VEGF internalization is not required for VEGFR-2 phosphorylation in bioengineered surfaces with covalently linked VEGF

Materials

(1-Mercapto-11-undecyl) tetra (ethylene glycol) (EG-OH) was purchased from Asemblon (Redmond, WA). (1-Mercapto-11-undecyl) septa (ethylene glycol) amine (EG-NH₂) was obtained from ProChimia (Evanston, IL). Heparin sodium salt from porcine intestinal mucosa was purchased from Alfa Aesar (Ward Hill, MA). Vascular Endothelial Growth Factor (VEGF) was kindly provided by Genentech (South San Francisco, CA). Human umbilical vein endothelial cells (HUVEC) were either a kind gift from Dr. Andrew Putnam of UC Irvine (now University of Michigan) or purchased from Lonza (Walkersville, MD). All other reagents and products were purchased from Fisher Scientific unless noted otherwise.

Cell culture

HUVECs were cultured in EGM-2 complete medium (Lonza, Walkersville, MD) at 37°C and 5% CO₂. The HUVECs were first obtained and cultured at passage 2. In order to provide enough cells for all of the experiments, the cells were expanded and frozen at passage 7. For each experiment, the cells were thawed and grown for 2 days in a T75 flask (Corning, Corning, NY). The cells were then split using standard technique and plated 1:2 onto PDMS sheets.

PDMS functionalization and culture with cells

Poly-dimethoxy silane (PDMS) sheets were made as previously described¹³. Briefly, base silane and curing agent were mixed in a 10:1 ratio, degassed and cast onto a 0.75 mm gel casting plate before being placed in a 170°C oven for 10 minutes to cure. After the cured PDMS sheets were removed, the sheets were placed in a 1:1 v/v methanol:hydrochloric acid solution for 30 minutes at room temperature, then dried and incubated in a 5% v/v amino-propyl triethoxysilane (APTES, Sigma-Aldrich, St. Louis, MO) solution in ethanol overnight at room temperature under inert atmosphere. The sheets were sterilized in 70% ethanol, then incubated with 10 μg/ml fibronectin solution (Millipore, Temecula, CA) for 1 h at 37°C. Harvested cells were plated and cultured for two days. Prior to the experiment, the cell culture was starved with serum-free media for 6 hours. Where indicated, the cells were pre-treated for 30 minutes with 80 μM dynasore, an inhibitor of dynamin, in order to investigate the requirement of receptor endocytosis for cell signaling^{9, 15}. Five minutes before exposure to the cells, sodium vanadate (0.1 mM) was added to the cell culture media^{9, 16}.

VEGF-functionalized gold slides and treatment with cells

VEGF-functionalized glass slides were made as previously described¹³. Briefly, standard laboratory glass slides were placed in an e-beam evaporator where 5 nm titanium and 30 nm of gold were deposited. The gold slides were incubated with alkanethiol solutions (99% EG-OH, 1% EG-NH₂) under inert atmosphere for 3 h at room temperature. A heparin solution was made by oxidizing 62.5 mg/ml heparin in 200 mM sodium periodate in 100 mM sodium acetate pH 4 for 30–60 minutes. The reaction was quenched with addition of glycerol, then diluted to 3 mg/ml heparin and adjusted to pH 7 with PBS. Heparin became photoactive by addition of azido-benzyl hydrazide (ABH, Pierce, Rockford, IL) for 2 hours at room temperature. The solution was then diluted to 1 mg/ml heparin and adjusted to pH 9–9.5 before incubating with the gold slides for 2 hours at room temperature. Then, the slides were incubated with 50 mM sodium cyanoborohydride in 100 mM sodium acetate pH 4 for 5 minutes at room temperature. Overnight incubation of VEGF at 4°C in 1% BSA-PBS at 200 ng/ml was followed by 365 nm wavelength UV light activation for 10 minutes to lock VEGF covalently to the surface at a density of 100 pg/cm². The VEGF functionalized slides were exposed to cells on PDMS surfaces for the indicated time points. A control, soluble VEGF at 2 ng/ml, was based on a mass transfer analysis that showed the boundary layer surrounding the cells did not have a significant effect in reducing the apparent VEGF concentration at the cell surface (Supplemental Figure 1). Given the VEGF density on the surface and the surface area of cells exposed, 2 ng/ml was an equivalent solution concentration.

Western blot analysis

After growth factor treatment at 37°C, the sheets were placed in Petri dishes on ice. The sheets were rinsed twice with ice cold PBS supplemented with 0.2 mM sodium vanadate⁹. After aspirating all remnants of liquid from the PDMS sheets, 100 μl of lysis buffer (1% Non-idet, 10 mM Tris-HCl, pH 7.6, 150 mM NaCl, 30 mM sodium pyrophosphate, 50 mM sodium fluoride, 2.1 mM sodium orthovanadate, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 2 μg/ml of aprotinin) was added to the surface and scraped⁹. Insoluble cell material was removed by centrifugation at 4°C for 10 min at 14,000 rpm (Beckman Coulter Microcentrifuge 22R, Brea, CA). Equal amounts of cell lysate (BCA assay, Bio-Rad, Hercules, CA) were diluted in 5X loading buffer (1 M Tris-HCL, pH 6.8, 20% SDS, 50% glycerol) supplemented with 5% (v/v) β-mercaptoethanol, boiled for 10 min at 70°C, separated by SDS-PAGE (8% resolving, 2 h at 130 V), and transferred to nitrocellulose membranes (2 h at 400 mA). The membranes were incubated in blocking buffer (5% milk in 0.1% Tween-20 in TBS) for 1 h at room temperature before overnight incubation with primary antibodies. Phosphorylated proteins were detected by immunoblotting using anti-phosphotyrosine antibodies (pVEGFR-2/1175 Cell Signaling, Boston, MA; pVEGFR-2/1214 Invitrogen, Carlsbad, CA; and phospho-p38 Cell Signaling, Boston, MA in blocking buffer) followed by secondary antibodies coupled with horseradish peroxidase (200 ng/ml, Invitrogen, Carlsbad, CA, 1 h at room temperature) and visualized by chemifluorescence (ECL detection reagents, GE Healthcare, Piscataway, NJ) using a Typhoon scanner (GE, Amersham Biosciences, Piscataway, NJ). Protein-loading control was assessed by Western blot using anti-VEGFR-2 or anti-p38 (Cell Signaling Technology, Boston, MA). Typhoon images were analyzed and normalized with ImageJ software. Statistical comparisons were done with 3 independent experiments. For p38 blot data, film exposure was utilized as previously described⁹.

Cdc42 GLISA

Cells were treated for 60 minutes following the same method mentioned earlier. After treatment, the cells were lysed and quantified with ProteinRed. The samples were assayed with the CytoSkeleton (Denver, CO) cdc42 GLISA kit following manufacturer’s instructions. Samples loaded at 0.5 mg/ml, and absorbance read at 490 nm.

¹²⁵I-VEGF internalization

VEGF labeled with Iodine-125 was purchased from Perkin Elmer (Santa Clara, CA) and diluted 1:100 in VEGF without Iodine-125. A standard curve was made to convert CPM to amount VEGF. The same procedure for gold slide synthesis was used, except with this solution of radiolabeled VEGF. The gold slides were exposed to cells on PDMS for the indicated time points. Both the gold slide and PDMS were read in a gamma counter. VEGF internalization curves were constructed based on the CPM readings and conversion via the standard curve.

A pulse-chase experiment was performed to begin to study the receptor-recycling dynamics of Vs, Ve and Vc. First, bound and soluble VEGF (cold) were exposed to cells for approximately 3 seconds, and then immersed in soluble ¹²⁵I-VEGF (hot) for 30 minutes. Second, bound and soluble VEGF (cold) were exposed to cells for 30 minutes, followed by immersion into ¹²⁵I-VEGF (hot) for 30 minutes. An additional control was inhibition of endocytosis by a 30-minute cell pre-treatment with 80 μM dynasore, a dynamin inhibitor^{9, 15}, and then treatment with hot and cold soluble VEGF.

Optical trap rupture force measurements

Cells were cultured on glass bottom culture dishes (MaTek, number 0, Ashland, MA) and supplemented with 80 μM dynasore (Sigma-Aldrich, St. Louis, MO) 30 minutes before the experiment. Amine-functionalized polystyrene beads 3 μm in diameter (Spherotech, Lake Forest, IL) were incubated with oxidized heparin, then reduced with sodium cyanoborohydride. Following a 5% BSA incubation, the beads were incubated with 200 ng/ml VEGF overnight at 4°C. The beads were exposed to 365 nm wavelength UV, then washed three times with PBS-T and stored in a 1 mg/ml heparin release solution for three days. On the day of the experiment, the beads were introduced to the cells.

Custom laser tweezers were integrated with an Olympus IX81 Inverted Microscope. A 1064 nm Ytterbium fiber laser (IPG) was focused by a PlanApo 60x 1.45NA oil immersion objective lens to trap a microsphere. Forward scattered light is collected by a high NA Olympus condenser and focused onto a quadrant photo diode (QPD) positioned conjugate to the back focal plane of the objective lens. Petri dishes were coarsely positioned by a stepper motor microscope stage and again with nanometer resolution by a piezoelectric insert. Samples were maintained at 37°C by an objective heater. Olympus Slidebook software and custom software coded in National Instruments LabView controlled data collection and microscope actuation.

Computational modeling and parameter fitting

Computational kinetic models were created using SimBiology (MathWorks) based on the radiometric assays of VEGF internalization over time. The purpose of this model was to generate hypotheses and support experimental results. A reaction mechanism describing the process of VEGF internalization into the cell was developed, with proposed ligand and receptor fates assigned unique rate constants. Differential equations for ligand-receptor binding, activation of the dimer, and subsequent ligand and receptor recycling were utilized in each simulation. Models were based on kinetic mass-action principles with the initial reaction rate parameters derived from literature where possible (k_bind, k_unbind, k_Rrecycle, k_Lrecycle)^{17, 18}. The models were then fit to the experimental data by parameter fitting of the rate constants. Rate constant fitting was based on reducing the deviation of the model from its matching experimental data for a given set of kinetic rate parameters:

Error was calculated as the sum of least squares of modeled versus experimental data for the total N recorded experimental data points. Minimization of the error was performed by a non-linear minimum search algorithm in MATLAB (MathWorks) for the kinetic rate parameters that generated models that best fit the data. Order of magnitude changes in the parameters were tested to be sure the residual error was at an absolute minimum and not one of many local minima. In the proposed reaction mechanism, ligand-dependent and ligand-independent receptor-mediated endocytosis was tested for the bound case to determine which event was more dominant based on the values of the kinetic constants controlling each scenario. This model was developed and executed in order to support the experimental observations made in the radiometric assays.

Statistical analysis

Data are presented as mean ± standard deviation. To identify significant trends in data, statistical comparisons were performed by one-way ANOVA with post-test using the Tukey method. Data were considered significantly different if p < 0.05.

Covalently bound VEGF sustains VEGFR-2 phosphorylation

Previously, we developed a method to covalently bind VEGF (Vc) or electrostatically bind VEGF (Ve) to a heparin-functionalized surface and showed its ability to phosphorylate VEGFR-2¹³. The same method was applied here and cells were exposed to the growth factor-coated surface from 0 to 60 minutes to determine the kinetics of phosphorylation (Figure 1A). At the indicated time points, cell lysates were collected and analyzed by Western blot for phosphorylation of VEGFR2 at Y1175 and Y1214 (Figure 1B–E). The soluble treatment (Vs), 2 ng/ml, was chosen based on a mass transfer calculation and data previously published to be an equivalent concentration (100 pg/cm²) as that initially exposed to the cells in the surface treatment¹³ (Supplemental Figure 1). The mass transfer model indicated that the boundary layer surrounding the cells had a negligible effect on the diffusion of soluble VEGF from the bulk solution to the surface of the cell. Thus, no gradients in VEGF concentration were present. The boundary layer ( < 10 μm¹⁹) was below the threshold for significantly affecting diffusion to cell calculated in the model ( > 20 μm, Supplemental Figure 1). Therefore, for the amount of cells exposed in the given surface area, 2 ng/ml was an equivalent concentration of VEGF from the surface exposure to the soluble treatment. This was done to ensure that the behavior observed was dependent on the presentation of the growth factor. While all treatments except for the negative control (Vn) induced significant levels of phosphorylation at Y1175 during early time points, only Vc treatment led to an extension of this phosphorylation at 60 minutes (p < 0.05, Figure 1B–C). Similar to what has been previously observed for matrix bound VEGF⁹, Vc and Ve both generated a stronger response from Y1214 of VEGFR-2 than Vs (Figure 1D–E). Vc, again, extended phosphorylation of VEGFR-2 at Y1214 up to 60 minutes over Vs. At 45 and 60-minutes of the treatment, Vc increased the phosphorylation intensity over Ve (p < 0.05, Figure 1E).

Covalently bound VEGF extends p38 pathway signaling

Activation of Y1214 on the carboxy-terminal domain of VEGFR-2 leads to activation of cdc42 and p38, members of the migration pathway in HUVEC signaling. Since Y1214 is activated more strongly in the bound cases, cdc42 and p38 activation were analyzed to study the effect of Vc on the migration pathway. Both cdc42 and p38 displayed increased and extended magnitude of activation through 60 minutes for Vc (Figure 2A–B, p < 0.001 for cdc42). Vc and Ve had greater than a two-fold increase in activation of p38 over Vs at 60 minutes of treatment (Figure 2B, Supplemental Figure 2).

Inhibition of receptor endocytosis decreased Y1175 activation and increased Y1214 activation

As recently reported, internalization of the receptor is required for receptor phosphorylation^{9, 15}. Intuitively, if VEGF is covalently locked to the surface, the receptor must either remain at the surface or internalize without VEGF. Covalently bound VEGF can be removed from the surface by protease action, but it was assumed that MMP concentration was low, especially in the early time points of activation after the cells were quiescent and serum starved for several hours. Also, it was thought unlikely that VEGF could be removed from the matrix at a rate and magnitude high enough to generate the responses we were observing. We hypothesized that another mechanism was responsible for the data. When inhibiting receptor endocytosis with addition of dynasore to the cell culture medium before activation, Y1175 phosphorylation was reduced (Figure 3A), with Vc, Ve, and Vs resulting in a 57.6%, 85.2%, and 81.3% decrease, respectively, after 5 minutes of treatment. However, both Vc and Ve resulted in 1.5-fold increase in Y1175 phosphorylation at 30 minutes (Figure 3A). Interestingly, Y1214 phosphorylation was increased with addition of dynasore (Figure 3B). For Vc and Vs, the fold increase in activation was at least 2-fold. In Ve, the increase in activation was 1.1-fold for 5 minutes, and 1.4-fold for 30 minutes. In these quantifications, the time point exposed to dynasore was normalized relative to the corresponding non-treated time point (no VEGF added), which was assigned a value of 1.

Covalently bound VEGF shifts VEGF affinity toward matrix and away from VEGFR-2

Covalent binding of VEGF to a biomaterial surface may disrupt the normal physiological interactions between the matrix, VEGF, and VEGFR-2. To quantify the mechanical changes induced by introduction of the covalent bond between VEGF and the matrix, we measured rupture forces. Optical trap technology allows a user to utilize lasers to move polystyrene beads next to cells and measure cell-bead interactions (Figure 4A). Heparin was attached to amine-functionalized 3 μm polystyrene beads using the same chemistry to form the surface¹³. VEGF was immobilized in the same manner as previously described, and the beads were placed next to HUVECs incubating in serum-free media to allow binding to occur. Once the bead began to interact with the cell, the optical laser oscillated for several minutes, and rupture force measurements were recorded. Vc resulted in a right shift of the histogram (6–12 pN) as compared to Ve (3–8 pN), indicating that the rupture force increased when the covalent bond was introduced (Figure 4B).

To answer the question of why Vc increased the rupture force, we turned to affinity data. Under normal physiological conditions, VEGF has a greater affinity for VEGFR-2 than for matrix components such as heparin, collagen, and fibronectin^{17, 20}. The dissociation constants are 0.1 and 40–160 nM for VEGF-VEGFR-2 and VEGF-matrix, respectively, in the absence of a covalent bond (Table 1)^{17, 20}. However, when a covalent bond is introduced between VEGF and the matrix, the theoretical affinity preference shifts toward VEGF and the matrix.

Internalization of covalently bound VEGF is reduced compared to soluble and heparin bound VEGF

The dynasore and rupture force data identified a change in VEGF/VEGFR-2 dynamics that is dependent on the nature of the growth factor presentation. We hypothesized that this was a result of ligand remaining on the biomaterial surface during receptor phosphorylation. The fate of the ligand was studied using radiolabeled VEGF, with a gamma count performed on both the cell surface and the growth factor surface in order to track the transfer of ligand. Confluent HUVECs were exposed to equal amounts of Vc, Ve, or Vs from 0 to 60 minutes. After the exposure time, the cell surface and growth factor surface were separated and analyzed for presence of radiolabeled VEGF. The data is presented as moles of ligand internalized. Of the 500 fmol VEGF supplied (Supplemental Figure 3), Vc resulted in internalization of approximately 0.4 fmol VEGF, while Ve and Vs had 0.8 fmol and 1 fmol VEGF internalized, respectively (Figure 5A). The 1 fmol VEGF internalized for the Vs condition matched the known density of receptors on the HUVECs^{18, 21}.

We next performed a receptor quenching experiment with radiolabeled VEGF to investigate the receptor-recycling dynamics of the system. In this experiment, the cells were exposed to cold VEGF in Vc, Ve or Vs forms for either a few seconds or 30 minutes, and then pulsed with hot, radiolabeled soluble VEGF for 30 minutes to allow for binding of radiolabeled ligand to residual receptors on the cell surface and internalization. There was a significant reduction in the internalized ligand for the Vc situation, suggesting less receptor availability because of a decrease in receptor recycling (Figure 5B, *p < 0.05, **p < 0.01, ***p < 0.001). Ve and Vs did not show a significant decrease in receptor availability, indicating that receptor recycling was not affected (Figure 5B). When treated with the endocytosis inhibitor dynasore, soluble VEGF (VsD) showed no decrease in receptor availability. The data here represented VEGF bound to the receptor on the cell surface. The 1 fmol of VEGF adhered to the cell surface in this case further supported the idea that Vs saturated the receptors in the internalization assay.

To better understand the mechanisms behind the kinetics of internalization, a mathematical model was developed to describe the trafficking of VEGF and VEGFR-2 during the internalization process. It was hypothesized that the soluble and bound cases undergo different binding and internalization mechanisms described by the following equations:

In this model, VEGF internalization in the soluble case begins by ligand (VEGF, L) binding to receptor (VEGFR-2, R). The constants k_bind and k_unbind are the forward and reverse reaction constants^{17, 18}. This binding event leads to formation of a complex (C) between ligand and receptor. Two complexes dimerize (D) with k_dimer and k_undimer as the forward and reverse reaction constants, which leads to internalization of the dimerized complex (D_i) and subsequent activation of signaling cascades controlled by the rate constant k_intern. During the dimerization and activation phase, the receptor undergoes conformational changes^{14, 22, 23} and internalization. Internalized dimers dissociate over some time scale into internalized receptors (R_i) and ligands (L_i) with rate constant k_dissoc. To avoid too many parameters being fit by the radiometric data, this step was described by a single kinetic rate constant (k_dissoc), which includes the rates of D_i→C_i→2R_i + 2L_i. The internalized ligands and receptors then have the potential to be degraded or recycled to the cell membrane surface (k_Rrecycle and k_Lrecycle ²⁴).

In the bound case, the initial steps are similar to the soluble case controlled by kinetic constants of the same values as their respective counterpart (k_bind, k_unbind, k_dimer, k_undimer). However, two options were proposed for the fate of the ligand based on the observations in the radiometric assay. Free ligand can either be regenerated at the surface by dissociation of the dimer prior to internalization (described by k_dissoc for the bound model) and receptor internalization without the ligand (described by k_break) or internalized into the cell, as is the case with soluble (described by k_nobreak). Also, the dissociation of the dimer in the bound case due to internalization has two options: dissociation in the usual sense after the entire complex is internalized (k_nobreak) or the dissociation occurring due to receptor internalization without the ligand (k_break). Receptor and ligand recycling constants for the bound case were determined using the parameter fitting algorithms.

The radiolabeled ligand internalization data was used to fit the unknown parameters of these three models (Vs, Ve, and Vc). Error between the predicted values and the data was reduced by adjusting the rate constants until a suitable fit was found. In the bound case, the constants governing the breakdown of the dimerization species gave insight into the fate of the ligand during the trafficking process. For the electrostatic situation, the rate at which the receptor and ligand would both be internalized (k_nobreak) was much greater than the rate at which the receptors and ligands unbind at the cell surface without being internalized (k_dissoc), indicating that it was much more likely for the dimer to be internalized as dissociated ligands (L_i) and receptors (R_i) than to unbind at the cell surface (L and R). The rate at which the receptor internalized and left the ligand at the cell surface (k_break) was greater than k_nobreak, suggesting that receptor internalization and activation without the ligand was possible. For the covalent case, k_break was an order of magnitude greater than k_nobreak, suggesting that ligand internalization was reduced compared to the electrostatic and the soluble case and that ligand was left at the surface (Table 2). The dissociation constant, k_dissoc, for Vc was two orders of magnitude smaller than the same constant for Vs, indicating that the dimer complex was longer lasting in the covalently bound case. Based on the constants of best fit, the model results suggested that recycling of the receptor was slowed with increasing affinity of growth factor for the matrix (Vs > Ve > Vc).

Covalently binding VEGF increases growth factor half-life

The internalization of VEGF was significantly reduced when covalently bound to the surface. Sustained activation of VEGFR-2 may be a result of ligand regeneration at the surface and stability of VEGF when bound to the biomaterial. To study if Ve and Vc have increased stability against degradation compared to Vs, we performed two types of experiments. In the first experiment (long term stability), HUVECs were exposed to Vc surfaces that were incubated at 4°C for up to 27 days, Vc surfaces that were incubated at 37°C for up to 11 days, and Vs solutions (2 ng/mL) that were stored at 4°C for up to 23 days. At the indicated time points, the slides were exposed to untreated serum starved cells for 5 minutes to assess activity through phosphorylation of VEGFR-2 at Y1175. Vc stored at 37°C had similar stability as Vs stored at 4°C (Figure 6A). Vc stored at 4°C was stable throughout the course of the experiment, still exhibiting 21% activity at day 27 (Figure 6A). The normalized quantifications showed stability curves that resembled autocatalytic degradation (Figure 6B). A fit equation derived from autocatalytic degradation was used to calculate the half-life²⁵. Vc at 37°C and Vs at 4°C have the same half-life (2 days). Vc at 4°C had a half-life of one week.

In the second experiment (repeat exposure), untreated serum starved cells were exposed to the same surface for 30 minutes each for a total of four rounds of exposure. Both Vc and Ve phosphorylated VEGFR-2 (Y1175 and Y1214) at sustained levels through four rounds of 30-minute cell exposures (Figure 6B). Vs declined in activity toward the third and fourth round of treatments for Y1175, and showed little activity at Y1214 compared to bound, as shown previously. To assure the 100-min half life of soluble VEGF at 37°C did not lead to misinterpretation of the data, refreshments of the growth factor maintained a constant concentration throughout the refresh experiment⁹ (Figure 6B).