
Searching for the Component of C. rhodostoma Venom That Inhibits the Interaction of Soluble ␣ 2 ␤ 1 Integrin with Collagen-Although rhodocytin/aggretin did not inhibit ␣ 2 ␤ 1 binding to collagen (Fig. 4), the whole snake venom hampered binding of soluble ␣ 2 ␤ 1 integrin to immobilized type I collagen. Taking advantage of the inhibition ELISA, the constituent of C. rhodostoma venom that is responsible for the inhibition of ␣ 2 ␤ 1 integrin binding to type I collagen was searched. In the first purification step, the venom proteins were separated according to their molecular masses by gel filtration on a Superose 6 column (Fig. 5A). When the eluate fractions were screened for their capability to inhibit ␣ 2 ␤ 1 binding to immobilized type I collagen, two peaks of inhibitory activity could be identified (Fig. 5B). Because of their different molecular masses, they were referred to as high molecular weight and low molecular weight Calloselasma inhibitor. Purification and identification of the LMW-CI activity was further pursued. Ion exchange chromatography both on Mono S and Mono Q could clearly separate rhodocytin/aggretin from the ␣ 2 ␤ 1 integrin inhibitory activity of LMW-CI. The Mono S column retained LMW-CI at pH 6.5 up to a ionic strength of 105 mM NaCl, whereas rhodocytin/aggretin barely bound to Mono S at pH 6.5 and was washed off the column at very low ionic strength. In the opposite elution order, LMW-CI was eluted from the Mono Q column at pH 8.5 at low ionic strength of about 100 mM NaCl, whereas rhodocytin/aggregetin remained bound to the Mono Q resin at NaCl concentrations of up to 300 mM NaCl. In conclusion, the isoelectric point of LMW-CI must be higher than the one of rhodocytin/aggretin, although the isoelectric points of both proteins must be in a pH range of 6.0 -8.5. As final purification step of LMW-CI, another gel filtration chromatography on a TSK G3000SWXL was performed, resulting in a highly purified band at 27 kDa in SDS-PAGE. Furthermore, coprecipitation experiments with ␣ 2 ␤ 1 integrin showed that the 27-kDa protein binds to the ␣ 2 ␤ 1 integrin, suggesting that the 27-kDa protein is the LMW-CI (data not shown). 2 Brackets indicate two possible amino acids that could not be clearly identified in the sequencing cycle. Parentheses indicate a less likely amino acid when the Edman degradation cycle did not give a clear identification but an option of two or more possible amino acids.  I collagen (B). In A, bovine type I collagen at 20 g/ml in 0.1 M acetic acid and rhodocytin/aggretin at 50 g/ml in TBS/MgCl 2 were coated onto a microtiter plate. After blockage with heat denatured BSA, soluble ␣ 2 ␤ 1 integrin was added for 2 h at room temperature. After the wells had been washed twice, the bound ␣ 2 ␤ 1 integrin was chemically fixed to the immobilized substratum and its amount measured by ELISA. In B, bovine type I collagen was immobilized on the microtiter plate at 40 g/ml in 0.1 M acetic acid. After blockage with heat denatured BSA, the soluble ␣ 2 ␤ 1 integrin was allowed to bind to the immobilized collagen in both absence and presence of various concentrations of soluble rhodocytin/aggretin for 2 h at room temperature. After the wells had been washed twice, the collagenbound ␣ 2 ␤ 1 integrin was chemically fixed, and its amount was determined by ELISA. Blank values measured in BSA-coated wells were subtracted from the measured absorbance values. In B, the absorbance values were normalized to the noninhibited binding value, taken as 100%. Each value was measured in duplicate. Standard deviations are indicated.
Characterization of LMW-CI-Under nonreducing conditions, LMW-CI shows an apparent molecular mass of 27 kDa in SDS-PAGE (Fig. 3, lane 2). LMW-CI is a heterodimer, which upon reduction falls apart in two subunits of 16 and 14 kDa (Fig. 3, lane 5). The N-terminal sequences of both LMW-CI subunits were identified by Edman degradation with the N terminus of the 16
On a molecular level, rhodocetin inhibited binding of soluble ␣ 2 ␤ 1 integrin to immobilized type I collagen in a dose-dependent manner (Fig. 6), thus proving that, in contrast to rhodocytin/aggretin, the effect of rhodocetin on whole platelets (16) can indeed be imitated on a molecular scale, i.e. on the interaction of isolated ␣ 2 ␤ 1 integrin to collagen. With increasing concentrations, LMW-CI/rhodocetin decreased the binding signal of the collagen receptor to its ligand and eventually abolished it entirely. From Fig. 6, an IC 50 value of about 30 nM could be determined.
Rhodocetin Is a Disintegrin That Directly and Specifically Binds to ␣ 2 ␤ 1 Integrin-Addition of various protease inhibitors to the inhibition ELISA ruled out the possibility that the decrease of ␣ 2 ␤ 1 binding to collagen was caused by proteolytic digestion of either binding partner by a snake venom protease. Therefore, a direct, yet nonenzymatic binding interaction of LMW-CI/rhodocetin with either ␣ 2 ␤ 1 integrin or with the integrin-binding site on type I collagen must be responsible for its inhibitory effect. To test a direct interaction of rhodocetin with the soluble ␣ 2 ␤ 1 integrin, rhodocetin was immobilized onto a microtiter plate, and binding of soluble ␣ 2 ␤ 1 was measured. As shown in Fig. 7, the soluble ␣ 2 ␤ 1 integrin directly bound to rhodocetin, thereby qualifying it to be a disintegrin. The binding signal could be increased slightly by addition of 1 mM MnCl 2 and the integrin-activating antibody 9EG7. However, in contrast to other integrin ligands, binding of ␣ 2 ␤ 1 to rhodocetin did not require any divalent cations, because addition of EDTA did not abolish ␣ 2 ␤ 1 binding to rhodocetin. A binding signal similar to the one of soluble ␣ 2 ␤ 1 integrin was obtained when detergent-extracted wild-type ␣ 2 ␤ 1 integrin from human platelets was applied (data not shown). Another soluble integrin, the laminin-5 receptor ␣ 3 ␤ 1 integrin (18), did not bind to immobilized, native rhodocetin, although it showed binding activity to laminin-5 (Fig. 7). The soluble ␣ 3 ␤ 1 integrin had been produced in our lab by insect cells similarly to the soluble ␣ 2 ␤ 1 integrin (18). Even more striking, another widespread collagen receptor, ␣ 1 ␤ 1 integrin, which had been isolated from human placenta according to Kern et al. (25) and was tested biologically active by its binding to type I and IV collagen, entirely fails to bind to rhodocetin (Fig. 7), proving the specificity of LMW-CI/ rhodocetin to recognize ␣ 2 ␤ 1 integrin selectively.
It is noteworthy that the ability of LMW-CI/rhodocetin to interact with ␣ 2 ␤ 1 integrin depended on its disulfide bridges, which stabilize both its quartenary and tertiary structure. Preincubation of LMW-CI at DTT concentrations higher than 0.016 mM without any thermal denaturation resulted in a strong decrease of ␣ 2 ␤ 1 binding (Fig. 8A). However, when scrutinized by SGS-PAGE (Fig. 8B), the partially reduced LMW-CI/rhodocetin run as stable heterodimer even up to 10 mM DTT. Amazingly, rhodocetin does not possess any intercatenary disulfide bridges (16), but its subunits stayed together even under the harsh denaturating condition of the SDS-PAGE sample buffer containing 2% SDS. Reduction of the intracatenary disulfide bridges at DTT concentrations higher than 10 mM made the rhodocetin heterodimer dissociate. Although lacking an intercatenary disulfide bridge, quaternary structure of rhodocetin is very stable and depends on the tertiary structure of . The eluate fractions were tested in the inhibition ELISA for their capability to inhibit binding of soluble ␣ 2 ␤ 1 to immobilized type I collagen (B). The inhibitory activity of the eluate fractions is shown as relative inhibition, which is defined as difference between noninhibited and inhibited binding signal normalized to the noninhibited binding signal. Note that two inhibitory peaks are separated on the size exclusion column that differ in their molecular masses. They are named high molecular mass Calloselasma inhibitor (HMW-CI) and low molecular mass Calloselasma inhibitor (LMW-CI), respectively.
FIG. 6. LMM-CI/rhodocetin inhibits binding of soluble ␣ 2 ␤ 1 integrin to immobilized type I collagen in a dose-dependent manner. Monomeric bovine type I collagen was coated onto a microtiter plate at 40 g/ml in 0.1 M acetic acid. After blockage of the wells with heat denatured BSA, soluble ␣ 2 ␤ 1 integrin at 6 g/ml was allowed to bind to immobilized type I collagen in both absence and presence of various concentrations of LMM-CI for 2 h at room temperature. After the wells had been washed twice, the collagen-bound ␣ 2 ␤ 1 integrin was detected by ELISA. Nonspecific binding signal as measured in the presence of 10 mM EDTA was subtracted from all values. The binding signals were then normalized to the noninhibited binding signal. Each value was measured in duplicate. Relative standard deviations are shown.
both subunits, which is stabilized by intracatenary disulfide bridges. As the binding signal of soluble ␣ 2 ␤ 1 integrin gradually decreased with increasing DTT concentrations higher than 0.016 mM and is entirely lost at 10 mM DTT, it can be envisioned that the intracatenary disulfide bridges are of paramount importance in maintaining the native tertiary structure of rhodocetin, which is essential for ␣ 2 ␤ 1 integrin binding. Furthermore, the tertiary structure of its subunits as evidenced by its integrin binding function seems to be even more sensitive to denaturation than its quaternary structure, i.e. dissociation into its two subunits.
To determine the binding affinity of rhodocetin to ␣ 2 ␤ 1 integrin, both native and denatured rhodocetin were immobilized onto a microtiter plate and titrated with soluble ␣ 2 ␤ 1 integrin (Fig. 9). Treatment of rhodocetin with 40 mM DTT in addition to thermal denaturation entirely abolished its binding activity to the integrin, again demonstrating that the specific interaction of rhodocetin with ␣ 2 ␤ 1 integrin requires the disulfide-stabilized native conformation of rhodocetin. For binding of soluble ␣ 2 ␤ 1 integrin to native rhodocetin, saturation was achieved at ␣ 2 ␤ 1 concentrations of about 100 nM. From such titration curves, an apparent K d value of LMW-CI/rhodocetin binding to ␣ 2 ␤ 1 integrin was calculated to be 10.3 nM.
Rhodocetin Does Not Contain a Triple Helical Collagen Domain-Being essential for its inhibitory activity, the native conformation of LMW-CI/rhodocetin was further studied by CD. We were especially interested in whether or not LMW-CI contains any triple helical collagenous motifs, because high affinity ligands of ␣ 2 ␤ 1 integrin are mostly collagenous molecules. Laminin-1, which lacks any collagenous structure, is bound by ␣ 2 ␤ 1 integrin with much lower affinity. Although LMW-CI/rhodocetin competed with the high affinity binding of Inactivation of LMW-CI was achieved by reduction of disulfide bridges with 40 mM DTT and heat denaturation at 95°C for 20 min. Reduced thiol groups were then blocked with 120 mM iodacetamide. Both native and inactivated LMW-CI was coated onto the microtiter plate at 40 g/ml and titrated with the indicated concentrations of soluble ␣ 2 ␤ 1 . Wells coated with heat denatured BSA were taken as blanks. The bound ␣ 2 ␤ 1 integrin was chemically fixed, and its amount was determined by ELISA. The blank values were subtracted from the binding signals. Each value was measured in duplicate. Standard deviations are indicated. FIG. 7. LMW-CI is a disintegrin which interacts directly and specifically with ␣ 2 ␤ 1 integrin. LMW-CI at 30 g/ml, monomeric bovine type I collagen (bCol-1) at 30 g/ml, monomeric human type IV collagen at 30 g/ml (hCol-4), and human laminin-5 (hLam-5) at 10 g/ml were coated onto the microtiter plate in TBS/Mg 2ϩ ions, except for the collagen, which were dissolved in 0.1 M acetic acid. After blocking with heat denatured BSA, the wells were incubated with octylglucosidesolubilized ␣ 1 ␤ 1 integrin (open, filled, and gray bars), with soluble ␣ 2 ␤ 1 integrin (thinly, thickly, and densely striped bars), or with soluble ␣ 3 ␤ 1 integrin (thinly, thickly, and densely hatched bars) for 2 h at room temperature either in the presence of 1 mM Mn 2ϩ ions (open, thinly striped, and hatched bars, respectively), or in the presence of both 1 mM Mn 2ϩ and a 3-fold molar surplus of integrin-activating 9EG7 antibody (filled, thickly striped, and hatched bars, respectively) or in the presence of 10 mM EDTA (gray, densely striped, and hatched bars, respectively). Wild-type ␣ 1 ␤ 1 had been extracted and isolated from human placental tissue according to Kern et al. (25). Both soluble human ␣ 2 ␤ 1 and ␣ 3 ␤ 1 integrin had been recombinantly expressed in Drosophila Schneider cells and isolated as described "Experimental Procedures" and according to Eble et al. (18), respectively. After the wells had been washed twice, substratum-bound integrin was chemically fixed, and its amount was determined by ELISA. The binding signals onto heat denatured BSA were taken as blanks. Each value was measured in duplicate, and standard deviations are shown. Binding of ␣ 1 ␤ 1 and ␣ 2 ␤ 1 integrin to laminin-5 and binding of ␣ 3 ␤ 1 integrin to type I and IV collagen were not determined (n.d.).
FIG. 8. The native tertiary structure of LMW-CI/rhodocetin, as evidenced by its inhibitory activity toward ␣ 2 ␤ 1 integrin (A), is destroyed at much lower concentration of reducing agents than its quartenary structure, i.e. association of its two subunits, as detected by SDS-PAGE under nonreducing conditions (B). A, LMW-CI was incubated for an hour with increasing concentrations of DTT at 37°C. After inactivation of DTT with a surplus of iodacetamide, LMW-CI was coated onto a microtiter plate at 40 g/ml. Soluble ␣ 2 ␤ 1 integrin at 6 g/ml was allowed to bind to the pretreated LMW-CI. After being chemically fixed, bound integrin was detected by ELISA. Binding signals were corrected for the nonspecific binding signal on heat denatured BSA and normalized to LMW-CI, which had been incubated without DTT and treated with iodacetamide. Values were measured in duplicate. Standard deviations are shown. Note that inibitory activity drops at DTT concentration higher than 0.016 mM, and is completely lost at 10 mM. B, DTT and iodacetamide-treated LMW-CI, which was used to coat the microtiter plate, was separated in a 12-18% polyacrylamide gradient gel under nonreducing conditions. Note that nonreduced LMW-CI heterodimer vanishes at DTT concentrations higher than 2 mM with a concomitant pronounced appearance of the two LMW-CI subunits. ␣ 2 ␤ 1 to type I collagen, the CD spectrum of LMW-CI (data not shown) did not reveal any collagenous triple helical structure within the disintegrin. Although lacking a collagen domain, rhodocetin possesses a distinct native structure, because deconvulation of the CD spectrum recorderd at 20°C disclosed a high content of 59.5% ␤-sheet and a minor amount of 10.3% ␣-helical secondary structure for rhodocetin. Heat denaturation abrogated any secondary structural signals in the CD spectrum, leaving a spectrum typical of random coil.
Rhodocetin Is an RGD-independent Integrin-Many disintegrins bind to RGD-dependent integrins in an RGD peptides inhibitable manner. However, the linear GRGDSP peptide failed to inhibit the interaction of ␣ 2 ␤ 1 integrin with immobilized LMW-CI/rhodocetin (Fig. 10). Even at concentrations of 4 mM, which represented an 800,000-fold molar surplus to the soluble ␣ 2 ␤ 1 integrin in the inhibition experiment, the GRGDSP peptide did not affect the ␣ 2 ␤ 1 disintegrin interaction, thus showing that LMW-CI/rhodocetin belongs to the small group of RGD-independent disintegrins.
Effect of Rhodocetin on ␣ 2 ␤ 1 -mediated Adhesion of Fibroblasts-Whether LMW-CI/rhodocetin can be used in vivo, e.g. to inhibit ␣ 2 ␤ 1 -mediated cell adhesion and migration or to influence other cellular reactions triggered by the ␣ 2 ␤ 1 -collagen interaction, the effect of the isolated rhodocetin on adhesion of HT1080 cells onto immobilized type I collagen was examined. HT1080 is a human fibrosarcoma cell line that abundantly expresses ␣ 2 ␤ 1 integrin on its surface and that adheres to immobilized type I collagen mainly via its ␣ 2 ␤ 1 integrin (32). When HT1080 cells were plated onto monomeric type I collagen in the presence of soluble rhodocetin, cell adhesion declined with increasing concentrations of the snake venom disintegrin and was eventually abolished completely (Fig. 11). At a cell density of 500,000 cells/ml, plated onto 0.2 mg/ml type I collagen, an IC 50 value for LMW-CI/rhodocetin of about 2 g/ml ϭ 60 nM was measured, showing that, even at low concentrations, this disintegrin is able to specifically affect ␣ 2 ␤ 1 -collagen interaction on a cellular level. DISCUSSION Here, we report that rhodocytin, an inducer of platelet aggregation from the hemorrhagic snake venom of C. rhodos-toma, does not interact with ␣ 2 ␤ 1 integrin, one of the collagen receptors on the surface of blood platelets. However, from the same snake venom, we have isolated and characterized LMW-CI, which is identical to rhodocetin, a recently published inhibitor of collagen-induced platelet aggregation (16). We show that LMW-CI/rhodocetin is a disintegrin that specifically and avidly binds to ␣ 2 ␤ 1 integrin. Independently of Wang et al. (16), we have purified LMW-CI/rhodocetin as a component of C. rhodostoma venom, which inhibits the binding of ␣ 2 ␤ 1 integrin to immobilized collagen on the molecular level. We used recombinantly expressed and purified, soluble ␣ 2 ␤ 1 integrin in an inhibition ELISA to screen the snake venom for components that specifically and nonproteolytically inhibit the interaction of ␣ 2 ␤ 1 integrin with collagen. Although commonly used, the method of using whole platelets to test for integrin agonists and antagonists may be biased by the presence of various other collagen receptors on the platelet surface. Furthermore, here we describe our detailed studies on the interaction of the novel disintegrin LMW-CI/rhodocetin with ␣ 2 ␤ 1 integrin.
Identification of LMW-CI/rhodocetin as disintegrin, which specifically binds to ␣ 2 ␤ 1 integrin, was made feasible by the recombinant production of a soluble ␣ 2 ␤ 1 integrin and its purification in sufficient amounts. Soluble ␣ 2 ␤ 1 integrin was generated by replacing the transmembrane and cytoplasmic domain of both integrin subunits ␣ 2 and ␤ 1 with the dimerizing motifs of the transcription factor Fos and Jun, respectively. Lately, a similar attempt to produce soluble ␣ 3 ␤ 1 had been successful (18). However, yields of soluble ␣ 2 ␤ 1 integrins were generally lower than with soluble ␣ 3 ␤ 1 integrin in compliance with the comparatively lower expression of membrane-bound ␣ 2 ␤ 1 on transfected mammalian cells, such as erythroleukemic K562 cells. 4 Although the human integrin ectodomains were heterologously expressed by Drosophila cells, both sunbunits were correctly processed proteolytically, because the leader sequences were cleaved off to give the N termini of both mature human subunits. Whereas the N-terminal amino acid sequence of mature human ␣ 2 subunit was accessible to Edman degra- FIG. 10. Binding of the disintegrin LMW-CI/rhodocetin to ␣ 2 ␤ 1 integrin does not depend on an RGD peptide sequence. LMW-CI was coated onto a mircrotiter plate at 40 g/ml. After being blocked with heat denatured BSA, wells were incubated with soluble ␣ 2 ␤ 1 integrin for 2 h in the absence and presence of the linear peptide GRGDSP at concentrations indicated in the plot. After wells had been washed twice, bound receptor was chemically fixed, and its amount was determined by ELISA. ␣ 2 ␤ 1 binding to heat denatured BSA was taken as blank and subtracted from the binding values. Binding signals were normalized to the noninhibited binding signal measured in the absence of peptide. Values were determined in duplicate. Standard deviations are indicated.
FIG. 11. LMW-CI/rhodocetin efficiently and entirely inhibits ␣ 2 ␤ 1 integrin-mediated cell adhesion of HT1080 fibrosarcoma cells to type I collagen. Monomeric type I collagen was coated at 0.2 g/ml in 0.1 M acetic acid. After the wells were washed and blocked with heat denatured BSA, HT1080 cells were plated onto the collagen substratum at a density of 50,000 cells/well for 35 min in the absence and presence of various concentrations of LMW-CI as indicated. Adhered cells were stained with crystal violet, which was solubilized from the cells after the wells had been destained. Absorbance was measured at 560 nm. Integrin-specific cell adhesion was completely abolished in the presence of 10 mM EDTA. This blank value was subtracted from the binding signals. The adhesion signals were normalized to the noninhibited adhesion of HT1080 cells. Each value was measured in triplicate. Standard deviations are shown. dation, the ␤ 1 subunit was N-terminally blocked. However, loss of signal sequence of the ␤ 1 subunit and its subsequent Nterminal blockage reaction to a pyroglutamate residue also occurs in the homologously expressed human ␤ 1 subunit, indicating a correct processing of the soluble human ␤ 1 integrin subunit by the insect cells. Furthermore, the ectodomain heterodimer of the collagen receptor ␣ 2 ␤ 1 integrin binds avidly to monomeric type I collagen, suggesting that the integrin is not only correctly processed but also correctly folded. The binding signals of soluble ␣ 2 ␤ 1 to different types of collagen and laminin-1 demonstrated a ligand preference similar to the wild-type ␣ 2 ␤ 1 integrin, with decreasing binding signals in the order of type I, type II, type IV, and type V collagen and laminin-1 (25,26). Binding affinities of soluble ␣ 2 ␤ 1 integrin could be increased with Mn 2ϩ ions and an activating antibody 9EG7, an observation that resembles the activity regulation of wild-type ␣ 2 ␤ 1 integrin.
A great advantage of the soluble ␣ 2 ␤ 1 integrin in comparison with the membrane-bound wild-type ␣ 2 ␤ 1 integrin, which was isolated by extracting blood platelets with the mild detergent octylglucoside, is its stability and the comparatively high yield. Whereas the detergent-extracted wild-type ␣ 2 ␤ 1 integrin lost activity within days after preparation, 2 the soluble ␣ 2 ␤ 1 integrin remained stable for several weeks. However, we have not found any explanation for this observation yet.
Having sufficient amounts of a stable, soluble ␣ 2 ␤ 1 integrin at hand, we could address the question of which component of the hemorrhagic snake venom of C. rhodostoma is responsible for inhibiting the interaction of ␣ 2 ␤ 1 integrin with type I collagen. This interaction is of major importance for platelet reactions to collagen. Collagen becomes accessible to platelets after damage of the blood vessels or tissue injury. It initiates the activation of platelets (8,30,33), which results in degranulation and an increase in number and/or affinity of other cell adhesion molecules on the platelet, such as the major platelet integrin ␣ IIb ␤ 3 (3), which eventually leads to platelet aggregation and blood clotting. RGD-containing disintegrins, such as rhodostomin (kistrin) from C. rhodostoma (27,34), inhibit the interaction of the RGD-dependent ␣ IIb ␤ 3 integrin with fibrin, thereby impairing a later step in the blood clotting mechanism. Moreover, snake venoms contain several proteases, such as the metalloprotease rhodostoxin (kistomin and major hemorrhagin) (27,28), and the serine protease ancrod (29) from C. rhodostoma, which cleave fibrinogen/fibrin or prothrombin, thus again interfering in the blood clotting cascade and resulting in bleeding and hemorrhages.
Being the initial step of collagen-triggered platelet activation and aggregation, ␣ 2 ␤ 1 integrin is of paramount importance in hemostasis (9,10,17). Previous studies with the hemorrhagic snake venom of C. rhodostoma were performed on whole platelets. However, platelets contain various collagen receptors on their surface with different characteristics, e.g. ␣ 2 ␤ 1 integrin (GPIa/IIa) and GPVI (9,10,17). Whereas GPVI mainly recognizes type I collagen molecules, which are bundled into collagen fibrils, in a divalent cation-independent manner, ␣ 2 ␤ 1 integrin mainly binds to monomeric type I collagen molecules in the presence of divalent cations (35)(36)(37). The importance of ␣ 2 ␤ 1 integrin in normal hemostasis is corroborated by severe bleeding disorders in patients, who either lack the ␣ 2 ␤ 1 integrin receptor on their platelets (11) or who have developed inhibiting autoantibodies against the ␣ 2 integrin subunit (12).
Possessing various collagen receptors and being easily activated by several stimuli other than collagen, such as ADP, thrombin, etc., whole platelets are rather coarse targets to screen hemorrhagic snake venoms for specific inhibitors to the ␣ 2 ␤ 1 integrin-collagen interaction, inasmuch as snake venoms by themselves contain a whole battery of various agents interfering with platelet activation and blood clotting. Based on such studies with whole platelets, rhodocytin/aggretin have been published to be an activator of ␣ 2 ␤ 1 integrin-mediated platelet aggregation (15,31). However, we have isolated rhodocytin/aggretin, proved its identity by N-terminal sequencing, and could not see any interaction of rhodocytin with ␣ 2 ␤ 1 integrin. Nor could we observe any influence of rhodocytin on the integrin-collagen interaction. Therefore, a direct interaction of rhodocytin/aggretin with the soluble ␣ 2 ␤ 1 integrin ectodomain on the protein level can be ruled out. Alternatively, its effects on platelets aggregation may be caused by protein-carbohydrate interactions. Wild-type ␣ 2 ␤ 1 integrin on the platelets surface may differ from recombinantly expressed soluble ␣ 2 ␤ 1 integrin in their carbohydrate side chains, because Drosophila Schneider cells are unable to process the N-linked carbohydrate side chains of proteins from high mannose type into complex type carbohydrate antennary structures (18). Because rhodocytin/aggretin belongs to the family of C-type lectins bearing homology to the carbohydrate recognition domains of Ca 2ϩ -dependent lectins (31), it can be surmised that rhodocytin cross-links several ␣ 2 ␤ 1 integrins on the platelet surface via their carbohydrate side chains, thereby imitating the recruitment of integrins into focal contact-like structures, which eventually leads to platelet activation and aggregation. However, because both recombinantly expressed soluble ␣ 2 ␤ 1 integrin and wild-type ␣ 2 ␤ 1 integrin extracted from platelets that are likely to bear high-mannose type and complex-type N-linked carbohydrate side chains, respectively, fail to interact with rhodocytin/aggretin in our binding tests, even in the presence of Ca 2ϩ ions, a mechanism involving a direct interaction of rhodocytin/aggretin with ␣ 2 ␤ 1 integrin to explain platelet activation and aggregation by rhodocytin can be considered very unlikely.
Without using whole platelets, we have established a cellfree inhibition assay as a tool to search for an inhibitor of the integrin-collagen interaction on the molecular level without any interfering cellular reactions that occur on platelets during or after their activation. Furthermore, the use of soluble ␣ 2 ␤ 1 integrin instead of detergent-extracted wild-type ␣ 2 ␤ 1 integrin even allows us to work not only in a cell-free but also in a detergent-free test system. Therefore, we could identify the disintegrin LMW-CI from C. rhodostoma venom, which specifically binds to ␣ 2 ␤ 1 integrin in an RGD-independent manner, thereby inhibiting the interaction of ␣ 2 ␤ 1 integrin with immobilized, monomeric type I collagen. N-terminal sequencing of LMW-CI revealed its identity with rhodocetin (16). We called this inhibitor low molecular weight Calloselasma inhibitor to distinguish it from another ␣ 2 ␤ 1 integrin inhibiting activity of the C. rhodostoma venom that was found in an earlier eluate fraction, i.e. higher molecular mass fraction, of a size exclusion column. However, we have not yet characterized the latter one, which we referred to as high molecular weight Calloselasma inhibitor. Although binding with high affinity and specificity, the LMW-CI/rhodocetin does not need any divalent cations to be bound by the integrin. Its native three-dimensional structure, which is stabilized by disulfide bridges, is essential for ␣ 2 ␤ 1 bindung. Unlike the other high affinity collagen ligands of ␣ 2 ␤ 1 integrin, LMW-CI/rhodocetin lacks a collagenous triple helical conformation. Nevertheless, it binds to ␣ 2 ␤ 1 integrin avidly and even competes with collagen efficiently. Because we had included protease inhibitors into the test assay and proved absence of protease activities in the LMW-CI/rhodocetin preparation by zymogram, the inhibitory effect of LMW-CI/rhodocetin is not caused by proteolytic activity.
LMW-CI/rhodocetin is a heterodimer with an apparent molecular mass of 27 kDa consisting of two subunits of 16 and 14 kDa, which are firmly attached. Despite lacking any covalent, intercatenary disulfide bridges (16), the two subunits remained associated under the harsh denaturation conditions of the SDS-PAGE sample buffer, containing 2% SDS. However, reduction of intracatenary disulfide bridges leads to destruction of the tertiary structure and subsequentially to the dissociation of the two subunits. Judging the native tertiary structure by its capability to bind to ␣ 2 ␤ 1 integrin, we found that the native tertiary structure of rhodocetin, which is required for integrin binding, is lost at much lower concentrations of reducing agents than the quartenary structure, detected as dissociation of the two subunits in SDS-PAGE under nonreducing conditions.
Interestingly, both rhodocytin/aggretin and LMW-CI/rhodocetin belong to a family of snake venom proteins that bear similarity to the carbohydrate recognition domain of C-type lectins (16,31). However, platelet-derived wild-type ␣ 2 ␤ 1 integrin and recombinantly expressed, soluble ␣ 2 ␤ 1 integrin bind equally well to immobilized LMW-CI/rhodocetin, although the two integrins may vary in their glycosylation pattern of Nlinked carbohydrate side chains, suggesting that the interaction of the novel disintegrin LMW-CI/rhodocetin bases on a protein-protein interaction. This direct interaction then causes the inhibition of collagen binding to ␣ 2 ␤ 1 integrin.
It is noteworthy that LMW-CI/rhodocetin does not bind to ␣ 1 ␤ 1 integrin, the other collagen-binding integrin with a widespread tissue distribution. Nor does this disintegrin interact with ␣ 3 ␤ 1 integrin. Therefore, LMW-CI/rhodocetin differs from other, mainly RGD-dependent disintegrins in its unique specificity toward ␣ 2 ␤ 1 integrin. Interestingly, LMW-CI/rhodocetin does not require divalent cations to bind to ␣ 2 ␤ 1 integrin, because, in contrast to other integrin ligands, deprivation of divalent cations by EDTA does not abolish ␣ 2 ␤ 1 binding to LMW-CI/rhodocetin. This suggests a binding mechanism distinct from the integrin binding mechanism to collagen. Nevertheless, LMW-CI can completely abolish ␣ 2 ␤ 1 integrin binding to collagen. Either LMW-CI binds at a site within ␣ 2 ␤ 1 integrin distinct from the ligand binding site that leads to a conformational change and to an allosteric inactivation of the integrin, or LMW-CI binds to a site within ␣ 2 ␤ 1 integrin, which is overlapping or even identical to the collagen binding site, thereby inhibiting collagen binding sterically. Future structural studies will help to answer this question.
As a prerequisite for its binding activity to ␣ 2 ␤ 1 integrin, LMW-CI indispensibly needs its native conformation, which is stabilized by disulfide bridges. Further structural insights into LMW-CI were gained by CD spectroscopy. The CD spectrum of LMW-CI is in good agreement with the CD spectrum provided by in Refs. 16. It clearly demonstrated that LMW-CI/rhodocetin does not bear any structural resemblance to a collagenous triple helix, which is typical of high affinity ligands of ␣ 2 ␤ 1 integrin. Still, LMW-CI/rhodocetin avidly binds to ␣ 2 ␤ 1 integrin and efficiently competes with type I collagen.
Having found and characterized LMW-CI/rhodocetin as snake venom disintegrin that specifically recognizes ␣ 2 ␤ 1 integrin and preventing it from binding to its collagen ligand, we eventually returned to whole cells to study the ␣ 2 ␤ 1 -related functions in the cellular context. ␣ 2 ␤ 1 integrin not only is a pivotal trigger in hemostasis, but its widespread distribution on other cell types also suggests a much broader biological role in the organism (17,38). The presence of ␣ 2 ␤ 1 integrin on endothelial cells of newly grown blood capillaries suggests a potential role in angiogenesis (39). Fibroblasts also bear ␣ 2 ␤ 1 integrin and use it to exert mechanical forces to a surrounding collagen gel, which in vivo takes place in connective tissue to maintain the shape of tissues and organs, during wound contraction and scar formation (40). Furthermore, ligand occupancy of ␣ 2 ␤ 1 integrin on fibroblasts and epithelial tumor cells elicits expression of various matrix metalloproteases (MMPs) (40), such as interstitial collagenase (MMP-1) (41,42), stromelysin-1 (MMP-3) (43), collagenase-3 (MMP13), and membranebound metallomatrixproteinase-1 (MT1-MMP, MMP14) (44). The latter one itself proteolytically activates gelatinase A (MMP-2) (40). ␣ 2 ␤ 1 integrin-triggered secretion of MMPs is a key point in tumor invasion and metastasis, because these proteases degrade extracellular matrix proteins, among them basal membrane proteins, thus opening the path for invading tumor cells. To manipulate such ␣ 2 ␤ 1 -triggered effects, LMW-CI/rhodocetin may be a valuable tool because of its unique specificity and high affinity toward ␣ 2 ␤ 1 integrin. Another advantage of LMW-CI/rhodocetin is its high solubility under physiological conditions compared with the poor solubility of collagen, which because of its high tendency to aggregate and precipitate cannot be applied as soluble inhibitor. Furthermore, because of its independence of divalent cations, LMW-CI is likely to bind in vivo as effectively as in the cell-free test, whereas ␣ 2 ␤ 1 integrin binds less avidly to collagen in vivo because of the presence of ␤ 1 -integrins attenuating Ca 2ϩ ions under physiological conditions. With HT1080 fibrosarcoma cells, which adhere to type I collagen mainly via ␣ 2 ␤ 1 integrin, we demonstrated that LMW-CI indeed inhibits cell adhesion to type I collagen as the initial step of integrin-mediated cell migration, gene activation, and anchorage-dependent growth. LMW-CI/rhodocetin efficiently and completely inhibits ␣ 2 ␤ 1mediated cell adhesion to type I collagen, proving its suitability as specific ␣ 2 ␤ 1 integrin inhibitor in vivo. Thus, LMW-CI may be a useful agent to study and influence ␣ 2 ␤ 1 integrin-triggered cell function, like cell adhesion, cell migration, or secretion of MMPs. Therefore, it may help not only in treating thrombosis but also in treatments aimed to prevent tumor invasion and metastasis.