Elastin-derived peptides enhance angiogenesis by promoting endothelial cell migration and tubulogenesis through upregulation of MT1-MMP

Angiogenesis involves the formation of capillaries from preexisting microvessels and therefore contributes to vascular remodeling and maturation (Carmeliet, 2000). It plays a pivotal function in a variety of normal and pathological conditions such as embryonic development, the menstrual cycle, hair cycle, wound healing, arthritis, psoriasis, proliferative diabetic retinopathy, atherosclerosis, post ischemic vascularization of the myocardium and tumor growth and metastasis (Folkman, 1995; Risau, 1997). The initiation of the physiopathological angiogenic response, known as the `angiogenic switch', depends on the dynamic balance between exogenous or endogenous stimuli (pro-angiogenic factors) and inhibitors (anti-angiogenic factors) acting in the immediate environment of endothelial cells (Liotta et al., 1993; Hanahan and Folkman, 1996; Distler et al., 2002). Disruption of such an equilibrium can favor the emergence of new vessel formation or lead to vessel quiescence or regression (Hanahan and Folkman, 1996). Several protein families can promote angiogenesis, among which are growth factors, cytokines, integrins, proteinases, extracellular matrix components and cell adhesive proteins (Folkman, 1995; Risau, 1997; Carmeliet and Jain, 2000). The contribution of matrix metalloproteinases (MMPs), especially MMP-2, MMP-9 and membrane-type metalloproteinase-1 (MT1-MMP, MMP-14) has been convincingly established by the use of natural or synthetic MMP inhibitors, both in vitro and in vivo (Moses, 1997; Maekawa et al., 1999; Hajitou et al., 2001). Genetic studies used mice deficient in those endopeptidases and showed reduced angiogenic responses (Itoh et al., 1998; Vu et al., 1998; Zhou et al., 2000) and in vitro studies used antisense oligonucleotides or antibodies directed against those MMPs (Fang et al., 2000; Lafleur et al., 2002). The participation of neutral proteinases on the cell invasive program is far more complicated than originally depicted (Hanahan and Folkman, 1996) as they can generate proteolytic fragments of the extracellular matrix (ECM), i.e. matrikines, or reveal cryptic sites within the matrix, i.e. matricryptines which behave as cytokine-like molecules, further influencing angiogenesis, either negatively or positively (Bellon et al., 2004). Overproduction of matristatins or downregulation of stimulatory angiogenic peptides, or both, might be involved in defective wound repair, ischemia and peripheral arterial occlusive disease (Maquart et al., 2004). Conversely, the opposite imbalance could lead to excessive angiogenesis, a hallmark of inflammatory disorders in many organs, and a prerequisite for tumor progression (Hanahan and Folkman, 1996; Zetter, 1998). Such matrikines, some of them being evidenced in the blood circulation, have been shown to exert their pro- or anti-angiogenic activity at distinct levels of the angiogenesis process either by interfering with cell proliferation or migration or apoptosis and through different mechanisms (Bellon et al., 2004).

Elastin occupies a privileged site in matrix biology, as it is the major component of elastic fibers, particularly abundant in tissues such as arteries and lung, but also present in skin, breast, cartilage and certain ligaments. It is also a long-lived macromolecule with no appreciable turn-over (Rosenbloom, 1987). However, its proteolysis by elastase-type proteinases belonging to the metallo, serine and cysteine families is linked to the genesis of several diseases affecting elastin-rich organs (Werb et al., 1982; Visse and Nagase, 2003). Besides altering the rheological property of those tissues, elastolysis through the generation of elastin-derived peptides (EDPs) might interfere strongly with tissue homeostasis. Thus, any sign of its degradation can represent a vital signal for organisms to initiate effective repair processes. Overall, those peptides were reported to display a wide range of biological activities, influencing cell migration (Senior et al., 1980; Hinek et al., 1992), differentiation (Grant et al., 1989), proliferation and chemotaxis (Kamoun et al., 1995; Ghuysen et al., 1992), tumor progression (Lapis and Timar, 2002; Timar et al., 1991; Huet et al., 2002; Ntayi et al., 2004), aneurysm formation and atherogenesis (Nackman et al., 1997; Hance et al., 2002; Robert, 1996). With regard to tumor invasion, soluble peptides from alkaline (κ-elastin) or elastase hydrolysis of insoluble elastin, as well as tropoelastin were shown to increase MMP-2 and MMP-3 production by human skin fibroblasts (Brassart et al., 2001; Huet et al., 2001); similarly, κ-elastin stimulated MMP-2, MT1-MMP and TIMP-2 expression in human HT-1080 fibrosarcoma cell lines and, as a consequence, could promote the invasive metastatic ability of tumor cells (Huet et al., 2002; Brassart et al., 1998). These numerous effects are mediated by the binding of EDPs to a 67-kDa multifunctional high affinity receptor with lectin-like properties named EBP (elastin binding protein) (Mecham et al., 1989; Hinek, 1994). EBP was further identified as an inactive spliced form of β-galactosidase known as S-Gal (Hinek et al., 1993; Privitera et al., 1998), expressed in elastin-producing and non-producing cells (Hinek, 1995; Yusa et al., 1989). S-Gal is associated at the cell surface with two other plasma membrane-anchored proteins, a 61-kDa moiety with neuraminidase activity and a 55-kDa protective protein corresponding to carboxypeptidase A or cathepsin A (Hinek, 1996). Cell responses to EDPs were often attributed to the binding of a VGVAPG hexapeptide sequence, repeated several times in tropoelastin, the soluble precursor form of elastin, to a unique sequence of S-Gal, encoded by the frame shifted exon 5 of β-galactosidase. Previous investigations indicated that the VGV tripeptide could represent the core sequence in EDPs, mediating their potent influence on vascular tone through increased intracellular calcium in endothelial cells (Faury et al., 1995; Faury et al., 1998a; Faury et al., 1998b). Conversely, a type VIII β-turn conformation adopted by elastin peptides with the GXXPG sequence was mainly involved in directing matrix metalloproteinase expression in fibroblasts and fibrosarcoma cells in culture (Brassart et al., 2001; Huet et al., 2002).

An angiogenic response has been correlated with extensive alterations and remodeling of elastic fibers and, in a rat aneurysm model, administration of VGVAPG hexapeptide or elastase was shown to induce adventitial angiogenesis (Nackman et al., 1997). In addition, EDPs, at a concentration averaging that present in the circulation (Kucich et al., 1983; Fülöp et al., 1990), activate low specificity calcium channels in human umbilical venous endothelial cells (HUVECs), leading to an enhancement of cytoplasmic and nuclear-free calcium concentration (Faury et al., 1998a). Such modifications in calcium flux induce an endothelium-dependent vasodilatation that can be suppressed by an inhibitor of nitric oxide production (Faury et al., 1995; Faury et al., 1998a; Faury et al., 1998b).

In the present study, we have shown that occupancy of EBP by VGVAPG motif-containing peptides on endothelial cells from vascular and microvascular origins triggered neoangiogenesis by promoting cell migration and tubulogenesis. The effects induced by EDPs were linked to upregulation of proMT1-MMP and proMMP-2 expression and activation. The contribution of MT1-MMP in EDP-mediated angiogenesis was demonstrated by the use of specific inhibitors of MT1-MMP and a siRNA approach for specifically silencing MT1-MMP expression in human microvascular endothelial cells (HMECs). We have demonstrated that two 21-bp siRNA duplexes targeting MT1-MMP mRNA at position 107-127 and 228-248 relative to the start codon, respectively, proved to decrease MT1-MMP expression fourfold after a 72-hour transfection. In parallel, siRNA_107-127 totally suppressed the effect of elastin peptides on pseudotubes formation by HMECs.

Insoluble elastin and soluble κ-elastin peptides (KE) were obtained as previously described (Jacob and Hornebeck, 1985). VGVAPG and VVGSPSAQDEASPL (V14) peptides were synthesized using 9-fluoromethoxycarbonyl (Fmoc) chemistry using a Fmoc-Val resin (0.22 meq/g). Couplings were performed with Fmoc-amino acid-pentafluorophenyl esters (Pfb) (4 molar excess). Each Fmoc deprotection step involved treatment with 20% (v/v) piperidine/dimethylformamide for 10 minutes. Cleavage of the peptide from the resin was achieved by a six-hour treatment with trifluoroacetic acid/water (95/5; v/v), followed by successive washings of the resin with ether. Purity of peptides was confirmed by HPLC and by fast atom bombardment mass spectrometry.

In vivo angiogenesis was performed according to an established shell-less culture technique, exposing the CAM to direct access for experimental handling (Averbar et al., 1974). At day six of embryonic development, angiogenic areas were delimited with a silicon ring (Weber Métaux, France) and PBS (as control) or κ-elastin (50 ng) or VGVAPG peptide (200 ng) in a final volume of 20 μl were placed inside the rings. The embryos were then placed in an incubator to induce spontaneous angiogenesis and were treated daily. Treated areas were photographed daily from day 6 to day 10 of embryonic development.

Human microvascular endothelial cell-1 line (HMEC) was provided by E. W. Ades (Center for Disease Control and Prevention, Atlanta, GA). HMECs are representative of the microvasculature and have properties similar to those of the original primary culture (Ades et al., 1992). HMECs were cultured in endothelial cell growth medium (ECGM) MV (PromoCell, Heidelberg, Germany) supplemented with 0.4% (w/v) ECGS/H, 2% (v/v) fetal calf serum (FCS), 10 ng/ml epidermal growth factor (EGF), 1 μg/ml hydrocortisone, 50 ng/ml amphotericin B and 50 μg/ml gentamicin. Human umbilical vein endothelial cells (HUVECs) were purchased from PromoCell (Heidelberg, Germany) and were used between the second and eighth passages. HUVECs were cultured in ECGM supplemented with 0.4% (w/v) ECGS/H, 2% (v/v) FCS, 0.1 ng/ml EGF, 1 ng/ml basic fibroblast growth factor (bFGF), 1 μg/ml hydrocortisone, 50 ng/ml amphotericin B and 50 μg/ml gentamicin.

HMECs (10⁴ cells) were seeded in 24-well culture plates and cultured for various periods in FCS-free ECGM MV in the absence (control) or presence of varying concentrations of κ-elastin (KE) or VEGF (20 ng/ml) or bFGF (20 ng/ml). At the end of each period of incubation, the cell number was determined using the violet crystal assay (Kueng et al., 1989). A standard absorbance curve was established.

Matrigel matrix (Sigma, France) was kept on ice for 24 hours. Then, 200 μl matrigel per well was added to a 24-well culture plate. After incubation at 37°C for 30 minutes, the gels were overlaid with 500 μl ECGM for HUVECs or ECGM MV for HMECs containing 4×10⁴ cells. Endothelial cells were then incubated in the absence or presence of KE, VGVAPG peptide or various agents at concentrations indicated in the text. Aiming to verify the contribution of S-Gal on EDP-mediated angiogenesis, we used lactose and V14 peptide as antagonists. Lactose was used to inhibit the binding of EDPs on EBP in keeping with the galactolectin-like property of this receptor (Hinek et al., 1998). V14 in turn, a 14mer peptide corresponding to part of the elastin binding sequence of EBP was used to compete with the binding of EDPs to EBP. The involvement of MMPs was demonstrated through the use of specific inhibitors: batimastat, a potent broad-spectrum MMP inhibitor, TIMP-2 known to inhibit MT1-MMP and MMP-2 without distinction, TIMP-1 that inhibits MMP-2 but not MT1-MMP and Dec-RVLR-cmk, a chloromethyl ketone peptide known to specifically impede proMT1-MMP activation by furin convertase, its intracellular activator. For experiments with siRNA (see below), 250 μl per well of matrigel was added to a 24-well culture plate. After incubation at 37°C for 15 minutes, the gels were overlaid with 500 μl ECGM MV containing 3×10⁴ HMECs pre-transfected with 25 nM siRNA₁₀₇ or scrambled siRNA₁₀₇. HMECs were then incubated in the absence or presence of 1 μg/ml KE in ECGM MV containing 2% (v/v) FCS. Capillary tube formation was observed at different times during a culture period of 24 hours with a phase-contrast microscope (Axiovert 25, Zeiss) equipped with a digital camera (Sony). Images were taken with a phase-contrast microscope with a 10× lens and semi-quantitative evaluation of pseudotube formation was performed in ten randomly selected fields (after black and white pixelization) by determining the number of black pixels relative to the total pixels. Experiments were performed in quadruplicate.

Capillary tube formation in three-dimensional type I collagen gel was performed as previously reported (Trochon et al., 1996). Briefly, HMECs (10,000 cells/well) were seeded on 2% (w/v) agarose gel in a 96-well plate. After an 18-hour incubation in complete ECGM MV, aggregate-forming cells were then incorporated into a three-dimensional type I collagen gel and incubated for 3 days at 37°C in FCS-free ECGM MV in the absence or presence of 100 ng/ml KE. Tube formation was observed by phase-contrast microscopy.

Endothelial cell migration assay was performed as previously described (Vincent et al., 2001). Briefly, HMECs (1×10⁵ cells/well) were seeded in 24-well plates and grown to confluence in ECGM MV. The cell monolayer was disrupted with a 1 mm cell scraper, and after washing with PBS, endothelial cells were incubated in basal ECGM MV supplemented with 2% (v/v) FCS (a concentration of FCS that allows cell survival but not cell proliferation) and stimulated in the absence or presence of KE (10, 100 and 200 ng/ml), VGVAPG peptide (10, 100, 200 ng/ml), bFGF (20 ng/ml) or VEGF (20 ng/ml). In some experiments, bFGF or VEGF were simultaneously added with KE or VGVAPG peptide. The influence of TIMP-2 or batimastat on elastin peptides-induced cell migration was evaluated at concentrations indicated in the text. After 24 hours of incubation, images were taken using a phase-contrast microscope with a 5× lens and cell migration was determined by measuring the number of cells invading a 0.5 mm² wounding area. Experiments were done in triplicate and ten fields from each well were randomly selected for cell counting. The percentage of migrating cells was determined relative to the control in the absence of effector.

Confocal microscopy analysis of actin filaments was performed on HUVECs and HMECs cultured on matrigel for 14 and 24 hours, respectively, in the absence or presence of KE (50 μg/ml). Actin filaments were visualized by tetramethyl-rhodamine isothiocyanate-labeled phalloidin as described (Manelli-Oliveira and Machado-Santelli, 2001). Confocal microscopy analysis of siRNAs was then performed. siRNAs were fluorescently labelled with Cy3 using the silencer™ siRNA labeling kit (Ambion, Huntingdon, UK) according to the manufacturer's instructions. HMECs were cultured on coverslips (LabTek, Nunc) in 24-well plates (Nunc, Roskilde, Denmark) in ECGM MV at 60% subconfluency and then transfected with 25 nM labeled siRNAs using 2 μl/well oligofectamine (Invitrogen, Cergy Pontoise, France) as transfection reagent in FCS-free medium for 72 hours at 37°C. Cells were observed by confocal laser-scanning microscopy using an MRC-1024 imaging system (Bio-Rad, Microscience, Hemel Hempstead, UK) coupled to an epifluorescence microscope (Olympus, Tokyo, Japan), equipped with a ×60 water immersion lens.

HUVECs were cultured on matrigel in the absence or presence of KE (50 μg/ml) for 14 hours and scanning electron microscopy analysis was performed as previously described (Anderson, 1966). The observations were carried out with a JEOL JSM 5400 LV scanning electron microscope. Changes in HUVEC shape were determined by semi-automatic and automatic quantitative image analysis as previously described (Godeau et al., 1986). Cell form factor was determined on 60 cells. For this purpose, the perimeter (P) and the surface (S) of a cell were determined from the outline of individual endothelial cells which were manually delimited with the pointer of the computer using the Imagenia 3000 Program (Biocom 200, Courtaboeuf, France). Evaluation of the form factor was determined by the formula 4πS/P², which is used as an index of the circularity of cells (Baak and Ort, 1983; Flook, 1987; Lorimier et al., 1998). As mentioned by Flook, the value of the circularity can decrease by elongation of the shape or by the addition of highly textured or serrated edge to an otherwise circular feature as can be observed when a cell progresses from an adhering state with a rounded aspect to a spreading or migrating state.

Gelatinolytic activities of MT1-MMP and EBP associated with plasma membranes were identified by gelatin zymography and western blot analysis as previously described (Brassart et al., 1998; Ntayi et al., 2004). Quantitative determination of total MT1-MMP was performed using the Biotrak MT1-MMP activity assay system (Amersham Biosciences Europe, Orsay, France) according to the manufacturer's instructions.

MMPs and TIMPs were analyzed by gelatin zymography and reverse zymography, respectively, as previously described (Huet et al., 2002). Molecular weight markers (Bio-Rad Laboratories) and recombinant MMP-2 (VWR International, Strasbourg, France) were added to each gel analyzed. Recombinant TIMP-1 and TIMP-2 (Calbiochem, Merck Eurolab, Fontenay-sous-Bois, France) were used as markers. Areas of gelatinolytic or TIMP activity were measured by automated image analysis, as described previously (Huet et al., 2002).

Protein samples were separated in 0.1% (w/v) SDS, 10% (w/v) polyacrylamide gel (Laemmli, 1970). After electrophoresis, proteins were transferred to Immobilon-P membranes (Millipore, Saint-Quentin en Yvelines, France) as previously described (Towbin et al., 1979). Then, the membranes were saturated with 5% (v/v) FCS, in 50 mM Tris-HCl, 150 mM NaCl, pH 7.5 (TBS) for 2 hours, and probed with either a mouse anti-human MT1-MMP monoclonal antibody (clone 114-1F2 from Calbiochem) (1:1000) or rabbit anti-human MT1-MMP (hinge region) polyclonal antibodies (1:5000) (Euromedex, Soufflewehersheim, France) or rabbit anti-human EBP polyclonal antibodies (Ntayi et al., 2004) directed against the EBP binding sequence of elastin (V14 peptide), overnight at 4°C. Membranes were then extensively washed in TBS containing 01% (v/v) Tween-20 (TBS/Tween), then probed with horseradish peroxidase (HRP)-conjugated sheep anti-rabbit or anti-mouse IgG antibodies (1:10,000) for 1 hour at 20°C. Membranes were washed in TBS/Tween and immunocomplexes were visualized by chemiluminescence using the ECL⁺ system according to the manufacturer's instructions (Amersham, France).

Total RNA from HUVECs or HMECs cultured as monolayers was harvested by using RNAzol B (Biogenesis LTD, Poole, UK), and extracted by the acid guanidinium/phenol/chloroform extraction method (Chomczynski and Sacchi, 1987). RNA concentration was determined by absorbance measurement at 260 nm and its integrity checked by 1.5% (w/v) agarose gel electrophoresis. RT-PCR was performed with 1 μg total RNA using the Hybaid Omnigen thermocycler (Teddington, Middx, UK) and two pairs of oligonucleotides (Invitrogen, France) as already reported (Giambernardi et al., 1997). Forward and reverse primers for human MT1-MMP and GAPDH were designed. MT1-MMP primers: forward, 5′-CGCTACGCCATCCAGGGTCTCAAA-3′; reverse, 5′-CGGTCATCATCGGGCAGCACAAAA-3′; and GAPDH primers: forward, 5′-ACCACAGTCCATGCCATCA-3′; reverse, 5′-TCCACCACCCTGTTGCTGT-3′.

Forward and reverse primers for human MT2-MMP, MT3-MMP and MT5-MMP were made as previously described (Ueda et al., 2003). cDNA products were amplified for 26, 28 and 32 cycles for all assays. PCR products were separated on 1.5% (v/v) agarose gel containing 1 μl/ml ethidium bromide and fragments of 497 bp, 578 bp, 461 bp, 565 bp and 452 bp lengths were obtained for MT1-MMP, MT2-MMP, MT3-MMP, MT5-MMP and GAPDH, respectively. The fluorescence of the bands was evaluated by scanning the gel at 312 nm and computerized using the Bio-Profile software (Vilbert Lourmat, Marne la Vallée, France).

The 21-nucleotide RNAs were synthesized by in vitro transcription using the silencer™ siRNA construction kit (Ambion). Four sets of specific siRNAs targeting human MT1-MMP mRNA (GenBank accession number Z48481) corresponding to the coding regions 107-127, 228-248, 949-969 and 1462-1482 relative to the start codon were tested but only the results obtained with siRNA_107-227 were shown. A leader sequence (CCTGTCTC) complementary to the T7 promoter primer was incorporated at the 3′ end of each sense and antisense oligonucleotide. In vitro transcription was carried out according to the manufacturer's instructions except for annealing of each siRNA oligotemplate to the T7 promoter primer, which was performed for 1 minute at 90°C followed by 1 hour at 37°C. The filling DNA step with Klenow DNA polymerase was performed for 45 minutes instead of 30 minutes. Antisense and sense siRNA_107-127 oligonucleotide templates were as follows: antisense, 5′-AAGCCTGGCTACAGCAATATGCCTGTCTC-3′; sense, 5′-AACATATTGCTGTAGCCAGGCCCTGTCTC-3′. Scrambled antisense and sense oligonucleotide templates were obtained by replacing four nucleotides (underlined in italics): GCTA was replaced by TCAG and TAGC by CTGA. Scrambled siRNAs were used as a negative control. SiRNAs were characterized by gel electrophoresis on 2% (w/v) agarose gel as well as by a nondenaturing 12% (w/v) polyacrylamide gel electrophoresis. The molar concentration (M) of the siRNAs was equal to the concentration of siRNA/14 in μg assuming that 1 nmol of an average 21-mer double strand (ds)RNA contained 14 μg RNA. HMECs were transfected with siRNAs at various concentrations and for various periods as indicated in the text. In some experiments, HMECs were incubated in the absence or presence of 1 μg/ml KE.

Statistical significance of the results was determined by using the Student's t-test. A value of P<0.05 was considered to be significant. Mean±s.d. values of three to four experiments are presented.

As EDPs were previously reported to influence the growth rate, the invasive potential or apoptosis of both normal and transformed cells, we suspected that such matrikines might similarly modify the angiogenic phenotype of endothelial cells. To validate those premises we initially evaluated the influence of EDP in vivo using the chick CAM model. From day 6 of embryonic development, membranes were treated daily with PBS (control), κ-elastin (KE) (50 ng) or VGVAPG (200 ng). As shown in Fig. 1A, in contrast to controls where angiogenesis was detected at day 8 of embryonic development and reached a maximum at day 9 (not shown), KE as well as VGVAPG peptide led to an enhanced rate of angiogenesis, occurring as early as 24 hours following peptide supplementation of embryos; it was also associated with an increased level of microvessel density reaching a maximum at day 8. In parallel to these in vivo experiments, the influence of KE on the ability of HMECs to reorganize in capillary tubes, when embedded in type I collagen lattice, was evaluated. KE considerably enhanced pseudotube formation (Fig. 1B). Similarly, when HUVECs or HMECs were cultured on matrigel, the surface area occupied by pseudotubes was increased by KE in a time- and concentration-dependent manner (Fig. 1C). Such an effect occurred within 6 hours of incubation and with peptide concentrations as low as 10 ng/ml, but was significant only after 14 hours of culture, with formation of numerous cellular extensions linking cellular aggregates at branch points of the array (Fig. 1C,D). Such effects of KE are similar to those obtained with VEGF.

Consistent with the intricate relationship between membrane calcium channel activation and cytoskeleton actin microfilaments, confocal microscopy analysis of HUVECs and HMECS revealed a main reorganization of actin bundles when endothelial cells were cultured on matrigel in the presence of EDPs or VEGF (Fig. 2A). In contrast to their spindle-shaped and sometimes rounded aspect in the control, endothelial cells adopted a more spread out shape in the presence of KE. Similar observations were made by scanning electron microscopy analysis of HUVECs cultured on matrigel (Fig. 2B). In the absence of KE, HUVECs reorganized into a sparse honeycomb network consisting of strings of rounded cells (Fig. 2Ba,c). By contrast, in the presence of KE, they formed a dense array of pseudotubes with numerous tight cell contacts and cellular protrusions within matrigel (Fig. 2Bb,d). Such changes in cell shape were then quantified on an image analysis basis. As shown in Fig. 2C, KE significantly decreased the cell form factor of HUVECs following just 4 hours of incubation, a phenomenon that was accentuated following 14 hours of incubation. As elongation and/or spreading of a cell is inversely proportional to the magnitude of such a factor, this result supported our proposal that cells adopted a more spread shape in the presence of KE.

The influence of EDPs on endothelial cell tubulogenesis might originate from either modulation of cell growth or increase of cell migration, or both. The kinetics of cell proliferation in the absence or presence of growth factor are shown in Fig. 3A,B. In the absence of growth factor, KE (200 ng/ml) did not significantly influence endothelial cell proliferation. However, in the absence of KE and growth factor, the number of cells drastically decreased with the time of culture. The kinetics of cell proliferation in the presence of KE and growth factor were similar to that obtained with growth factor alone (Fig. 3Ab,c). As shown in Fig. 3B, in the absence of growth factor, survival of endothelial cells after 5 days of culture was improved following treatment of cells with KE at concentrations ranging from 10 to 200 ng/ml, whereas no difference was observed in the presence of growth factor (Fig. 3Bb,c). By contrast, KE induced a concentration-dependent increase of endothelial cell migration in an in vitro wound healing assay (Fig. 3C). A 2.5- and 1.6-fold increase of cell migration was obtained with 200 ng/ml KE (2.67 nM) and 200 ng/ml VGVAPG peptide (400 nM), respectively. Such increments were higher than those obtained with 20 ng/ml bFGF (1.2 nM) or VEGF (0.6 nM). EDPs and growth factors acted independently of each other in terms of endothelial cell migration (Fig. 3Cb,c). These data suggested that EDPs triggered an increased rate of tubulogenesis through enhancing influence on endothelial cells migration.

As already mentioned, most effects of EDPs on cell phenotype are consecutive to the binding of those peptides to a cell-associated elastin-binding protein (EBP) with lectin-like property, which proved to be identical to a spliced form of β-galactosidase (S-Gal). To assess whether EDP-mediated influence on endothelial cell tubulogenesis could be attributed to the binding of such peptides to their cognate receptor, we first attempted to reproduce endothelial cell morphotype using VGVAPG peptide, the known repeated motif in tropoelastin interacting with EBP. VGVAPG peptide reproduced the effect of KE on pseudotube formation (Fig. 4A,B). Such effects were also reproduced by several peptides displaying an (X)GXXPG consensus sequence with type VIII β-turn conformation that favors S-Gal binding (data not shown). As a second set of experiments, aiming to verify the contribution of S-Gal on EDP-mediated angiogenesis, we used lactose or V14 peptide as antagonists. Both compounds attenuated the effect of EDPs on pseudotube formation (Fig. 4). Interestingly, polyclonal antibodies directed against the V14 sequence of EBP involved in the binding of elastin peptides reproduced the ligand effect on tubulogenesis (data not shown). As shown in Fig. 4B, EBP was expressed in HUVECs and HMECs as a 67-kDa protein as in fibroblasts.

As EDPs were previously shown to enhance MMP expression, acting as key players in angiogenesis and the invasive property of several cell types, we evaluated the influence of those matrikines on MMPs expression by HUVECs and HMECs. EDPs had no influence on MMP-1, MMP-3 and MMP-9 expression by HUVECs and HMECs (data not shown). Considering that MTI-MMP has recently been shown to play a major role in endothelial cell migration and tubulogenesis in several endothelial cell culture models, we next examined the effect of EDPs on the proMT1-MMP/proMMP-2 system in HMECs. Gelatin zymography analysis was used to evaluate gelatinase A secreted into the culture medium and associated with cell extracts in HMECs cultured as monolayer, following a 24-hour cell exposure to different concentrations of KE. As illustrated in Fig. 5A, HMECs were found to constitutively release latent proMMP-2 into the conditioned medium (Fig. 5Aa). KE also induced an increase of both pro and active MMP-2 forms in cell extracts (Fig. 5Ab). Such effects were concentration-dependent and were observed with 100 ng/ml KE, a concentration shown above to influence cell morphogenesis and migration. At the highest KE concentration (100 μg/ml), secreted and membrane-associated MMP-2 levels were increased two- and 19-fold, respectively, and active MMP-2 represented 31.8% of total cell-bound enzyme. It suggested that elastin peptides could trigger major quantitative alterations in partners involved in proMMP-2 activation process i.e. MT1-MMP and TIMP-2. However, levels of TIMP-2, TIMP-1 (Fig. 5Ac) as well as αVβ3 integrin (data not shown) were not substantially modified by supplementing HMEC culture medium with KE. However, those peptides enhanced proMT1-MMP expression and activation, similar to MMP-2, at the protein (Fig. 5Ad) and mRNA levels (Fig. 5B). Similar data were obtained with HUVECs instead of HMECs and could also be reproduced by VGVAPG peptide (data not shown).

To investigate the contribution of the MT1-MMP/MMP-2 system in the elastin peptide-mediated influence on HMEC tubulogenesis and migration, experiments were performed in the presence of agents known to interfere with the activity of those enzymes. As shown in Table 1, batimastat nearly totally abolished the effect of KE on pseudotube formation by HMECs cultured on matrigel. A similar level of inhibition was obtained with TIMP-2 whereas TIMP-1 displayed only a partial repressive effect on tubulogenesis. In addition, Dec-RVLR-cmk suppressed the effect of KE on tubulogenesis, in a similar manner to batimastat. Overall, these data argue for the main involvement of MT1-MMP in the EDP-mediated influence on endothelial cell behavior. Indeed, TIMP-2, as well as batimastat, also inhibited the effect of KE on cell migration as determined by the wound healing assay, thus further emphasising the central role of MT1-MMP in such phenomena (Table 1).

To address this issue in a less ambiguous way, we used a siRNA approach for silencing proMT1-MMP mRNA expression. Four independent sets of 21-bp siRNA duplexes were chosen. In preliminary experiments, we showed that only siRNA₁₀₇ and siRNA₂₂₈ proved to be efficient in suppressing proMT1-MMP expression in endothelial cells (data not shown). Further experiments were performed with siRNA₁₀₇. Confocal microscopy analysis was first performed to determine the subcellular distribution of siRNA₁₀₇ within HMECs. For this purpose, siRNA₁₀₇ was labeled with Cy3, a fluorescent marker of siRNA duplex, and HMECs were transfected for 72 hours with 25 nM Cy3-siRNA₁₀₇ or scrambled Cy3-siRNA₁₀₇. As shown in Fig. 6A, scrambled Cy3-siRNA₁₀₇ localized preferentially within the cytosol of cells, whereas Cy3-siRNA₁₀₇ appeared mostly perinuclear. This localization was often associated with morphological changes, with cells exhibiting a more condensed shape and less pseudopode formation compared to HMECs transfected with scrambled siRNA₁₀₇. The influence of siRNA107 on proMT1-MMP mRNA expression was time- (Fig. 6C) and concentration-dependent (Fig. 6D). A 75-80% decrease of proMT1-MMP RNA level was reached by transfecting 25 nM and 50 nM siRNA107, respectively, for 72 hours. Such siRNA transfection led to a complete inhibition of EDP-induced proMT1-MMP expression and proMMP-2 activation as determined by the Biotrak ELISA assay (Fig. 7A) and gelatin zymography analysis (Fig. 7B). Moreover, it abolished the effect of EDPs on pseudotube formation (Fig. 7C,D). SiRNA₁₀₇ is specific to MT1-MMP as no effect was observed on MT2-MMP, MT3-MMP and MT5-MMP expression (data not shown).

Altogether these data demonstrate that upregulation of proMT1-MMP expression and activation was the main proteolytic determinant involved in the EDP-mediated accelerated rate of angiogenic phenotype.

Initial in vivo experiments, revealing an enhancement of angiogenesis by EDPs prompted us to investigate, in vitro, the EDP-mediated mechanism involved in such a process. Data obtained indicated that elastin fragments and peptides containing GXXPG sequence (VGVAPG) found as repeats in tropoelastin could accelerate the angiogenic phenotype of endothelial cells from micro and macrovasculature in the matrigel assay. VGVAPG-containing elastin peptides were shown to be generated in vivo from lung elastin by human leukocyte elastase in bronchoalveolar lavages of patients with emphysema (Maccioni and Moon, 1993); in addition, such a hexapeptide motif was shown not to be cryptic in human skin elastin and liberated in the vicinity of vertical melanoma by elastin-degrading enzymes (Ntayi et al., 2004). Moreover, SGVAPG, AGGLPG and MGGIPG sequences as found in α1 chain of type XV collagen, α2 chain of type V collagen and fibrillin covalent structures, respectively, mimicked the effects of EDPs in in vitro angiogenesis (data not shown). However, their action as cryptic sites or liberated matrikines is currently purely speculative. We show that EDPs induced cytoskeleton reorganization in endothelial cells when cultured on matrigel, an effect that had already been reported for smooth muscle cells (Mochizuki et al., 2002). This result is also in keeping with the known effect of EDPs on calcium flux and microtubular cytoskeleton refolding, properties that are associated with marked activation of the MT1-MMP/MMP-2 system in several cell models (Sakim and Yana, 2003). Particularly, MT1-MMP activity was reported to control cell geometry of tumor cells within the confines of the 3D ECM (Hotary et al., 2003). We have demonstrated that, under our experimental conditions, the EDP-mediated angiogenic phenotype was not associated with plasminogen activators or MMP-1 and MMP-9 upregulation (data not shown). However, at concentrations found in the circulation and lower (10^-6 to 10^-2 mg/ml) (Kucich et al., 1983; Fülöp et al., 1990), these peptides could significantly enhance proMT1-MMP and proMMP-2 expression and activation in HUVECs and HMECs.

The EDP-mediated angiogenesis phenotype was triggered by EBP as lactose and V14 completely abolished the effect of EDPs on endothelial cell migration, capillary tube formation and proMT1-MMP and proMMP-2 expression and activation. Such results are in accordance with a previous report indicating that the binding of galactosidase-bearing moieties like lactose to EBP causes the loss of its ability to bind elastin and induces its dissociation from the complex (Hinek et al., 1998). In addition, the inhibition of EDP-mediated tubulogenesis by batimastat, TIMP-2 but not TIMP-1, or by an inhibitor of furin convertase, is consistent with the putative contribution of MT1-MMP in the effects observed. To gain more convincing evidence of its direct involvement, we developed a siRNA approach for selectively silencing MT1-MMP in HMECs. SiRNAs were synthesized by in vitro transcription as, according to a previous report (Brown et al., 2002), siRNA produced by in vitro transcription is as much as 20-fold more potent than chemically synthesized siRNA with a similar sequence. The perinuclear localization of siRNA₁₀₇ was consistent with previous reports showing similar subcellular distribution of siRNAs used to silence other genes (Byron et al., 2002; Nykanen et al., 2001). This localization could represent sites of siRNA processing or sites where the RNA-induced silencing complex (RISC) resides (Montgomery et al., 1998). By contrast, scrambled siRNA₁₀₇ localized essentially in the cytoplasm. HMECs transfected with siRNA₁₀₇ adopted a condensed morphology that is consistent with the role of MT1-MMP in the control of cell shape. Consequently EDP-mediated pseudotube formation on matrigel was drastically inhibited. These data further emphasized that MT1-MMP plays a pivotal function in EDP-mediated angiogenic phenotype. During the course of the present investigation, Ueda and colleagues, using a similar approach to silence MT1-MMP expression, demonstrated that downregulation of this endopeptidase reduced the invasive capacity of HT-1080 fibrosarcoma and gastric carcinoma cell lines in matrigel and also decreased their motility on hyaluronan (Ueda et al., 2003). Given the importance of MT1-MMP in promoting EDP-mediated angiogenic phenotype and also tumor progression (Brassart et al., 1998; Huet et al., 2002; Ntayi et al., 2004), such a siRNA-MT1-MMP silencing strategy might represent a more suitable alternative to the use of MMP inhibitors in treating tumors in clinical settings.

Angiogenesis estimated by microvessel density or other endothelial cell markers is an independent prognostic variable in vertical growth phase melanomas and is associated with intense dermal elastolysis (Labrousse et al., 2004). The importance of MT1-MMP and MMP-2 in angiogenesis and tumor invasion has been firmly established through several investigations. Angiogenesis impairment has been consistently observed in MMP-2-deficient mice (Itoh et al., 1998) and MT1-MMP-deficient mice display reduced activation of latent MMP-2 and vascular inversion of calcified cartilage, and fail to respond to bFGF in the corneal angiogenesis assay (Zhou et al., 2000). Moreover, MT1-MMP has been shown to be mainly involved in endothelial cell migration, invasion and capillary tube formation in various angiogenesis models including tubulogenesis in 3D collagen and fibrin gel invasion assays (Lafleur et al., 2002; Galvez et al., 2001). According to previous reports, MT1-MMP may promote cell migration not only through degradation and remodeling of cell-associated ECM, but also through its ability to process and activate several cell surface molecules, such as CD44, integrin αVβ3 and tissue transglutaminase (tTG), that are known to regulate cell migration (Seiki, 2003). Indeed, CD44H is shed by MT1-MMP further promoting cell migration (Kajita et al., 2001). However, the precise mechanism involved in the stimulation of cell motility is still unknown but it was speculated that the processing of CD44 by MT1-MMP could promote cell detachment from the ECM or generate cell signalling by activating the ERK pathway, further inducing cell migration. Alternatively, it could trigger transcriptional activation of target genes through the cleaved cytoplasmic portion of the processed CD44 (Kajita et al., 2001; Gingras et al., 2001; Okamoto et al., 2001). Moreover, the alpha V chain of αVβ3 integrin is alternatively processed by MT1-MMP into a functional form that stimulates migration through phosphorylation of focal adhesion kinase (FAK) (Deryugina et al., 2002; Ratnikov et al., 2002; Baciu et al., 2003), and cleavage of tTG by this enzyme also promoted cell migration on type I collagen (Belkin et al., 2001). On the other hand, the matrix-degrading activity of MT1-MMP is required for the formation of an endothelial cell tubular network in fibrin and collagen gels (Hiraoka et al., 1998; Lafleur et al., 2002; Haas et al., 1998; Galvez et al., 2001).

Importantly, the link between elastin degradation, MT1-MMP upregulation and angiogenesis, may be related to several cardiovascular diseases where enhanced angiogenesis (Einstein, 1991; Thompson et al., 1996), degradation of elastic fibers (Robert and Robert, 1980; Nakashima et al., 1990; Nakashima and Sueishi, 1992; Jacob, 2003) and/or MT1-MMP overexpression (Rajavashist et al., 1999; Hong et al., 2000) are associated. As a key example, atherosclerotic abdominal aortic aneurysm (AAAA) is associated with neovascularization in all three layers (Thompson et al., 1996; Holmes et al., 1995; Ferrara et al., 1991), VEGF overexpression (Kobayashi et al., 2002) and tropoelastin accumulation (Krettek et al., 2003). Neovascularization in AAAA could also be related to enhanced elastin degradation, as EDP (VGVAPG) was shown to induce several characteristic features of aneurysmal disease such as an increase of vessel density in a rat aneurysmal model (Nackman et al., 1997) and an increased concentration of soluble elastin fragments in human serum from patients with aneurysmal, ulcerative manifestations of atherosclerosis (Petersen et al., 2002) and acute aortic dissection (Shinoara et al., 2003). In addition, as MT1-MMP is overexpressed by proinflammatory molecules, it might also contribute to the enhanced local matrix degradation in human atherosclerotic plaques as previously reported (Rajavashisth et al., 1999; Hong et al., 2000). On the other hand, decreased arteriolar and capillary density in a spontaneous hypertensive rat (SHR) model correlated with decreased MT1-MMP and VEGF receptor expression (VEGFR-2) (Wang et al., 2004). These data were in keeping with previous reports showing an increase of aortic elastin content in hypertensive animals (Keeley and Alatawi, 1991); hypertension was also associated with elastin haploinsufficiency in human and mice (Faury et al., 2003). By contrast, angiogenesis was restored by increasing the levels of VEGFR-2 and MT1-MMP using a sponge implantation model associated to a VEGF gene transfer technology in SHR (Wang et al., 2004). However, the aortic elastin content was not appreciated in this study.

Recently, MT1-MMP overexpression has been associated with increased VEGF expression in breast carcinoma cell lines (Sounni et al., 2004). Under our experimental conditions EDPs were found not to influence VEGF or bFGF production by endothelial cells (data not shown). However, as EDP-mediated MT1-MMP upregulation triggered the PI3-kinase/Akt pathway in endothelial cells (data not shown), we cannot exclude the fact that those peptides might transactivate VEGFR-1 through EBP as this receptor was reported to inhibit VEGFR-2-mediated proliferation but not migration via PI3-kinase activation (Zeng et al., 2001). Such a receptor transactivation mechanism was reported in aortic smooth muscle cells in cultures where EBP occupancy by EDPs activates PDGF receptors (Mochizuki et al., 2002). Experiments to test this hypothesis are currently underway.

Thus, EDP-mediated MT1-MMP expression on the cell surface of endothelial cells, by triggering matrix pericellular proteolysis, MT1-MMP endocytosis, MMP proteolytic activation cascades and/or cell surface receptor shedding, may be of importance in processes where those mechanisms are required for angiogenesis. Overall, our results ascribed a new function to EBP as a mediator of angiogenesis.

We thank Hervé Kaplan for technical assistance with microscopy. This work is part of an Interregional Research Program Régulation de la Matrice Extracellulaire et Pathologie on angiogenesis and was supported by a grant of the `Ministère de l'Education Nationale et de la Recherche', and by CNRS (Centre National de la Recherche Scientifique), France.