Fluid shear stress modulates endothelial cell invasion into three-dimensional collagen matrices

Abstract

Endothelial cells are subjected to biochemical and mechanical stimuli, which regulate their angiogenic potential. We determined the synergistic effects of sphingosine-1-phosphate (S1P) and fluid wall shear stress (WSS) on a previously established model of human umbilical vein endothelial cell invasion into three-dimensional collagen matrices. Collagen matrices were incorporated into a parallel-plate flow chamber to apply controlled WSS to the surface of endothelial monolayers over a period of 24 h. Cell invasion required the presence of S1P, with the effects of S1P being enhanced by shear stress to an extent comparable with S1P combined with angiogenic growth factor stimulation. The number of invading cells depended on the magnitude of shear stress, with a maximal induction at a shear stress of ∼5 dyn/cm², whereas the invasion distance was proportional to the magnitude of shear stress. The enhancement of invasion by 5.3 dyn/cm² shear stress coincided with elevated phosphorylation of Akt and matrix metalloproteinase (MMP)-2 activation. Furthermore, invasion induced by the combined application of WSS and S1P was attenuated by inhibitors of MMPs (GM6001) and the phosphatidylinositol 3-kinase/Akt signaling pathway (wortmannin). These results provide evidence that shear stress is a positive modulator of S1P-induced endothelial cell invasion into collagen matrices through enhanced Akt and MMP-2 activation.

angiogenesis, the development of new blood vessels from preexisting vessels, is a critical step in physiological and pathological events such as wound healing and tumor vascularization (18). Sprouting angiogenesis in vivo involves endothelial cell (EC) degradation of the basement membrane, proliferation, and migration toward angiogenic stimuli. These events are coordinated with eventual formation of a lumen within the endothelial sprout and the joining of sprouts to form a capillary bed (4, 19). Importantly, both biochemical and mechanical forces influence these newly developing structures.

Several biochemical factors are recognized to enhance angiogenesis. Basic fibroblast growth factor (FGF-2) and vascular endothelial growth factor (VEGF) are the best characterized inducers. In vitro, FGF-2 and VEGF induce EC proliferation, matrix proteolytic activity, invasion into three-dimensional (3-D) collagen matrices, and formation of tubular structures (12, 20). More recently, sphingosine-1-phosphate (S1P) has been identified as a potent proangiogenic factor (25). S1P can act as an intracellular signaling molecule and is also deposited by activated platelets during wound healing (5, 24). Exogenous S1P administration or endogenous S1P production by sphingosine kinase overexpression promotes postischemic angiogenesis and blood flow recovery in mouse ischemic hind limb models (49). Bayless and Davis (2) previously reported that S1P-induced EC invasion and lumen formation required integrin cell surface receptors and membrane-associated metalloproteinase activity. Membrane-type matrix metalloproteinases (MMPs), a family of transmembrane zinc and calcium-dependent proteases, degrade extracellular matrix proteins and are integral for sprouting angiogenesis (7, 50, 52). Multiple angiogenic events are mediated by the serine/threonine kinase Akt, including EC migration, survival, and tube formation (54). Collectively, biochemical signals, including VEGF, FGF-2, and S1P coordinate to promote EC outgrowth by driving intracellular signaling events during angiogenesis.

Although the roles of biochemical inducers of angiogenesis have received much attention, the signals downstream of hemodynamic forces that regulate new blood vessel growth are less well understood. Blood flow has been implicated in the growth of blood vessels for nearly a century. In the tails of frog larvae, Clark (8) observed increased capillary growth by sprouting in vessels with high blood flow, whereas capillaries regressed when flow ceased. In later studies using a rabbit ear chamber model, Clark et al. (9) observed that capillaries with faster flow developed into venules. Using the same model, Ichioka et al. (29) increased vascular growth by long-term administration of the vasodilator prazosin, which caused elevated wall shear stress (WSS) lasting up to 13 days postoperatively. With the use of in vitro models of EC morphogenesis, preexposure to WSS enhanced the development of cord-like networks in a two-dimensional (2-D) Matrigel (10) and FGF-2 release, which increased EC sprouting in fibrin matrices (21). Using a 3-D matrix model, Ueda and colleagues (56) demonstrated that WSS, in the presence of serum and FGF-2, increased the formation of EC networks within collagen matrices compared with static culture. Although these data indicate that flow is a proangiogenic stimulus, the mechanism remains unclear. Identifying the salient aspects of flow-induced angiogenesis from published results is difficult due to the variety of techniques used and the lack of definition of the perfusate composition. To elucidate the mechanism of flow-induced EC invasion into 3-D collagen matrices, we have developed a system in which both WSS and media composition were carefully controlled. Using this system, we evaluated the ability of S1P, growth factors, and WSS to promote invasion of ECs into 3-D collagen matrices. We also explored the mechanism by which WSS regulated invasion, namely through WSS-induced Akt phosphorylation and MMP-2 activation.

MATERIALS AND METHODS

Cell culture.

Unless otherwise indicated, all reagents were obtained from Sigma (St. Louis, MO). Human umbilical vein ECs were purchased from Lonza BioProducts (San Diego, CA) and were used in all experiments at passages 4–6. ECs were cultured in medium 199 (M199; Invitrogen) supplemented with 20% FBS (Lonza BioProducts), 50 μg/ml ascorbic acid, 0.1 mg/ml heparin, 50 mg/ml gentamicin (Invitrogen), and antibiotic-antimycotic (Invitrogen) and lyophilized bovine hypothalamic extract purified as previously described (39) at 37°C in a humidified 5% CO₂ incubator. Cells were passaged once weekly on flasks coated with 1 mg/ml porcine gelatin in PBS. Only confluent flasks were used in experiments.

Collagen matrix invasion assay.

Rat tail collagen type I was purified as described (3) and used to prepare collagen matrices at 3.75 mg/ml as previously described (2, 13). D-erythro-S1P (Avanti Polar Lipids) was added to the collagen matrices (unless otherwise noted) to a final concentration of 1 μM and thoroughly mixed. Circular wells (7 mm diameter) were formed by attaching a silicone rubber gasket (Specialty Manufacturing), of 1-mm thickness and perforated with eight holes, onto glass microscope slides (Fig. 1). The collagen matrices (45 μl) were added to the wells and allowed to polymerize for 30 min at 37°C and 5% CO₂.

Fig. 1.

Apparatus to apply controlled wall shear stress (WSS) to human umbilical vein endothelial cells (ECs). A: ECs were cultured on 8 (4 shown) circular collagen gels that were polymerized in the bottom plate of a parallel-plate flow chamber. Medium 199 media was perfused through the chamber to subject the cell monolayers to steady, uniform WSS. B: multiple flow chambers (2 shown) were connected in series into a pressure-driven flow circuit in which media pH was controlled with humidified 5% CO₂-95% air (dashed lines).

ECs were fed 24 h before seeding cells onto collagen matrices. The cells were rinsed with 14 ml PBS and removed with 2 ml of 0.25% trypsin-EDTA (Invitrogen) at 37°C for 1 min. The trypsin was neutralized with 2 ml FBS. Cells were counted with a hemocytometer, washed once in 10 ml serum-free M199, and plated as a confluent monolayer at a density of 1.2 × 10⁵ cells/cm² in M199 containing 5% FBS and 50 μg/ml ascorbic acid. The cells were allowed to attach for 60 min and then washed twice with serum-free M199 before applying the defined media and WSS for the experiment.

Shear stress experiments.

The plates containing ECs seeded on collagen matrices were assembled into parallel-plate flow chambers designed to apply uniform steady WSS to the cell monolayer (Fig. 1A). The WSS magnitude was calculated as τ = μQ/wh², where τ is WSS, μ is fluid viscosity (0.7 cP), Q is flow rate, w is the width of the flow channel (29.21 mm), and h is the height of flow channel. The flow of the culture medium was provided by a sterile continuous-flow loop, with the flow rate controlled with a pulse-free gear pump (Ismatec) and monitored with an ultrasonic tubing flow sensor (Transonic Systems). The perfusion medium consisted of M199 containing reduced serum II (RSII) and 50 μg/ml ascorbic acid. RSII consisted of 2 mg/ml bovine serum albumin (BSA), 20 ng/ml human holo-transferrin, 20 ng/ml insulin, 17.1 ng/ml sodium oleate, and 0.02 ng/ml sodium selenite. Media pH was maintained at 7.4 by perfusing 5% CO₂-95% air first through a sparger containing sterilized water and then through the head gas of the media reservoir. Bubbles were removed from the media using a bubble trap. The entire system was enclosed by an acrylic box in which temperature was maintained at 37°C using a heat gun with feedback temperature control (Omega Engineering, Stamford, CT).

Imaging and analysis.

Following each experiment, collagen matrices containing invading cells were washed briefly in PBS, fixed in 3% glutaraldehyde in PBS for 2 h, stained with 0.1% toluidine blue in 30% methanol for 12 min, and washed with deionized water to clearly identify the invading cells. Cross sections were prepared using a razor blade and imaged using an Olympus CKX41 inverted microscope equipped with an Olympus Q-Color 3 camera. From the digital images of the cross sections, the invasion distance was measured for individual sprouts as the distance from the bottom of the cell monolayer and the point of deepest penetration into the matrix. Pixel values from invasion distance measurements were converted to microns.

To quantify invasion density, en face images were observed using bright-field illumination with a 10× objective on an Olympus BH-2 upright microscope. The microscope was focused on the invading cells, which were located immediately below the EC monolayer. Each data point represents a field in the center of a well, where the number of invading cells was counted manually using an eyepiece equipped with an ocular grid covering an area of 1 mm². A single measurement was recorded for each well.

Gelatin zymography.

To subject ECs to WSS magnitudes of 0.12, 5.3, and 12 dyn/cm², three flow systems were run simultaneously. Each flow system contained two identical parallel-plate flow chambers (i.e., identical channel dimensions) arranged in series. Conditioned media was sampled from the reservoir at the time points indicated, centrifuged at 500 g for 5 min, and frozen at −20°C. Aliquots of conditioned medium were concentrated ∼10-fold using a Centricon centrifugal filter unit containing an Ultracel YM-10 membrane (Millipore) at 5,000 rpm in a Beckman J2-21M centrifuged at 4°C. The concentrated media (30 μl) were prepared under nonreducing conditions and loaded in 8.5% acrylamide gels containing a final concentration of 1 mg/ml porcine gelatin. Following electrophoresis, the matrices were rinsed three times in 100 ml of 2% Triton X-100 in water for 1 h and rinsed twice in distilled water before being placed in 25 mM Tris·HCl (pH 7.5) containing 5 mM CaCl₂ overnight. The matrices were stained with 0.1% Amido Black in 30% methanol and 10% acetic acid for 15 min at room temperature and destained in 30% methanol and 10% acetic acid before image analysis as described (2).

Western blots.

To subject ECs to WSS magnitudes of 0.12, 5.3, and 12 dyn/cm², three flow systems were run simultaneously with each flow system containing two identical parallel-plate flow chambers arranged in series; individual chambers were removed at various times. The collagen matrices containing invading ECs were washed briefly with PBS before lysing the cells with heated 1.5× Laemmli sample buffer containing 15% glycerol, 3.45% SDS, 371 mM Tris (pH 6.8), and 0.0013% Bromphenol Blue. The lysed cells were boiled for 10 min, incubated on ice for 5 min, and frozen at −20°C. Proteins were loaded onto 10% SDS-polyacrylamide gels and transferred to polyvinylidene fluoride membranes (Fisher Scientific). After blocking in 5% milk at room temperature for 1 h, the membranes were incubated with monoclonal rabbit anti-phospho-Akt (Ser473) primary antibody (1:1,000; Cell Signaling) at 4°C overnight in Tris-Tween 20 saline containing 5% BSA to detect phosphorylated Akt. Membranes were warmed to room temperature for 30 min and washed three times in Tris-Tween 20 saline before incubation with goat anti-rabbit secondary antibody (1:4,000; Dako) in Tris-Tween 20 saline containing 5% milk for 1 h. The same Western blotting procedure was used with anti-Akt primary antibody (1:4,000; Cell Signaling) to verify equal loading between samples or polyclonal antisera to cleaved caspase-3 fragments (Asp 175) primary antibody (1:1,000; Cell Signaling) to quantify apoptosis.

Image analysis and quantification.

Images of the Akt blots were scanned into a computer with a desktop scanner, whereas polyacrylamide gels (MMP-2) were scanned with a FluorChem 8900 digital imaging system (Alpha Innotech, San Leandro, CA). The band intensities were measured using ImageJ image analysis software (National Institutes of Health). Akt phosphorylation was expressed as the ratio of phosphorylated Akt to total Akt protein. Akt phosphorylation was normalized relative to the level of the static control at 0 h. The band intensities for active MMP-2 and MMP-2 pro-peptide (pro-MMP-2) were expressed relative to their maximal values.

Inhibition experiments.

Wortmannin and GM6001 (EMD Biosciences) were dissolved in DMSO at 5 and 2.5 mM concentrations, respectively. ECs were preincubated with wortmannin (10 nM) or GM-6001 (500 nM) for 60 min during attachment to collagen matrices before shear stress exposure, and this was maintained throughout the application of WSS for 24 h.

Proliferation assay.

To subject ECs to WSS magnitudes of 5.3 and 12 dyn/cm², a flow system was run with six identical parallel-plate flow chambers arranged in series; individual chambers were removed after 1, 12, and 24 h. Collagen matrices containing invading cells were fixed in 4% paraformaldehyde and immunostained with polyclonal rabbit anti-Ki-67 primary antibody and FITC-conjugated (goat anti-rabbit) secondary antibody. The collagen matrices were then mounted with 4′,6-diamidino-2-phenylindole (DAPI) mounting medium to stain DNA for identification of cell nuclei. To quantify proliferation density, en face images were observed at 20× focusing on the FITC signal in cell nuclei of cell in the monolayer. Each data point represents a field in the center of a well, where the number of proliferation cells was counted manually. The number of nuclei that stained positively and negatively for Ki-67 was quantified.

Statistical analysis.

Data are presented as the means and SD for each group of samples. Statistical analysis was performed using SAS software (Cary, NC). Comparisons between two groups were performed using Student's t-tests. Comparisons between three or more groups were performed by one-way ANOVA followed by post hoc pairwise comparison testing using Tukey's method. Two-way ANOVA was performed to determine the effects of time and WSS magnitude on MMP-2 activation and pro-MMP-2 expression.

RESULTS

WSS enhanced S1P-induced cell invasion.

We have previously shown that S1P synergizes with VEGF and FGF-2 to potently induce EC invasion in 3-D collagen matrices (2). VEGF and WSS have been reported to increase the expression of the S1P receptor, S1P₁, and augment S1P-stimulated EC migration in 2-D culture (28, 31); thus we hypothesized that WSS would enhance S1P-induced cell invasion. To test this hypothesis, ECs cultured on collagen matrices containing or lacking S1P (1 μM) were treated with either growth factors (10 ng/ml each of VEGF and FGF-2), WSS (5.3 dyn/cm²), or no treatment for 24 h (Fig. 2). No invasion was observed in control collagen matrices (Fig. 2A), and this was not changed by the application of either growth factors (Fig. 2B), S1P (Fig. 2C), or WSS (Fig. 2D). The simultaneous application of growth factors and WSS induced a slight but nonsignificant amount of cell invasion (Fig. 2E). Simultaneous application of S1P and growth factors (Fig. 2F) or S1P and WSS (Fig. 2G) induced significant invasion. Finally, application of growth factors, S1P, and WSS all together resulted in the highest level of invasion (Fig. 2H), although this was not significantly greater than the combined application of S1P and WSS. These results indicate that S1P synergized with growth factors and WSS to promote EC invasion and that no individual stimulus (i.e., S1P, growth factors, or WSS) alone was capable of promoting a response.

Fig. 2.

Sphingosine-1-phosphate (S1P) synergizes with WSS and growth factors to induce cell invasion. Representative cross-sectional micrographs are shown for ECs left untreated (A) and treated with 10 ng/ml each VEGF + bFGF (GFs; B), S1P (1 μM; C), 5.3 dyn/cm² WSS (D), VEGF + bFGF + WSS (E), S1P + GFs (F), S1P + WSS (G), and GFs + S1P + WSS (H). The cell monolayer is located at the top of each figure. Invasion density (I) for each condition was quantified (means ± SD; n = 4 fields) as described in materials and methods from samples observed en face. Conditions that were not significantly different (P < 0.05) as tested by ANOVA and multiple comparison testing are grouped in brackets. Scale bar, 100 μm. bFGF, basic FGF.

Of potential concern with the WSS studies are the release of S1P into the perfusate and the generation of other proangiogenic soluble factors by the ECs subjected to WSS. With the assumption that all the S1P is released from the collagen matrices into the perfusate, the maximal concentration of S1P in the perfusate would be 0.16 μM, which is higher than the dissociation constant (K_d) for S1P to its receptors (10 to 30 nM) (46). WSS has also been shown to induce the expression of FGF-2 and VEGF (21, 55). Applying freshly collected conditioned media from cultures under the combined S1P and WSS condition did not upregulate invasion after 24 h in separate static cultures of ECs on S1P-incorporated gels (data not shown). Together, these results indicate that the effects of WSS on invasion were not due to the generation of proangiogenic soluble factors and that S1P incorporation within the matrix is required to promote directed invasion of ECs into collagen matrices. For all subsequent flow experiments, S1P was incorporated into collagen matrices and growth factors were not added to the perfusion media.

WSS enhanced EC invasion in a time- and magnitude-dependent manner.

The effects of WSS magnitude on EC invasion were next characterized by quantifying invading cell depth and invading cell density. Both invasion distance and density increased with time in ECs subjected to 5.3 dyn/cm² WSS (Fig. 3). To compare the effects of WSS magnitude in the same experiment, multiple chambers were connected in series and subjected to WSS for 24 h. Since flow rate was the same in each chamber, the WSS within individual chambers was modified by varying chamber height to subject the cells to WSS ranging from 0.12 to 12 dyn/cm². The invasion distance increased monotonically with increasing WSS (Fig. 4, A “–C” and D), with cells invading approximately twice the distance at 12 dyn/cm² than at 0.12 dyn/cm². The density of cell invasion showed a biphasic response, however, with invasion density initially increasing with WSS up to a maximum value of 64 ± 23 cells/mm² at 5.3 dyn/cm² but then decreasing to 15 ± 5 cells/mm² at 12 dyn/cm² (Fig. 4, A′–C′ and E).

Fig. 3.

WSS-induced cell invasion progresses steadily over time. ECs were subjected to 5.3 dyn/cm² WSS for 0, 10, 20, and 24 h in the presence of S1P, fixed, and stained with toloudine blue for morphometric analysis. The invasion distance (A; n = 50 sprouts) and density (B; n = 15 fields) are plotted as a function of time (means ± SD). *Significant difference from 0 h (P < 0.05).

Fig. 4.

Invasion density and distance are dependent on the magnitude of WSS. A-E: ECs were subjected to steady WSS ranging from 0.12 to 12 dyn/cm² for 24 h in the presence of S1P, fixed, and stained with toloudine blue for morphometric analysis. Representative cross-sectional micrographs are shown of en face preparations focusing on the EC monolayer (A–C) and immediately below the monolayer (A′–C′) and of cross sections (A‴–C‴) for ECs subjected to WSS magnitudes of 0.12 (A), 5.3 (B), and 12 (C) dyn/cm². The invasion distance (D; n = 150 sprouts) and density (E; n = 20 fields) of invading ECs are plotted as a function of WSS magnitude (means ± SD). F: collagen matrices containing S1P were subjected to different WSS magnitudes for 48 h before seeding and measuring the extent of cell invasion after 17 h under static conditions with 10 ng/ml VEGF and FGF-2 in the media. *Significant difference from value at 0.12 dyn/cm² WSS (P < 0.05). Scale bar, 100 μm.

To test whether WSS influenced invasion through direct effects on collagen matrices, we applied 0, 0.12, 5.3, and 12 dyn/cm² WSS to collagen matrices containing S1P for 48 h and then seeded ECs on the presheared matrices in media containing VEGF and FGF-2 to quantify invasion. The invasion density was unchanged by preconditioning the matrices with WSS regardless of the WSS magnitude (Fig. 4F), indicating that the matrices were not altered by perfusion in a manner that affected cell invasiveness.

Low invasion responses observed with 12 dyn/cm² WSS could potentially be explained by alterations in cell proliferation or cell viability. Low WSS has been associated with increasing EC turnover through elevated rates of proliferation and apoptosis (11), and high WSS can reduce EC proliferation in a dose-dependent manner (37). We examined the expression of Ki-67 [a protein expressed in cell nuclei of dividing, but not resting, cells (53)] in ECs subjected to different magnitudes of WSS for 24 h. An average of 23.1 ± 6.0% of ECs subjected to 0.12 dyn/cm² WSS stained positively for Ki-67 compared with 13.0 ± 6.8% (at 5.3 dyn/cm²) and 14.9 ± 1.2% (at 12 dyn/cm²). These data are consistent with reports that high WSS (>1.5 dyn/cm²) suppresses DNA synthesis (14, 38); however, there was no significant difference in the levels of Ki-67 staining between samples subjected to 5.3 and 12 dyn/cm² WSS. Based on these data, proliferation rates cannot explain reduced invasion observed with 12 dyn/cm² WSS compared with 5.3 dyn/cm² WSS. An alternate possibility is that apoptosis rates are altered between treatment groups. However, we did not detect any significant levels of cleaved caspase-3 in response to any of the WSS treatments (data not shown). These results indicate that the effects of WSS magnitude on EC invasion do not occur through changes in the rates of cell proliferation or apoptosis but rather occur through biochemical changes as a result of different WSS magnitudes, which we investigate below.

Akt activation.

A role for Akt has previously been demonstrated during EC sprouting in vitro and angiogenesis in vivo (5a). Both S1P and WSS have been reported to induce dose-dependent Akt phosphorylation (16, 30). To determine whether S1P or WSS stimulated Akt phosphorylation in ECs seeded onto 3-D collagen matrices, cultures were exposed to static conditions or 5.3 dyn/cm² WSS on collagen matrices in the absence or presence of S1P (Fig. 5A). In the absence of S1P (Fig. 5A, left), WSS did not induce a noticeable increase in Akt phosphorylation compared with static controls (0 dyn/cm²). There was a significant increase in Akt phosphorylation, however, when ECs were seeded on collagen matrices containing S1P for 1 h that subsided after 8 h (Fig. 5A, right). The level of Akt phosphorylation at the 1-h time point was noticeably enhanced by the addition of 5.3 dyn/cm² WSS for 1 h. We next determined whether the WSS-induced Akt phosphorylation varied depending on the magnitude of WSS applied to ECs on S1P-containing collagen matrices using Western blot and densiometric analyses (Fig. 5, B and C, respectively). Consistent with the results in Fig. 5A, 5.3 dyn/cm² WSS increased Akt phosphorylation relative to the static control. This robust increase was not observed with 0.12 and 12 dyn/cm² WSS. All together, these results suggest that WSS-induced invasion may involve Akt phosphorylation and that the increase in Akt phosphorylation was maximal at an intermediate WSS magnitude of ∼5 dyn/cm², which also induced the maximal density of invasion shown in Fig. 4. To demonstrate a functional requirement for phosphorylated Akt in invasion responses, ECs were treated with the specific PI3K inhibitor wortmannin (Fig. 6). Although Akt was transiently phosphorylated in ECs sheared in the presence of the vehicle (control), the increase in Akt phosphorylation after the application of WSS was completely blocked in the presence of 10 nM wortmannin (Fig. 6A). Furthermore, wortmannin treatment significantly attenuated EC invasion, as demonstrated by the photographs shown in Fig. 6, B and C. Quantification of these responses demonstrated that invasion distance (Fig. 6D) and density (Fig. 6E) were significantly reduced by wortmannin treatment in response to 5.3 dyn/cm² WSS. These results reaffirm that Akt phosphorylation is involved in mediating S1P and WSS-induced invasion responses.

Fig. 5.

S1P-induced Akt activation is enhanced by 5.3 dyn/cm² WSS. A: ECs were subjected to WSS magnitudes of 0 (static control) and 5.3 dyn/cm² for 0, 1, and 8 h, in the absence or presence of S1P. The cells were then lysed, and proteins were then collected as described in materials and methods. Western blot analysis was performed with an antibody directed against the phosphorylated form (p) of Akt (top blot) and an antibody directed against all forms of Akt (bottom blot). Each blot is representative of 3 independent experiments. B and C: ECs were subjected to steady WSS magnitudes of 0 (static control), 0.12, 5.3, or 12 dyn/cm² for the indicated durations in the presence of S1P. Western blots for the phosphorylated form of Akt (top blot) and total Akt (bottom blot) are representatively shown (B) and quantified by densitometric analysis from 3 independent experiments (C; means ± SD). *Significant difference from static control (P < 0.05). Fold Exp, fold experiment.

Fig. 6.

Wortmannin inhibits the effects of WSS on Akt phosphorylation and attenuates cell invasion. ECs were subjected to 5.3 dyn/cm² WSS for 0, 1, 8, and 24 h in the presence of S1P and either DMSO vehicle or 10 nM wortmannin. Western blot analysis was performed with an antibody directed against the phosphorylated form of Akt (A; top blot) and an antibody directed against all forms of Akt (A; bottom blot). Representative cross-sectional micrographs are shown for ECs that were subjected to 5.3 dyn/cm² WSS for 24 h in the presence of S1P and either DMSO vehicle (B) or 10 nM wortmannin (C). The invasion distance (D; n = 150 sprouts) and density (E; n = 12 fields) of invading ECs are plotted as a function of WSS magnitude (means ± SD). *Significant difference between groups (t-test; P < 0.05). Scale bar, 100 μm.

MMP-2 activation.

Because MMP-2 activation has been previously correlated with S1P and growth factor-induced EC invasion (2), we determined the effects of WSS magnitude on MMP-2 expression and activation. ECs were subjected to 0.12, 5.3, and 12 dyn/cm² WSS in the presence of S1P before performing gelatin zymography using conditioned media collected from the flow system at time points indicated (Fig. 7A). MMP-2 levels increased with time in cultures subjected to 0.12 and 5.3 dyn/cm² WSS, whereas high WSS (12 dyn/cm²) resulted in suppression of pro- and activated forms of MMP-2. Quantification and two-way ANOVA of the zymograms (Fig. 7, B and C) indicated that the pro-MMP-2 band intensities were statistically lowest in the perfusate from the cells subjected to 12 dyn/cm² WSS, whereas the activated MMP-2 band intensities were statistically highest in perfusate from cells subjected to 5.3 dyn/cm² WSS. These results suggested that MMP activity is important for EC invasion in response to WSS + S1P and that MMP-2 expression and activation are maximal at 5.3 dyn/cm² WSS, whereas MMP-2 expression and activation is largely suppressed by 12 dyn/cm² WSS.

Fig. 7.

Matrix metalloproteinase (MMP)-2 activation is maximal in response to 5.3 dyn/cm² WSS. ECs were subjected to steady WSS ranging from 0.12 to 12 dyn/cm² in the presence of S1P, and samples of the conditioned media were collected at the indicated time points. The concentrations of MMP-2 pro-peptide (pro-MMP-2) and active MMP-2 are representatively shown by gel zymography (A) and quantified by densitometric analysis from 2 independent experiments (B and C, respectively; means ± SD). The band intensities were normalized relative to the highest intensity for each form of MMP-2. *Two-way ANOVA indicates a significant difference from the other WSS magnitudes (P < 0.05).

To confirm that metalloproteinase activity mediates invasion induced by WSS and S1P, we tested the effect of the hydroxamate-based metalloproteinase inhibitor GM6001 along with vehicle (control) on the invasion response of ECs subjected to 5.3 dyn/cm² WSS. Photographs illustrate that GM6001 treatment (500 nM) completely blocked EC invasion (Fig. 8, A vs. B), which is quantified in Fig. 8C. Gelatin zymography confirmed complete blockade of MMP-2 activation by GM6001 (data not shown). These results confirm a requirement for MMP activity in mediating S1P and WSS-induced invasion responses.

Fig. 8.

Invasion in response to the combined applications of S1P and WSS is blocked by MMP inhibition. Representative cross-sectional micrographs are shown for ECs that were subjected to 5.3 dyn/cm² WSS for 24 h in the presence of S1P and either DMSO vehicle (A) or 500 nM GM6001 (B). The invasion density (C; n = 24 fields) of invading ECs was quantified for each condition (means ± SD). *Significant difference from vehicle control (t-test; P < 0.05). Scale bar, 100 μm.

DISCUSSION

Normally, ECs are maintained in a quiescent monolayer. However, under certain circumstances, ECs can be stimulated to escape from the monolayer and invade the underlying extracellular matrix to produce new vessels. In wound healing, for example, platelets from damaged blood vessels accumulate into the wound (40) where they can create a provisional matrix and release their abundant stores of proangiogenic factors including S1P, VEGF, and FGF-2 (41, 44, 59). The high concentration of these factors in the wound stimulates EC invasion from nearby preexisting vessels and provides a directional cue for their migration toward the provisional matrix. We speculate that the WSS experienced by ECs in these neighboring vessels will modulate the extent of invasion and subsequent formation of new vessels. Specifically, our results support the notion that increasing WSS will enhance the rate of sprout growth; however, the density of invasion will be diminished if the WSS magnitude is too high. Thus WSS may provide a negative feedback on the recruitment of new sprouts under conditions of high tissue perfusion (i.e., high WSS). In contrast, when tissue perfusion is inadequate (i.e., low to intermediate WSS), the enhanced recruitment of new vessels provides a mechanism to improve blood perfusion to wounded tissue.

In our experimental model, invasion requires the presence of S1P in the collagen matrix. We confirmed our previous result that FGF-2 and VEGF synergize with S1P to stimulate EC invasion into collagen matrices (2). Importantly, we demonstrated that WSS is also a proangiogenic stimulus that can also enhance the effects of S1P on EC invasion. This is supported by the observation that WSS alone did not induce invasion, but WSS did enhance invasion in the presence of S1P to an extent comparable with angiogenic growth factor stimulation. Furthermore, WSS alone did not induce Akt phosphorylation, but WSS did increase Akt phosphorylation when applied in the presence of S1P. Synergism between S1P, VEGF, and WSS has been observed in EC migration and proliferation in 2-D cultures. Specifically, Hughes et al. (28) showed that WSS and VEGF independently enhanced S1P-induced EC migration into a wound, and these effects were not further increased by combining WSS and VEGF. In their study, Hughes et al. (28) found that WSS or VEGF alone induced a greater degree of wound healing than S1P alone. In the present study, EC invasion was not observed when either VEGF/FGF-2 or WSS alone was applied. Our model requires a directional cue for invasion into the matrix, which is provided by S1P incorporation in the collagen matrix. The VEGF and FGF-2 were added to the media and hence did not provide a concentration gradient directing cell migration into the matrix. Immobilization of VEGF and FGF-2 to collagen matrices via covalent binding has been shown to promote cell invasion (53A). It remains to be determined whether WSS enhances invasion of ECs into matrices containing immobilized growth factors in the absence of S1P.

Our results show similarities to, as well as important differences from, previous studies that explored the role of WSS on angiogenesis responses in vitro. Cullen and colleagues (10) reported that subjecting bovine aortic ECs to WSS caused Transwell migration and tubule formation to increase monotonically with increasing WSS ranging from 0 to 20 dyn/cm². In these studies, Cullen et al. (10) applied WSS to bovine aortic ECs before placing them onto Transwell chambers to assay migration or onto 2-D Matrigels to measure tubule formation. In our experiment, where WSS was applied during the 3-D invasion assay, invasion distance increased monotonically with increasing WSS; however, EC invasion density shows a bimodal dependence on WSS magnitude. It is possible that the inhibitory effect of high WSS observed in our study only occurs during the application of WSS and hence would not occur when ECs are presheared and then angiogenic assays are later performed under static conditions. Furthermore, the assays used to characterize angiogenic responses in their study are significantly different from those used in the present study. For example, migration through a Transwell filter would not require MMP activity, whereas migration through a collagen matrix does. Gloe and colleagues (21) presheared porcine aortic ECs with 16 dyn/cm² WSS for 6 h and demonstrated the formation of capillary-like structures after 24 h of static culture, which they attributed to shear-induced FGF-2 expression and secretion into the culture media. We were unable to stimulate invasion of ECs in a static system using conditioned media from WSS-treated invasion cultures, indicating the enhancement of invasion caused by WSS was not due to the generation of soluble proangiogenic factors. It is worth noting that these two studies (10, 21) were performed using media containing 10–20% FBS or FCS. Given that FBS and FCS contain 141–180 nM S1P, and that the K_d for S1P to its receptors is ∼20 nM (46), the possibility exists that S1P receptor activation may have contributed to the angiogenic events observed in these studies.

The 3-D model utilized in these studies provides a realistic environment to study angiogenic sprouting events and observe the complex interactions between invading cells and their surrounding matrix proteins. Importantly, these studies are carried out in collagen type I, the most abundant extracellular matrix protein encountered by ECs during sprout formation. Ueda et al. (56) developed the first perfusion chamber containing a 3-D collagen matrix, allowing simultaneous application of WSS and observation of angiogenic events. They demonstrated that the extent of EC network formation from invading cells is enhanced in the presence of 3 dyn/cm² WSS (56), which is within the range of WSS for maximal invasion in our study (3 to 5 dyn/cm²). The networks generated in their study did not extend more than 50 μm beneath the monolayer but rather extended parallel to the matrix surface. In our system, the ECs did not form shallow networks but rather migrated predominantly orthogonal to the monolayer surface to depths >100 μm. In this study, S1P was incorporated in the collagen matrix and resulted in the ECs migrating orthogonal to the monolayer in a manner that more closely represents the morphology of sprouting ECs in vivo.

The correlation between invasion density and WSS magnitude (Fig. 4E) and the attenuation of invasion by wortmannin (Fig. 6) indicate that Akt activation contributes to EC invasion induced by WSS in the presence of S1P. Akt plays a major role in EC invasion and sprouting, as demonstrated by a 40% reduction in the average length of sprouts from Akt1 knockout mice compared with wild-type mice in an ex vivo aortic ring invasion assay (5a). The effects of VEGF and S1P on EC migration and tubulogenesis in vitro have also been shown in separate studies to be dependent on Akt activity (17, 33, 45). Dimmeler et al. (15) demonstrated that Akt phosphorylation increases monotonically with increasing WSS in the range of 0 to 45 dyn/cm² in 2-D cultures. Here, ECs were cultured on a 3-D matrix, and we observed that the level of S1P/WSS-induced Akt activation was maximal at 5.3 dyn/cm² WSS and decreased at 12 dyn/cm² WSS. Activation of S1P₁ receptors in ECs has been reported to stimulate the activation of Akt via pertussis toxin-sensitive G proteins (22, 36). S1P-induced migration (36) and 3-D invasion (2) are blocked by pertussis toxin. Interestingly, EC migration and tubulogenesis following preshearing are also attenuated by pertussis toxin (10). Akt has several downstream targets implicated in angiogenesis, including endothelial nitric oxide synthase (eNOS) (16), hypoxia-inducible factor (23), and the transcription factor forkhead box transcription factor (FOXO) (6). Of note, both eNOS activity and FOXO expression are regulated by WSS in ECs (6, 16). Whether these signals are affected by the S1P signaling pathway in either 2-D or 3-D settings remains to be investigated.

Our data indicate that WSS can modulate proteolytic events during EC invasion. Matrix proteolysis is an established mediator of cell invasion in 3-D collagen and fibrin matrices (2, 7, 35, 48, 52). S1P/growth factor-induced invasion correlates with activation of MMP-2 (2). In the present study, the rate of invasion correlated temporally with increased rates of pro-MMP-2 expression and MMP-2 activation (Figs. 3 and 8). Interestingly, EC invasion density and the level of MMP-2 activation were maximal at an intermediate WSS magnitude of 5.3 dyn/cm². These data are consistent with a previous report by Milkiewicz and colleagues (42) that demonstrated 16 dyn/cm² WSS downregulated both MMP-2 protein and mRNA levels (vs. static control) in cultured microvascular ECs, whereas 5 dyn/cm² WSS did not. Increased MMP-2 activation has previously been identified as an indicator of membrane type 1 metalloprotease (MT1-MMP) activity in ECs (34). Our data observed at 5.3 dyn/cm² demonstrate a correlation between MMP-2 activation and cell invasion density, consistent with a role for MT1-MMP mediating movement through 3-D matrices in vitro (52) and in vivo (7). The effects of 12 dyn/cm² WSS on the morphology of invading cells are intriguing. Although 12 dyn/cm² WSS reduced the density of invading ECs, the invading structures extended more rapidly into collagen matrices and generally had narrower lumens than sprouts from ECs subjected to lower magnitudes of WSS. The ability of ECs to invade into collagen matrices despite a decrease in MMP-2 levels is also noteworthy and demonstrates MMP-2 is not required for EC invasion into 3-D collagen matrices. These data are consistent with previous reports that tissue inhibitor of metalloproteinase-1 (TIMP-1) (2, 52), an endogenous inhibitor of soluble MMPs, including MMP-2 and MMP-9 (1), does not alter invasion responses. The differential invasion responses to 5.3 and 12 dyn/cm² WSS illustrate a complex regulation of sprout architecture by WSS; however, further studies are needed to determine how shear stress regulates proteolytic events and related angiogenic intracellular signaling cascades.

It is intriguing that invasion distance and density show different dependencies on WSS magnitude (Fig. 4, D and E). It is well established that low and high WSS have distinct effects on cell signaling, migration, morphology, proliferation, and apoptosis in ECs cultured in 2-D culture (11, 27, 43). This is the first study to investigate the effects of different WSS magnitudes on EC invasion into 3-D matrices. Although Akt phosphorylation and MMP-2 activation correlated with invasion density, it is unclear why invasion distance was highest at 12 dyn/cm² WSS. WSS levels of 5.3 and 12 dyn/cm² induced similar levels of Ki-67 staining and caspase-3 cleavage. Thus other mechanisms unrelated to cell proliferation and apoptosis are involved in regulating the sensitivity of invasion to the magnitude of WSS.

Recently, Venkataraman et al. (57) reported data supporting a role for ECs in maintaining plasma levels of S1P. Furthermore, these authors demonstrated that 15 dyn/cm² WSS increased the secretion of S1P from cultured ECs with concomitant upregulation of sphingosine kinase expression and downregulation of the expressions of sphingosine lyase and sphingosine phosphatase. With the consideration of the results of this new report and the present study, it appears that WSS regulates both the generation of S1P as well as the ability of S1P to induce EC invasion. The role of WSS-induced S1P generation in our system may not be critical in our system due to the high concentration present in the matrix; however, such localized S1P generation may be important in regulating EC invasion in vivo.

Although flow has long been recognized as a regulator of angiogenesis (8), few studies have directly evaluated the role of flow-induced WSS on angiogenesis in vivo. Nasu et al. (47) studied the growth and flow rate distribution of microvessels in tumors and concluded their observations indicated that vessel growth in tumors depends more on local hemodynamics than on vascular growth factors. Typical WSS magnitudes in the microvasculature are reported to be in the range of 5 to 150 dyn/cm² (26, 51); however, WSS levels in postcapillary venules in mouse cremaster muscle have been estimated to be in the range of 1 to 5 dyn/cm² (32). Ichioka et al. (29) measured changes in blood flow and vascular growth in response to long-term administration of the vasodilator prazosin in a rabbit ear chamber. They estimated that WSS in postcapillary venules increased from 3.7 to 5.3 dyn/cm², and this coincided with an increase in the rate of tissue vascularization within the chamber (29). Our results support the finding that increasing WSS within the range of 3 to 5.3 dyn/cm² enhances the angiogenic response. It is important to note that sprouting angiogenesis is primarily localized to postcapillary venules, which under quiescent conditions are exposed to relatively lower WSS than blood vessels in other areas of the vasculature. The model presented here is consistent with studying angiogenic responses in a wound environment where increased deposition of S1P by platelets and immune cells and increased WSS would be expected to occur. The findings presented here establish a model to study the critical role of S1P in sprouting angiogenesis in 3-D collagen matrices, along with a complementary role for alterations in blood flow in modulating S1P-induced sprouting.

GRANTS

This work was supported by American Heart Grants 0730238N (to R. Kaunas) and 0530020N (to K. Bayless).