Angiogenesis 1998/1999; 2; 331344

ORIGINAL ARTICLE

The formation of tubular structures by endothelial cells is under the control of brinolysis and mechanical factors
one Tranqui1 Bruno Vailhe1, Marc Lecomte2, Nicolas Wiernsperger2 and Le
Laboratoire de Bioenergetique Fondamentale et Appliquee, Universite Joseph Fourier, B.P.53X, 38041 Grenoble Cedex 9, France; 2 Lipha-Inserm U 352, Insa Lyon, Bldg 406, 69269 Villeurbanne, France
1

This study highlights the importance of several factors involved in the formation of capillary-like structure formation (CLS) using Human Umbilical Vein Endothelial Cells (HUVEC) and Bovine Retinal Endothelial Cells (BREC) cultured on brin gels. The brin concentration inducing (CLS) was 0.5 mg/ml for HUVEC and 8 mg/ml for BREC. The high brin concentration required for the latter cells appeared necessary to counterbalance the extensive brinolysis of the gel by the BREC. Fibrin degradation products measured in the culture media showed that brin degradation was mandatory but not sufcient for CLS formation. Fibrin degradation acted in concert with the mechanical, concentration dependent properties of the gels to induce CLS. For example, HUVEC did not form CLS on a rigid brin of 8 mg/ml in spite of brinolysis. As cell reorganisation occurred, the brin was disrupted (HUVEC) or pleated (BREC) giving indirect proof of the development of mechanical forces. During CLS formation, an increasing amount of latent TGFb1 was measured in the medium (10001700 pg/ml). The active form of TGFb1 was not, however, detected and the addition of anti-TGF-b1 antibody to the medium did not inuence the formation of the CLS network. Yet, added activated TGF-b1 led to the formation of less organised structures, that were completely abolished by the concomitant addition of the same anti-TGF-b1 antibody. Thus, it is likely that TGF-b1 secreted by the endothelial cells remained in its latent form. In conclusion, a balance between the mechanical properties of brin and the brinolytic activity of each cell type may regulate CLS formation in our models. We think that the high brinolitic activity of the BREC may represent a defense mechanism to protect the retina against thrombosis-induced damage in vivo.

Key words: bovine retinal endothelial cells, endogenous TGF-b1, brin, brin degradation products, human vein endothelial cells, in vitro angiogenesis, in vitro tubule formation

Abbreviations: BREC bovine retinal endothelial cell(s); CLS capillary-like structure(s); FDP brin degradation products; HS human serum; HUVEC human umbilical vein endothelial cell(s); MMP-matrix metallo proteinases; PAI-1- plasminogen activator inhibitor-1; TGF-b1 -transforming growth factor beta1; u-PA urokinase-type plasminogen activator; u-PAR urokinase-type plasminogen activator receptor

(Received 2 December 1997; accepted in revised form 14 October 1998)

Introduction
Angiogenesis is a multifactorial process dened as the formation of new blood vessels by sprouting from pre-existing vessels.1,2 The list of compounds involved in angiogenesis grows but little is known about the relative importance of each factor.25 In an attempt to clarify and identify the factors involved, many in vitro models of neovascularization have been developed.69 Under conditions specic to each model, the endothelial cells undergo a morphological change resulting in the formation of a network of capillary-like tubes (CLS) which may exhibit lumina. Endothelial cells of different vascular origin have been used to develop these models: human umbilical vein
Angiogenesis Vol 2 No 4 1998/1999 331

Correspondence to Leone Tranqui, Laboratoire de Bio energetique, Universite Joseph Fourier, B.P.53X, 38041 Grenoble Cedex 9, France. Tel: (+33) 4 76 63 58 21; Fax: (+33) 4 76 51 42 18; E-mail: leone.tranqui@ujf-grenoble.fr

1999 Kluwer Academic Publishers

B. Vailhe et al.

endothelial cells (HUVEC) or microvascular endothelial cells which display organ specicity and morphological heterogeneity. The fundamental importance of the biochemical composition of the matrix has been demonstrated with in vitro angiogenesis models.1012 All these models suggest that the tension-dependent biomechanical interaction between endothelial cells and matrix proteins may serve to regulate capillary development.1316 The angiogenic properties of brin as a provisional matrix has been recognised, and we previously suggested that the mechanical stresses resulting from the balance between cell traction and the mechanical resistance of the brin gel are likely to play a critical role.17 As brin plays a prominent role in wound healing, tumour growth, and other events involving angiogenesis, it occupies a place of special interest among the matrices used to test in vitro angiogenic compounds. Matrix remodelling coupled to the proteolytic activity of the cells are also two fundamental processes which regulate capillary formation.5,2 Among the proteolytic activities already described, the brinolytic system plays a central role.5,18 It is mainly modulated by the tripartite complex formed by the urokinase receptor (u-PAR), urokinase (u-PA), and its inhibitor (PAI-1), which controls plasmin formation.1921 The role of brinolysis could be dual : to remodel the matrix and to activate latent cytokines either linked to the matrix or secreted by the cells. Among the cytokines identied to date, basic broblast growth factor (bFGF),22 vascular endothelial growth factor (VEGF) and transforming growth factor beta (TGF-b) are the most extensively characterised.2,23,24 Transforming growth factors-b constitute a family of pluripotent regulators of cell growth and differentiation. They are secreted as latent complexes that need to be converted into active forms before interacting with their ubiquitous receptors.25,26 The mechanism regulating the activation of these latent forms into active factors in vivo are not fully dened but may involve processes such as proteolysis.27,28 The roles of TGF-bs are not fully understood, but they may inhibit the secretion of proteases, and stimulate the secretion of protease inhibitors, especially PAI-1.24,29 Thus, proteases play a key role in angiogenesis as they intervene at several steps in this
332 Angiogenesis Vol 2 No 4 1998/1999

process, such as matrix degradation and regulation of other factors implicated in angiogenesis.5 The aim of this study was to describe two in vitro angiogenesis models using bovine retinal endothelial cells (BREC) and HUVEC, in order to identify the specic conditions under which each cell type formed CLS. The well-characterised HUVEC were used as a reference. Retinal endothelial cells were selected as they can provide a novel physiological model to study factors involved in angiogenic eye diseases. The respective contribution of the brin matrix concentration and of the cellular proteolytic activity in inducing the reorganization of the cells were studied in these two endothelial cell models. Analysis of the temporal morphological changes of the endothelial cells and measurements of brin degradation have been performed in order to establish a correlation between matrix remodelling and CLS formation. Moreover, the potential role of the endogenous and exogenous TGF-b1 in the formation of CLS in these culture systems were investigated.

Materials and methods

Matrix protein purication

Human bronectin was puried by afnity chromatography according to the method of Engvall and Ruoslahti.30 Human brinogen was puried according to the method of Keckwick et al.31 The purity of the brinogen used in our experiments, including the absence of contaminating plasminogen, thrombin, factor VIII, von Willebrand factor and bronectin was checked as previously described.32 We tested the absence of plasmin activity as follows: after coagulation of the brinogen solution, the brin clot was incubated for 3 days at 37C in phosphate buffer saline (PBS) containing 1 mM Ca2+, 1 mM Mg2+, and no degradation was observed.
Cell culture

Unless otherwise indicated, all the reagents were purchased from Sigma Chemical Co., Saint Louis, MO, USA.

The formation of tubular structures by endothelial cells

HUVEC, isolated according to the method of Jaffe et al.,33 were seeded on bronectin and grown in 199 medium supplemented with 20% heat-inactivated human serum (HS) from (Sigma) glutamine and antibiotics, until conuence. Cells were maintained at 37C in 5% CO2 humidied atmosphere and the medium was changed every 2 days. Cells were subcultured on a bronectin coat, seeded at a density of 105 cells/ml and used within three passages. The endotoxin content of the human serum furnished was less than 0.1 ng/ml. BREC were cultured essentially as described previously.34 Two retinas were processed and homogenized in ice-cold oxygenated Ca2+/Mg2+ free Hank's buffer saline solution (HBSS). After centrifugation, the pellet was digested for 1 h at 37C with a mixture containing collagenase/dispase (2 mg/ml) (Boehringer Mannheim), DNAse type 1 from bovine pancreas (20 U/ml) and Tosyl Lysine Chloromethyl ketone (TLCK) (50 ng/ml) in oxygenated Ca2+/Mg2+ free HBSS solution containing penicillin, streptomycin, and 10 mM HEPES, pH 7.4. After digestion, the suspension was ltered through a sterile 70 lm nylon sieve and the microvessels retained on the lter were washed twice with Ca2+/Mg2+-free HBSS. Microvessels were then centrifuged at 600 g for 5 min at 4C, washed twice and resuspended in 3 ml of DMEM containing 10% fetal bovine serum, glutamine (2 mM) and antibiotics. The microvessels were seeded in 60 mm tissue culture dishes precoated with 100 lg of bronectin and 50 lg of collagen IV in 2 ml of HBSS for 3 h. After a period of 3 h for attachment, the medium with debris was removed and replaced by fresh DMEM containing 15% heat-inactivated HS, glutamine and antibiotics. The medium was changed every 2 days until cells reached conuence after 810 days. Cells were subcultured on a bronectin coat and seeded at a density of 105 cells/ml. They were used within three passages. The endotoxin content of the human serum furnished by Sigma was less than 0.1 ng/ml. HUVEC and BREC phenotype was evaluated on the basis of the cobblestone appearance of the cell monolayer and by positive staining with a Von Willebrand's factor polyclonal antibody furnished by Sigma. Contamination of BREC by pericytes was evaluated after staining with anti-a-

smooth muscle actin antibody and determined to be less than 5%.

In vitro angiogenesis assays on brin gels

One ml of brinogen solutions at concentrations varying between 0.5 and 8 mg/ml were mixed with a constant quantity of thrombin (0.2 units/ ml; Sigma Chemical Co.) in Ca2+/Mg2+ free HBSS, in 3.5 cm diameter Petri dishes, in order to obtain brin gels of 1 mm thickness. Fibrin was allowed to polymerise overnight at 37C. For the assays, 1.5 105 cells suspended in 1 ml of 199 medium containing 1% heat-inactivated HS, glutamine and antibiotics, were seeded on brin gels. Under these conditions CLS formed within 12 days. We veried that plasmin was the major protease responsible for brin degradation in our culture system, by adding alpha2-antiplasmin (Sigma) in the culture medium at concentrations varying between 0.1 and 2 lm. We routinely used aprotinin (Sigma Chemical Co.), an inhibitor of serine proteases, with the exceptions of thrombin and active factor 10, to modulate the degradation of the gels by plasmin. Aprotinin inhibits proteases by interacting in a reversible manner with the proteases. An inactive enzyme-inhibitor complex is formed in an equimolar ratio.
Measurement of the length of the capillary-like structures

The quantication of the formation of the CLS in our assay was determined by measuring the length of the CLS. We used 20 micrographies issued from 3 different experiments to obtain the average length of the tubular structures formed under dened culture conditions.
Microscopy

For confocal microscopy, cells were seeded at 1.5 105 cells/ml and grown on 22 22 mm brin-coated glass coverslips (Poly Labo, Paul Block & Cie, Strasbourg, France) and maintained in culture until the formation of CLS. Glass coverslips were placed at the bottom of Petri dishes of
Angiogenesis Vol 2 No 4 1998/1999 333

B. Vailhe et al.

33 mm diameter. The brinogen solution was added followed by the thrombin. The coverslip was maintained in place while the solutions were poured by holding them down with sterile tweezers. The gel lling the Petri dish was carefully cut around the coverslips which were then removed and immersed in freshly prepared 3% paraformaldehyde in PBS at 37C for 15 min. The glass coverslips were washed with PBS and stained with 0.2% phloxine in water for 2 min (Sigma Chemical Co.) to localize the whole cell body. The glass coverslips were mounted in a mixture of Mowiol 4-88 (Hoechst) and glycerol.34,35 Confocal photometry was carried out with a LSM 410 Zeiss confocal imaging system equipped with a planapo X63 oil immersion, NA 1.40 objective lens. An helium neon laser (k 633 nm) was used to excite phloxine and the emission signal above 650 nm was recorded. The micrographs represent a longitudinal optical section of the CLS. Phase-contrast photographs of living cells were recorded using an inverted microscope (Diaphot Nikon).
Culture media processing

antibody recognizes brinogen, brin monomer and FDP. Conditioned culture media were adsorbed onto 96-well Nunc-immuno plates overnight at 4C. Then 100 ll of 1% bovine serum albumin in PBS were added for 35 min at 37C to saturate the wells. The plate was washed at room temperature for 5 min with 200 ll PBS containing 0.05% Tween 20; then 100 ll of primary antibody were added to each well. After a 2-hour incubation at room temperature, wells were washed three times for 5 min at room temperature with 200 ll of PBS-Tween 20 and 100 ll of a secondary antibody linked to peroxidase were added (goat anti-rabbit IgG, Bio-Rad Laboratories, Richmond, CA, USA). Two hours later, the wells were washed three times for 5 min with 200 ll of PBS-Tween 20 and once with 200 ll of PBS, before addition of the peroxidase substrate (TMB Peroxidase Substrate Kit, Bio-Rad Laboratories). After 5 min, the absorbance was measured at 450 nm using an ELISA microplate reader.
Calculation. All measurements were performed in

Cells were seeded on gels consisting of 0.5 mg/ml (HUVEC) or 8 mg/ml (BREC) brin, with or without aprotinin, in 199 medium containing 1% HS. Culture media were withdrawn either at the end of the angiogenesis process or at specic times when morphological changes occurred, i.e. 2, 12, 18, 24, 28 2 hours after seeding for HUVEC and 2, 22, 32, 46, 50 2 hours after seeding for BREC. Both conditioned culture media and brin gels were collected from the dishes. Then the mixture of conditioned culture media and brin gels, was centrifuged 10 minutes at 5000 g. The supernatants were withdrawn and used to quantify both the brin degradation products (FDP) and TGF-b1.
Immunoassay for brin degradation products

duplicate. The standard curves were constructed by plotting the mean absorbance on the y-axis against the FDP concentrations on the x-axis. The standard curve was linear over an FDP concentration range of 0 to 5 ng/ml. The FDP mixture was obtained by digesting a 1 mg/ml brin gel by human plasmin (610 units/mg; Sigma Chemical Co.) for a period of 24 hours at 37C.
Amount of brinolytic activity in human serum. We routinely tested all the serum batches for their brinolytic activity as follows : brin gels were incubated at 37C in the absence of cells with medium containing serum concentrations of 1 to 5%; after 24 h the amount of FDP was measured. TGF-b1 assays

to detect the brin degradation products in culture supernatants (rabbit anti-human, Dako A/S Glostrup, Danemark; reference A 0080). This
334 Angiogenesis Vol 2 No 4 1998/1999

Assay procedure. A polyclonal antibody was used

trations of TGF-b1 were determined in the culture media with a commercially available kit (QuantikineTM, `Human TGF-b1 Immunoassay', R&D sytems, Abingdon, Ox, UK). The minimal detectable dose was 5 pg/ml. Latent TGF-b1 was

Detection of TGF-b1 in cell media. The concen-

The formation of tubular structures by endothelial cells

also detected in the media by acid treatment to activate latent TGF-b1 to immunoreactive TGFb1. The activation procedure was performed according to the manufacturer's instructions. the amount of TGF-b1 in a 1% serum-containing medium after incubation at 37C in the absence of cells remained constant over time at about 1000 pg, after heat activation of the serum.
Neutralization of potentially active endogenous TGF-b1. To neutralize the biological activity of

TGF-b1, a chicken polyclonal anti-human TGFb1 antibody (R&D systems; AB101NA) was used. According to the manufacturer, it has the ability to neutralize the biological activity of recombinant human TGF-b1 and porcine TGF-b1.2. The Neutralization Dose50 ranged between 0.2 and 0.6 lg/ml. In our assay, this antibody was added at different concentrations (5, 10 or 20 lg/ml) to the culture medium, either at cell seeding (HUVEC and BREC), or 12 h (HUVEC) and 20 h after seeding (BREC).

the brin concentration rose from 0.5 to 8 mg/ml (Figures 1AD), while BREC formed CLS only at high brin gel concentration. At lower concentrations the gels were degraded by BREC before CLS could form (Figures 1EG). Nevertheless, for both cell types, when the angiogenesis process took place, the gel-induced morphological changes led to the formation of hollow capillary-like tubes, as shown by confocal microscopy (Figure 2B, arrow). During this process the brin gel was remodelled: CLS formed by HUVEC delimited areas in lacunae where brin gel was disrupted, (Figure 1B) while BREC were able to pleat the brin gel (arrowhead in Figure 1H). This prompted us to investigate whether this gel remodelling by cell forces depended on brin degradation and was paralleled by the morphological changes of the cells.
The degradation of the brin gels was paralleled in time by cellular morphological changes

effect of exogenous TGF-b1 (Genzyme, Cambridge, MA, USA; reference TG010), active TGFb1 was added in culture media (1 ng/ml) at cell seeding.

Effect of exogenous TGF-b1 on endothelial cell reorganisation into CLS on brin. To determine the

Results
HUVEC and BREC form capillary-like structures when seeded on a brin gel of appropriate concentration

HUVEC and BREC form CLS when seeded on a brin gel in medium 199 supplemented with 1% human serum. However, the brin gel concentration inducing CLS formation depended on the cell type. HUVEC formed CLS on a brin gel concentration of 0.5 mg/ml (Figure 1A, arrow) and the average length of the tubular structure was 550 50 lm). The BREC required a brin gel concentration of 8 mg/ml (Figure 1H, arrow) and the length of the tubes was shorter, 400 50 lm. HUVEC lost their ability to form CLS when

Figure 3 shows the kinetic of CLS formation either by HUVEC seeded on a 0.5 mg/ml brin gel (AD) or by BREC seeded on an 8 mg/ml brin gel (EH) in the absence of protease inhibitors. During CLS formation, BREC and HUVEC exhibited the same sequence of morphological changes. Cells rapidly rearranged themselves in elongated and thin tubes. The following steps were observed: cells adhered (Figures 3A and 3E), then areas devoid of cells called lacunae appeared (Figures 3B and 3F), and the lacunae increased in number and size (Figures 3BD and 3FH). The tube structures were always formed at the junction of two lacunae (arrows in Figures 3D and 3H). The length of the tubes increased from 450 50 lm (Figure 3C) to 550 50 lm (Figure 3D) for the HUVEC and from 100 50 lm (Figure 3G) to 400 50 lm for BREC (Figure 3H). The most obvious difference observed between the two cell types was the kinetics of formation of each step leading to CLS formation. As cells degraded the brin gel, we measured the brin FDP present in the culture medium at different times during the network formation. Figures 4A and 4B show the results obtained from one typical experiment. A slight degradation was already meaAngiogenesis Vol 2 No 4 1998/1999 335

B. Vailhe et al.

Figure 1. Effect of brin concentrations on the ability of brin gels to support capillary-like structure formation by HUVEC and BREC. HUVEC (AD) or BREC (EH) were seeded on brin gels at concentrations ranging from 0.5 mg/ml to 8 mg/ml. CLS (arrows) were observed 24 2 h after HUVEC (A) were seeded on 0.5 mg/ml brin gel while 8 mg/ml gels were required for CLS formation by BREC (H) within 48 2 h. The average length of the tubes was 550 50 lm in A and 400 50 lm in H. Gel disruption by HUVEC was observed in the lacunae. Asterix in B in the lacunae points out the brin depleted area; it appears clearer than the remaining brin located at the right edge of this lacunae. BREC induced pleats in the brin gels (arrowheads in H). Bar, 450 lm.

surable 2 h after seeding, a time at which cells adhered to the matrix. An increase in degradation occurred between 10 and 18 2 h after HUVEC seeding and between 22 and 32 2 h after BREC seeding, i.e. when the rst lacunae formed. This
336 Angiogenesis Vol 2 No 4 1998/1999

brinolysis led to cell detachment at the end of the process if protease inhibitors were not added to the medium. Both cell types degraded approximately the same quantity of brin by the end of the CLS formation. The quantity of FDP

The formation of tubular structures by endothelial cells

Figure 2. Longitudinal confocal micrographies of a capillarylike structure (HUVEC). Cross-sectional photographs were taken at 0.5 lm intervals from the basal to the apical surface. A representative area of two confocal planes (A and B) are shown, with distance intervals of 5 lm. Cells were xed and stained by phloxine. Phloxine was excited at 633 nm and emission signal above 650 nm was recorded to localise the cell body. The arrow points to the lumen-like zone of the tubular structure, which appears in black on this photograph. Bar, 50 lm.

measured in the medium was 0.2 mg/dish of HUVEC (Figure 4A), and 0.18 mg/dish of BREC (Figure 4B). These results show that matrix degradation was paralleled by intense morphological changes of the cells. In the following set of experiments, we veried that matrix degradation was a critical step required for CLS formation.
Fibrin remodelling and specic mechanical properties of the brin gel were necessary for the formation of capillary-like structures

Table 1 shows the relationship between the degradation of brin gels by either HUVEC or BREC and the formation of CLS. Firstly, in vitro angiogenesis was always accompanied by brinolysis; however, it could take place within a large brin degradation range. We never observed angiogenesis when the brin degradation had been completely inhibited by adding aprotinin to the medium. The aprotinin concentration required to inhibit the process depended on the cell type used. HUVEC stopped to form CLS at a concentration of aprotinin equal to or greater than 10A4 lg/ml, whereas as much as 10 lg/ml of aprotinin was needed to block the process in BREC culture. This inhibitor's requirements may reect the high brinolytic activity of BREC. The plasminogen plasmin system must be in-

volved since 0.1 lM of a2-antiplasmin was sufcient to inhibit HUVEC'S CLS formation, while 10 times that amount had a similar effect on BREC. No brinolytic activities were detected in human sera-supplemented media without cells, nor FDP. Thus the plasminogen/plasmin system was activated when the cell reorganised into CLS. Secondly, a high brinolytic activity hampered the reorganisation of the cells. These levels of brinolytic activity could be decreased by adding aprotinin to the culture medium and/or modifying the brin gel concentrations. BREC seeded on 0.5 mg/ml brin gels degraded the gel within 3 h without forming CLS. However, CLS formation was promoted either by adding 1 lg/ml aprotinin to the BREC's culture medium or by increasing the brin concentration to 8 mg/ml. Neither brin degradation products nor plasmin per se can trigger in vitro CLS formation as shown by the following experiments: rst the cells were incubated under conditions conductive to CLS formation; then a2-antiplasmin was added and we observed that the process was blocked. Finally, we added both a2-antiplasmin and increasing concentrations (from 50 ll to 200 ll) of FDP issued from a digest of brin gel, 1 mg/ml, but CLS formation did not resume. Although the medium of HUVEC cultured on 8 mg/ml brin gel contained plasmin in sufcient quantity to partially degrade the gel, there was no CLS. Remodelling of brin gel itself must trigger CLS formation. Thirdly, the mechanical properties of brin gels depending on their concentrations inuenced the formation of the CLS. In fact, when HUVEC were seeded on 8 mg/ml brin gels, they degraded the matrix as extensively as did BREC, although they were not able to form CLS while BREC did (Table 1). This observation could be related to the role of the mechanical factors resulting from the balance between the viscoelasticity of the brin gels and cell tractions. Thus even partially degraded, an 8 mg/ml brin gel is probably too rigid to allow its invasion by the HUVEC. Conversely, BREC formed CLS when cultured on brin gels ranging from 0.5 mg/ml to 8 mg/ml only if brinolysis was limited either by the addition of aprotinin or by increasing the concentration of the brin clot. Thus, the brinolytic capacity depended both on the brin concentraAngiogenesis Vol 2 No 4 1998/1999 337

B. Vailhe et al.

Figure 3. Kinetics of CLS formation by HUVEC and BREC on dened brin gels. AD illustrate the morphological behaviour of the HUVEC photographed, respectively, 2, 12, 18, and 24 2 h after cell seeding on 0.5 mg/ml brin gels, while EH represent the morphological behaviour of BREC photographed, respectively, 2, 22, 32 and 46 2 h after seeding on 8 mg/ml brin gels. These sequences are representative of at least ve different experiments with similar results. The rst lacunae appeared as early as two hours after seeding (arrows in A and E). Their number and size increased in time, leading to the formation of a tubular structure at the junction of two lacunae (arrow in D and H). The length of the tubes increased from 450 50 lm (C) to 550 50 lm (D) and from 100 50 lm (G) to 400 50 lm (H). Bar, 450 lm.

tion and on the cell type. It is possible that force transfer across the integrins modulate the activity of the brinolytic activity.
Endogenous TGF-b1 is not involved in CLS formation

Since it has been assumed that plasmin may activate endogenous TGF-b1 during angiogenesis,27
338 Angiogenesis Vol 2 No 4 1998/1999

we veried whether this process contributed to the formation of CLS in our model. Figures 5A and 5B show that both cell types secreted latent TGF-b1 during the angiogenic process. The amount of TGFb1 in 1% serumcontaining medium incubated at 37C, in the absence of cells remained constant for 24 h. Similarly, in the absence of CLS formation, neither HUVEC nor BREC secreted TGF-b1 in the medium over time and the level was 1000 pg

The formation of tubular structures by endothelial cells

Figure 4. Accumulation of brin degradation products during the formation of CLS. Culture media either from HUVEC (A) cultured on brin 0.5 mg/ml or from BREC (B) cultured on brin 8 mg/ml were collected at different times. Fibrin degradation products accumulation were then measured by ELISA as described in Materials and Methods. The rst points in curves A and B represent the amount of FDP measured in 1% human serum incubated without cells. Each point represents the mean (SD) of three separate experiments performed in duplicate.

(Figure 5). The level of latent TGF-b1 in culture medium increased with time during the formation of CLS from approximately 1000 pg to 1700 pg per dish. For each cell type, a large increase was observed when the cells reorganized into CLS (between 18 and 24 h for HUVEC, and between 32 and 48 h for BREC; Figure 5). No active TGFb1 was found without prior acid-activation of the culture media. However, there are known instances where the active form of TGF-1 could not be detected in the cell medium because it interacted rapidly.27 To verify this hypothesis, we added an anti-TGF-b1 antibody raised against the active form of TGF-b1 to the culture medium in an attempt to neutralize the putative activity of the cytokine. This antibody added at different concentrations (5 lg/ml, 10 lg/ml or 20 lg/ml) to cell cultures, either at the time of cell seeding or when cells started to reorganize into CLS, did not modify the angiogenic activity of these two cell types. These indicated that either there was no active endogeneous TGF-b1, or active TGF-b1 was not implicated in the formation of CLS. To elaborate on these observations we performed the following experiments. Active TGF-b1 (1 ng/ml) was added to culture media at the time of seeding. As shown in Figure 6, 24 h after seeding untreated cells reorganised into CLS with an average tubular structure length of 550 50 lm, while TGF-b1-treated cells retained their less organized morphology (no

Table 1. Modulation of the formation of CLS on brin gel aprotinin. One ml of endothelial cells (1.5 105 cell/ml) were seeded on brin gels in the presence of aprotinin (0 to 10 lg/ml). FDP in the conditioned culture media were measured by ELISA (as described in `Materials and methods'). The media were collected either at the end of the angiogenesis process (i.e. between 24 2 and 48 2 h after seeding) or 48 h after seeding when angiogenesis did not occur. Results are expressed as the means ( SD) of at least four different experiment performed in duplicate Aprotinin (lg/ml) HUVEC Fibrin gel (mg/dish) 0.5 0.5 0.5 0.5 8 8 F.D.P. (mg/dish) Angiogenesis BREC Fibrin gel (mg.dish) 0.5 0.5 0.5 0.5 8 8 F.D.P. (mg/dish) Angiogenesis

0 10A4 1 10 0 10

0.2 ( 0.1) 0.025 ( 0.01) 0.005 ( 0.002) Non-detected 0.160 ( 0.04) 0.06 ( 0.02)

Yes Yes No No No No

0.47 ( 0.05) 0.37 ( 0.07) 0.01 ( 0.005) 0.002 ( 0.001) 0.184 ( 0.05) 0.016 ( 0.005)

No* No* Yes No Yes No

* In this experiment, the media was collected 3 h after seeding, since BREC completely degraded the brin gel within this time interval.

Angiogenesis Vol 2 No 4 1998/1999

339

B. Vailhe et al.

Discussion
The two models described here reveal that each cell type has its own physiological behaviour in forming CLS in vitro. Our results show that one of the major differences between the two cell types involved in CLS formation was the protein concentration in the matrix. HUVEC form an extended CLS network on a 0.5 mg/ml brin gel, whereas BREC formed CLS on a much higher brin gel concentration (8 mg/ml). Indeed, we observed that when BREC were seeded on brin gels polymerised from brinogen solutions below 8 mg/ml, the cells degraded the brin gels too quickly to adhere and formed an extended CLS network. However, brinolysis was a necessary step before CLS could occur, since high doses of aprotinin (10 lg/ml) hampered this process (Table 1). In each case, brinolysis was always associated with angiogenesis when in vitro angiogenesis took place, extracellular proteolysis always partially occurred, leading to the remodelling of the brin network into lacunae. Schnaper et al.37 found that both serine protease inhibitors and matrix-metalloproteinase inhibitors were required to inhibit endothelial cell reorganization on Matrigel. As Matrigel is constituted of several glycoproteins including laminin A and collagen IV, it is likely that both matrix-metalloproteinase (MMP) and PA pathways are necessary for appropriate ECM remodelling by endothelial cells before CLS take place on this matrix. Indeed, in vivo, collagen basement membrane is specically degraded by MMP. Thus, it is not surprising that aprotinin alone, or a2-antiplasmin was sufcient to inhibit this process in our in vitro model, since brin was the only constituent of the matrix. The anti-FDP antibody we used recognised brinogen as well as all its degradation products. The only means whereby brinogen or its products could be released into the medium was activation of the plasminogen/plasmin system as endothelial cells reorganized themselves into CLS. Incubated sera did not contain any FDP as shown in Figures 4A and 4B. Furthermore we established that brin gels were not degraded by 1% and even 5% serum-containing medium after a 24-h incubation in the absence of cells. Thus our system, as well as those of others5,38,39 favour the

Figure 5. Measurements of latent TGF-b1 accumulation during the formation of CLS. Culture media either from HUVEC (A) cultured on brin 0.5 mg/ml or from BREC (B) cultured on brin 8 mg/ml were collected at different times and TGF-b1 was measured by ELISA as described in methods. To transform latent TGF-b1 secreted in the culture media into immunoreactive TGF-b1 detectable by the Quantikine TGF-b1 assay, we followed the procedure recommended by the manufacturers. The lled squares in (A) and (B) represent TGF-b1 measured in culture dishes when no CLS were formed. The rst experimental point of the curves A and B represent the amount of TGFb1 in 1% human serum incubated without cells, this amount did not increase over time. Each point represents the mean (SD) of three separate experiments performed in duplicate.

tube was formed). The inhibitory effect of exogenous TGF-b1 was accompanied by a 50% decrease in brin degradation. Furthermore, the simultaneous addition of TGF-b1 and a specic antibody against the active form of the TGF-b1 suppressed this inhibitory effect (data not shown). From these experiments we conclude that exogenous active TGF-b1 had an inhibitory effect on the formation of CLS in our model; however, no active TGF-b1 was formed from the latent cytokine secreted by either HUVEC or BREC, despite plasminogen activation by these two types of cells.
340 Angiogenesis Vol 2 No 4 1998/1999

The formation of tubular structures by endothelial cells

Figure 6. Inhibitory effect of TGF-b1 during the formation of CLS. Micrographs show that control HUVEC (A) reorganise into capillary-like structures (length of the tubes is 550 50 lm) while cells treated by 1 ng/ml TGF-b1 (B) retained their less-organised morphology, no tubular structures were observed. Observations were performed 24 2 h after seeding. Bar, 250 lm.

plasminogen/plasmin system as a major factor in angiogenesis on brin gels. It has been shown that FDP, and especially fragment E, is able to activate angiogenesis in vivo.40 However, in our model it is unlikely that FDP played an important role. Indeed, depending on the culture conditions, FDP concentration could vary over 10-fold without affecting CLS formation. Furthermore, HUVEC never formed CLS on 8 mg/ml brin gels, yet FDP accumulated in culture medium. The differences observed between BREC and HUVEC behaviours may be related to their tissue origin,41 i.e. to their specic physiological roles as molecular sieves, and suggest regional specicity in the regulation of the brinolysis. The high brinolytic activity of BREC may well represent a defense mechanism against thrombosis in the highly vascularised retina. Occlusion of retinal capillaries may lead to ischemia, deleterious to the retinal nervous tissue which is highly dependent on adequate oxygenation. It is clear from our results that brinolysis is required to form CLS. However, ECM proteolysis alone is not sufcient to induce cell re-arrangement. In fact, HUVEC seeded on 8 mg/ml brin

gels did not form CLS although signicant brinolysis occurred, probably because the matrix remained too rigid to initiate an angiogenic signal in HUVEC.17 Conversely, BREC form CLS under the same conditions. In this latter case, the brin gel was pleated into dense zones, giving indirect proof of the implication of cell forces. Mechanical forces play a widely recognised role in cellular morphological changes and angiogenesis and result from the balance between cell forces and the mechanical properties of the matrix.14,15 It has previously been shown that the mechanical properties of the brin gels are related to the concentration in the gel.42,43 In this work,17 the number and the size of lacunae increased during in vitro angiogenesis on brin. Using time-lapse video-microscopy, we observed that during the increase of lacunae tractions exerted by the cells on the matrix led to brin disruption.17 These mechanical tensions appear to favour the CLS network formation. The activity of mechanotransducers, i.e. receptor types, number of receptors, and cytoskeleton proteins specic to each cell type, may be responsible for the different behaviours of
Angiogenesis Vol 2 No 4 1998/1999 341

B. Vailhe et al.

HUVEC and BREC. Integrins are known to be involved in mechanotransduction. These transmembrane receptors are connected intracellularly with the cytoskeleton and extracellularly with the matrix and other factors as well.44 The integrin avb3 is expressed on endothelial cells cultured either on brinogen, vitronectin45 or brin46 substrates. Furthermore, it has been shown that uPAR, a component of the brinolytic system, colocalizes with avb347 and interacts with vitronectin.48 This avb3/u-PAR/vitronectin complex appears to increase the stiffness of the cytoskeletal forces.21 We can speculate that avb3/u-PAR/brin ternary complexes form in our model and are implicated in matrix degradation and remodelling through the forces exerted on the gel by the cytoskeleton. Thus the number of receptors implicated in mechanotransduction could be different for HUVEC and BREC. The stiffness developed by the cytoskeleton of BREC may be more important than that developed by HUVEC, permitting BREC to reorganise into CLS on stiffer brin gels. In addition to extracellular matrix degradation, the proteolytic activity of the cells during angiogenesis may activate latent cytokines in their environment, as has been suggested by several authors.2 The role of cytokines in in vitro angiogenesis models has been mostly studied by adding exogenous cytokines; but a few investigations deal with the role of endogenous cytokines. In this study, we observed that cells secreted latent TGFb1 during the in vitro angiogenesis process. It has been postulated that TGF-b1 was activated by proteases responsible for brin degradation, especially by plasmin. However, we could not detect any active TGF-b1 in the extracellular medium while plasmine was present. It may be argued that TGF-b1 was activated in a spacelocalised, tightly regulated manner.26 Yet, the addition of an antibody, specic to the active form of TGF-b1, to the culture medium, in an attempt to neutralise the putative endogenously produced biologically active TGF-b1, did not inuence the formation of CLS. In contrast, added active TGFb1 (1 ng/ml) reduced the number of CLS formed. Taken together, these results suggest that plasmin was unable to activate latent TGF-b1 under our culture conditions, as described by Bizik et al.28
342 Angiogenesis Vol 2 No 4 1998/1999

We observed that the inhibition due to exogenous TGF-b1 was related to a 50% decrease in matrix degradation, compared to the control cultures in which the CLS formed. The inhibition induced by exogenous TGF-b1 could be related to a modication of the balance between activators and inhibitors of the cell proteolytic activities. In fact, partial degradation of the matrix by the cells is necessary for the formation of CLS. These results raise questions on the eventual role of secreted TGF-b1 in our in vitro model. Opposite effects of TGF-b1 on CLS formation have been described49,50 and a biphasic effect has been proposed by Pepper et al.,51 while we have shown, in previous work, that TGF-b1 effects depended on serum concentration in the culture medium.52 It is known that TGF-b1 increases PAI-1 secretion53,54 and stimulates the synthesis of extracellular matrix proteins, such as laminin and bronectin, by endothelial cells.5557 In our model, the secretion of latent TGF-b1 increased after CLS formed, and was maximal at the end of the process. However, we never detected active TGFb1. It is clear that in our systems plasmin did not activate latent TGFb1. In conclusion, our results show that there are differences between BREC and HUVEC with respect to their ability to remodel the matrix via proteolytic activities and mechanical forces, in order to form CLS on brin gels. The role of latent TGF-b1 secreted during the reorganisation of the cells into capillary-like structures remains to be determined.

Acknowledgements
We thank Dr. Alice Barrieux for her comments on the manuscript and Dr. Yves Usson for the confocal microscopy.
References 1. Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nature Med 1995; 1, 27 31. 2. Madri JA, Sankar S, Romanic AM. Angiogenesis. In: Clark RAF (ed), The Molecular and Cellular Biology of Wound Repair, 2nd edn. New York: Plenum Press, 1996; 355371.

The formation of tubular structures by endothelial cells

3. O'Reilly MS, Holmgren L, Chen C, Folkman J. Angiostatin induces and sustains dormancy of human primary tumors in mice. Nature Med 1996; 2, 689692. 4. Brooks PC. Cell adhesion molecules in angiogenesis. Cancer Metastasis Rev 1996; 15, 187194. 5. Pepper MS, Montesano R, Mandriota SJ, et al. Angiogenesis: a paradigm for balanced extracellular proteolysis during cell migration and morphogenesis. Enzyme Protein 1996; 49, 138162. 6. Folkman J, Haudenschild C. Angiogenesis in vitro. Nature 1980; 288, 551556. 7. Montesano R, Vassalli JD, Bird A, et al. Basic broblast growth factor induces angiogenesis in vitro. Proc Natl Acad Sci USA 1986; 83, 72977301. 8. Nicosia RF, Ottinetti A. Growth of microvessels in serum-free matrix culture of rat aorta. Lab Invest 1990; 63, 115122. 9. Vernon RB, Lara SL, Drake CJ, et al. Organized type I collagen inuences endothelial patterns during `spontaneous angiogenesis in vitro', planar cultures as models of vascular development. In vitro Cell Dev Biol 1995; 31, 120131. 10. Chalupowicz G, Chowdhury ZA, Bach TL, et al. Fibrin II induces endothelial cell capillary tube formation. J Cell Biol 1995; 130, 207215. 11. Kubota Y, Kleinman HK, Martin GR, Lawley TJ. Role of laminin and basement membrane in the morphological differentiation of human endothelial cells into capillary-like structures. J Cell Biol 1988; 107, 15891598. 12. Nicosia RF, Bonanno E, Smith M, Yurchenco P. Modulation of angiogenesis in vitro by laminin-entactin complex. Dev Biol 1994; 164, 197206. 13. Davis GE, Camarillo CW. Regulation of endothelial cell morphogenesis by integrins, mechanical forces, and matrix guidance pathways. Exp Cell Res 1995; 216, 113123. 14. Ingber DE, Folkman J. How does extracellular matrix control capillary morphogenesis? Cell 1989; 58, 803805. 15. Ingber DE. Cell shape, cytoskeletal mechanics, and cell cycle control in angiogenesis. J Biomech 1995; 28, 14711484. 16. Vernon RB, Sage EH. Between molecules and morphology. Extracellular matrix and creation of vascular form. Am J Pathol 1995; 147, 873883 17. Vailhe B, Ronot X, Tracqui P, et al. In vitro angiogenesis is modulated by the mechanical properties of brin and is related to avb3 integrin localisation. In Vitro Cell Dev Biol 1997; 33, 763773. 18. Dubois-Stringfellow N, Jonczyck A, Bautch VL. Perturbations in the brinolytic pathway abolish cyst formation but not capillary-like organisation of

19. 20. 21.

22. 23.

24.

25. 26. 27.

28.

29.

30. 31. 32.

cultured murine endothelial cells. Blood 1994; 83, 32063217. Vassalli JD, Sappino AP, Belin D. The plasminogen activitor/plasmin system. J Clin Invest 1991; 88, 10671072. Lijnen HR, Bachmann F, Collen D, et al. Mechanisms of plasminogen activation. J Int Med 1994; 236, 415424. Wang N, Planus E, Pouchelet M, et al. Urokinase receptor mediates mechanical force transfer across the cell surface. Am J Physiol 1995; 268, C1062 C1066. Friesel RE, Maciag T. Molecular mechanisms of angiogenesis: broblast growth factor signal transduction. FASEB J 1995; 9, 919925. Montesano R, Vassali JD, Orci L, Pepper MS. The role of growth factors and extracellular matrix in angiogenesis and epithelial morphogenesis. In Sizonenko PC, Augert ML and Vassali JD (eds), Developmental Endocrinology, Frontiers in Endocrinology. Rome: AresSerono Symposia Publications, 1994; 6, 4366. Mandriota SJ, Menoud PA, Pepper MS. Transforming growth factor b1 down-regulates vascular endothelial growth factor receptor 2/k-1 expression in vascular endothelial cells. J Biol Chem 1996; 271, 1150011505. Sporn MB, Roberts AB. Transforming growth factor-b: recent progress and new challenges. J Cell Biol 1992; 119, 10171021. Lawrence DA. Transforming growth factor-b: a general review. Eur Cytokine Netw 1996; 7, 363374. Lyons RM, Gentry LE, Purchio AF, Moses HL. Mechanism of activation of latent recombinant transforming growth factor b1 by plasmin. J Cell Biol 1990; 110, 136167. Bizik J, Felnerova D, Grofova M, Vaheri A. Active transforming growth factor-b in human melanoma cell lines: no evidence for plasmin-related activation of latent TGF-b. J Cell Biochem 1996; 62, 113122. Pepper MS, Vassalli JD, Orci L, Montesano R. Angiogenesis in vitro, cytokines interactions and balanced extracellular proteolysis. In: Maragoudakis ME (ed), Angiogenesis: Molecular Biology, Clinical Aspects. New-York: Plenum Press, 1994; 149170. Engvall E, Ruoslahti E. Binding of soluble form of broblast surface protein, bronectin, to collagen. Int J Cancer 1977; 20, 15. Keckwick RA, McKay ME, Nance MH, Record BR. The purication of human brinogen. Biochem J 1955; 60, 671683. Marguerie GA, Plow EF, Edgington TS. Human platelets possess an inducible and saturable receptor specic for brinogen. J Biol Chem 1979; 254, 5357 5363.

Angiogenesis Vol 2 No 4 1998/1999

343

B. Vailhe et al.

33. Jaffe EA, Nachman ARL, Becker CG, Minnick CR. Culture of human endothelial cells derived from umbilical veins, identication by morphologic and immunologic criteria. J Clin Invest 1973; 52, 2745 2756. 34. Lecomte M, Paget C, Ruggiero D, et al. Docosahexaenoic acid is a major n-3 polyunsaturated fatty acid in bovine retinal microvessels. J Neurochem 1996; 66, 21602167. 35. Heimer GV, Taylor CED. Improved mountant for immunouorescence preparations. J Clin Pathol 1974; 27, 254256. 36. Osborn M, Weber K. Immunouoresence and immunocytochemical procedures with afnity puried andibodies: tubuline containing structures. Methods Cell Biol 1982; 24, 97132. 37. Schnaper HW, Barnathan ES, Mazar A, et al. Plasminogen activators augment endothelial cell organization in vitro by two distinct pathways. J Cell Physiol 1995; 165, 107118. 38. Pepper MS, Vassalli JD, Montesano R, Orci L. Urokinase-type plasminogen activator is induced in migrating capillary endothelial cells. J Cell Biol 1987; 105, 23532541. 39. Koolwijk P, Van Erck MG, De Vree WJ, et al. Cooperative effect of TNF alpha, bFGF and VEGF on the formation of tubular structures of human microvascular endothelial cells in a brin matrix. Role of urokinase activity. J Cell Biol 1996; 132, 11771188. 40. Thompson WD, Stirk CM, Melvin WT, Smith EB. Plasmin, brin degradation and angiogenesis. Nat Med 1996; 2, 493. 41. Rymaszewski Z, Szymanski PT, Ablanalp W, et al. Human retinal vascular cells differ from umbilical cells in synthetic functions and their response to glucose. Proc Soc Exp Biol Med 1992; 199, 183191. 42. Ferry JD. Structure and Rheology of brin networks. In: Kramer O (ed), Biological and Synthetic Polymer Networks. London: Elsevier, 1988; 4155. 43. Ferrenq I, Tranqui L, Vailhe B, et al. Modelling biological gel contraction by cells: mechanocellular formulation and cell traction force quantication. Acta Biotheoretica 1997; 45, 267293. 44. Ingber DE. Integrins as mechanochemical transducers. Curr Opin Cell Biol 1991; 3, 841848. 45. Cheresh D. Human endothelial cells synthesize and express an Arg-Gly-Asp directed adhesion receptor involved in attachment to brinogen and von Willebrand factor. Proc Nat Acad Sci USA 1987; 84, 64716475. 46. Henke CA. Intra-alveolar brosis. Alpha v beta 3 integrin and chondroitin sulfate proteoglycan me-

47.

48.

49.

50.

51. 52.

53.

54.

55.

56.

57.

diate endothelial cell adhesion, migration, and invasion into the provisional matrix. Chest 1994; 105 (suppl), 121S. Myohanen HT, Stephens RW, Hedman K, et al. Distribution and lateral mobility of urokinase-receptor complex at the cell surface. J Histochem Cytochem 1993; 41, 12911301. Kanse SM, Kost C, Wilhelm OG, et al. The urokinase receptor is a major vitronectin-binding protein on endothelial cells. Exp Cell Res 1996; 224, 344 353. Tada K, Fukunaga T, Wakabayashi Y, et al. Inhibition of tubular morphogenesis in human microvascular endothelial cells by co-culture with chondrocytes and involvement of transforming growth factor b: a model for avascularity in human cartilage. Biochem Biophys Acta 1994; 1201, 135 142. Choi ME, Ballerman BJ. Inhibition of capillary morphogenesis and associated apoptosis by dominant negative mutant transforming growth factor-b receptors. J Biol Chem 1995; 270, 2114421150. Pepper MS, Vassalli JD, Orci L, Montesano R. Biphasic effect of transforming growth factor-b1 on in vitro angiogenesis. Exp Cell Res 1993; 204, 356363. Vailhe B, Tranqui L. The role of transforming growth factor-b1 (TGF-b1) and of vascular endothelial growth factor (VEGF) on the in vitro angiogenesis process. CR Acad Sci Paris, Life Sciences (serie III) 1996; 319, 100310. Pepper MS, Belin D, Montesano R, et al. Transforming growth factor-beta1 modulates basic broblast growth factor-induced proteolytic and angiogenic properties of endothelial cells in vitro. J Cell Biol 1990; 111, 743755. Pepper MS, Montesano R, Orci L, Vassalli JD. Plasminogen activator inhibitor-1 is induced in microvascular endothelial cells by a chondrocytederived transforming growth factor-beta. Biochem Biophys Res Commun 1991; 176, 633638. Merwin JR, Anderson JM, Kocher O, et al. Transforming growth factor beta 1 modulates extracellular matrix organization and cell-cell junctional complex formation during in vitro angiogenesis. J Cell Physiol 1990; 142, 117128. Madri JA, Pratt BM, Tucker AM. Phenotypic modulation of endothelial cells by transforming growth factor-b depends upon the composition and organization of the extracellular matrix. J Cell Biol 1988; 106, 13751384. RayChaudhury A, D'Amore PA. Endothelial cell regulation by transforming growth factor-beta. J Cell Biochem 1991; 47, 224229.

344

Angiogenesis Vol 2 No 4 1998/1999