Endorepellin laminin-like globular 1/2 domains bind Ig3–5 of vascular endothelial growth factor (VEGF) receptor 2 and block pro-angiogenic signaling by VEGFA in endothelial cells

Authors

  • Chris D. Willis,

    1. Department of Pathology, Anatomy and Cell Biology, and the Cancer Cell Biology and Signaling, Program, Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, PA, USA
    Current affiliation:
    1. Thomson Reuters, IP & Science, Philadelphia, PA, USA
    Search for more papers by this author
  • Chiara Poluzzi,

    1. Department of Pathology, Anatomy and Cell Biology, and the Cancer Cell Biology and Signaling, Program, Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, PA, USA
    Search for more papers by this author
  • Maurizio Mongiat,

    1. Department of Molecular Oncology and Translational Research, National Cancer Institute, Aviano, Italy
    Search for more papers by this author
  • Renato V. Iozzo

    Corresponding author
    • Department of Pathology, Anatomy and Cell Biology, and the Cancer Cell Biology and Signaling, Program, Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, PA, USA
    Search for more papers by this author

  • C.D.W. and C.P. contributed equally to this work

Correspondence

R. V. Iozzo, Department of Pathology, Anatomy and Cell Biology, Thomas Jefferson University, 1020 Locust Street, Suite 336 JAH, Philadelphia, PA 19107, USA

Fax: +1 215 923 7969

Tel: +1 215 503 2208

E-mail: iozzo@KimmelCancerCenter.org

Abstract

Endorepellin, a processed fragment of perlecan protein core, possesses anti-angiogenic activity by antagonizing endothelial cells. Endorepellin contains three laminin G-like (LG) domains and binds simultaneously to vascular endothelial growth factor receptor 2 (VEGFR2) and α2β1 integrin, resulting in dual receptor antagonism. Treatment of endothelial cells with endorepellin inhibits transcription of VEGFA, the natural ligand for VEGFR2, attenuating the pro-survival and migratory activities of VEGFA/VEGFR2 signaling cascade. Here, we investigated the specific binding site of endorepellin within the ectodomain of VEGFR2. Full-length endorepellin was not capable of displacing VEGFA binding from VEGFR2 and LG3 domain alone did not bind VEGFR2. This suggested different binding mechanisms of the extracellular Ig domains of VEGFR2. Therefore, we hypothesized that endorepellin would bind through its proximal LG1/2 domains to VEGFR2 in a different region than VEGFA. Indeed, we found that LG1/2 did not bind Ig1–3, but did bind with high affinity to Ig3–5, distal to the known VEGFA binding site, i.e. Ig2–3. These results support a role for endorepellin as an allosteric inhibitor of VEGFR2. Moreover, we found that LG1/2 blocked the rapid VEGFA activation of VEGFR2 at Tyr1175 in endothelial cells. In contrast, LG1/2 did not result in actin cytoskeletal disassembly in endothelial cells whereas LG3 alone did induce cytoskeletal collapse. However, LG1/2 did inhibit VEGFA-dependent endothelial migration through fibrillar collagen I. These studies provide a mechanistic understanding of how the different LG domains of endorepellin signal in endothelial cells while serving as a template for protein design of receptor tyrosine kinase antagonists.

Structured digital abstract

[Structured digital abstract was added on 3 May 2013 after original online publication]

Abbreviations
GFP

green fluorescent protein

HIF-1α

hypoxia-inducible factor 1 α

HRP

horseradish peroxidase

LG

laminin-like globular

PAE

porcine aortic endothelial

PAE-VEGFR2

PAE overexpressing VEGFR2

SHP-1

Src homology phosphatase-1

VEGF

vascular endothelial growth factor

VEGFR2

VEGF receptor 2

Introduction

Angiogenesis, or the proliferation of new networks of blood vessels, is an essential process for tumor development and for over 40 years has been considered a target for cancer therapy [1]. In addition to playing an essential role in development, neovascularization allows solid tumors to receive oxygen and essential nutrients while removing waste products to promote tumor progression. Once the ‘angiogenic switch’ is turned on tumor invasion and metastasis are not far behind, ultimately resulting in poor clinical outcomes [2].

The soluble cytokine mainly responsible for the signaling of growth and survival of blood vessels is vascular endothelial growth factor (VEGFA) together with its main transcriptional controller hypoxia-inducible factor 1α (HIF-1α). During abnormal vascular remodeling, the tumor microenvironment experiences hypoxia which stabilizes HIF-1α levels to promote VEGFA expression [3]. Vascular endothelial growth factor receptor 2 (VEGFR2) is the main signaling receptor for VEGFA ligand which binds between the extracellular Ig domains 2 and 3 [4]. The resulting signaling cascade following dimerization of this receptor tyrosine kinase includes endothelial cell proliferation, migration, actin remodeling and vascular permeability [5]. The tumor microenvironment consisting of these recruited stromal cells that line the blood vessels in addition to the neoplastic cells make the pro-growth feedback loop a viable therapeutic target [6].

Several approaches have been taken to block this pathway including neutralizing VEGFA and VEGFR2. In spite of promising results showing anti-angiogenic capabilities in mice, agents targeting VEGFA or VEGFR2 in humans have not been as impressive [7]. Despite initial regressions in tumor size and neovascularization, the blockade of VEGF eventually overcomes the treatment resulting in tumor progression marked by the remodeling of blood vessels [8]. One explanation is the observation that following VEGF blockade tumors begin abnormal accumulation of perlecan among other basement membrane proteins [9].

Perlecan is a ubiquitous, modular heparan sulfate proteoglycan which has also been directly implicated in regulating endothelial cell adhesion [10], chondrogenesis [11], vascular injury and remodeling [12-15], lipid metabolism and fibrosis [16, 17], corneal and epidermal structure [18, 19], and developmental and tumor angiogenesis [20-27]. Indeed, proteoglycans have been associated with cancer biology for over 50 years based on their deregulated expression in various types of cancer [28-31], their propensity to act as depot for cytokines and growth factor, and their ability to present cytokine/growth factor complexes to their cognate receptors in a bioactive form [32-37]. The significance of perlecan regulating the angiogenic switch results from its opposing terminal angiogenic activities [38]. On one end, the N-terminus, which includes three heparan sulfate chains, is responsible for harboring a number of angiogenic growth factors within the basement membrane and extracellular matrix, including FGF2, FGF18, progranulin and VEGFA, in close proximity to their functional receptors [36, 39-41]. Furthermore, 15-fold increases in perlecan mRNA have been observed in melanomas [42], consistent with the overabundance of this pro-angiogenic proteoglycan in the matrix of tumor cells [43]. In contrast, at its C-terminus, perlecan is proteolytically processed by BMP-1 [44] and cathepsin L [45]. The latter enzymatic processing liberates a bioactive angiostatic fragment named endorepellin to designate its intrinsic repulsive activity against endothelial cells [46]. Endorepellin is among a class of liberated C-terminal fragments derived from parental extracellular matrix proteins that along with other angio-inhibitors such as arresten, canstatin, endostatin and tumstatin bind to integrin receptors to regulate intracellular pathways of endothelial and tumor cells [47]. Endorepellin contains three laminin-like globular (LG) domains for which LG3 has been well studied and is known to act in a dominant negative fashion to endothelial cells, similar to full-length endorepellin [48]. Notably, LG3 has an anti-apoptotic effect on fibroblasts [49], suggesting that its activity might be cell-specific.

Previously, we have shown that much of endorepellin's angiostatic activity exists through binding to α2β1 integrin [23, 50-52], a key receptor involved in angiogenesis [53-55]. The interaction between endorepellin and the α2β1 integrin leads to a signaling cascade that induces collapse of the actin cytoskeleton and activation of the tyrosine phosphatase SHP-1, which in turn inactivates several receptor tyrosine kinases, including VEGFR2 [56]. More recently we have put forward a new concept of dual receptor antagonism in which endorepellin interacts with both VEGFR2 and α2β1 in microvascular and macrovascular endothelial cells [57]. This mechanism of VEGFR2 antagonism extends to direct extracellular receptor binding resulting in transcriptional repression of HIF-1α and VEGFA along with inhibition of nuclear factor of activated T-cells 1 (NFAT1) activation [58].

In the present study, we show the distinct domains of endorepellin that specifically engage VEGFR2 as well as the differential mechanisms of anti-angiogenic activity for the two bioactive fragments LG1/2 and LG3. These results further support the emerging theme of dual receptor antagonism as a therapeutically viable mechanism for blocking receptor crosstalk [57, 59].

Results

Laminin G domains 1 and 2 of endorepellin bind specifically and with high affinity to Ig domains 3–5 of VEGFR2

To determine the mechanism of binding of various LG domains of endorepellin at the surface of endothelial cells, we engineered recombinant plasmids to express the different LG and Ig domains of endorepellin and VEGFR2, respectively. These constructs included LG1/2 and LG3 from endorepellin (Fig. 1A) and the immunoglobulin-like repeats Ig1–3 and Ig3–5 from the ectodomain of VEGFR2 (Fig. 1B). All proteins contained a C-terminal His6 tag which allowed for purification through nickel affinity chromatography, whereas the VEGFR2 Ig mutants contained an additional Myc tag just proximal to the His6 tag. Western blotting with a rabbit antibody against endorepellin confirmed the proper expression of LG1/2 and LG3 proteins (Fig. 1C). Similarly, immunoblotting with an antibody against the Myc epitope tag confirmed proper Ig1–3 and Ig3–5 expression (Fig. 1D). Consistently we observed a smeared band suggesting the possibility of post-translational modifications such as glycosylation for these Ig domain proteins. These purified recombinant proteins were then used for the studies presented below.

Figure 1.

Domain architecture for constructs of endorepellin and the ectodomain VEGFR2 as predicted by smart algorithms and immunoblotting analysis. (A) Endorepellin fragments. Two recombinant DNA plasmids were engineered to express endorepellin derivative proteins containing either LG1/2 or LG3. (B) The extracellular ectodomain of VEGFR2 was broken down into smaller fragments containing either Ig1–3 or Ig3–5. Note the different subtypes of Ig domains (see 'Discussion'). Both sets of proteins contain C-terminal His6 tags. In addition, Ig1–3 and Ig3–5 contain Myc epitope tags, immediately upstream to the His6 tag. (C) Western immunoblotting using the rabbit antibody against endorepellin shows proper expression of the purified LG1/2 and LG3 proteins following Ni-nitrilotriacetic acid affinity chromatography at the indicated molecular weights. (D) Western immunoblotting using an anti-Myc antibody against purified Ig1–3 and Ig3–5 proteins.

To test the specificity of the interaction between endorepellin and VEGFR2, we performed blot overlays (far western blots) as an initial qualitative assessment of binding. We separated the Ig1–3 and Ig3–5 proteins by gel electrophoresis and then transferred the recombinant proteins to nitrocellulose membranes. This was followed by overlaying the membranes with soluble LG1/2. This region of endorepellin bound preferentially to Ig3–5 upon detection with a rabbit antibody against endorepellin (Fig. 2A). As a positive control we also performed the same immobilizing ligand step but overlaid the membrane with VEGFA. As expected, VEGFA preferentially bound to Ig1–3 (Fig. 2B) which contains its known binding motif embedded between Ig2 and 3 [5].

Figure 2.

Qualitative confirmation of the specificity of LG1/2 of endorepellin binding for Ig3–5 of VEGFR2. (A) Blot overlays using immobilized Ig1–3 and Ig3–5 showing the greater affinity of recombinant endorepellin LG1/2 for VEGFR2 ectodomain Ig3–5. Following migration, the gel was transferred to nitrocellulose membrane, incubated with LG1/1 and detected with anti-endorepellin (α-ER) antibody, followed by HRP-labeled secondary antibody and chemiluminescence detection. (B) Similar blot overlay protocol as in (A) but overlaid with recombinant VEGFA followed by HRP-labeled secondary antibody and chemiluminescence detection. Notice that VEGFA exclusively binds to Ig1–3, the established binding site with VEGFR2 ectodomain. (C) Co-immunoprecipitation (IP) experiments using Protein A-bound anti-endorepellin antibody and soluble LG1/2 and Ig1–3. The input, precipitate and wash were subjected to immunoblotting (IB) with anti-Myc (α-Myc) to detect Ig1–3. Notice that Ig1–3 does not bind in solution to endorepellin LG1/2. The presence of the IgG heavy chain (Ig-HC) in the precipitate serves as a control for confirmation of coupling to the Protein A beads used. (D) Co-immunoprecipitation of Ig3–5 and LG1/2 using the rabbit antibody against endorepellin (α-ER) antibody for the IP and α-Myc for the IB, as indicated. Notice the ability of Ig3–5 to form a stable complex with endorepellin LG1/2.

To further corroborate these binding assays, we performed co-immunoprecipitation experiments using the same constructs. Using the rabbit antibody against endorepellin bound to beads coated with Protein A, we incubated the complex with LG1/2 followed by addition of either Ig1–3 or Ig3–5. Following washing, we performed western immunoblotting to confirm positive interactions within the precipitates. Notably, Ig3–5 but not Ig1–3 bound to the LG1/2 complex as detected in the precipitate (Fig. 2C,D).

Collectively, these results demonstrate qualitatively that LG domains 1 and 2 of endorepellin bind to the region containing Ig3–5 of VEGFR2.

Next, we set out to determine the binding kinetics associated with this receptor–ligand interaction. Using solid phase ligand binding assays, we used Ig3–5 of VEGFR2 as the immobilized substrate and either LG1/2, LG3 or VEGFA as the soluble ligands. We saw no absorbance increases, consistent with no associated binding interactions, with LG3 or VEGFA (Fig. 3B,C). In contrast, LG1/2 bound immobilized Ig3–5 with high affinity (Fig. 3A). The interaction between Ig3–5 and LG1/2 was saturable at high concentrations of ligand, and produced a dissociation constant (Kd) of 6.6 nm. This apparent Kd is consistent with previously published binding constants of full-length perlecan and endorepellin associating with the full-length ectodomain of VEGFR2 in the low nanomolar range [57]. To further prove specificity of binding, we performed displacement assays using constant molar amounts of LG1/2 (50 nm) and increasing molar amounts of VEGFA. We found that VEGFA did not appreciably displace LG1/2 from Ig3–5 of VEGFR2 (Fig. 3D). These results corroborate those reported above using overlay and co-immunoprecipitation studies and further strengthen the concept that the LG1/2 domains of endorepellin engage Ig domains of VEGFR2 in a different region than VEGFA.

Figure 3.

Affinity interaction between endorepellin LG1/2 and Ig3–5 domains of VEGFR2. (A)–(C) Ligand-binding assays using LG1/2, LG3 or VEGFA as soluble ligands and the Ig3–5 of VEGFR2 receptor as immobilized substrate. (D) Displacement assay using constant molar amounts of LG1/2 (50 nm) and increasing molar amounts of VEGFA as indicated. LG1/2 binds specifically and with high affinity to Ig3–5 of VEGFR2 but neither VEGFA nor LG3 are capable of binding to Ig3–5. VEGFA does not displace LG1/2 from Ig3–5 of VEGFR2 suggesting that LG1/2 domains are binding to VEGFR2 in a different region than VEGFA. The results represent the mean ± SEM of three separate experiments run in triplicate.

LG1/2 blocks receptor activation by VEGFA to attenuate signaling through Tyr1175 phosphorylation and downstream transcriptional activation of VEGFA promoter

Based on the observation that LG1/2 has a different mechanism of binding than VEGFA to the ectodomain of VEGFR2, we hypothesized that LG1/2 might act as an allosteric inhibitor to attenuate the pro-angiogenic signaling of VEGFA. To address this question we used a porcine aortic endothelial (PAE) cell line which overexpresses VEGFR2 (PAE-VEGFR2). These cells have been extensively utilized for various investigations regarding the biology of VEGFA/VEGFR2 and co-receptors such as neuropilin-2 [57, 60-63]. We treated confluent monolayer PAE-VEGFR2 cells with either VEGFA (2.6 nm) or LG1/2 (150 nm) both for 10 min. Additionally, cells were pretreated for 1 h with LG1/2 followed by VEGFA for 10 min. We then probed treated cell lysates for phosphorylation of Tyr1175, a VEGFR2 activation marker that initiates pro-proliferative pathways through PLCγ [64]. VEGFA caused a rapid induction of Tyr1175 phosphorylation whereas LG1/2 had no effect on activation of this residue (Fig. 4A). Notably, pretreatment with LG1/2 followed by the addition of VEGFA completely blocked VEGFA-evoked phosphorylation at Tyr1175 (Fig. 4A). These experiments were repeated three times with similar results.

Figure 4.

Attenuation of VEGFA evoked intracellular Tyr1175 phosphorylation by LG1/2 in endothelial cells. (A) PAE-VEGFR2 cells were serum starved for 2 h followed by treatment with 10 ng·mL−1 VEGFA for 15 min, 150 nm LG1/2 for 15 min, or pretreated for 1 h with LG1/2 and then subjected to VEGFA for 15 min. Cell lysates were prepared for western blotting using an antibody specific for phosphorylation at Tyr1175 of VEGFR2 or total VEGFR2. Notice the robust VEGFR2 activation upon treatment with VEGFA (lane 2), in contrast to LG1/2 alone (lane 3). Upon pretreatment with LG1/2, VEGFA was not capable of inducing phosphorylation of Tyr1175 (lane 4). (B) Luciferase reporter assays assessing the effects of LG1/2 and VEGFA on downstream VEGFA transcription. PAE-VEGFR2VEGFA-Luc cells were serum starved for 2 h, followed by 6 h of VEGFA (10 ng·mL−1), LG1/2 (150 nm), or 1 h pretreatment with LG1/2 followed by 6 h with VEGFA. VEGFA treatment resulted in approximately a twofold induction of VEGFA transcription whereas pretreatment with LG1/2 significantly blunted this effect as shown in column 4; = 10 for each column, ***< 0.001. (C)–(E) Representative fluorescence images of PAE-VEGFR2-GFP cells in the presence or absence of VEGFA or LG1/2 pretreatment followed by VEGFA treatment as indicated. The cells were immunoreacted with anti-Phospho-Tyr1175 (red). Notice the overlap of the signal (white arrows in D) following a 15-min treatment with 10 ng·mL−1 VEGFA. Pretreatment with LG1/2 blocks completely VEGFR2 phosphorylation at Tyr1175 in agreement with the biochemical data shown in (A). Bar ~ 10 μm.

Next, we assessed the effects of LG1/2 ± VEGFA treatment on downstream VEGFA transcription. To this end, we utilized PAE-VEGFR2 stably transfected with a 2.6 kb genomic fragment of the human VEGFA promoter driving firefly luciferase reporter cassette [57, 58]. These PAE-VEGFR2VEGFA-Luc cells were serum starved for 2 h, followed by a 6-h incubation with either VEGFA or LG1/2 or pretreated with LG1/2 for 1 h followed by a 6-h incubation with VEGFA. As expected, VEGFA treatment resulted in a twofold induction of VEGFA transcription (< 0.001, Fig. 4B), whereas pretreatment with LG1/2 markedly blunted this effect (< 0.001, Fig. 4B). In contrast, LG1/2 had no significant effects on VEGFA promoter luciferase activity in these cells (= 0.75, Fig. 4B).

To further prove the antagonistic properties of LG1/2 on VEGFR2 activity, we engineered a new PAE cell line stably overexpressing VEGFR2 tagged with green fluorescent protein (GFP). Upon using the same treatment protocol as in Fig. 4A, PAE-VEGFR2-GFP cells were fixed and subjected to immunohistochemistry to probe for Tyr1175. Under basal conditions, VEGFR2 molecules were not actively phosphorylated at Tyr1175 as demonstrated by undetectable immunological levels (lack of fluorescent red staining, Fig. 4C). Upon treatment with VEGFA, Tyr1175 phosphorylation was markedly induced (Fig. 4D) as shown by the formation of yellow overlapping immune deposits (white arrows in Fig. 4D). This VEGFA-evoked induction of Tyr1175 phosphorylation was markedly suppressed by pretreatment with LG1/2 (Fig. 4E). Collectively, these results indicate that binding of LG1/2 to the ectodomain of VEGFR2 blocks intracellular receptor activation at Tyr1175 by the natural ligand VEGFA.

LG1/2 affects VEGFA-driven endothelial cell migration but not actin cytoskeleton dynamics

We have previously shown that endorepellin and its C-terminal LG3 module induce endothelial cell disassembly of actin cytoskeleton and focal adhesions and this process is mediated by a specific interaction with the α2β1 integrin [50]. Blocking monoclonal antibodies targeting the I domain of the α2 integrin subunit, the site where endorepellin and LG3 bind, or small interfering RNA (siRNA) mediated knockdown of the α2 integrin subunit block this process completely [44, 56]. Given the binding results shown above for LG1/2 versus LG3, we hypothesized that LG1/2 would not cause dissolution of the actin cytoskeleton. Control cells treated with vehicle showed intact actin filaments similar to those observed with LG1/2 treatment (Fig. 5C). In contrast, both endorepellin and LG3 treatment caused the breakdown of actin stress fibers and collapsing of the actin filaments into large irregular aggregates (white arrows in Fig. 5B and D, respectively). Notably, quantification of 50 cells for each treatment showed that endorepellin and LG3 caused a significant decrease in stress fibers per cell (< 0.001, Fig. 5E), in contrast to LG1/2 (= 0.85, Fig. 5E).

Figure 5.

Diverse functional effects of LG1/2 on actin disassembly and endothelial cell migration. (A–D) Representative images of PAE-VEGFR2 cells treated with vehicle (control), endorepellin, LG1/2 or LG3 (150 nm each for 2 h) as indicated. Notice the collapse of actin stress fibers evoked by endorepellin or LG3 (white arrows in B and D, respectively). In contrast, LG1/2 has no appreciable effects (C). Confluent four-chamber slides of PAE-VEGFR2 cells were serum starved for 2 h, treated, fixed and stained with rhodamine-phalloidin along with DAPI to show the actin cytoskeletal structure. Bar ~ 20 μm. (E) Quantification of actin stress fibers per cell. The values represent the mean ± SEM with 50 cells per condition (***< 0.001). (F) Quantification of PAE-VEGFR2 cell migration through a fibrillar collagen I as evoked by VEGFA (40 ng·mL−1 in the lower chamber). The values represent the mean ± SEM of three independent experiments and are expressed as the average number of migrated cells as a percentage of control (***< 0.001).

Next we performed endothelial cell migration assays through a fibrillar collagen type I matrix. In this case, all the constituents of endorepellin evoked a marked suppression of VEGFA-evoked endothelial cell migration (< 0.001, Fig. 5F). Thus, the two functions embedded within endorepellin can be functionally dissociated, one in LG1/2 that affects VEGFR2 and one in LG3 that affects the biology of the α2β1 integrin receptor. This dual receptor antagonism provides a mechanistic explanation for the multiple biological activities of endorepellin and related compounds.

Discussion

The localization and large modular nature of perlecan within basement membranes underlying epithelial and endothelial cells makes this proteoglycan a regulator of key biological processes within the tumor microenvironment. Five distinct domains exist within perlecan and the importance of this molecule is highlighted by the ubiquitous extracellular matrix expression across tissue and cell types [65-67] as well as genetic studies showing death during the 10–12 day embryonic stage due to cardiac tamponade and cardiovascular defects [68, 69]. The C-terminal module endorepellin also allows for cell surface targeting of the parent proteoglycan due to high affinity interactions with the α2β1 integrin receptor [47] and VEGFR2 [57]. We recently discovered that perlecan knockdown causes a paradoxical increase and distribution of VEGFA in zebrafish and that the perlecan morphants could be partially rescued by microinjections of VEGFA [24]. Consistent with these findings, combined administration of perlecan and VEGFA to human umbilical vein endothelial cells enhances VEGFR2 phosphorylation at Tyr951 [24]. Similarly, a soluble form of perlecan domain I harboring heparan sulfate chains enhances VEGFA activity on VEGFR2 Tyr951 [70]. Collectively, these results support a dual role for perlecan as a regulator of ‘proper’ VEGFA targeting to endothelial cells and as a central mediator of signaling through VEGFR2. We have formulated a hypothesis in which the anti-angiogenic therapeutic potential of endorepellin derives from the ‘dual receptor antagonism’ exhibited through intracellular endothelial cell signaling [58] and rapid receptor internalization and degradation upon treatment [57]. Important for specific targeting of endorepellin is the fact that endothelial cells are the only cell types to express these two receptors, while crosstalk between integrins and receptor tyrosine kinases is a key step in angiogenesis through controlling cell migration and proliferation [71]. The proof of principle for our novel concept of dual receptor antagonism has been recently provided by a study which has developed a single chain VEGF (scVEGF) mutant that acts as dual-specific antagonist for both VEGFR2 and αVβ3 integrin [59]. This dual scVEGF mutant concurrently binds to both receptors with antibody-like affinity, similar to endorepellin binding affinity in the low nanomolar range. Importantly, when compared with monospecific scVEGF, the dual-specific scVEGF more efficiently inhibits VEGF-mediated receptor phosphorylation, capillary tube formation and in vivo angiogenesis [59]. These data provide robust support to the hypothesis that recombinant proteins with dual affinity for VEGFR2 and an angiogenic integrin receptor could be biologically active similar to a portion of perlecan that has evolved for over 500 million years.

Endorepellin also activates the Tyr phosphatase SHP-1 by enhancing its recruitment to the intracellular domain of the α2β1 integrin [56]. SHP-1 then dephosphorylates several receptor tyrosine kinases, including VEGFR2, thereby blocking pro-angiogenic endothelial cell signaling through migratory, survival and proliferative pathways. This dual receptor interaction leads to a rapid recruitment of these receptors and to their internalization and degradation, which together with the deactivation of VEGFR2 by SHP-1 causes transcriptional repression of VEGFA gene and attenuation of VEGFA protein production and secretion [57, 58]. The repression of VEGFA transcription would further contribute to the overall anti-angiogenic activity of endorepellin through inhibition of the positive feedback loop within the tumor stroma. The signaling mechanisms associated with anti-angiogenic activity by the LG domains of endorepellin are summarized in Fig. 6. Notably, TIMP-2, another angiostatic protein [72], attenuates VEGFA-evoked activation of VEGFR2 [73] and activates SHP-1 [74], all properties shared by endorepellin/LG1/2.

Figure 6.

Mechanism of endorepellin dual receptor antagonism and block of angiogenesis. LG domains 1 and 2 of endorepellin bind to VEGFR2 between Ig domains 3–5, distal to the known binding site of VEGFA (between Ig1–3). The interaction between endorepellin and VEGFR2 inhibits activation of VEGFR2 by VEGFA as verified experimentally through Tyr1175 phosphorylation. Downstream results of this inhibition include blockage of endothelial cell migration and transcription of VEGFA, both of which contribute to angiostasis. The effects of the LG3 domain of endorepellin also contribute to anti-angiogenic signaling in this pathway through direct binding to α2β1 integrin. Anti-angiogenic signaling of LG3 involves the activation of SHP-1, which targets Tyr1175 of VEGFR2, as well as downstream collapse of the actin cytoskeleton through previously identified intracellular regulators. An additional contribution to angiostasis is the endorepellin-evoked dual receptor internalization and degradation via a caveosome-mediated pathway.

The mechanism of VEGFA activation of VEGFR2 has been well studied, and through mutational analysis the second and third Ig domains were identified [5]. Furthermore, electron microscopy has confirmed that VEGFA induces VEGFR2 dimerization through the membrane-proximal Ig domain 7 [4]. The present work extends previous findings showing the biological relationship between endorepellin and the VEGFA/VEGFR2 axis [24, 75]. By creating fusion constructs containing different regions of the Ig domains of VEGFR2, we were able to show that endorepellin binds to a different region, more distal than the endogenous pro-angiogenic signaling ligand VEGFA. In particular, it is the LG1/2 region that shows the specificity of binding Ig domains 3–5. The LG1/2 region of endorepellin has not been studied in as much detail as the angiostatic fragment LG3 and here we define a novel binding mechanism for this region. Important to note is the measured low nanomolar affinity of LG1/2 compared with the low picomolar binding of VEGFA [5]; thus, LG1/2 should not physiologically compete for VEGFR2 ectodomain binding. Interesting to consider also for future studies, the LG2 module has been shown to bind to endostatin [46], another C-terminal angiostatic fragment processed from the parental proteoglycan collagen XVIII. In addition, endostatin binds to VEGFR2 [76] while the anti-angiogenic activities of endorepellin and endostatin are neutralized in the presence of each other [46]. Therefore it will be important to compare the circulating concentrations during normal versus disease states of these two fragments to assess their potential bioactivity.

As mentioned above, blockade of Tyr1175 phosphorylation induction of VEGFR2 has been shown to result from endorepellin evoking an increase in SHP-1 tyrosine phosphatase activity [56]. Here we present a second layer of Tyr1175 regulation as we show that LG1/2 is a potent inhibitor of VEGFA activation to mediate downstream signaling pathways. Tyr1175 is a key regulator of VEGFR2 signaling and is essential for proper vasculogenesis [77]. Phosphorylation of this key Tyr residue regulates a number of downstream activities by mediating the binding of adaptor proteins such as SHB (SH2-domain containing adaptor protein), Sck [SHC (Src homology and collagen homology) related adaptor protein], SHC and GRB2 (growth factor receptor bound protein 2) [64, 78]. Thus, intervention with LG1/2 could represent an endogenous biologic to block these pathways with less potential to initiate an immune response.

In our study we have been able to dissociate the bioactivity of endorepellin and its two main fragments on VEGFR2 and α2β1 integrin. Indeed LG1/2 was able to inhibit VEGFA-evoked endothelial cell migration through fibrillar collagen I, which resembles the provisional matrix formed during in vivo endothelial cell sprouting [43]. In contrast, LG1/2 was not able to induce dissolution of actin cytoskeleton, a property that was specifically mediated by the interaction of LG3 with α2β1 integrin. The specificity of the different LG domains for their functional receptors is shown in Fig. 6 with the downstream signaling cascades ultimately converging on anti-angiogenesis, thus highlighting the role of endorepellin as a dual receptor antagonist for endothelial cells.

In conclusion, in our current working model we propose that endorepellin would act as an allosteric inhibitor of VEGFR2 by binding to a region distal to Ig2–3 where VEGFA engages VEGFR2 [79]. We demonstrate here that this binding occurs via the two proximal LG1/2 domains, whereas LG3 would bind to the α2β1 integrin. We provide a new paradigm for anti-angiogenic fragments derived from large precursors, i.e. a dual receptor antagonism. We predict that a similar bioactivity could be operational for other processed forms of angiostatic matrix molecules acting as intrinsic microenvironmental barriers to tumorigenesis [80]. Indeed, powerful anti-angiogenic proteins such as thrombospondin-1 and endostatin bind both VEGFR2 and at least one integrin [76, 81, 82]. Thus, the paradigm of dual receptor antagonism may be the future framework for novel cancer therapeutics to maximize angiostatic potency.

Materials and methods

Antibodies

The following primary antibodies were used in this study: rabbit monoclonal anti-Phospho-VEGFR2 (Tyr1175), rabbit anti-hVEGFR2 and mouse monoclonal IgG (light-chain specific) were from Cell Signaling Technology Inc. (Danvers, MA, USA); mouse monoclonal anti-Myc conjugated to horseradish peroxidase (HRP) was from Invitrogen (Carlsbad, CA, USA). The rabbit antibody against endorepellin (α-ER) was described previously [50]. The secondary HRP-conjugated goat anti-rabbit antibody was from Millipore (Billerica, MA, USA) and the goat anti-rabbit Alexa Fluor568 used was from Invitrogen.

Generation of recombinant DNA constructs

Endorepellin fragments LG1/2 and LG3 were cloned into the NheI and XhoI sites of pCep-Pu with the BM40 signal peptide as previously described [46]. The DNA template for human VEGFR2 gene was obtained from Addgene (plasmid 23925) as described before [83]. Ig domains 1–3 of VEGFR2 ectodomain were amplified with the following primers: forward 5′- CCCCTAAGCAGCTTGCCCAGGCTCAGCATACA-3′; reverse 5′-AATGGGTGACCTCGAGAGACCCTGACAAATGTGCT-3′. Ig domains 3–5 were amplified using forward 5′-CCCCTAAGCAGCTTGCTGAGTCCGTCTCATGG-3′ and reverse 5′-AATGGGTGACCTCGAGAGGTCACGTGGAAGGAGAT-3′. The PCR inserts were gel extracted with a kit from Qiagen (Valencia, CA, USA) and quantified with a NanoDrop 2000 (Thermo Scientific, Wilmington, DE, USA). The destination mammalian expression vector used was pcDNA 3.1 Myc-His C (Invitrogen) which was digested with KpnI and XhoI restriction enzymes New England Biolabs Inc. (Beverly, MA, USA). This vector also had the BM40 signal peptide engineered into the HindIII upstream restriction site. Digested vector and purified inserts were ligated using the Infusion Kit from Clontech which uses the overhangs built into the primers for directional specificity. The same KDR template was also used to amplify full-length hVEGFR2 with the following primers: forward 5′-ATTCTGCAGTCGACGATGCAGAGCAAGGTGCTGCTG-3′; reverse 5′-GATCCCGGGCCCGCGGAACAGGAGGAGAGCTCAGTGTG-3′. The purified insert was ligated into pDsRed1-n1 (Clontech, Palo Alto, CA, USA) which was digested with Kpn1 and cloned in frame with the C-terminal red fluorescent protein (RFP) tag. The RFP tag was then swapped out for GFP using inverse PCR. All constructs were sequenced at the Kimmel Cancer Center, Cancer Genomics Shared Resource Facility at Thomas Jefferson University.

Protein production and purification

293F suspension cells (Invitrogen) were transiently transfected when concentrations reached ~ 106 cells·mL−1 and cells were > 90% viable as determined with Trypan Blue (Sigma-Aldrich, St Louis, MO, USA). In general, 60 μg of DNA was delivered via Lipofectin-based 293fectin (Invitrogen) to ~ 3 × 107 cells and the medium containing recombinant protein was collected and sterile filtered through a 0.22 μm filter unit (Millipore) 72 h post transfection. The conditioned medium was dialyzed followed by overnight 4 °C rotating incubation with Ni-nitrilotriacetic acid agarose beads (Qiagen). The beads were washed with gravity chromatography followed by a 500 mm imidazole elution step. Positive fractions as detected with Coomassie staining (Bio-Rad, Richmond, CA, USA) were collected and dialyzed against NaCl/Pi. All steps included 2 mm phenylmethylsulfonyl fluoride to inhibit proteases and final yields were quantified at 10–20 mg protein·mL−1 of medium. Highly concentrated recombinant human VEGFA (VEGF165) was obtained from the NIH repository.

Blot overlay assays

Approximately 1 μg of purified Ig1–3 or Ig3–5 protein was loaded into each well and subjected to SDS/PAGE. Gels were transferred to nitrocellulose membranes (Bio-Rad) and stained with Ponseau Red to confirm equal loading amounts. Membranes were incubated in overlay buffer (50 mm Tris, pH 8, 120 mm NaCl, 1% BSA, 2 mm dithiothreitol, 0.5% Nonidet P-40, 0.1% Tween 20) overnight at 4 °C. The following day, 1 μg of LG1/2 or VEGFA was added to fresh overlay buffer and allowed to incubate rocking with the blot at room temperature for 4 h. Membranes were washed at room temperature in clean overlay buffer three times for 5 min and were then washed three times for 5 min in Tris-buffered saline (NaCl/Tris) with 0.1% Tween 20 (NaCl/Tris/Tween) and 1% BSA. Following the washes, the membranes were blocked overnight at 4 °C in NaCl/Tris/Tween supplemented with 5% BSA. Primary antibodies were diluted in NaCl/Tris/Tween with 1% BSA and incubated at room temperature for 4 h. Following another washing step, membranes were either detected using chemiluminescence or incubated in secondary antibody diluted in NaCl/Tris/Tween 1% BSA before being developed using SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific), an enhanced chemiluminescence substrate for detection of HRP enzyme on an image Quant LAS 4000 (GE Healthcare, Piscataway, NJ, USA).

Co-immunoprecipitation

Magnetic beads coated with Protein A-Sepharose from GE Healthcare were washed with NaCl/Tris supplemented with protease inhibitors (Thermo Scientific) and incubated while rotating overnight with 2 μg rabbit primary antibody against endorepellin at 4 °C. The following day, beads were washed three times with the NaCl/Tris solution followed by overnight 4 °C incubation with 2 μg LG1/2. Beads were once again washed three times and then incubated overnight at 4 °C with 2 μg of Ig1–3 or Ig3–5. Following three washes, the beads were boiled in reducing buffer for 5 min. Supernatants were subjected to SDS/PAGE and, following separation and transfer to nitrocellulose, probed with the indicated antibodies. Blots were developed using the same protocol as the blot overlays.

Solid-phase binding assays

Ligand-binding assays were performed following a sandwich ELISA protocol essentially as described before [57, 84] with minor modifications. Briefly, VEGFR2 Ig3–5 ectodomain (100 ng·well−1) was allowed to adhere overnight at 4 °C in the presence of carbonate buffer, pH 9.6. Plates were washed with NaCl/Pi, blocked with 1% BSA, and then incubated for 2 h with serial dilutions of LG1/2, LG3 or VEGFA. In the quantitative competition experiments, LG1/2 was kept at constant concentration (50 nm) and incubated with increasing molar amounts of recombinant VEGFA. After ligand incubation, plates were extensively washed with NaCl/Pi, blocked with 1% BSA, and incubated with primary and HRP-conjugated secondary antibodies. The immune complexes were revealed using SIGMA-FAST™ O-phenylenediamine dihydrochloride. Absorbance at 490 nm was measured in a VICTOR3 multilabel reader (Perkin Elmer Life Sciences, Waltham, MA, USA).

Western blotting and stable transfection

Following treatment, PAE cells were washed and then lysed in RIPA buffer (50 mm Tris, pH 7.4, 150 mm NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mm EDTA, 1 mm EGTA, 1 mm Na3VO4, 10 mm β-glycerophosphate), with a protease inhibitor (1 mm phenylmethanesulfonyl fluoride). Cell lysates were collected and boiled in reducing buffer along with 5 × sample buffer. Supernatant lysates were separated by SDS/PAGE, transferred to nitrocellulose and probed and developed as described before [85]. For stable transfection, PAE cells were electroporated using a Gene Pulser II (Bio-Rad). Parental PAE cells were washed, trypsinized, spun down and resuspended in a total volume of 500 μL which included 2 μg VEGFR2-GFP DNA per 106 cells. Dulbecco's modified Eagle's medium (DMEM) was added to a 0.4 cm electrode gap cuvette (Bio-Rad) and a single pulse of 200 V was applied with a 960 Ohms resistance resulting in a 25 ms pulse time. Transfected cells were re-plated and fluorescence activated cell sorting was used to select the top 5% of GFP-expressing cells 72 h after transfection to use for downstream experiments.

Immunofluorescence microscopy

Immunofluorescence microscopy was performed using a Leica DM5500B microscope. Approximately 5 × 104 PAER2-eGFP cells were plated on four-well chamber slides (BD Biosciences, San Jose, CA, USA) that were coated with 0.2% gelatin from Sigma. Cells were grown to confluence in DMEM supplemented with 10% fetal bovine serum at 37 °C and 5.0% CO2 followed by switching the DMEM to serum-free 2 h prior to each treatment. Slides were washed with DMEM and fixed with ice-cold 4% paraformaldehyde for 10 min. Fixed cells were washed with ice-cold NaCl/Pi followed by permeabilization with 0.1% Tween 20 for 2 min. Following another washing step, fixed cells were blocked overnight with NaCl/Pi containing 3% BSA. Both primary and secondary antibodies (4 h and 1 h incubations, respectively, at 25 °C) were diluted in NaCl/Pi with 1% BSA. Following another wash step, slides were mounted with Vectashield from Vector Laboratories (Burlingame, CA, USA) and images were processed using leica application suite, Advanced Fluorescence software (version 1.8; Leica Microsystems Inc., Wetzlar, Germany).

Luciferase reporter assays

We have recently described the mass culture of PAE-VEGFR2VEGFA-Luc cells which contain a 2.6 kb genomic fragment of the human VEGFA promoter [57, 58] driving firefly luciferase reporter cassette. These cells were grown to confluence in 24-well dishes, serum starved for 2 h, treated, washed with NaCl/Pi, and then lysed in a buffer containing 50 mm potassium phosphate buffer, 2% Triton X-100, 20% glycerol and 4 mm dithiothreitol. Cells were lysed at 25 °C for 10 min and centrifuged at 2000 g for 2 min, and ~ 100 μL of cleared cell lysate was dispensed into a 96-well ELISA plate, together with 100 μL luciferase assay buffer (100 mm potassium phosphate, 2 mm dithiothreitol, 8 mm MgSO4, 175 μm coenzyme A, 750 μm ATP) and 0.5 mm d-luciferin. Luciferase activity was measured using a VICTOR3 multilabel reader (Perkin Elmer Life Sciences) and normalized on total cell number.

Actin staining and endothelial cell migration through fibrillar collagen

Actin staining with rhodamine-phalloidin (Life Technologies, Gaithersburg, MD, USA) was performed as described before [50]. For migration assays, ~ 105 PAE-VEGFR2 cells were pre-incubated for 30 min with either vehicle, endorepellin (150 nm), LG1/2 (150 nm) or LG3 (150 nm) and allowed to migrate through 8-μm nucleopore, polyvinylpyrrolidine-free polycarbonate filters (Corning, Cambridge, MA, USA), pre-coated overnight with 100 μg·mL−1 fibrillar collagen type I. The bottom chambers contained as a chemoattractant recombinant VEGFA (40 ng·mL−1) in DMEM supplemented with BSA (50 μg·mL−1). Migration was allowed to occur for 4–5 h at 37 °C. The filters were stained/fixed with 0.1% crystal violet in 20% ethanol (v/v) for 30 min. Cells from the top layer of the filters were then scraped off by using a cotton swab. Images of the transmigrated cells were acquired using a Leica DMIL LED microscope equipped with a Leica D-LUX3 camera and processed for further analysis. Quantification of the migrated cells was performed as described before [46].

Statistical analysis

Each experiment was repeated three times or more with comparable patterns of responses. All data were expressed as means ± SEM. Results were statistically analyzed with the Student's t-test or paired t-test using sigma-stat software 11.0 (SPSS Inc.) as described before [85, 86]. < 0.05 was considered statistically significant.

Acknowledgements

We would like to thank L. Claesson-Welsh (Uppsala University, Sweden) for generously providing the PAE-VEGFR2 cells, E. Andreuzzi for help with the cloning of LG1/2, and J. Casulli for excellent help with the illustrations. This work was supported in part by the National Institutes of Health grants RO1 CA39481, RO1 CA47282 and RO1 CA120975 (to R.V.I.). C.D.W. was supported by National Institutes of Health training grants T32 AR060715.

Ancillary

Advertisement