Non-cytotoxic Cobra Cardiotoxin A5 Binds to αvβ3 Integrin and Inhibits Bone Resorption

Severe tissue necrosis with a retarded wound healing process is a major symptom of a cobra snakebite. Cardiotoxins (CTXs) are major components of cobra venoms that belong to the Ly-6 protein family and are implicated in tissue damage. The interaction of the major CTX from Taiwan cobra, i.e. CTX A3, with sulfatides in the cell membrane has recently been shown to induce pore formation and cell internalization and to be responsible for cytotoxicity in cardiomyocytes (Wang, C.-H., Liu, J.-H., Lee, S.-C., Hsiao, C.-D., and Wu, W.-g. (2006) J. Biol. Chem. 281, 656-667). We show here that one of the non-cytotoxic CTXs, i.e. CTX A5 or cardiotoxin-like basic polypeptide, from Taiwan cobra specifically bound to αvβ3 integrin and inhibited bone resorption activity. We found that both membrane-bound and recombinant soluble αvβ3 integrins bound specifically to CTX A5 in a dose-dependent manner. Surface plasmon resonance analysis showed that human soluble αvβ3 bound to CTX A5 with an apparent affinity of ∼0.3 μm. Calf pulmonary artery endothelial cells, which constitutively express αvβ3, showed a CTX A5 binding profile similar to that of membrane-bound and soluble αvβ3 integrins, suggesting that endothelial cells are a potential target for CTX action. We tested whether CTX A5 inhibits osteoclast differentiation and bone resorption, a process known to be involved in αvβ3 binding and inhibited by RGD-containing peptides. We demonstrate that CTX A5 inhibited both activities at a micromolar range by binding to murine αvβ3 integrin in osteoclasts and that CTX A5 co-localized with β3 integrin. Finally, after comparing the integrin binding affinity among CTX homologs, we propose that the amino acid residues near the two loops of CTX A5 are involved in integrin binding. These results identify CTX A5 as a non-RGD integrin-binding protein with therapeutic potential as an integrin antagonist.

and constitute ϳ50% of the weight of cobra venom. CTXs are believed to play a critical role in cobra venom toxicity (5). We have shown that CTXs bind to glycosaminoglycans with specificity and are retained on the membrane surface for action (6 -8). Interestingly, one of the CTXs, CTX A3, interacts with sulfatide to form a pore and becomes internalized to further target mitochondria in cardiomyocytes and H9C2 myoblasts (9 -12). The mechanisms for CTX-induced perturbation of the wound healing process and severe tissue damage are unknown. It is not clear whether there are cellular receptors for CTXs, although other cobra venom components such as secretory phospholipase A 2 are known to have diverse targets for their actions by involving glycosaminoglycans, protein receptors, and membrane lipids (13,14).
CTXs are all ␤-sheet basic polypeptides of 60 -62 amino acid residues with a three-fingered loop-folding topology and are members of the Ly-6 protein family. Members of this family share one or several repeat units of the Ly-6 domain, which is defined by a distinct disulfide bonding pattern of between 8 and 10 cysteine residues (15). This protein family can be divided into two subfamilies. One subfamily includes the secreted single domain snake cytotoxins (e.g. CTXs and neurotoxins), which possess only eight cysteines and no glycosylphosphatidylinositolanchoring signal sequence. Another subfamily comprises glycosylphosphatidylinositol-anchored glycoprotein receptors with 10 cysteine residues (e.g. urokinase-type plasminogen activator (uPA) receptor with three Ly-6 domains (16).
Recent advances in understanding the structure and mechanism of toxin interaction with integrins (29 -31) have allowed the development of small molecule antagonists with therapeutic potential (32)(33)(34). For instance, echistatin has been shown to be a potent inhibitor of bone resorption both in culture and in an animal model by directly interacting with ␣v␤3 integrin via the RGD sequence (35). ␣v␤3 in osteoclasts, multinucleate cells (MNCs) formed by the fusion of mononuclear pro-genitors of the macrophage family, is the key integrin in mediating the formation of the fused polykaryon in the late stage of osteoclast differentiation and in osteoclast adhesion during bone resorption (36,37). A peptidomimetic antagonist of ␣v␤3 based on the RGD sequence inhibits bone resorption in vitro and prevents osteoporosis in vivo (38).
We have previously reported that the uPA receptor binds to several integrins and that this interaction plays a critical role in signal transduction from uPA and the uPA receptor (39). Based on the structural similarity between CTX and the uPA receptor, we hypothesized that integrins may be involved in the action of CTXs. In the present study, we demonstrate that several CTXs specifically bound to ␣v␤3 integrins. Of all CTX homologs with known three-dimensional structures tested, non-cytotoxic CTX A5 from Taiwan cobra (Naja atra) exhibited the strongest binding to human ␣v␤3 at a K d of ϳ0.3 M. We also show that CTX A5 binding to ␣v␤3 effectively inhibited bone resorption and differentiation of murine osteoclasts. Comparison of the binding affinity of all studied CTXs for ␣v␤3 revealed a potential role for the amino acid residues located at the two loops of the three-fingered CTXs. These results suggest a potential role for integrins in the actions of CTXs and that non-RGD CTX A5 has therapeutic potential as an integrin antagonist.

EXPERIMENTAL PROCEDURES
Materials-CTXs were purified from the crude venom of N. atra, Naja mossambica, and Naja nigricollis by SP-Sephadex C-25 ion exchange chromatography, followed by high pressure liquid chromatography on a reverse-phase C 18 column as described previously (40,41). The purity of the studied CTXs, including rhodamine-labeled CTX A5 (12), was checked by mass spectrometric analysis routinely during the purification process to be consistent with the known protein sequences. The fibrinogen ␥-chain C-terminal globular domain (␥C; amino acids 151-411) was synthesized in bacteria as an insoluble protein and refolded as described previously (42). Recombinant soluble ␣v␤3 was synthesized in CHO-K1 cells using the soluble ␣v and ␤3 expression constructs provided by Dr. Tim Springer (Center for Blood Research, Boston, MA) and purified by nickel-nitrilotriacetic acid affinity chromatography as described (20). Mouse anti-human ␣v␤3 integrin monoclonal antibody LM609 was from Chemicon International, Inc. (Temecula, CA). Medium 200 (catalog no. M-200-500) and low serum growth supplement (catalog no. S-003-10) were purchased from Cascade Biologics for testing cell proliferation under low serum conditions. Rabbit anti-human ␤3 integrin polyclonal antibody was purchased from Chemicon International, Inc. Recombinant mouse RANKL and macrophage colony-stimulating factor (M-CSF) were obtained from R&D Systems (Minneapolis, MN). The BD BioCoat TM Osteologic TM bone cell culture multitest system was obtained from BD Biosciences. The GRGDSP and GRGESP peptides were purchased from American Peptide Co. Inc. (Sunnyvale, CA). Naphthol AS-MX phosphate and fast red violet LB salt were obtained from Sigma.
Cells-Chinese hamster ovary (CHO) cells expressing human ␤3 integrin (designated ␤3-CHO cells) have been described previously (43). ␤3-CHO cells express a hamster ␣v/human ␤3 integrin hybrid. As a control, CHO cells were transfected with the pBJ-1 vector together with the neomycin gene and were selected for G418 resistance (designated mock-transfected CHO cells). Calf pulmonary artery endothelial (CPAE) cells (CCL-209, between passages 17 and 23) were obtained from American Type Culture Collection and cultured in minimal essential medium supplemented with 10% fetal bovine serum, and penicillin/ streptomycin. RAW 264.7 murine monocytic cells (American Type Culture Collection TIB-71) were maintained in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal bovine serum, 100 units/ml penicillin G, and 100 g/ml streptomycin.
Binding of Soluble ␣v␤3 Integrin to CTXs-Binding assays were performed as described previously (43). Briefly, 96-well microtiter plates were coated with 100 l of 0.1 M NaHCO 3 (pH 9.4) containing CTXs at 1ϳ5 M and incubated for 16 h at 4°C. The remaining protein-binding sites were blocked by incubation with 0.1% bovine serum albumin (BSA; Sigma) for 1 h at room temperature. Soluble horseradish peroxidaseconjugated ␣v␤3 in 50 l of HEPES/Tyrode's buffer supplemented with 10 mM cation or 5 mM EDTA was added to the wells and incubated at room temperature for 1 h. After non-bound soluble integrins were removed by rinsing the wells with the same buffer, bound integrins were quantified by measuring the absorbance at 450 nm and developed by adding the 3,3Ј,5,5Ј-tetramethylbenzidine substrate for horseradish peroxidase.
Cell Adhesion, Proliferation, and Viability Assays-The 96-well microtiter plates were coated with CTXs as described above. In adhesion assays, cells (10 5 cells/well) in 100 l of HEPES/Tyrode's buffer supplemented with 1 mM cation were added to the wells and incubated at 37°C for 1 h. After non-bound cells were removed by rinsing the wells with the same buffer, bound cells were quantified by measuring endogenous phosphatase activity (44). For CTX-induced cytotoxicity, we defined cell proliferation for the long time effect (24 or 48 h) and cell viability for the short time effect (2 h). CPAE cells (1 ϫ 10 5 cells/well) were incubated with CTXs at the indicated concentrations in 96-well cell culture plates at 37°C for 48 h. Proliferation of CPAE cells was measured using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. In brief, after culture, MTT solution (2 mg/ml) was added to each well and incubated for 1 h. The precipitated dye was solubilized by dimethyl sulfoxide (Sigma), and the absorbance at 570 nm was measured. The absorbance measured at 570 nm was proportional to the number of living cells in culture.
Inhibition of CTX-␣v␤3 Interaction by ␥C-␥C (10 g/ml) (45) was immobilized on the 96-well plates. Soluble ␣v␤3 integrin was added to the wells in the presence of CTXs at various concentrations. Bound integrin was detected in binding assays as described above.
Surface Plasmon Resonance Binding Studies-Surface plasmon resonance (SPR) experiments were performed on Biacore X instruments (Biacore International AB, Uppsala, Sweden). ␣v␤3 integrin was immobilized on a Biacore CM5 chip at a constant flow rate of 5 l/min by the amine coupling method (46). In general, the dextran surface was activated by injection of a 1:1 mixture of 0.1 M N-hydroxysuccinimide and 0.39 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride for 7 min ␣v␤3 (in 10 mM acetate (pH 4.0) and preincubated with 1 mM Mn 2ϩ ) was injected over the activated surface for 7 min. Finally, 35 l of 1 M ethanolamine was used to deactivate the surface. The difference in the base line before and after immobilization accounts for the integrin immobilized. One resonance unit corresponds to 1 pg of protein accumulated per mm 2 . The reference surface was activated by the same activation/deactivation procedure except without injection of the integrin solution. Because of the relatively small size of CTXs (ϳ7 kDa) compared with integrin (ϳ200 kDa), a high surface density integrin surface (ϳ5000 resonance units) was prepared to observe the dominant binding response. All of the observed binding responses were found to be fast; and therefore, there was no significant mass transport effect under the experimental conditions.
All binding experiments were performed at 25°C at a continuous flow rate of 40 l/min with HEPES-buffered saline (10 mM HEPES-NaOH (pH 7.4), 150 mM NaCl, and 0.005% (v/v) Tween 20). The equilibrium binding response of CTX, which was the response difference between the integrin-immobilized surface and the reference surface, was used to estimate the binding affinity. Upon titration with various CTX concentrations, the binding response was not a simple Langmuir equilibrium binding isotherm. The binding affinities were determined from the slopes of the Scatchard plots at the reported concentration range.
Osteoclast Differentiation Assays-RAW 264.7 cells (2 ϫ 10 3 cells/ well) were cultured in 96-well plates in the presence of RANKL (50 ng/ml) and M-CSF (10 ng/ml) for 5 days with a change to fresh medium every 2 or 3 days. Osteoclast-like cells were evaluated by tartrate-resistant acid phosphatase (TRAP) staining. TRAP staining was performed according to Evans et al. (47) with a slight modification. After the culture period, adherent cells were fixed in 10% formaldehyde for 10 min. Cells were then stained for 30 min for TRAP activity with 0.1 mg/ml naphthol AS-MX phosphate as a substrate and 0.6 mg/ml fast red violet LB salt as a stain. This staining was performed in 0.1 M sodium acetate buffer (pH 5.0) containing 50 mM sodium tartrate, followed by counterstaining with hematoxylin. TRAP-positive MNCs with more than three nuclei were considered to be osteoclast-like cells and were counted under a microscope. For measurement of TRAP intensity, the plates were scanned by a transparent light scanner, and the red color image was extracted from the scanned image using ImageJ and is represented here as TRAP intensity (48).
Bone Resorption Assays-Bone resorption assays were performed according to Ariyoshi et al. (49) with a slight modification. RAW 264.7 cells were cultured for 10 days with RANKL (50 ng/ml), M-CSF (10 ng/ml), and various concentrations of CTX A5 on BD BioCoat TM Osteologic TM multitest slides, which are calcium phosphate ceramic thin film-coated quartz disks, to quantify osteoclastic cell-mediated mineral resorption. After the culture period, cells were removed using 6% NaOCl and 5.2% NaCl. The resorption area was observed under a light microscope and counted using ImageJ.
Osteoclastic Cell Viability-Cell viability was evaluated by the MTT assay. RAW 264.7 cells were plated in 96-well plates at a concentration of 2 ϫ 10 3 cells/well 1 day before the experiment and then stimulated FIGURE 1. Binding of CTXs from Taiwan cobra venom as measured by enzymelinked immunosorbent and SPR binding assays. A, soluble ␣v␤3 integrin was incubated with various CTXs (5 M) coated on 96-well microtiter plates for 1 h at room temperature in HEPES/Tyrode's buffer supplemented with 10 mM Mg 2ϩ . Flavoridin (FLV; 20 l/ml) and BSA (1%) were used as controls. B, the SPR experiments were performed using a Biacore X instrument. The Scatchard plots demonstrate CTX binding to immobilized ␣v␤3 (10,000 resonance units (RU)) with preincubation with 1 mM Mn 2ϩ before immobilization procedures. C, shown are representative response traces of the indicated CTXs (2, 5, and 10 M) after passage over an immobilized ␣v␤3 surface (5000 resonance units). Data are expressed as the relative response after subtraction of the background signal recorded on a reference surface. Note that CTXs A2 and A4 exhibited larger responses compared with CTX A3 even though they bound more weakly to integrin according to the Scatchard plots.   with RANKL and M-CSF. Stimulated cells were cultured in the presence or absence of CTX A5 for 5 days.
Confocal Microscopy-Rhodamine-labeled CTX A5 was prepared as described previously (12). Macrophages were cultured with RANKL and M-CSF for 5 days. To monitor the location of CTX A5, 5 M rhodamine-labeled CTX A5 was applied to the culture in Hanks' balanced salt solution (HBSS) for 5 min at 37°C before fixation in formaldehyde. After CTX A5 was applied and then washed with 1% BSA in HBSS (10), ␤3 integrin was detected using anti-␤3 integrin polyclonal antibody dissolved in HBSS with 1% BSA and washed twice with HBSS with 1% BSA and once with HBSS. The fluorescein isothiocyanate-conjugated secondary antibody was then applied. Co-localization is depicted in yellow. Images were recorded using a Zeiss LSM 510 confocal microscope with a full-width half-maximum of 0.6 m to set a pinhole size.
Scanning Electron Microscopy-After removing cells with NaOCl and NaCl as described above, a BD BioCoat TM Osteologic TM multitest slide was treated by ultrasonication in water for 0.5 h. It was then allowed to dry in vacuum before being coated with gold and examined by scanning electron microscopy (Hitachi S-4700) at an accelerating voltage of 5 kV and an emission current of 9500 nA (50).

RESULTS AND DISCUSSION
We tested whether recombinant soluble ␣v␤3 integrin binds to immobilized CTXs from Taiwan cobra venom in an enzyme-linked immunosorbent assay-type assay. As shown in Fig. 1A, all CTXs, i.e. CTX A1 to A6, exhibited detectable binding compared with the positive control flavoridin and the negative control BSA. Soluble ␣v␤3 bound to immobilized CTX A5 most efficiently. We determined how strongly the soluble CTXs bound to immobilized ␣v␤3 in SPR studies. We immobilized soluble ␣v␤3 by amine coupling on a biosensor chip and measured the CTX binding to the chip surface. SPR measurement (Fig. 1B) confirmed the high affinity of CTX A5 for ␣v␤3 with an apparent dissociation constant of ϳ0.3 M. As shown in the biosensor response and related Scatchard plots (Fig. 1, B and C), CTXs bound to ␣v␤3 with different affinities (CTX A5 Ͼ A3 Ͼ A2 and A4 Ͼ A1 and A6) in a dose-dependent manner. For CTX A6, which exhibited the weakest response, the Scatchard plot at low concentration range indicated a binding affinity of ϳ15 M. Therefore, there is a difference of ϳ2 orders of magnitude in terms of CTX affinity for ␣v␤3. Although other CTXs such as CTX A3 target sulfatides and induce CTX pore formation in the plasma membrane of cardiomyocytes, CTX A5 lacks observable cytotoxicity (40 -41) and is thus known as a cardiotoxin-like basic polypeptide. We examined whether CTX A5 acts as an integrin ligand in more biological systems.
We first studied whether CHO cells expressing recombinant ␣v␤3 integrin (designated ␤3-CHO cells) adhere to the CTX A5-coated surface. We found that ␤3-CHO cells adhered to the CTX A5-coated surface in a dose-dependent manner at higher levels compared with control CHO cells transfected with vector only (designated mock-transfected CHO cells) (Fig. 2). These results suggest that ␣v␤3 is involved in the binding of CTX A5 to the cell surface. Because mock-transfected CHO cells showed low level adhesion, it is possible that other integrins or non-integrin receptors may also be involved in the binding of CTX A5 to the cell surface. Soluble ␣v␤3 integrin bound to immobilized CTX A5, but not significantly to CTX A6 (Fig. 3B; for a sequence comparison between CTXs A5 and A6, see Fig. 3A). Saturated binding was observed for CTX A5 with an apparent affinity of ϳ1 M. Mn 2ϩ (10 mM) or Mg 2ϩ (10 mM) enhanced its binding to CTX A5 by ϳ5and 2-fold, respectively, and EDTA significantly reduced the binding of ␣v␤3 integrins to CTX (Fig.  3C). Mn 2ϩ and Mg 2ϩ enhanced the binding of soluble ␣v␤3 to CTX A6 and the flavoridin disintegrin as well. These results suggest that the cation dependence of CTX A5 is similar to that of other known integrin ligands.
We tested whether CTXs compete for binding to soluble ␣v␤3 integrin with ␥C. As shown in Fig. 4, CTX A5 competed with ␥C for binding to ␣v␤3, but CTX A6 did not, consistent with the observation that CTX A6 shows weak binding to ␣v␤3. These results suggest that CTXs share a common binding site in ␣v␤3 with ␥C.
We investigated whether the difference in integrin binding among CTXs is related to their cytotoxicity. CPAE cells constitutively express ␣v␤3 integrin on their surface. As shown in Fig. 5A, CPAE cells bound to a CTX-coated plate in a manner similar to ␤3-CHO cells. This suggests that ␣v␤3 is indeed a predominant receptor for CTX binding to CPAE cells. In this experiment, we also included CTX A3 for comparison because we have shown recently that CTX A3-induced cell death of cardiomyocytes depends on the binding of CTX A3 to sulfatides, a glycosphingolipid located at the outer leaflet of plasma membranes (10). Interestingly, CTX A3 and flavoridin blocked CPAE cell proliferation in a dose-dependent manner, but CTXs A5 and A6 did not (Fig. 5, B and  D). Under low serum conditions, however, CTX A6, but not CTX A5, exhibited observable cytotoxicity (Fig. 5B, inset). Although CTX A5 induced the strongest cell attachment, it did not show detectable cytotoxicity, consistent with the non-cytotoxic property of CTX A5 despite its binding to integrin.
We should emphasize that CTX A3-induced cytotoxicity is cell-dependent and that its targets might also vary depending on the cell system studied. For instance, anti-sulfatide antibody did not block the cytotoxicity of CTX A3 in CHO cells (data not shown) even though the same antibody works well with the effect of CTX A3 in cardiomyocytes and H9C2 myoblasts (10). Surprisingly, despite the binding of CTX A3 to ␣v␤3 integrin in CPAE cells, anti-␣v␤3 integrin antibody LM609 exhibited only a slight protection effect on CPAE cell viability (Fig. 5C). Recent progress in understanding the function of integrins has revealed a complicated regulation mechanism among integrins, the glycosphingolipid domain, and endocytosis (51,52). Considering that CTX A3 binds to integrin and the glycosphingolipid domain in the plasma membrane and that it can also be internalized into mitochondria via a still unknown mechanism, future investigations on CTX A3-induced CHO cell death should shed light on the cell signaling process involving integrins and the lipid domain.
Although CTX A5 does not have an RGD motif, we investigated whether CTX A5 affects bone resorption via its binding to ␣v␤3 integrins. Using RANKL to stimulate murine osteoclast differentiation, we show in Fig. 6 that CTX A5 inhibited osteoclast formation by perturbing the polykaryon fusion required in the late differential stage of osteoclast formation. CTX A5 (10 M) reduced the formation of TRAP-positive MNCs to the basal level (Fig. 6A), but did not show significant cytotoxicity (Fig. 6B). The blocking effect could be observed only when CTX A5 was present in the late stage of osteoclast formation, i.e. when CTX A5 was present either during the entire period of osteoclast formation or during days 4 -6 ( Fig. 6C, lower panel), when the expression of significant ␣v␤3 integrin starts for MNC formation. There was no significant difference in TRAP intensity in the early stage of osteoclast formation (Fig. 6C, upper panel). Because there was no inhibitory effect when CTX A5 was present during the initial 1-3 days, the result is consistent with the idea that CTX A5 binds to ␣v␤3 integrin to block differentiation.
␣v␤3 integrin plays a role not only in osteoclast differentiation, but also in osteoclast adhesion during bone resorption. We studied the effect of CTX A5 binding to osteoclasts on their bone resorption activity. Fig. 7 shows the activity as measured by the resorption area determined by light microscopy and scanning electron microscopy. Significant inhibition (ϳ50%) was detected at a CTX A5 concentration of ϳ2.5 M. Finally, a confocal microscopic study was performed to determine whether CTX A5 can indeed co-localize with ␣v␤3 as visualized using anti-␤3 integrin antibody. We found that the integrin overlapped with rhodamine-labeled CTX A5 (Fig. 8B), suggesting that CTX A5 indeed binds to ␣v␤3 in osteoclasts. We propose that CTX A5 inhibits murine osteoclast differentiation and resorption by binding to ␣v␤3.
These results also predict that CTX-integrin interaction may be a potential therapeutic target for tissue degeneration induced by CTXs even though the role that integrin binding plays in the CTX-perturbed wound healing process is not clear at this time. CTXs constitute ϳ50% of the weight of the Taiwan cobra venom toxin. The local concentration of CTXs easily reaches a micromolar range at the bitten area because each bite will inject ϳ20 -570 mg of venom (53). Thus, the binding of CTXs to ␣v␤3 integrin at K d values of ϳ0.3 to 15 M should be biologically relevant in the wound tissue. However, CTX A5 induced cell adhesion of CPAE cells (which expressing ␣v␤3) without detectable cytotoxic effects. Thus, unlike many other disintegrins that are known to perturb cell proliferation by binding to integrin on the membrane surface, the binding of CTXs to cells through integrins is not sufficient to induce their cytotoxic effect.
We detected only weak binding of CTX A6 to ␣v␤3 integrin. It is interesting that CTX A6, a CTX homolog identified only cobras caught in the eastern part of Taiwan (54), exhibited weak ␣v␤3 binding activity. CTX A6 was not cytotoxic in cardiomyocytes, CHO-K1 cells, and H9C2 myoblasts. 4 We tested the effect of CTXs on CPAE cell proliferation as a measure of their cytotoxicity. CTX A6 did not have detectable inhibitory effects on CPAE cell proliferation. However, under low serum conditions, the cytotoxicity of CTX A6 could be detected (Fig. 5). Sequence comparison between CTXs A6 and A3 (Fig. 9A) showed that the amino acid residues located at loop I (Lys-5, Val-7, and Leu-9), loop II (Thr-29), or the tight turn (Pro-15) of CTX A3 might be involved in  CTX-integrin interaction, explaining their difference in integrin binding affinity.
It is interesting to point out that the loop I conformation of CTX A6 adopts a type VI turn with a cis-peptide bond between the two prolines of the conserved Leu-Ile-Pro-Pro-Phe sequence for group I CTXs (54). In contrast, CTX A5 was previously classified as a group II CTX based on the absence of a proline residue at position 9. As a consequence of the conformational difference at position 9, the tip of loop I protrudes to the FIGURE 9. Structure and activity relationship of CTX-integrin interaction. A, role of loop I in the differential binding of CTXs A3 and A6 to immobilized integrin as detected by SPR binding measurements. The concentration ranges used were 0.25-10 and 1-30 M for CTXs A3 and A6, respectively. The shaded amino acid residues in the sequence and the side chain in the three-dimensional structure represent the residues possibly involved in the interaction. RU, resonance units. B, role of loop II in the differential binding of T␥ (CTX Tr) and CTX M2 to immobilized integrin as detected by SPR binding measurements. The concentration ranges used were 0.5-10 and 0.2-10 M for T␥ and CTX M2, respectively. C, stereo view of the three-dimensional structure of CTX A5 emphasizing the locations of amino acid residues suggested to affect CTX-integrin interaction. The backbone of CTX A6 is also shown overlaid on CTX A5 to indicate the effect of the cis-Pro peptide bond on the loop I conformation. The boxed amino acid residues in the sequence represent residues of CTX A5 overlapped by residues of CTX A3 and T␥ proposed to be involved in CTX-integrin interaction. Three additional residues, i.e. Glu-17, Asp-59, and Arg-38, are also highlighted in the three-dimensional structure to show possible sites responsible for the metal dependence. The three-dimensional structures of CTX A3 (monomer A; code 1XT3), T␥ (monomer A; code 1TGX), and CTX A5 (monomer A; code 1KXI) are from the Protein Data Bank, and the sequences of CTX A3, CTX A6 (code 1UG4), T␥, CTX M2, and CTX A5 are aligned based on their structures. opposite side of the slightly concave flat molecules between group I and II CTXs (Fig. 9C). We predict that the structure of loop I of CTXs may be responsible for integrin binding.
Group I CTXs with a cis-Pro peptide bond in the loop I region are usually purified from African cobras. To test whether the loop I region is mainly responsible for CTX binding to ␣v␤3 integrin, we performed a SPR binding study on two group I CTXs, i.e. T␥ (CTX ␥ purified from N. nigricollis) and CTX M2 (Fig. 9B). We select these CTXs because their amino acid sequences are different only in the specific region at the tip of loop II. In contrast to the SPR binding exhibited by other group II CTXs such as CTXs A5 and A3, group I CTXs bound to ␣v␤3 with two-phase behavior. Such binding behavior could be fitted by a twostage reaction with a conformational change model (BIAevaluation Version 4.1, Biacore International AB) (data not shown). It is therefore tempting to speculate that the binding of integrin to group I CTXs at loop I might induce a conformational change in the CTX molecules, e.g. cis,trans-isomerization of Pro-9, to allow additional binding at the loop II region. The interpretation is also consistent with the fact that CTX M2 bound to integrin in a significantly different manner compared with T␥. The CTX M2 and T␥ molecules show a structural difference only in the loop II region. To reconcile the weak binding of CTX A6 (group I), we observed that those group I CTXs with detectable two-phase binding behavior have a positively charged residue (Arg-28) near the loop II region, whereas CTX A6 has a hydrophobic residue (Val-28). The apparent low affinity of CTX A1 can also be explained by the presence of a negatively charged residue (Asp-30) in this region. Interestingly, the strongest integrin-binding CTX, CTX A5, also has positively charged residues (Lys-28 and Lys-29) in this region. Based on these observations, we suggest that the hydrophobic domain near the tip of loop I and the charged residues flanking the hydrophobic loop II region may be involved in CTX-integrin interaction.
It should be noted that we did not find any acidic residues near the two loop regions of CTX A5. Given the observed metal dependence and inhibition by fibrinogen, it is possible that another acidic residue of CTX A5 such as Glu-17 or Asp-59, which is ϳ15 Å away from Arg-38 (Fig.  9C), might form a three-dimensional motif to substitute for the linear RGD motif to account for the effect (30). Although such a model can be tested in future studies by site-directed mutagenesis, our result indicates that there is indeed a correlation between the structure and integrin binding functions.
We have shown that CTX A5 has a potential biomedical application as an integrin antagonist in the bone resorption model in this study. ␣v␤3 integrin is expressed in tumor cells, wounds, and inflammatory tissues and in angiogenic endothelial cells. Therefore, it will be interesting to see whether CTX A5 affects inflammation, tumor growth, and angiogenesis in future studies (55).