Receptor protein tyrosine phosphatase beta/zeta is a functional binding partner for vascular endothelial growth factor
- 2k Downloads
Receptor protein tyrosine phosphatase beta/zeta (RPTPβ/ζ) is a chondroitin sulphate (CS) transmembrane protein tyrosine phosphatase and is a receptor for pleiotrophin (PTN). RPTPβ/ζ interacts with ανβ3 on the cell surface and upon binding of PTN leads to c-Src dephosphorylation at Tyr530, β3 Tyr773 phosphorylation, cell surface nucleolin (NCL) localization and stimulation of cell migration. c-Src-mediated β3 Tyr773 phosphorylation is also observed after vascular endothelial growth factor 165 (VEGF165) stimulation of endothelial cells and is essential for VEGF receptor type 2 (VEGFR2) - ανβ3 integrin association and subsequent signaling. In the present work, we studied whether RPTPβ/ζ mediates angiogenic actions of VEGF.
Human umbilical vein endothelial, human glioma U87MG and stably transfected Chinese hamster ovary cells expressing different β3 subunits were used. Protein-protein interactions were studied by a combination of immunoprecipitation/Western blot, immunofluorescence and proximity ligation assays, properly quantified as needed. RPTPβ/ζ expression was down-regulated using small interference RNA technology. Migration assays were performed in 24-well microchemotaxis chambers, using uncoated polycarbonate membranes with 8 μm pores.
RPTPβ/ζ mediates VEGF165-induced c-Src-dependent β3 Tyr773 phosphorylation, which is required for VEGFR2-ανβ3 interaction and the downstream activation of phosphatidylinositol 3-kinase (PI3K) and cell surface NCL localization. RPTPβ/ζ directly interacts with VEGF165, and this interaction is not affected by bevacizumab, while it is interrupted by both CS-E and PTN. Down-regulation of RPTPβ/ζ by siRNA or administration of exogenous CS-E abolishes VEGF165-induced endothelial cell migration, while PTN inhibits the migratory effect of VEGF165 to the levels of its own effect.
These data identify RPTPβ/ζ as a cell membrane binding partner for VEGF that regulates angiogenic functions of endothelial cells and suggest that it warrants further validation as a potential target for development of additive or alternative anti-VEGF therapies.
KeywordsChondroitin sulphate Endothelial cells Migration Pleiotrophin Tyrosine phosphatases Vascular endothelial growth factor
Bovine serum albumin
Fetal bovine serum
Phosphate buffered saline
Proximity ligation assay
Receptor protein tyrosine phosphatase beta/zeta
Vascular endothelial growth factor
Vascular endothelial growth factor receptor 2
Vascular endothelial growth factor A (VEGF) is a growth factor that activates several functions of endothelial cells, thus triggering angiogenesis and vascular permeability. It exists as different isoforms of 121, 145, 165, 189 and 206 amino acids, among which VEGF165 is dominant in terms of amount and biological activity. VEGF165 is over-expressed in a variety of human tumors, and its over-expression is correlated with progression, invasion, and metastasis of tumors. VEGF165 cell signaling leading to increased endothelial cell migration and tubular formation is mediated via vascular endothelial growth factor receptor 2 (VEGFR2), which interacts with co-receptors, such as ανβ3 integrin . It has been shown that c-Src-mediated phosphorylation of β3 cytoplasmic tail tyrosine residues occurs in response to VEGF165 and is essential for VEGFR2-β3 integrin association and subsequent signaling [2,3]. Up to date, it remains unclear how c-Src is activated by VEGF165. The first step in c-Src activation requires dephosphorylation of its carboxy-terminus Tyr530 , suggesting that a tyrosine phosphatase may be involved.
Receptor protein tyrosine phosphatase beta/zeta (RPTPβ/ζ) is a member of the family of receptor-type transmembrane protein tyrosine phosphatases that interacts with several cell adhesion molecules, such as neutral cell adhesion molecule, neuron-glia cell adhesion molecule, tenascin and contactin. It also acts as a receptor for the soluble, heparin-binding growth factors midkine and pleiotrophin (PTN) . RPTPβ/ζ and PTN are expressed in endothelial cells  and over-expressed in several types of cancer [5,7]. PTN binding to RPTPβ/ζ on endothelial cells leads to Tyr530 dephosphorylation and activation of c-Src, β3 Tyr773 phosphorylation, and PTN-induced endothelial cell migration and tube formation on matrigel [5,8]. Besides PTN, RPTPβ/ζ has been shown to be the putative receptor for the vacuolating cytotoxin VacA produced by Helicobacter pylori , as well as a functional receptor for interleukin-34 , suggesting that it acts as a functional binding partner for several soluble molecules.
We have recently shown that RPTPβ/ζ-induced, c-Src-mediated β3 Tyr773 phosphorylation is also required for PTN-induced cell surface nucleolin (NCL) localization . NCL is over-expressed on the plasma membrane of cancer and activated endothelial cells and has been shown to play critical roles in the modulation of tumorigenesis and angiogenesis through its interaction with a variety of ligands, among which tumor homing peptide F3, endostatin, P-selectin and PTN . VEGF165 induces NCL localization on the surface of endothelial cells and this effect is considered important for its angiogenic actions [13,14]; however, the receptors and pathways involved have not been elucidated.
In the present work, we explored the possibility that RPTPβ/ζ is involved in the stimulatory effect of VEGF165 on endothelial cell signaling leading to cell migration. Our data show that VEGF165 directly interacts with RPTPβ/ζ to induce c-Src-mediated β3 Tyr773 phosphorylation. The latter is required for both cell surface NCL localization and increased interaction of ανβ3 with VEGFR2, leading to VEGF165-induced endothelial cell migration.
Results and discussion
Phosphorylation of β3 Tyr773 is required for VEGF165-induced cell migration and cell surface NCL localization
RPTPβ/ζ plays a role in VEGF165-induced endothelial cell signaling that leads to cell surface NCL localization
To further investigate the involvement of RPTPβ/ζ in VEGF actions, we studied the role of several signaling molecules known to be activated by both VEGF165 and RPTPβ/ζ on cell surface NCL localization. Inhibition of c-Src and phosphatidylinositol 3-kinase (PI3K) abolished VEGF165-induced cell surface NCL localization, while inhibition of ERK1/2 had no effect (Figure 2D). In order to investigate whether PI3K lays up- or downstream of ανβ3, the effect of PI3K inhibition on VEGF165-induced β3 Tyr773 phosphorylation was studied. The c-Src inhibitor PP1 was used as a positive control, since c-Src is known to lay upstream of β3 Tyr773 phosphorylation [2,8,11]. The p38 inhibitor (SB203580 10 μΜ, BioSource Europe, Nivelles, Belgium) was used as a negative control, since it has been previously shown not to affect RPTPβ/ζ-mediated β3 Tyr773 phosphorylation and PI3K activation . PI3K inhibition did not affect VEGF165-induced β3 Tyr773 phosphorylation (Figure 2E), suggesting that it lays downstream of ανβ3. By using an ELISA for activated PI3K in HUVEC, it was found that inhibition of c-Src abolished VEGF165-induced PI3K activation (Figure 2F). Using the same assay in CHO cells over-expressing wild-type β3, β3Y773F or β3Y785F, it was found that VEGF165 significantly induced PI3K activation in CHO cells over-expressing wild-type β3 or β3Y785F, while it had no effect in cells over-expressing β3Y773F (Figure 2G), suggesting that PI3K lays downstream of ανβ3 and requires β3 Tyr773 phosphorylation. This pathway resembles the one we have recently shown for PTN-induced cell surface NCL localization , further supporting the notion that RPTPβ/ζ mediates this effect of VEGF165. Similarly to what has been previously discussed for PTN, it remains unclear how PI3K affects cell surface NCL localization. Co-immunoprecipitation of NCL with PI3K [15,16] favors a direct regulation of NCL by PI3K and one possibility is by regulating the trafficking of exocytotic vesicles. NCL has been detected in cytoplasmic vesicles fused with the plasma membrane , while inhibition of PI3K interferes with the trafficking of such exocytotic vesicles, affecting the number of several receptors on the plasma membrane, such as the transferrin receptor , the glucose transporter GLUT4  or β integrin . Alternatively, PI3K may indirectly regulate cell surface NCL localization, by regulating recruitment of proteins containing pleckstrin homology domains onto the cell membrane . NCL does not possess such sequences but may act as a ligand for proteins containing pleckstrin homology domains through its acidic motifs .
RPTPβ/ζ plays a role in VEGF165-induced interaction of VEGFR2 with ανβ3
RPTPβ/ζ plays a role in VEGF165-induced endothelial cell migration
VEGF directly interacts with RPTPβ/ζ
Bevacizumab used at a concentration that inhibits VEGF-induced cell migration and binding to both VEGFR1 and VEGFR2 , did not affect interaction of VEGF with RPTPβ/ζ (Figure 5C), suggesting that this interaction involves a different region on the VEGF molecule than the VEGFR binding site. Supporting this notion, bevacizumab did not inhibit VEGF165-induced cell surface NCL localization (Figure 5D), which is also mediated by RPTPβ/ζ as discussed above. Bevacizumab at the same concentration significantly inhibited interaction of VEGF with VEGFR2 (Additional file 4). These data clearly indicate that although binding to VEGFR2 is essential for VEGF-induced endothelial cell migration , some of the angiogenic actions of VEGF, such as cell surface NCL localization, are not inhibited by bevacizumab. Taking into account that angiogenic factors, such as PTN, hepatocyte growth factor and even VEGF itself act through cell surface NCL , the lack of effect of bevacizumab on VEGF165-induced cell surface NCL localization may explain at least some of the cases of resistance development to bevacizumab, e.g. in glioblastomas, where classical VEGF signaling through VEGFRs has been found to remain inhibited . Moreover, our data provide a mechanistic support to the notion that RPTPβ/ζ is a valuable target for glioblastoma therapies [25,26].
There are two transmembrane isoforms of RPTPβ/ζ, the long isoform that has been described as a CS proteoglycan, and the short isoform considered a glycoprotein. Up to date, it is still unclear which of the two RPTPβ/ζ isoforms are responsible for each of its actions and it is not known whether glycosylation of these isoforms differs among different types of cells. Efforts to identify its CS glycanation suggest that it is regulated both developmentally and in different pathophysiological situations  and our unpublished observations based on Western blot analyses of RPTPβ/ζ expressed in different cell types support the notion of a cell-type specific RPTPβ/ζ glycosylation. PTN binding to RPTPβ/ζ involves both low (K d = 3 nM) and high (K d = 0.25 nM) affinity binding sites , which represent more than one sites of interaction that involve both the protein core and the CS chains of the receptor . Oversulfation of CS is essential for PTN affinity and PTN-mediated functions [29-32], which is in line with our observation that CS-E inhibits interaction of PTN with RPTPβ/ζ in both HUVEC and U87MG cells (Additional file 5). It should be noted, however, that although inhibition in U87MG cells was at the level of 70% (total signals per cell in control: 23 ± 7 and in CS-E treated cells: 8 ± 1), in HUVEC it was smaller, at the level of 50% (total signals per cell in control: 6 ± 0.4 and in CS-E treated cells: 3 ± 0.3).
The idea that the interaction of VEGF with RPTPβ/ζ does not involve its heparin-binding properties is further supported by the observation that VEGF121, which induces HUVEC migration but does not contain the heparin-binding site of VEGF , also induces cell surface NCL localization (Figure 6D). Collectively, our data suggest that interaction of VEGF with RPTPβ/ζ may involve a part of the molecule that is distinct from those involved in VEGFR or glycosaminoglycan binding. This notion is similar to the recent finding that the high affinity neuropilin-1 binding of VEGF-A involves the exon 8-encoded C-terminal Arg . Tuftsin, a naturally occurring TKPR peptide with sequence similarity to the sequence coded by exon 8 of VEGF, blocks VEGF165-induced autophosphorylation of VEGFR2, without inhibiting VEGF binding to VEGFR2 . Taking into account that PTN binding to RPTPβ/ζ involves PTN's carboxy terminal region that is rich in basic amino acids , and that PTN inhibits VEGF binding to RPTPβ/ζ (see below), one can speculate that binding of VEGF to RPTPβ/ζ might be mediated by its exon 8-encoded sequence.
PTN stimulates human endothelial cell migration by approximately 40% [6,8,11], an effect that is smaller than that of other growth factors, such as VEGF. Interestingly, PTN decreased VEGF165-induced HUVEC migration to the levels of its own, smaller stimulatory effect (Figure 7C), in agreement with a previous study showing that PTN did not abolish but decreased VEGF165-induced endothelial cell infiltration of matrigel in vivo to the levels of its own stimulation . These data favor the notion that PTN and VEGF compete for binding to RPTPβ/ζ, as discussed above, and highlight a possible role for PTN as a regulator of angiogenesis, by limiting the aberrant effect of VEGF, while it still induces a smaller, significant stimulatory effect.
It is important to note that although RPTPβ/ζ is required for cell migration induced by PTN  and VEGF165 (present study), it is not sufficient by itself to induce cell migration, based on our previous data showing that midkine  or PTN112–136 , which lead to c-Src Tyr530 dephosphorylation and β3 Τyr773 phosphorylation through RPTPβ/ζ, do not induce endothelial cell migration. It seems that other receptors/signaling pathways are activated in parallel with the RPTPβ/ζ/c-Src/ανβ3 pathway to complementarily induce cell migration. Such a pathway for VEGF165 involves at least VEGFR2 and the cross-talk of signaling molecules activated by VEGF165 through RPTPβ/ζ, ανβ3 and VEGFR2 or other cell surface binding molecules in different cell types is under further investigation.
Human recombinant VEGF165 was prepared as previously described . VEGF121 was purchased from RELIATech GmbH (Wolfenbüttel, Germany). Bevacizumab (AVASTIN) was from Roche Applied Science (Indianapolis, IN, USA). Human recombinant PTN was from PeproTech, Inc. (Rocky Hill, NJ, USA) or prepared as previously described . PTN from both sources was equally active in all cases. CS-E was either from Seikagaku Corporation (Tokyo, Japan) (CS-E I), or prepared from a crude preparation of squid cartilage CS (purchased from Yantai Changsen Chemical Co., Ltd., Shandong, China) (CS-E II), both sources being equally active. PP1, wortmannin and U0126 were from TOCRIS (Minneapolis, MN, USA) and SU1498 was from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). All secondary horseradish peroxidase-conjugated antibodies were from Cell Signaling Technology Inc. (Beverly, MA, USA). Human IgG and other reagents were from Sigma (St. Louis, MO, USA).
HUVEC, human glioma U87MG cells and CHO cells expressing αν but deficient in endogenous β3 integrin or stably transfected to express wild type β3, β3Y773F, β3Y785F or β3Y773F/Y785F were cultured as previously described [8,11]. Cell culture reagents were from BiochromKG (Seromed, Germany). All cultures were maintained at 37°C, 5% CO2, and 100% humidity. When cells reached 70-80% confluence, they were serum starved for 16 h (where indicated) before performing migration assays, lysed for immunoprecipitation/Western blot assays or fixed for immunofluorescence assays and in situ PLA.
Cells were fixed with 4% formaldehyde in phosphate-buffered saline (PBS) pH 7.4 for 10 min and permeabilized with 0.1% Triton in PBS for 15 min. After being washed 3 times with PBS, the cells were blocked with PBS containing 3% BSA and 10% fetal bovine serum (FBS) for 1 h at room temperature. Cells were stained with the following primary antibodies: rabbit anti-VEGF (1:250; Santa Cruz Biotechnology Inc.), mouse anti-RPTPβ/ζ (1:250; BD Biosciences, San Diego, CA, USA), mouse anti-ανβ3 (1:500; Merck Millipore, Darmstadt, Germany), and rabbit anti-NCL (1:1,000, Sigma). Cells were then incubated with fluorescent Alexa secondary antibodies (1:500; Molecular Probes, Carlsbad, CA, USA). Nuclei were stained with Draq5 (Biostatus Limited, Leicestershire, UK). Cells were mounted with Mowiol 4–88 (Merck Millipore) and visualized at room temperature with Leica SP5 (X63 objective with a numerical aperture of 1.4) confocal microscope.
In situ PLA
For detection of protein-protein interactions, in situ PLA was performed. The components used (Sigma) were as follows: anti-mouse PLA plus probe, anti-rabbit PLA minus probe and Detection Reagents Orange. HUVEC or U87MG cells were grown on μ-Chamber 12 well on glass slides (Ibidi©, Martinsried, Germany). After reaching 80% confluence or after appropriate treatment of cells, the assay was performed according to the manufacturer’s instructions. Briefly, after fixation and blocking, cells were incubated with the primary antibodies: mouse anti-VEGF (1:250), rabbit anti-VEGF (1:250), rabbit anti-Flk-1 (1:250), mouse anti-Flk-1 (1:250), mouse anti-NCL (1:50), goat anti-RPTPβ/ζ (1:250) (all from Santa Cruz Biotechnology Inc.), mouse anti-ανβ3 (1:500; Merck Millipore), mouse anti-PTN (1:500; Abnova, Heidelberg, Germany) and mouse anti-RPTPβ/ζ (1:250; BD Biosciences). Subsequently, cells were incubated with secondary antibodies conjugated with oligonucleotides. After hybridization and ligation of the oligonucleotides, the DNA was amplified. A detection mixture detected the amplicons, resulting in red fluorescence signals. Nuclei were counterstained with Draq5; cells were mounted with Mowiol 4–88 and visualized at room temperature with Leica SP5 confocal microscope.
Quantification of in situ PLA signals
Estimation of nuclei and cytoplasm size was performed using the Duolink ImageTool software (Olink Bioscience). In order to calculate the total number of spots per cell, an algorithmic procedure was developed and implemented in the Matlab environment (The MathWorks Inc., Massachusetts, USA). “I” denotes an immunofluorescence image of m × n size. Such an image can be seen as a “stack” of three matrices of the same size; representing the red, green and blue values for each pixel indicated as IR, IG, IB for the three color bands, respectively. For every pixel at position (i, j) correspond three values; IR(i, j), IG(i, j), IB(i, j), ranging from 0 to 255. Image pre-processing was performed to identify pixels of red intensities less than a determined threshold value θ, in order to exclude background noise. Several values were tested and evaluated and finally θ was set to 50. In order to detect a red dot, the algorithm searches the red band (i.e., the IR matrix) for a group of at least c connected pixels. The value of the parameter c, which controls the minimum number of pixels that constitute a dot, c was set to 10. Moreover, in the case of overlapping dots, large groups of pixels were identified. In such cases, the true number of dots was calculated as [n /100] since a typical dot contains approximately 100 pixels, where n is the total number of pixels of the initial group and the operator [α] is used to round the number α into its nearest integer.
Cells were grown to 50% confluence in medium without antibiotics. Transfection was performed in serum-free medium for 4 h using annealed RNA for RPTPβ/ζ (siRPTPβ/ζ#1, VBC Biotech Services, Vienna, Austria), as previously described . Another siRNA sequence for RPTPβ/ζ (siRPTPβ/ζ#2, Hs_PTPRZ1_1 FlexiTube siRNA, Qiagen GmbH, Germany) was also used as a control for off-target effects. JetSI-ENDO (Polyplus Transfection, Illkirch, France) or Lipofectamine® RNAiMAX (Life Technologies) were used as transfection reagents. Double-stranded negative control siRNA (Ambion, Austin, TX, USA) was used in all experiments. Cells were incubated for another 48 h in serum-containing medium and lysed, serum starved or fixed before further experiments.
Migration assays were performed as previously described [8,11] in 24-well microchemotaxis chambers (Corning, Inc., Lowell, MA, USA) using uncoated polycarbonate membranes with 8 μm pores. Serum-starved cells were harvested, resuspended at a concentration of 105 cells/0.1 ml in serum-free medium containing 0.25% bovine serum albumin (BSA) and loaded in the upper chamber. The bottom chamber was filled with 0.6 ml of serum-free medium containing 0.25% BSA and the tested substances. Cells were incubated for 4 h at 37°C. After completion of the incubation, filters were fixed and stained with 0.33% toluidine blue solution. The cells that migrated through the filters were quantified by counting the entire area of each filter, using a grid and a microscope with a X20 objective (Optech Microscope Services Ltd., Thames, UK).
PI3K p85 ELISA
The levels of total and phosphorylated PI3K p85 were quantified using Fast Activated Cell-based ELISA assays (Active Motif, Carlsbad, CA, USA) according to the manufacturer's instructions. Briefly, cells were cultured in 96-well plates one day prior to manipulation. Serum starved CHO cells or HUVEC were treated with 10 ng/ml VEGF165 for 10 min (in the presence or absence of inhibitors of signaling molecules where appropriate), fixed and incubated with anti–phospho and anti-total p85 antibodies.
Cells were lysed with RIPA buffer, as previously described . Three mg of total protein were incubated with primary antibody for 16 h at 4°C under continuous agitation. The primary antibodies used were: mouse anti-VEGF (3 μg), mouse anti-Flk-1 (3 μg), goat anti-RPTPβ/ζ (3 μg), goat anti-PTN (3 μg) (Santa Cruz Biotechnology Inc.) and goat anti-β3 (1.5 μg; Merck Millipore). Protein A- and protein G-agarose beads (Merck Millipore) were added, samples were further incubated for 2 h at 4°C, and beads with bound proteins were collected by centrifugation (5,000 g for 5 min at 4°C) and washed twice with ice-cold PBS pH 7.4. Immunoprecipitated proteins were resuspended in SDS loading buffer and analyzed by Western blot.
Western blot analysis
Proteins were analyzed by SDS-PAGE and transferred to Immobilon P membranes. Blocking was performed by incubating the membranes with Tris-buffered saline (TBS) pH 7.4 with 0.05% Tween (TBS-T), containing either 5% nonfat dry milk or 3% BSA. Membranes were incubated with primary antibodies for 16 h at 4°C under continuous agitation, washed 3 times with TBS-T, and incubated with secondary antibodies for 1 h at room temperature. Primary antibodies used were mouse anti-Flk-1 (1:500), goat anti-β3 (1:500), mouse anti-VEGF (1:500), rabbit anti-phospho-β3(Y773) (1:1,000; Santa Cruz Biotechnology Inc.), mouse anti-RPTPβ/ζ (1:500; BD Biosciences), rabbit anti-c-Src (1:1,000; Merck Millipore), rabbit anti-phospho-c-Src(Y418) (1:1,000; Acris Antibodies GmbH) and rabbit anti-non-phospho-c-Src (1:1,000; Cell Signaling Technology Inc.). Detection of immunoreactive bands was performed using the enhanced chemiluminescence detection kit (Pierce Biotechnology, Rockford, IL, USA). Protein levels were quantified using the ImagePC image analysis software (Scion Corp., Frederick, MD, USA).
Results are expressed as mean ± s.e.m. or by using box plots, where the box is determined by the 25th and 75th percentiles, the whiskers are determined by the 5th and 95th percentiles, the line in the box marks the median and the empty square in the box marks the mean. Outliers have been plotted as individual points. Where applicable, variability between the results from each group and the corresponding control was determined by unpaired t test.
This work was supported by the European Union (European Social Fund – ESF, Heracleitus II (to M. Koutsioumpa - E. Papadimitriou), Thales (to V. Megalooikonomou) and ΙΚΥ Fellowship of Excellence for Postgraduate studies in Greece - Siemens Program (to E. Pantazaka). This work was also supported in part by a Grant-in-Aid for Scientific Research for Challenging Exploratory Research 25670018 (to K. Sugahara) from the Japan Society for the Promotion of Science (JSPS). The authors thank the Advanced Light Microscopy facility of the Medical School, University of Patras for using the Leica SP5 confocal microscope and the PhD candidate Margarita Lamprou for technical support.
- 5.Koutsioumpa M, Papadimitriou E: PG receptors with phosphatase action in cancer and angiogenesis. In Extracellular Matrix: Pathobiology and Signalling. Edited by Karamanos N. Berlin/Boston: Walter de Gruyter GmbH and Co KG; 2012:813–823.Google Scholar
- 27.Pantazaka E, Papadimitriou E. Chondroitin sulfate-cell membrane effectors as regulators of growth factor-mediated vascular and cancer cell migration. Biochim Biophys Acta. 1840;2014:2643–50.Google Scholar
- 32.Deepa SS, Umehara Y, Higashiyama S, Itoh N, Sugahara K. Specific molecular interactions of oversulfated chondroitin sulfate E with various heparin-binding growth factors. Implications as a physiological binding partner in the brain and other tissues. J Biol Chem. 2002;277:43707–16.CrossRefPubMedGoogle Scholar
- 33.ten Dam GB, van de Westerlo EM, Purushothaman A, Stan RV, Bulten J, Sweep FC, et al. Antibody GD3G7 selected against embryonic glycosaminoglycans defines chondroitin sulfate-E domains highly up-regulated in ovarian cancer and involved in vascular endothelial growth factor binding. Am J Pathol. 2007;171:1324–33.CrossRefPubMedCentralPubMedGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.