Aggretin Venom Polypeptide as a Novel Anti-angiogenesis Agent by Targeting Integrin alpha2beta1

VEGF and VEGFR antibodies have been used as a therapeutic strategy to inhibit angiogenesis in many diseases; however, frequent and repeated administration of these antibodies to patients induces immunogenicity. In previous studies, we demonstrated that aggretin, a heterodimeric snake venom C-type lectin, exhibits pro-angiogenic activities via integrin α2β1 ligation. We hypothesised that small-mass aggretin fragments may bind integrin α2β1 and act as antagonists of angiogenesis. In this study, the anti-angiogenic efficacy of a synthesised aggretin α-chain C-terminus (AACT, residue 106–136) was evaluated in both in vitro and in vivo angiogenesis models. The AACT demonstrated inhibitory effects on collagen-induced platelet aggregation and HUVEC adhesion to immobilised collagen. These results indicated that AACT may block integrin α2β1−collagen interaction. AACT also inhibited HUVEC migration and tube formation. Aortic ring sprouting and Matrigel implant models demonstrated that AACT markedly inhibited VEGF-induced neovascularisation. In addition, induction of FAK/PI3K/ERK1/2 tyrosine phosphorylation and talin 1/2 associated with integrin β1 which are induced by VEGF were blocked by AACT. Similarly, tyrosine phosphorylation of VEFGR2 and ERK1/2 induced by VEGF was diminished in integrin α2-silenced endothelial cells. Our results demonstrate that AACT is a potential therapeutic candidate for angiogenesis related-diseases via integrin α2β1 blockade.

Scientific RepoRts | 7:43612 | DOI: 10.1038/srep43612 VEGF has been established to be involved in many stages of angiogenesis in malignant diseases via its multi-functional effects in activating and integrating signalling pathway networks 14 . VEGF signalling blockade reduces new vessel growth and induces endothelial cell apoptosis. Thus, the use of tyrosine kinase inhibitors or VEGF/VEGF receptor (VEGFR) antibodies to inhibit crucial angiogenic steps represents a practical therapeutic strategy for the treatment of neovascularisation diseases 15 . E7820, a potent angiogenesis inhibitor, has been shown to reduce integrin α 2 mRNA expression and inhibit basic fibroblast growth factor/VEGF-induced HUVEC proliferation and tube formation 16,17 . Integrin α 2β 1/α 1β 1 expression is reportedly regulated by VEGF, and an inhibitory antibody against α 2β 1/α 1β 1 has been shown to inhibit angiogenesis and tumour growth in VEGF-overexpressing tumour cells 12,18 . Therefore, we hypothesised that peptide-based integrin α 2β 1 blockade may have potential anti-tumour effects by inhibiting angiogenesis.
In this study, we demonstrate that aggretin α -chain C-terminal (AACT, 31 amino acid residues) inhibits collagen-induced platelet aggregation and HUVEC adhesion predominantly via integrin α 2β 1 ligation. The ability of endothelial cells to adhere to collagen was also diminished by integrin α 2 silencing. Thus, we hypothesised that aggretin-derived integrin α 2 antagonism may inhibit angiogenesis in response to VEGF. In this study, we unveiled the anti-angiogenic activities of AACT by demonstrating its inhibitory effects on HUVEC migration, Matrigel-induced capillary tube formation and aortic ring sprouting in ex vivo assays and reducing neovascularisation in Matrigel implant angiogenesis assays in vivo. VEGF-stimulated focal adhesion kinase (FAK), Phosphoinositide 3-kinase (PI3K) and Extracellular Signal-regulated Kinase 1/2 (ERK 1/2) phosphorylation were attenuated by AACT. The talin1/2 associated with integrin β 1 was also abolished by AACT. Similarly, VEGF-induced VEFGR2 and ERK1/2 activation were abolished by integrin α 2 siRNA transfection. These results demonstrate that AACT inhibits angiogenesis in response to VEGF via α 2β 1 integrin blockade.

Effects of AACT on collagen-induced platelet aggregation and HUVEC-collagen interaction.
Since the integrin β 1/C-type lectin-like receptor 2 (CLEC-2) were demonstrated as the binding targets of AACT 19 and there are lack of CLEC-2 expression in HUVECs 20 , the integrin α 2β 1 may be the binding target in HUVECS. To investigate the inhibitory effect of AACT on integrin α 2β 1 activation, we examined the effect of AACT on collagen-induced platelet aggregation. As shown in Fig. 1A, AACT (25 and 50 μ g/ml, equivalent to 6.75 and 13.5 μ M, respectively) significantly inhibited collagen-induced aggregation (approximately 50% inhibition). Furthermore, to confirm integrin α 2β 1 as the major target for AACT-mediated HUVEC-collagen attachment, we next examined the involvement of integrin α 2 in cell adhesion. Endothelial cell adhesion to collagen was inhibited by integrin α 2 mAb and AACT (50 μ g/ml). Similarly, knockdown of α 2 also inhibited cell adhesion to collagen (Fig. 1B). Moreover, we investigated the binding of AACT to integrin α 2. HUVECs treated with or without AACT (50, 100 and 300 μ g/ml) were cultured with anti-α 2 antibodies. As shown in Fig. 1C, AACT inhibited the binding of integrin α 2 mAb to endothelial cells as measured by flow cytometry, but not the binding of anti-Glycoprotein VI (GPVI) or anti-Glycoprotein Ib (GPIb)(AP1) antibodies ( Fig. 1D and E). We also used the HUVECs membrane receptor to explore the binding site of AACT on HUVECs. HUVECs membrane proteins bound to biotinylated AACT were isolated and eluted. Only one membrane receptor was recognized by integrin β 1 (Fig. 1F). These results indicate that AACT inhibits platelet and HUVEC-collagen adherence, predominantly via integrin α 2 blockade.
Effects of AACT on HUVEC viability and proliferation. As integrin α 2β 1 activation is involved in endothelial cell growth, we evaluated the inhibitory effects of AACT on cell viability using 3-[4, 5-dimethylthiazol-2-yl]-2, 5-diphenyltetrazolium bromide (MTT) assays. As shown in Fig. 2A, AACT reduced serum induced HUVEC viability by 63.8% at a concentration of 50 μ g/ml. Furthermore, in order to confirm the inhibitory effects of AACT on endothelial cell growth, we performed bromodeoxyuridine assays. As expected, AACT was found to significantly inhibit HUVEC proliferation (Fig. 2B).

AACT inhibits migration of HUVECs in vitro and ex vivo.
As HUVECs migration is essential for angiogenesis, the effect of AACT (10, 25 and 50 μ g/ml) on HUVEC haptotaxis migration with Transwell was assayed. As shown in Fig. 3A, a 4.72-fold increase was observed in the number of HUVECs in the lower filter membrane coated with collagen. Under similar conditions, AACT significantly inhibited HUVEC migration. Furthermore, we evaluated chemotactic migration with Transwell to determine the effect of AACT on HUVEC migration in response to VEGF. As shown in Fig. 3B, a 7.45-fold increase in the number of HUVECs was observed following VEGF stimulation, with AACT found to inhibit HUVEC migration. In addition, the vessels sprouting of the rat aortic ring induced by VEGF was also significantly decreased in AACT treated group (Fig. 3C). These results showed that AACT is capable of inhibiting VEGF-induced HUVECs migration in vitro and ex vivo.
Effects of AACT on Matrigel tube formation. HUVECs had significantly greater numbers of branching tube networks after 16 h of 20% FBS incubation (20% FBS treatment Fig. 4B as compared to serum free Fig. 4A), and this tube branching was attenuated by VEGF Ab treatment (Fig. 4C). AACT (10, 25 and 50 μ g/ml) also attenuated serum-induced HUVEC tube formation ( Fig. 4D-F). Moreover, to confirm the involvement of integrin α 2 in the tube formation process, we examined the inhibitory effect of integrin α 1 and α 2 Ab in our in vitro angiogenic model. Integrin α 2 mAb treatment, but not integrin α 1 mAb treatment, significantly decreased VEGF-induced tube formation ( Fig. 4G-J). These results indicate that AACT inhibits VEGF-stimulated angiogenesis predominantly via integrin α 2 blockade, as shown in Fig. 4K.
Effect of AACT on angiogenesis in response to Matrigel implantation. An in vivo model containing Matrigel premix with VEGF (200 ng/ml) was used to determine the inhibitory effect of AACT on angiogenesis. Matrigel (in the presence or absence of AACT (10, 25 and 50 μ g/ml)) was then subcutaneously Scientific RepoRts | 7:43612 | DOI: 10.1038/srep43612 injected into mice. At 7 days after inoculation, capillary network formation was observed in implanted plugs. In the AACT-treated group, less vessel growth and less red blood cell infiltration was observed in implanted plugs (Fig. 5). Haemoglobin levels were significantly lower in AACT-treated mice. These results suggest that AACT also inhibits angiogenesis in vivo.
FAK and PI3K phosphorylation are involved in many cell responses to VEGF. In this study, we found that VEGF-induced FAK and PI3K p85α phosphorylation were significantly inhibited by AACT (Fig. 6). AACT

Figure 1. Effects of AACT on collagen-induced platelet aggregation and HUVEC-collagen interation.
(A) Washed platelets were preincubated with AACT (25, 50 μ g/ml) at 37 °C for 3 min, and then collagen (3 μ g/ml) was added to trigger platelet aggregation. Platelet aggregation was measured turbidimetrically (Δ T, change in transmission). All experiments were conducted in triplicate at least four times and similar results were obtained. (B) HUVECs were seeded onto plates coated with collagen or negative control (gelatin). Cells were labeled with fluorescent dye BCECF-AM for 30 min and then preincubated with integrin α 2 Ab, AACT or pre-transfected with integrin α 2 siRNA. Attached cells were read by Cytofluor microplate reader with fluorescence excitation and emission wavelength at 485 nm and 530 nm, respectively, and were quantified as the percentage of fluorescence intensity of control. HUVECs were preincubated with vehicle or AACT (50, 100, 300 μ g/ml in C; 50 μ g/ml in D and E) for 30 min, then the binding as probed with anti-integrin α 2 (C), anti-GPVI (D) or anti-GPIb (E) mAb and subjected to flow cytometric analysis by using FITC-conjugated anti-IgG mAbs as the second antibody. (F) HUVEC proteins eluted from biotinylated AACT-bound streptavidin-Sepharose beads were blotted with anti-integrin β 1, GPIbα and CLEC-2 antibodies. Results are presented as cell numbers vs. binding fluorescence intensity. Data are presented as mean ± S.E.M. (n = 4). **P < 0.01, ***P < 0.001 compared with control. The pattern shown is a representative of one of at least three similar results.
Role of integrin α2 involved in VEGF signalling. The integrin α 2-subunit siRNA was used to evaluate the involvement of integrin α 2β 1 in VEGF signalling. Expression levels of integrin α 2 markedly decreased in siRNA-transfected cells compared to non-targeted control siRNA-transfected HUVECs (Fig. 7). VEGF-induced VEGF Receptor 2 and its downstream signalling molecular ERK1/2 phosphorylation were significantly inhibited in both integrin α 2 siRNA transfected endothelial cells and cells pre-treated with AACT (Fig. 7). Collectively, these results indicate that integrin α 2 is involved in VEGF signalling and that AACT inhibits angiogenesis via endothelial integrin α 2 ligation.

Discussion
HUVEC expression of integrin α Vβ 3, α Vβ 5 and α 2β 1 has been implicated in angiogenesis 21,22 . Several snake venoms interact with platelets via α 2β 1, GPVI and GPIb; however, few demonstrate significant specificity. Many snake venoms have binding sites for several platelet targets. Recombinant techniques allow the production of specific polypeptides which bind integrin α 2β 1 without involving GPIb or GPVI. In previous studies, we demonstrated that aggretin induces platelet activation through binding integrin α 2β 1 and GPIb leading to FAK and PLCγ 2 phosphorylation 23 . Aggretin also induces HUVEC-dependent angiogenesis by interacting with integrin α 2β 1 through the PI3K/Akt/ERK1/2 pathways, with increased expression of VEGF 11 . Therefore, in this study we synthesised a small-mass aggretin fragment, AACT, and examined its effects on platelet aggregation and angiogenesis. Interestingly, we demonstrated that AACT blocks platelet aggregation induced by collagen (Fig. 1A), HUVEC-collagen attachment and HUVEC-integrin α 2 mAb binding (Fig. 1B,C). These results indicate that this aggretin fragment acts as α 2β 1 antagonist rather than an intrinsic α 2β 1 agonist such as intact aggretin. Similarly, AACT was found to inhibit collagen/VEGF-induced HUVEC migration, FBS-induced Matrigel tube formation and VEGF-induced aortic ring sprouting (Figs 2-4). Since integrin α 2β 1 but not GPIb, GPVI and CLEC-2 is the major target of aggretin on endothelial cells (Fig. 1D-F), we hypothesised that AACT blocks in vitro or ex vivo angiogenesis via integrin α 2β 1 ligation. Furthermore, AACT abolished VEGF-induced angiogenesis in a Matrigel plug implant assay, suggesting that AACT may be utilised as an anti-angiogenic peptide for inhibiting angiogenesis in vivo (Fig. 5). Although the AACT exhibited a potent antiangiogenic activity, most results in our study is not dose-dependent. According to the previous studies, there are several sites for collagen/snacle binding and induced α 2β 1 activation [24][25][26] . It may be the reason why AACT was failed to show dose-dependent effect via competitive inhibition. We also found other native peptides derived from integrin α 2β 1 inhibitor were more likely through the on/off pattern to inhibit integrin α 2β 1 activity 27,28 . The detail integrin α 2β 1 binding site for AACT still needs further investigation.
Tumour-induced angiogenesis is critical for nutrition and oxygen supply via blood to local tumour and tumour metastases to other organs. Anti-angiogenesis therapy provides many benefits, including broad applicability to different tumour types, less tumour cell resistance and reduced chemotherapeutic dosages. A C-type lectin-like selective α 2β 1 integrin inhibitor, vixapatin, has demonstrated the ability to inhibit angiogenesis 29 . E7820 inhibits the proliferation and tube formation of HUVECs through suppression of endothelial integrin  30 . These studies indicate that integrin α 2β 1 mediated angiogenesis may represent a novel pharmacological target.
VEGF is the major angiogenesis regulatory factor and induces neovascularisation via interaction with endothelial cells 31 . To regulate angiogenesis-related processes, VEGF activates many signal transduction networks, such as FAK, PI3K/AKT, ERK1/2, Src and PLCγ . The inhibitory effect of AACT may be through ligation to integrin α 2β 1, resulting in blockade of the VEGF signal transduction pathway. VEGF-induced PI3K and ERK1/2 activation were markedly inhibited by pre-treatment of HUVECs with AACT in this study. Several β 1 integrin have been found to regulate angiogenesis and VEGFR2 activity 32,33 . β 1 integrins and VEGFR2 interaction plays a role in infantile hemangiomas pathogenesis and matrix-bound VEGFA signalling 34,35 . The CD36 and β 1 integrin association was reported to cooperate with VEGFR2 in promoting angiogenesis 36 . Integrin α 2β 1 also physically associates with EGFR and functions regulate EGFR activation 37 . These findings support a role for integrin α 2β 1 as a mediator for VEGFR2 signalling. To confirm the action of integrin α 2β 1 in VEGF induced signalling, we transfected integrin α 2 siRNA and evaluated the effect of VEGF, where VEGF-induced ERK1/2 activation was found to be significantly inhibited by integrin α 2 siRNA (Fig. 7A). These results suggest that the existence of a crosstalk or a positive feedback loop between integrin β 1 and VEGFR signaling.
The modulatory role of integrin α 2β 1 in angiogenesis observed in in vitro experiments illustrates its involvement in supporting VEGF signalling and HUVEC migration. Studies from other researchers have provided additional support for the involvement of α 2β 1 12,13 . Although the significance of the integrin α 2β 1 in tissue angiogenesis remains undetermined, our findings support the concept that integrin α 2β 1 contributes to the regulation of VEGF signalling. Integrin-linked kinase (ILK) has been identified as an integrin β 1 tail binding protein 38 . Silencing ILK with siRNA significantly suppressed tube formation and reduced the tube lengths 39 . Loss of ILK signalling may be involved in decreasing the responses of integrin α 2β 1-silenced HUVEC to VEGF.
Integrin/ECM interactions are one of the major mediators of cell adhesion-mediated drug resistance 6 . β 1 integrins reportedly play a crucial role in head and neck squamous cell carcinoma cell radioresistance 40 . Kanda et. al. also reported increased β 1, α 2 and/or α 5 integrin expression in refractory tumours following treatment with gefitinib and/or erlotinib 41 . Furthermore, they demonstrated the integrin β 1/Src/Akt signalling pathway as a key mediator of acquired resistance to EGFR-targeted anti-cancer drugs (gefitinib or erlotinib) 41 . Combined treatment with E7820, an integrin α 2 expression blocker, and erlotinib significantly decreased microvessel density and increased apoptosis of tumour-associated endothelial cells compared with use of only one of the agents 42 . Ligation of integrin α 2β 1 may modulate EGFR-mediated endothelial cell functions. The combination of an integrin α 2β 1 blocker with a growth factor inhibitor may represent an alternative strategy for overcoming drug resistance in cancer treatment.
In summary, we identified an important functional cooperativity between integrin α 2β 1 and VEGF. In particular, the findings of this study indicate that integrin α 2β 1 provides crucial support not only for endothelial cell migration but also for VEGF signal transduction and pro-angiogenic functions. In contrast to the intact aggretin, a pro-angiogenic α 2β 1 agonist, AACT, was found to act as an antagonist of integrin α 2β 1 and suppresses VEGF-driven angiogenesis in ex vivo and in vivo models. Many angiogenesis inhibitors have been approved for clinical use and have been studied in clinical trials; however, side effects are reportedly associated with increased risk of arterial thromboembolism 43 . AACT has anti-angiogenic potential and exhibits anti-thrombotic activity by blocking collagen-induced platelet activation. This may also represent an alternative advantage of AACT for patients receiving anti-angiogenesis therapy at a risk of thromboembolism. Recently, our group found that AACT also blocks the interaction of platelet receptor CLEC-2 with the tumour receptor, podoplanin 19 . CLEC-2, a type II transmembrane receptor, is highly and selectively expressed in the liver and some myeloid subsets 44 . As CLEC-2 is not expressed on the surface of HUVEC 20 , the possibility of AACT interacting with CLEC2 can be excluded, indicating that integrin α 2β 1 is the major target of AACT affecting angiogenesis. Thus, optimisation of small-mass CLP-derived integrin α 2β 1 antagonists contributes to future drug development.
Preparation of human platelet suspension and platelet aggregation assay. Platelet suspensions were prepared as previously described 45 . HUVECs culture. HUVECs were provided by the National Research Program for Biopharmaceuticals, Taiwan and approved by its Institutional Review of Board. HUVECs were maintained in M199 (with FBS (20%), ECGS (30 μ g/mL), L-glutamine (4 mM), penicillin (100 U/mL), and streptomycin (100 μ g/mL)) and incubated at 37 °C in 5% CO2. The cells were used between second to fourth passages. Adhesion Assay. HUVECs were labeled BCECF-AM and then incubated with or without AACT (50 μ g/ml) for 37 °C 30 min. HUVECs were applied to plates which were pre-coated with 100 μ l matrices (collagen or gelatin, 50 μ g/ml) and were subjected to adhesion as previously described 11 .
Flow cytometry. The inhibitory effect of AATC on integrin α 2, GPIb and GPVI interaction with HUVECs was measured by flow cytometry. HUVECs were pretreated with AACT, and incubated with anti-integrin α 2, anti-GPVI or anti-GPIb antibody at room temperature, then were analyzed byFACS Calibur (Becton Dickinson, San Diego, CA, USA).
MTT assay. HUVECs were starved with M199 contain 2% FBS and then grown in M199 with 20% FBS in absence or presence of AACT (10, 25 or 50 μ g/ml). HUVECs were subjected to adhesion as previously described 11 . BrdU proliferation assay. HUVECs were starved and then treated with or without AACT (10, 25 or 50 μ g/ml) for 48 hours. Cell proliferation was measured by BrdU Cell Proliferation Assay Kit (Chemicon, Temecula, U.S.A), and followed by the manufacturer's protocol. Binding assays. HUVECs were fixed with 1% paraglutaldehyde and then pretreated with AACT. Treated cells then incubated with integrin α 2 antibody and were subjected to adhesion as previously described 11 .
Matrigel capillary tube formation. HUVECs (1.2 × 10 5 cells) treated with or without AACT (10, 25, 50 μ g/ml) in presence of 20% FBS. Cells were subjected to tube formation as previously described 11 . Total length in each condition was quantified by using Kurabo Angiogenesis Image Analyzer (Kurabo, Japan). Figure 6. Effect of AACT on FAK, ERK1/2 and PI3K p85α activation and talin1/2 associated with integrin β1. (A) HUVECs were pretreated with or without indicated concentrations of AACT for 30 min and vehicle or VEGF (20 ng/ml) was added to the cells as a basal or control for another 10 min. Cells were lysed with lysis buffer. FAK, PI3K and ERK1/2 activation were detected by Western blotting using anti-phospho-FAK, antiphospho-PI3K p85α and anti-phospho-ERK1/2 mAb, whereas beta actin was used as an internal control. (B) HUVECs were pretreated with or without indicated concentrations of AACT for 30 min and vehicle or VEGF (20 ng/ml) was added to the cells as a basal or control for another 10 min. Cells were lysed with lysis buffer and immunopreciptated with integrin β 1 mAb. Talin 1/2 associated with integrin β 1 were detected by Western blotting using anti talin 1/2 mAb. Blot images were cropped for comparison and all relevant gels have been run under similar experimental conditions. The pattern is one example of three independent experiments with similar results. Quantitative analyses of FAK (C), PI3K p85α (D) and ERK1/2 (E) tyrosine phosphorylation are presented as mean density for the ratio between phosphorylated protein and total protein as determined by a densitometer. Quantitative analyses of talin 1/2 association are presented as mean density for the ratio between associated protein and total protein as determined by a densitometer (F). Intensities were normalized to static basal, and fold decreases were calculated. Data are presented as mean ± S.E.M. (n = 4). ***P < 0.001 compared with control. The pattern shown is representative of 1 of at least 3 similar results.
Scientific RepoRts | 7:43612 | DOI: 10.1038/srep43612 Aortic ring sprouting assay. Sprague-Dawley rats aortic rings were treated with or without AACT (10, 25 or 50 μ g/ml) and placed in pre-coated Matrigel. Aortic rings were subjected to sprouting assay as previously described 46 . The sprouting area was measured by the area of endothelial cell out-growth.
Matrigel-implant angiogenesis assay. Matrigel implant assay was modified from previously study 47 .
Matrigel containing VEGF (200 ng/ml) was premix with or without AACT were subjected to matrigel implant assay as previously described 48 . Migration assay. Transwell (8.0 μ m pore size, Costar) were coated with type I collagen (0.2 μ g)/BSA (20 μ g) or filled absence or presence of VEGF (20 ng/ml) as a chemoattractor in lower chamber. HUVECs were subjected to migration assay as previously described 49 . Protein-based affinity pulldown. These studies were performed as previously described 50 .
Transfection of small interfering RNA (siRNA). Integrin α 2 siRNA or non-targeting siRNA (negative siRNA) were transfected into HUVECs. The integrin α 2 sense sequences of these three siRNA sequences were as follows: UGAAUUGUCUGGCGUAUAATT, CAACUGGGAUCUGUUCUGATT and GCCAAUGAGCCGAGAAUUATT. The sequence for the negative siRNA is UAAGGCUAUGAAGAGAUAC Immunoprecipitation and immunoblot analysis. HUVECs were pretreated with or without AACT (25 or 50 μ g/ml) for 30 min, and VEGF (20 ng/ml) was added to the cells as basal or control for another 10 min. Cells were subjected to immunoprecipitation and immunoblot as previously described 11 . Statistical analysis. Data are presented as mean ± SEM. Groups were assessed by one-way ANOVA and Newman-Keuls multiple comparison test. P values less than 0.05 (P < 0.05) were considered to be significantly different. Endothelial cells were treated with or without VEGF (20 ng/ml) for 10 min. Cells were lysed with lysis buffer. VEGFR2 and ERK1/2 activation was detected by Western blotting using anti-phospho-VEGFR2 mAb or anti-phospho-ERK mAb, whereas beta actin was used as an internal control. Integrin α 2 expression after siRNA interference was detected by Western blotting using anti-integrin α 2 mAb. Blot images were cropped for comparison and all relevant gels have been run under similar experimental conditions. The pattern is one example of three independent experiments with similar results. Quantitative analyses of VEGFR2 (B) and ERK1/2 (C) tyrosine phosphorylation are presented as mean density for the ratio between phosphorylated protein and total protein as determined by a densitometer. Intensities were normalized to static basal, and fold decreases were calculated. Data are presented as mean ± S.E.M. (n = 4). ***P < 0.001 compared with the control.