ADAM23 Negatively Modulates α_vβ₃ Integrin Activation during Metastasis

Integrins are membrane-spanning heterodimers composed of α and β subunits, both of which having a large extracellular domain, a single transmembrane domain, and a short cytoplasmic tail (1). Many integrins are expressed with their extracellular domains in a default low-affinity ligand-binding state (nonactive state; ref. 2). However, in response to cellular stimulation, the conformation of these receptors is changed in a process often termed integrin activation (3).

Changes in integrin expression level and activation state have been extensively documented in tumor cells and are thought to contribute to neoplastic progression (4). Altered expression of α_vβ₃ integrin has been detected in different types of tumors, including breast (5, 6), prostate (7), ovary (8), melanomas (9, 10), and gliomas (11), and this expression has been correlated with an aggressive phenotype and metastatic dissemination. Moreover, expression of activated, but not nonactivated, α_vβ₃ integrin in the MDA-MB-435 cell line strongly promotes metastasis formation by enhancing the interaction between circulating tumor cells and platelets and promoting tumor cell arrest in the vasculature (12, 13).

The ADAMs (a disintegrin and metalloprotease domain) constitute a family of type I transmembrane glycoproteins with a common structural organization, which includes a metalloprotease and disintegrin domain (14–16). The disintegrin domain of ADAM proteins is ∼ 90 amino acids long and is named for its presence in the snake venom metalloproteases, where it is involved in binding to integrins present on the surface of platelets, thereby blocking platelet aggregation (14, 15). Recent reports analyzing ADAM-integrin interaction in a cellular context indicate that ADAM proteins negatively modulate integrin-mediated cell proliferation, adhesion, and migration in patterns dictated, in part, by the integrin binding profiles of their disintegrin domains (17–20).

ADAM23 exhibits the typical structure of ADAM family members, although its metalloprotease domain is inactive (21). Analysis of ADAM23 binding to integrins has revealed a specific interaction with α_vβ₃ mediated by the disintegrin domain (22). The ADAM23 gene is frequently silenced in breast (23), gastric (24), pancreatic (25), and head and neck (26) tumors as well as in gliomas (27). In breast tumors, we showed that the ADAM23 gene is frequently silenced by promoter hypermethylation and that primary breast tumors in more advanced stages show a higher degree of DNA methylation, suggesting that silencing of ADAM23 might be associated with the acquisition of a metastatic phenotype (23).

Here, we investigated the possibility that interaction between ADAM23 and α_vβ₃ might negatively modulate α_vβ₃ integrin activation during metastatic progression. According to our hypothesis, loss of ADAM23 expression, frequently observed in advanced tumors, would promote α_vβ₃ activation and contribute to the acquisition of a metastatic phenotype by enhancing cell migration and adhesion mediated by α_vβ₃ integrin activation. Our results strongly support a functional role of ADAM23 during metastatic progression by negatively modulating α_vβ₃ integrin activation and suggest that silencing of the ADAM23 gene in primary breast tumors can be used as a complementary marker for metastasis risk assessment.

Cell lines and cell culture. MDA-MB-435 cell line was obtained from the American Type Culture Collection and grown at 37°C, under 5% CO₂, using RPMI 1640 (Invitrogen) supplemented with 10% fetal bovine serum (Cultilab) and 1% l-glutamine (Sigma).

Proteins, peptides, and antibodies. Vitronectin and fibronectin were purified from human plasma of healthy donors. Peptides (Multiple Peptide Systems) were RGD-containing (KPQVTRGDVFTMPE) and control (AEEELCSGKPFDAF). Experiments were done with the α_vβ₃ integrin-specific monoclonal antibodies BV4 (Abcam), LM609 (Chemicon), and 23C6 (Santa Cruz Biotechnology), with a conformational sensitive anti-β₃ monoclonal antibody AP5 (GTI Diagnostics) and with normal mouse IgG (Sigma). Antibodies against β₁ integrin (JB1A; Chemicon) and α_vβ₅ integrin (P1F6; Chemicon) were also used in control experiments.

GST pull-down assay. Cells were lysed in PBS containing 100 mmol/L n-octyl-β-d-glucopyranoside, 1 mmol/L CaCl₂, 1 mmol/L MgCl₂, 1 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L Nem, and 50 mmol/L Tris-HCl (pH 8.0). Lysates were cleared by centrifugation and aliquots containing 500 μg protein were incubated for 2 h at 4°C with Sepharose beads covalently linked with recombinant GST or ADAM23 disintegrin-GST (22). Beads carrying bound proteins were recovered by centrifugation, washed with lysis buffer, suspended in sample buffer [100 mmol/L Tris-HCl (pH 6.8), 4% SDS, 20% glycerol, and 0.2% bromophenol blue], and loaded on 6% SDS-PAGE gels. Integrin subunits were detected with polyclonal antibodies specific for β₃ or β₁.

Generation of short hairpin RNAs. A plasmid vector-based strategy was used to suppress ADAM23 expression. Short hairpin RNA (shRNA) constructs were designed targeting ADAM23 (GenBank NM_003812) coding regions 692 to 710 nucleotides (shRNA-04) and 1,717 to 1,735 nucleotides (shRNA-25). Pairs of complementary oligonucleotides (Operon) were annealed and ligated into the Sal I and Xba I restriction sites of the pAVU6+27 vector as described (28). Cells were transfected using Lipofectamine 2000 (Invitrogen) and clones were selected with 1 mg/mL G418 (Promega).

Reverse transcription-PCR. Total RNA was extracted using Trizol (Invitrogen). Reverse transcription was done with 2 μg DNA-free RNA using Superscript II reverse transcriptase (Invitrogen). ADAM23 and GAPDH expressions were analyzed by PCR as described previously (23).

Flow cytometry. Cells (5 × 10⁵) were incubated on ice for 45 min with the primary antibody at the indicated concentrations diluted in HBSS buffer with 0.5% bovine serum albumin and either 1.3 mmol/L CaCl₂ or 500 μmol/L MnCl₂. After incubation with the Alexa 488-linked anti-mouse secondary antibody (Molecular Probes; 1:500) for 30 min on ice, cells were washed and suspended in 0.5 mL HBSS before analysis on a FACSCalibur cytometer using the Cell Quest software (Becton Dickinson).

Migration assay. Haptotatic migration assays were done in Transwell (Corning) plates with 8 μm polycarbonate filters coated with 12.5 μg/mL vitronectin for 2 h at 37°C and blocked with 2.5% bovine serum albumin in RPMI 1640. Cells were seeded in serum-free RPMI 1640 in duplicate (10⁵ per well) and allowed to migrate for 18 h at 37°C and 5% CO₂. The top side of the filters was scraped with cotton swabs, fixed, and stained with 0.1% crystal violet. Cells were counted at ×100 magnification in 20 different optical fields. For inhibition studies, cells were incubated at 4°C for 30 min with the monoclonal antibodies (10 μg/mL).

Adhesion assay. Cell adhesion was done in 48-well plates (Costar) coated with different amounts of purified human vitronectin or fibronectin for 2 h at 37°C and blocked with 2.5% bovine serum albumin (Sigma) for 1 h at 37°C. Cells were plated at 2 × 10⁵ per well. After 90 min at 37°C, nonadherent cells were washed and removed, and remaining adherent cells were fixed with 3.7% formaldehyde in PBS for 15 min at room temperature and stained with 0.1% crystal violet. Quantification of adhesion was carried out using a light microscope with a coupled DP-70 camera (Olympus) at ×200 magnification and cell counting of 20 different optical fields. For inhibition studies, cells were pretreated for 20 min at room temperature with antibodies (10 μg/mL) or synthetic peptides and adhesion was evaluated on 2 μg/mL vitronectin by counting cells at ×100 magnification as described above.

Pulmonary cell arrest. The in vivo tumor cell arrest assay was done by injecting 2 × 10⁶ cells in the lateral tail vein of 9-week-old female CB-17/SCID mice (CEMIB-UNICAMP) and chasing human DNA in the lungs 4 h later by quantitative real-time PCR as described previously (29). Alu sequences, as tracer of human tumoral DNA, and the mouse cellular prion (PrPc) gene, as endogenous control, were analyzed by the absolute quantification method. Lung DNA was amplified in the presence of SYBR Green PCR Master Mix and analyzed using the 7300 Real-time PCR System (Applied Biosystems). All experiments with mice were conducted according to ethic standards and approved by the Animal Experimentation Ethics Committee of the Antonio Prudente Foundation. Statistical analysis employed one-way ANOVA with Bonferroni's comparison test using the GraphPad Prism software.

Tumor samples and patients. Ninety-four ductal invasive breast tumors were obtained from patients treated consecutively at the Hospital do Cancer A.C. Camargo from 1998 to 2001. Tumor samples were collected after informed consent and the institution's ethics committee approved the study. Median age at the time of diagnosis was 59.2 years and the median follow-up time was 78.2 months. Of the 94 patients analyzed, 20 developed distant metastases and 15 died during the follow-up period.

DNA methylation analysis. DNA was subjected to sodium bisulfite treatment using the “CpG Modification Kit” (Intergen, Serologicals). Aberrant DNA methylation was determined by the methylation-specific PCR (30). Methylation-specific PCR primers for region 1 were designed within a region comprising 22 CpG dinucleotides located −1,025 to −549 bp from the transcription start site. Primer sequences and amplification conditions are available upon request. Methylation-specific PCR primers and amplification conditions for region 2 were described previously (24). χ² test or Fisher's exact test was used to examine the association between ADAM23 hypermethylation and clinicopathologic parameters. Disease-specific survival (from the date of initial diagnosis until death due to breast cancer or date of the last follow-up) and distant metastases–free survival (from the date of initial diagnosis until the date of diagnosis of metastases or date of the last follow-up) curves were calculated with Kaplan-Meier method. Log-rank test was used to assess statistical differences between groups. Multivariate analysis was carried out using Cox proportional hazards model (stepwise forward selection). All variables presenting P < 0.20 on the univariate analysis were selected for building a multiple model. For all tests, type I error (α) was established as 0.05 and results were considered statistically significant when P < 0.05.

Interaction between ADAM23 and α_vβ₃ negatively modulates α_vβ₃ integrin activation. The MDA-MB-435 cell line was selected to investigate whether ADAM23/α_vβ₃ interaction negatively modulates integrin activation, because this cell line expresses high levels of this integrin and has been extensively used as a model for α_vβ₃ activation (12, 13). Physical interactions between ADAM23 and α_vβ₃ integrin were originally described in neuroblastoma and astrocytoma human cell lines (22). To confirm that this interaction is also occurring in MDA-MB-435 cells, we used the disintegrin domain of the ADAM23 protein fused to GST to pull-down α_vβ₃ integrin from MDA-MB-435 lysates. A band corresponding to β₃ integrin (90-100 kDa) was detected in extracts from beads containing ADAM23 disintegrin-GST but not in those derived from beads containing GST alone (Supplementary Fig. S1). Because α_vβ₃ is the only β₃ integrin expressed in MDA-MB-435 cells, these results confirm the physical association between ADAM23 and α_vβ₃ integrin in this cell line.

ADAM23 expression was then knocked down in the MDA-MB-435 cell line using shRNA and integrin activation was then measured by flow cytometry using a monoclonal antibody, which specifically recognizes α_vβ₃ integrin in the activated state. The MDA-MB-435 cell line was transfected with two different hairpin RNA constructs (shRNA-04 and shRNA-25) and stable clones for each construction were selected. Significant reduction of ADAM23 mRNA expression was observed for different clones transfected with the two different constructs when compared with the parental MDA-MB-435 cell line and with cells transfected with an empty vector (Fig. 1A). The morphology and growth rate of the two clones (shRNA-04 and shRNA-25) studied herein presented no differences that could be correlated with ADAM23 expression level and remained phenotypically stable in relation to ADAM23 expression (Supplementary Fig. S2).

Integrin activation was then measured by flow cytometry using the AP5 monoclonal antibody, which, in normal extracellular calcium levels, specifically recognizes α_vβ₃ integrin in the activated state. As a positive control for integrin activation, ADAM23-positive and knockdown cells were incubated with Mn²⁺, a global integrin activator, before flow cytometry analysis. The LM609 monoclonal antibody, which recognizes α_vβ₃ in both conformational states, was also used to evaluate whether ADAM23 knockdown could also interfere with α_vβ₃ expression levels. Flow cytometry analysis using LM609 and AP5 antibodies showed that ADAM23 knockdown cells display equivalent amounts of α_vβ₃ integrin at the cell surface, when compared with the parental cell line or with cell lines transfected with the empty vector, but have higher levels of activated α_vβ₃ (Fig. 1B and C). Interestingly, the Mn²⁺ effect on integrin activation was more pronounced in the ADAM23 knockdown clones. Taken together, these results confirm our hypothesis that interaction between ADAM23 and α_vβ₃ negatively modulates α_vβ₃ integrin activation.

Loss of ADAM23 expression promotes α_vβ₃ integrin-mediated cell migration. Activation of α_vβ₃ integrin in the MDA-MB-435 model has been shown to enhance tumor cell motility, causing a drastic increase in metastatic activity (13). We next determined whether α_vβ₃ activation in the ADAM23 knockdown cells was correlated with increased migration using haptotactic migration assays in which the bottom of the membrane was coated with vitronectin, a classic α_vβ₃ ligand. Ablation of the ADAM23 promoted a 4-fold increase in cell migration (Fig. 2A and B). Enhanced migration of ADAM23 knockdown cells was significantly inhibited by an adhesion-blocking monoclonal antibody (23C6) specific to α_vβ₃ integrin, confirming that enhanced migration is primarily mediated by this receptor (Fig. 2A and B). Similar results were also observed when fibronectin, another classic α_vβ₃ ligand, was used to coat the membrane (data not shown).

Loss of ADAM23 expression enhances α_vβ₃ integrin-mediated cell adhesion. Because integrin activation is known to enhance cell adhesion to specific substrates by regulating ligand affinity, we next evaluated stationary adhesion of ADAM23 knockdown cells to vitronectin. ADAM23 knockdown cells adhered to vitronectin more efficiently than the parental cell line or cells transfected with an empty vector (Fig. 3A). The difference in the efficiency of adhesion between the cells was particularly evident at lower vitronectin concentrations, suggesting that ADAM23 knockdown cells display a higher avidity for vitronectin. Adhesion of ADAM23 knockdown cells was significantly inhibited by RGD peptides (Fig. 3B) and by an adhesion-blocking anti-α_vβ₃ integrin antibody (LM609; Fig. 3C), confirming that enhanced adhesion of ADAM23 knockdown cells is primarily mediated by α_vβ₃ integrin. Similar results were also observed for fibronectin (data not shown).

Loss of ADAM23 expression favors tumor cell arrest in the lungs of immunocompromised mice. Integrin-mediated binding of tumor cells to platelets represents an efficient mechanism for tumor cell arrest in the bloodstream and is a prerequisite for their extravasation. The activated, but not the nonactivated, state of α_vβ₃ integrin has been shown to enhance the interaction between tumor cells and platelets, promoting emboli formation and tumor cell arrest within the vasculature. Because integrin activation is enhanced in the absence of ADAM23, one would expect ADAM23 knockdown to present enhanced interaction with platelets and, as a consequence, to be more efficiently arrested in the bloodstream. To verify whether ADAM23 knockdown clones were more efficiently arrested in the bloodstream, ADAM23 knockdown cells and controls were injected into the lateral tail vein of immunocompromised mice. Tumor cell arrest was then indirectly quantified by real-time PCR detection of human specific Alu sequences (29) in the lungs of these animals 4 h after injection (Fig. 4). The number of ADAM23 knockdown cells arrested in the lungs of the immunocompromised mice was significantly higher than that of the control cell lines, indirectly showing that loss of ADAM23 expression enhances tumor cell aggregation with platelets and arrest in secondary organs during circulation.

ADAM23 silencing by promoter hypermethylation in primary breast tumors is associated with the development of distant metastases and a worse disease outcome. We have shown previously that the ADAM23 gene is frequently silenced by promoter hypermethylation in advanced-stage primary breast tumors (23). If loss of ADAM23 expression promotes α_vβ₃ activation, which, in turn, confers metastatic potential to tumor cells, one would expect silencing of ADAM23 gene in primary breast tumors to be associated with the development of metastasis. To address this issue, and to complement our functional analysis with clinical observations, the methylation status of two CpG islands (regions 1 and 2) located at the upstream regulatory region of theADAM23 gene was determined by methylation-specific PCR in 94 primary breast tumors. Hypermethylation was observed in 29 (30.9%) tumors for region 1 and in 14 (14.9%) tumors for region 2. A total of 8 (8.5%) tumors presented hypermethylation in both regions.

We then investigated the association between ADAM23 hypermethylation and well-established clinicopathologic parameters used for breast cancer. As shown in Supplementary Table S1, ADAM23 hypermethylation within region 2 was associated with tumor size (P = 0.030), the number of positive lymph nodes (P = 0.039), and p53 protein overexpression (P = 0.011). No statistically significant association was observed for region 1.

Kaplan-Meier analysis was then used to estimate the relationship between silencing of ADAM23 gene by promoter hypermethylation, development of metastasis, and disease outcome. The 5-year distant metastases–free survival of patients with tumors harboring ADAM23 gene hypermethylation in both regions was significantly shorter than that of patients with a single hypermethylated region or with an unmethylated tumor (37.5% versus 74.1% versus 91.0%, respectively; P < 0.001; Fig. 5A). Similarly, patients with primary tumors exhibiting ADAM23 gene hypermethylation in both regions had a shorter 5-year disease-specific survival when compared with patients with a single methylated region or with an unmethylated tumor (50.0% versus 92.4% versus 93.1% respectively; P < 0.001; Fig. 5B).

A multivariate analysis was then done to determine whether ADAM23 gene hypermethylation was an independent factor in predicting patient outcome. All variables presenting P < 0.20 on the univariate analysis (Supplementary Table S2), including the use of adjuvant chemotherapy and hormonal therapy, were selected to build the multiple model. The number of positive lymph nodes, the use of tamoxifen, and ADAM23 gene hypermethylation were considered as independent prognostic factors for distant metastases–free and disease-specific survival. Patients with hypermethylation in both regions of the ADAM23 gene regulatory sequence had a higher risk of developing distant metastases [hazard ratio (HR), 8.84; 95% confidence interval (95% CI), 2.31-33.76; P = 0.001] than patients with a single methylated region (HR, 6.53; 95% CI, 2.07-20.57; P = 0.001) or with an unmethylated tumor (reference group; Table 1). Similarly, patients with hypermethylation in both regions of the ADAM23 gene upstream regulatory sequence had a higher risk of dying from the disease (HR, 9.97; 95% CI, 1.96-50.61; P = 0.006) than patients with a single methylated region (HR, 2.30; 95% CI, 0.52-10.07; P = 0.270) or with an unmethylated tumor (reference group; Table 1). These results suggest that, in breast cancer patients, silencing of the ADAM23 gene by promoter hypermethylation can be used as a complementary marker for metastasis risk assessment.

The disintegrin domain of ADAM proteins can bind to various integrins via a short amino acid sequence that is functionally equivalent to the classic RGD motif present in many extracellular matrix proteins. Recent reports indicate that ADAM proteins negatively modulate integrin-mediated cell proliferation, adhesion, and migration in patterns dictated, in part, by the integrin binding profiles of their disintegrin domains (17–20).

There are many reports showing that members of the ADAM family are differentially expressed in human tumors and are frequently associated with tumor progression and poor disease outcome (31, 32); however, until recently, these studies were not accompanied by functional analyses (33–35). Here, we provide additional evidence for the involvement of ADAM family members in metastatic progression. Using the MDA-MB-435 cell line, we show that interaction of ADAM23 and α_vβ₃ integrin in this cell line negatively modulates integrin activation. To our knowledge, this is the first report suggesting the involvement of an ADAM family member in the modulation of integrin activation in tumors.

The MDA-MB-435 cell line was selected for our studies because it expresses high levels of ADAM23 and α_vβ₃ integrin and has been extensively used as a model for α_vβ₃ activation. Additional breast tumor cell lines (MCF-7, MDA-MB-231, MDA-MB-436, MDA-MB-468, and SKBR3) were screened for concomitant expression α_vβ₃ and ADAM23, but we were unable to find a cell line with such characteristics. These results are not unexpected and are in agreement with the data presented in this article, which show that loss of ADAM23 expression favors metastatic dissemination by enhancing the activation of α_vβ₃ integrin expressed in metastatic cells. α_vβ₃ integrin is not expressed in normal mammary epithelium and aberrant expression of this integrin in its activated form is frequently observed in metastatic tumors. On the contrary, ADAM23 is expressed in normal breast tissue and is frequently silenced in metastatic tumors. Thus, the concomitant expression of ADAM23 and α_vβ₃ integrin is not expected to occur very frequently in breast tumors.

Integrin activation primarily involves conformational changes in the extracellular domain of the αβ heterodimer that modulates receptor affinity to different substrates (2). Integrin activation also involves facilitation of lateral diffusion and/or clustering of heterodimers into membrane microdomains, which modulate receptor avidity and endocytosis (2). Based on our results, we envisage two possible mechanisms by which anchorage of ADAM23 protein at the cell surface would negatively modulate integrin activation. First, interaction between ADAM23 and α_vβ₃ integrin would keep the integrin in the nonactivated form by inhibiting conformational changes in the extracellular domain of the αβ heterodimer that modulates receptor affinity to different substrates. In the second mechanism, interaction with ADAM23 would affect lateral diffusion and/or clustering of α_vβ₃ integrin into membrane microdomains, which would in turn modulate receptor avidity and endocytosis. Here, we provide solid evidence that the absence of ADAM23 protein at the cell surface allows conformational changes in the integrin structure leading to its activation. The effect of ADAM23 on integrin activation is particularly evident when cells are incubated with Mn²⁺. α_vβ₃ integrin present at the surface of ADAM23 knockdown cells seems to be more prone to activation than integrin present at the surface of MDA-MB-435 cells. However, these mechanisms are not mutually exclusive, and at this point, we cannot exclude the possibility that ADAM23 is also regulating integrin clustering and endocytosis.

There is some specificity in the binding of ADAM proteins to integrins, although our understanding of the binding specificity and overlap between the different ADAM family members is limited at present (16). Analysis of ADAM23 binding to integrins has revealed a specific interaction with α_vβ₃ integrin mediated by the disintegrin domain (22); however, a systematic analysis of ADAM23 binding to other integrins is not available at present. As shown in Figs. 2 and 3C, residual migration and adhesion of ADAM23 knockdown cells is observed after incubation with an adhesion-blocking monoclonal antibody specific to α_vβ₃ integrin, suggesting that ADAM23 might also be modulating other integrins. Adhesion of MDA-MB-435 cells to vitronectin is predominantly mediated by α_vβ₃ integrin, but other α_v integrins, such as α_vβ₅ integrin, also contribute. To address a possible role of ADAM23 in modulating α_vβ₅ integrin activity, adhesion to vitronectin was done in the presence of a combination of anti-α_vβ₃ and anti-α_vβ₅ blocking antibodies. As shown in Supplementary Fig. S3, adhesion of ADAM23 knockdown cells to vitronectin was completely abolished in the presence of both antibodies, suggesting that the modulatory effect of ADAM23 is not restricted to α_vβ₃ integrin and should be further explored.

We have also shown that silencing of the ADAM23 gene by promoter hypermethylation in primary breast tumors is significantly associated with a higher incidence of metastasis and reduced overall survival. Because we were unable to address α_vβ₃ integrin expression and activation in the ADAM23 hypermethylated tumor samples, we cannot exclude the possibility that the poor prognosis observed for these patients could be associated to additional mechanisms other than α_vβ₃ integrin activation.

The risk of developing metastatic disease in breast cancer patients is currently estimated by the number of positive lymph nodes and tumor size complemented by information on tumor grade, steroid hormone receptor status, and c-erbB2 overexpression (36). However, lymph node information and tumor size are insufficient to accurately assess individual risk, as 20% to 30% of women with node-negative breast cancer still develop metastatic disease (37). The identification of prognostic biomarkers to guide patient treatment is clearly needed (38), and ideally, these biomarkers should be highly specific and sensitive and reflect the presence of tumor-specific alterations functionally related to the acquisition of metastatic potential. We have herein shown that ADAM23 gene hypermethylation in primary breast tumors is exactly such a marker for assessing the risk of metastasis for breast cancer patients.

No potential conflicts of interest were disclosed.

Grant support: FAPESP (04/09088-9). This work was conducted under the auspices of the Hilton Ludwig Cancer Metastasis Initiative.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. David R. Engelke, Vilma R. Martins, and Santiago Cal for providing plasmid pAVU6+27, RGD peptides, and ADAM23 disintegrin-GST, respectively, and Drs. Sandro de Souza and Sarah White for critical reading of this article.