Heparan Sulfate Regulates VEGF₁₆₅- and VEGF₁₂₁-mediated Vascular Hyperpermeability

Abstract

VEGF was first described as vascular permeability factor, a potent inducer of vascular leakage. Genetic evidence indicates that VEGF-stimulated endothelial proliferation in vitro and angiogenesis in vivo depend on heparan sulfate, but a requirement for heparan sulfate in vascular hyperpermeability has not been explored. Here we show that altering endothelial cell heparan sulfate biosynthesis in vivo decreases hyperpermeability induced by both VEGF₁₆₅ and VEGF₁₂₁. Because VEGF₁₂₁ does not bind heparan sulfate, the requirement for heparan sulfate suggested that it interacted with VEGF receptors rather than the ligand. By applying proximity ligation assays to primary brain endothelial cells, we show a direct interaction in situ between heparan sulfate and the VEGF receptor, VEGFR2. Furthermore, the number of heparan sulfate-VEGFR2 complexes increased in response to both VEGF₁₆₅ and VEGF₁₂₁. Genetic or heparin lyase-mediated alteration of endothelial heparan sulfate attenuated phosphorylation of VEGFR2 in response to VEGF₁₆₅ and VEGF₁₂₁, suggesting that the functional VEGF receptor complex contains heparan sulfate. Pharmacological blockade of heparan sulfate-protein interactions inhibited hyperpermeability in vivo, suggesting heparan sulfate as a potential target for treating hyperpermeability associated with ischemic disease.

Introduction

The endothelium lining blood vessels forms an essential selective barrier that regulates the passage of plasma macromolecules between blood and tissues. The barrier function of the endothelium is well maintained under normal physiological conditions, but ischemia and inflammation can lead to its breakdown and vascular hyperpermeability (1, 2). The primary molecular event that initiates hyperpermeability involves binding of VEGF to VEGF receptor complexes, resulting in VEGF receptor phosphorylation and activation of Src, endothelial NOS (eNOS),² and other downstream signal transduction proteins (3). Expression of VEGF is rapidly induced under ischemic conditions by activation of the transcription factor, hypoxia-inducible factor 1 α (4, 5). Many types of tumors constitutively express VEGF as well, and the enhanced vascular leakage in tumors increases the risk of metastasis and raises tumor interstitial pressure while reducing the efficacy of chemotherapeutic drug delivery (6–11). Thus, understanding the mechanism by which VEGF interacts with its receptors might provide insights into new therapeutic strategies for the treatment of ischemic disease or cancer.

Three subgroups of cell surface proteins bind VEGF. The first group is composed of the receptor tyrosine kinases, VEGF receptor-1 (VEGFR1), VEGFR2, and VEGFR3. The second group includes neuropilin-1 (Nrp1) and Nrp2, which lack intrinsic kinase activity. The last group consists of one or more heparan sulfate proteoglycans (HSPGs), which consist of a core protein and one or more covalently attached heparan sulfate chains. Various combinations of these receptors form the functional signaling complexes in endothelial cells lining the blood and lymphatic vasculature. In general, VEGFR2 plays the central role in mitogen-activated signaling in blood endothelial cells. Nrp1 and one or more HSPGs are thought to act as co-receptors to promote more efficient binding of the ligand to VEGFR2 and enhance the stability of the receptor complex (3).

All VEGF-A isoforms (VEGF₁₂₁, VEGF₁₄₅, VEGF₁₆₅, VEGF₁₈₉, VEGF₂₀₆) except VEGF₁₂₁ also bind to heparan sulfate or to heparin, a highly sulfated form of heparan sulfate (12, 13). Heparin augments the interaction of VEGF₁₆₅ with VEGFR2 (14) and binds to VEGFR1, Nrp1, and Nrp2 (15–17). Because heparin/heparan sulfate can interact with both ligands and receptors, it might serve as a template to facilitate approximation of the components, similar to its role in the assembly of FGF-FGF receptor complexes (18, 19) and antithrombin-thrombin complexes (20, 21). Evidence for the co-receptor activity of heparan sulfate proteoglycans in VEGF signaling has been provided in the context of tumor angiogenesis and vasculogenesis (22–24), where heparan sulfate potentiates the duration and magnitude of VEGF signaling and regulates the localization and quantity of VEGF-VEGFR2 complexes in the tumor vasculature. VEGF₁₂₁ does not bind heparan sulfate except at low pH (24, 25), but heparan sulfate apparently can regulate the interaction of VEGF₁₂₁ to VEGFR1 (26), possibly by binding independently to the receptor. In contrast, binding of VEGF₁₂₁ to VEGFR2 does not appear to be modulated by heparan sulfate or heparin (27).

In the present study, we investigated the role of heparan sulfate in the VEGF-induced vascular hyperpermeability, which is mediated by VEGFR2. We found that mice deficient in endothelial heparan sulfate had reduced dermal and tumor vascular permeability, providing the first genetic evidence that heparan sulfate is directly involved in vascular leakage in vivo. Surprisingly, the deficiency of heparan sulfate not only affected hyperpermeability induced by VEGF₁₆₅ but also equally affected responses mediated by VEGF₁₂₁. Mechanistically, we identified a direct interaction between heparan sulfate and VEGFR2 in situ, strongly suggesting that heparan sulfate regulates the assembly of functional VEGF receptor complexes. Blockade of heparan sulfate by the heparin-neutralizing agent, surfen (28), significantly attenuated VEGF-induced dermal vascular hyperpermeability, thus providing pharmacological evidence that altering heparan sulfate-VEGF receptor interactions might prove beneficial for treatment of disorders associated with ischemia, such as stroke, myocardial infarction, or cancer.

EXPERIMENTAL PROCEDURES

Reagents

Mouse VEGF₁₆₄ (catalogue number 450-32) and VEGF₁₂₁ (catalogue number 100-20A) from PeproTech were used for the Miles assay (32) and in vitro signaling studies. Human VEGF₁₆₅ (catalogue number 100-44) and VEGF₁₂₁ (catalogue number 100-116) from Shenandoah Biotechnology were also used in some experiments. Histamine was from Calbiochem, and protamine sulfate and Evan's Blue were from Sigma. Surfen (bis-2-methyl-4-amino-quinolyl-6-carbamide) was obtained from the Open Chemical Repository in the Developmental Therapeutic Program at the NCI, National Institutes of Health (NSC12155). Antibodies to phospho-Tyr-1175-VEGFR2 (catalogue number 2478), VEGFR2 (catalogue number 2497), phospho-Ser-1177-eNOS (catalogue number 9571), and eNOS (catalogue number 9572) were from Cell Signaling. Monoclonal antibody to heparan sulfate (H1890) from U. S. Biological and goat polyclonal antibody to mouse VEGFR2 (catalogue number AF644) from R&D systems were used for proximity ligation assays. Polyclonal antibodies against VEGFR2 from Santa Cruz Biotechnology (catalogue number sc-48161) were used for immunoprecipitation. FITC-conjugated antibodies to endothelial markers CD31 (catalogue number 11-0311), CD34 (catalogue number 11-0341), and phycoerythrin-conjugated antibodies to CD105 (catalogue number 12-1051) and CD166 (catalogue number 12-1661) were from eBioscience. Heparin lyases I, II, and III were generous gifts from Dr. Jian Liu (University of North Carolina at Chapel Hill).

Mutant Mouse Strains

Ndst1^f/fTie2Cre⁺ mice were generated and maintained as described previously (29). Vegfr2^+/− mice were purchased from The Jackson Laboratory. Ext1^f/+Tie2Cre⁺ mice were generated by crossing Ext1^f/f mice (30) to Tie2Cre transgenic mice (31). All animals were fully backcrossed on C57Bl/6 background.

Miles Assay

For the Miles assay (32), mice (6–10 weeks of age) were injected intravenously with 200 μl of 0.5% Evan's Blue, and 15 min later 200 ng of VEGF in 20 μl of PBS with 0.1% BSA was injected intradermally into shaved dorsal skin. An equal volume of PBS-BSA was injected into a separate dorsal skin patch in each mouse as a negative control. Histamine (1 μg) in PBS-BSA was used as positive control. Mice were sacrificed 40 min after VEGF injection, and a 6-mm diameter skin biopsy was obtained at the injection site using a dermal punch (Acuderm, Inc.). The sample was then extracted with 1 ml of formamide at 55 °C for 24 h. Evan's Blue was quantified by measuring absorbance at 605 nm and converted to mass by using a standard curve.

Protamine inhibition was measured by intradermal injection (20 μl) of a mixture of 12.5 μg of protamine and 200 ng of VEGF in PBS-BSA. When surfen was tested, 18 μl of 60 μm surfen (in PBS-BSA and 0.2% DMSO) was mixed with 2 μl of PBS-BSA containing 200 ng of VEGF. As a positive control, 18 μl of PBS-BSA with 0.2% DMSO was mixed with 2 μl of PBS-BSA containing 200 ng of VEGF.

LLC Tumor Vascular Permeability Model

Lewis lung carcinoma cells (4 × 10⁵ cells, LL/2 line from ATCC) were inoculated subcutaneously into the hindquarters of 12–14-week old mice. Mice were used for experiments when tumors first became visible, 9–14 days after tumor inoculation. Subcutaneous Lewis lung carcinoma cells (LLC) tumors with an average volume of 50–100 mm³ (∼50 mg of wet weight) were processed and stained with anti-CD31 mAb, and microvascular density was measured as described previously (23). These small, early tumors were vascularized to the same extent in all animals regardless of genotype. Mice were injected intravenously with 200 μl of 0.5% Evan's Blue and sacrificed 1 h later. Tumors were carefully removed without any attached skin, weighed, and extracted in 0.5 ml of formamide solution at 55 °C for 24 h. Extracted dye was quantitated as described above in the Miles assay. For the assessment of VEGF expression, tumors were homogenized in 1 ml of radioimmune precipitation buffer (100 mm Tris, 150 mm NaCl, 1% Triton X-100, 0.1% SDS, pH 7.4) containing a protease inhibitor mixture, and 20 μg of tumor homogenate was resolved by SDS-PAGE and immunoblotted with a rabbit polyclonal antibody to VEGF (Abcam, catalogue number ab46154).

Isolation of Primary Mouse Microvascular Endothelial Cells

Murine brain microvascular endothelial cells (mBMEC) were isolated from cerebral cortex essentially as described (33), except that cells were selected with 5 μg/ml puromycin for 4 days in DMEM medium (Invitrogen) supplemented with 20% FBS (Atlanta Biologicals), 50 μg/ml endothelial growth supplement (BTI, Inc.), 50 μg/ml heparin, nonessential amino acids, penicillin, and streptomycin. Cells generally reached confluence in 7 days and were used without passage. Murine lung microvascular cells were isolated and cultured as described previously (23). Disaccharide analysis of heparan sulfate derived from the cells was determined by liquid chromatography/mass spectrometry as described (34).

VEGF Binding Assay

Murine lung endothelial cells were treated with (or without) 10 milliunits/ml heparan lyase I for 20 min at room temperature. Endothelial cells (2 × 10⁵) were incubated with VEGF₁₆₄ and VEGF₁₂₁ (40 ng) for 1 h at 4 °C. After rinsing, the cells were incubated with 1 μg/ml biotinylated rabbit anti-hVEGF (catalogue number 500-P10Bt, PeproTech) for 1 h at 4 °C. Cells were stained with streptavidin-phycoerythrin (eBioscience) at a dilution of 1:400 and analyzed by flow cytometry.

Immunoblotting

mBMEC were serum-starved in DMEM for 5 h prior to treatment with VEGF. In some experiments, mBMEC were incubated with a mixture of heparin lyases I (2 milliunits/ml), II (2 milliunits/ml), and III (5 milliunits/ml) in serum-free DMEM at 37 °C for 15 min. Cells were then stimulated with 20 ng/ml mVEGF₁₆₄ or 50 ng/ml hVEGF₁₂₁ at 37 °C for 2 or 15 min. Cells were lysed in radioimmune precipitation buffer (Cell Signaling) with a protease and phosphatase inhibitor mixture (Pierce). Samples (10 μg) were resolved on 4–12% Bis-Tris NuPAGE gel and blotted onto polyvinylidene difluoride membrane. After incubation with primary and secondary antibodies, reactive bands were visualized by SuperSignal West Femto (for anti-phospho-Tyr-1175-VEGFR2 and anti-phospho-Ser-1177-eNOS mAbs) or West Pico (for all other antibodies) chemiluminescent substrate (Pierce).

Proximity Ligation Assay

mBMEC were treated with heparan lyases as described above in growth medium lacking serum. Cells were then stimulated with 20 ng/ml mVEGF₁₆₄ or 50 ng/ml hVEGF₁₂₁ at 37 °C for 2, 5, and 15 min followed by fixation in 2% paraformaldehyde and blocking with 5% donkey serum in PBS. To study heparan sulfate/VEGFR2 interaction, cells were incubated overnight at 4 °C with 1 μg/ml anti-heparan sulfate (mAb 10E4), a mouse monoclonal antibody that recognizes N-sulfated glucosamine units, and 2 μg/ml affinity-purified goat polyclonal antibody against mouse VEGFR2. Proximity ligation was performed using a Duolink assay kit (Olink) according to the protocol provided by the manufacturer, except that the oligonucleotide-conjugated anti-mouse and anti-goat probes were used at 1:10 dilution. All images were taken with a DeltaVision deconvolution microscopy system with a 20× objective, deconvolved with SoftWoRx software, and analyzed with NIH ImageJ software. Five images were taken from each sample well (0.2 cm²). Each experiment represents a fresh preparation of mBMEC.

Statistics

All data were analyzed by unpaired one-tailed Student's t test to assess statistical significance, except that the paired test was used for protamine and surfen inhibition assays. p values <0.05 were considered statistically significant.

RESULTS

Ndst1-deficient Mice Show Reduced Dermal Vascular Hyperpermeability in Response to VEGF₁₆₅ and VEGF₁₂₁

To investigate whether heparan sulfate is involved in VEGF-induced vascular hyperpermeability, we tested the response to VEGF in vivo using mutant mice with altered endothelial heparan sulfate induced by the conditional inactivation of Ndst1 (Ndst1^f/fTie2Cre⁺). Ndst1 encodes the enzyme N-deacetylase-N-sulfotransferase-1, which is partially responsible for the N-deacetylation and subsequent N-sulfation of subsets of N-acetylglucosamine residues in heparan sulfate. Offspring of genotype Ndst1^f/fTie2Cre ⁺ show ∼50% decrease in overall sulfation of heparan sulfate due to coupling of N-sulfation to other downstream modifications (O-sulfation and uronic acid epimerization) of the chain (23, 29). Because endothelial cells also express a second Ndst isozyme (Ndst2), incomplete reduction of sulfation of endothelial heparan sulfate occurs in Ndst1-deficient mice (29). Compounding these two mutations, however, results in loss of viability (24, 35).

Dermal vascular hyperpermeability was measured by the leakage of Evan's Blue in response to local intradermal injection of VEGF (modified Miles assay) (32). As shown in Fig. 1A, mutant mice showed significantly reduced leakage of Evan's Blue into sites injected with VEGF₁₆₅ when compared with wild-type littermates (1.5 ± 0.2 in Ndst1^f/fTie2Cre⁺ versus 2.1 ± 0.2 μg in Ndst1^f/fTie2Cre⁻). After subtraction of the basal level of permeability (i.e. when mice were injected intradermally with PBS-BSA instead of VEGF), we calculated that the extent of leakage was reduced 2-fold in the mutant. In contrast, mutant and wild-type mice showed no difference in leakage induced by histamine, an inflammatory factor that does not depend on heparan sulfate (Fig. 1A). We also assayed leakage in Ext1^f/+Tie2Cre ⁺ mice, which lack one copy of the copolymerase Ext1 responsible for the formation of the backbone of heparan sulfate. This mutation results in shortened heparan sulfate chains (∼50% reduction in length) (36). Altering chain length, however, did not affect VEGF₁₆₅-induced vascular hyperpermeability (Fig. 1B), emphasizing the importance of sulfation of the heparan sulfate in the hyperpermeability response.

FIGURE 1.

Dermal vascular hyperpermeability induced by VEGF is reduced in Ndst1-deficient mice. A, dermal vascular permeability was measured by Miles assay, in which Evan's Blue was injected intravenously. The amount of dye that leaked into the injection site was measured by absorbance after extraction. Ndst1^f/fTie2Cre⁻ (WT) and Ndst1^f/fTie2Cre⁺ (labeled as Ndst1 ^−/−) mice were injected with 200 ng of VEGF₁₆₅ (n = 20) and VEGF₁₂₁ (n = 10). Histamine (1 μg, n = 12) was injected as positive control, and PBS with 0.1% BSA (n = 20) was injected as negative control. Error bars indicate S.E. B, dermal vascular hyperpermeability of Ext1^f/+Tie2Cre⁺ (labeled as Ext1^+/−) and Ext1^f/+Tie2Cre⁻ (WT) mice was measured as in A. n = 6–9. C, binding of VEGF₁₆₅ and VEGF₁₂₁ to primary lung endothelial cells measured by flow cytometry. VEGF binding to untreated cells is shown as a black line, binding to heparin lyase I-treated cells is shown as a gray line, and the control sample, incubated only with biotinylated antibody and streptavidin-phycoerythrin, is shown as the filled gray histogram. % of Max, percentage of maximum.

As a control, we examined the response to VEGF₁₂₁, a splice isoform that does not bind to heparin or heparan sulfate (12, 13, 24). As shown in Fig. 1C, VEGF₁₂₁ bound poorly to isolated endothelial cells when compared with VEGF₁₆₅. (Under these conditions, binding of VEGF₁₆₅ was primarily mediated by heparan sulfate because treatment of the cells with the degradative enzyme, heparin lyase I, reduced binding by >90%, whereas binding of VEGF₁₂₁ was not affected.) Surprisingly, Ndst1-deficient mice also showed reduced hyperpermeability in response to VEGF₁₂₁ (Fig. 1A; 1.3 ± 0.2 in Ndst1^f/fTie2Cre⁺ versus 2.0 ± 0.1 μg in Ndst1^f/fTie2Cre⁻). Thus, the ability of VEGF to induce hyperpermeability in this system depends less on the interaction of heparan sulfate with the ligand than on its interaction with one or more signaling receptors activated by VEGF. Note that the binding of VEGF₁₂₁ to endothelial cells was about 2.5-fold lower than the binding of VEGF₁₆₅ to heparin lyase-treated cells. Experiments with mAb 10E4, which reacts with native heparan sulfate, showed complete removal of the epitope by heparin lyase digestion (data not shown). Thus, the difference in binding of the two VEGF isoforms might reflect a lower affinity of VEGF₁₂₁ for VEGF receptors when compared with VEGF₁₆₅ as reported previously (37).

Ndst1-deficient Mice Show Reduced Tumor Vascular Permeability

To test whether endothelial heparan sulfate plays a similar role in regulating tumor vascular hyperpermeability, we injected LLC subcutaneously into Ndst1^f/fTie2Cre⁺mice and wild-type littermates (Ndst1^f/fTie2Cre⁻). When tumors were visually discernable, mice were injected with Evan's Blue, and the extent of leakage was measured. As shown in Fig. 2A, less dye extravasation occurred in tumors grown in Ndst1-deficient mice when compared with tumors grown in wild-type mice (21.3 ± 2.1 versus 30.7 ± 2.3 ng/mg of wet tumor mass in mutant and wild-type mice, respectively). In previous studies, we showed decreased LLC tumor growth and vascular density in Ndst1-deficient mice due to deficient sprouting of endothelial cells (23). However, that study examined much larger tumors (volume = 150–400 mm³) taken 15 days after tumor injection. In the experiments reported here, care was taken to ensure that the tumors used had comparable mass (48 ± 4 versus 50 ± 5 mg of wet weight for mutant and wild-type mice, respectively, corresponding to 50–100 mm³, Fig. 2B) and similar tumor microvascular density (Fig. 2C). Western blotting showed comparable levels of VEGF expression in tissue extracts prepared from mutant and wild-type tumors (Fig. 2D). Analysis of tumor sections for infiltration of leukocytes, which might contribute to the tumor vascular hyperpermeability, also showed no differences (Fig. 2E). The results of the skin assays and tumor studies provide the first in vivo evidence that endothelial heparan sulfate regulates VEGF-induced vascular permeability.

FIGURE 2.

Tumor vascular permeability is reduced in Ndst1-deficient mice. A, LLC (4 × 10⁵) were injected into the hindquarters of Ndst1^f/fTie2Cre⁻ (WT) and Ndst1^f/fTie2Cre⁺ (labeled as Ndst1 ^−/−) mice. Mice with comparably sized tumors were injected i.v. with 1 mg of Evan's Blue (EB) and sacrificed 1 h later. The amount of dye extravasation into the tumor was measured spectrophotometrically and normalized to the amount tumor mass that was analyzed (wet weight). B, mass (wet weight) of the tumors used in the experiment shown in panel A. C, multiple sections from tumors (n = 4) were stained with antibody to CD31. The density of tumor microvasculature was estimated by counting the number of CD31+ microvessels per 400× microscopic field. D, VEGF expression in tumors was assessed by Western blotting after separation of 20 μg of tumor lysate by SDS-PAGE. Each lane represents a separate tumor. E, sections from tumors (n = 4) were stained with antibody to CD45. The number of CD45+ cells per 400× microscopic field was determined. Error bars indicate S.E. HPF, high power field.

Loss of Endothelial Heparan Sulfate Attenuates VEGFR2 Phosphorylation in Response to VEGF₁₆₅ and VEGF₁₂₁

VEGFR2 is the primary VEGF receptor expressed in blood endothelial cells and is thought to be essential for mediating the hyperpermeability response induced by VEGF (38, 39). Consistent with these observations, we found that dermal permeability of Vefgr2^+/− mice was abrogated in the Miles assay (Fig. 3A). VEGFR2 expression was readily detectable by Western blotting in isolated neonatal dermal endothelial cells (data not shown). The complete loss of response in the heterozygous mice was surprising, a finding that might suggest that a critical threshold of receptors is needed for hyperpermeability response.

FIGURE 3.

Loss of heparan sulfate attenuates VEGFR2 phosphorylation in response to both VEGF₁₆₅ and VEGF₁₂₁. A, dermal vascular hyperpermeability induced by VEGF₁₆₅ was blocked in Vegfr2^+/− mice. Wild-type or Vegfr2^+/− mice were injected intradermally with 200 ng of VEGF₁₆₅ (n = 6) or PBS with 0.1% BSA (n = 6) as negative control. B, chemical analysis of heparan sulfate derived from the mBMEC. Heparan sulfate was isolated by anion-exchange chromatography and digested with heparin lyases I, II, and III, and the resulting disaccharides were analyzed by liquid chromatography/mass spectrometry (34). The number of sulfate groups at each position was determined and expressed per 100 disaccharides. Ndst1^−/−, Ndst1^f/fTie2Cre⁺. C, immunoblot analysis of VEGFR2 phosphorylation in mBMEC isolated from mutant and wild-type mice after stimulation with 20 ng/ml VEGF₁₆₅ for 2 and 15 min (n = 3). D and E, wild-type mBMEC were treated with heparan lyases I, II, and III (Hep I–III) for 15 min and then stimulated with 20 ng/ml VEGF₁₆₅ (n = 4) or 50 ng/ml VEGF₁₂₁ (n = 3). Representative immunoblots are shown. The bar graphs represent the relative signal intensity after correction for the background and normalization to the loading control. The average values ± S.E. were determined from the indicated number of replicates. pY1175, phospho-Tyr-1175.

To establish that the alteration in vascular leakage in Ndst1-deficient mice was caused by alterations in VEGFR2 signaling, we isolated primary mBMEC, which, in contrast to dermal endothelial cells, can be readily obtained from adult animals. Cells derived from Ndst1-deficient and wild-type animals showed reactivity for characteristic markers of mBMEC (CD31, CD34, CD105, and CD166, supplemental Fig. S1). The level of N-sulfation, 2-O-sulfation, and 6-O-sulfation of heparan sulfate derived from mutant mBMEC was reduced when compared with the wild type (Fig. 3B), and the extent of reduction was similar to that observed previously in lung microvascular endothelial cells (29).

Treatment of mBMEC with VEGF₁₆₅ induced rapid and robust phosphorylation of VEGFR2 in wild-type cells, based on reactivity to a phospho-specific antibody that recognizes phospho-Tyr-1175 (Fig. 3C). After normalizing the band intensity to the total amount of VEGFR2 in each extract, we found that phosphorylation of VEGFR2 in Ndst1-deficient cells was reduced ∼50% at 2 min, with attenuation of the effect 15 min after stimulation. Removal of endothelial heparan sulfate by treatment with heparin lyases caused a more substantial reduction of VEGFR2 phosphorylation (73% at 2 min and 58% at 15 min, Fig. 3D). Thus, the incomplete reduction of VEGFR2 phosphorylation in Ndst1 mutant cells was likely due to partial reduction in N-sulfation. VEGF₁₂₁ activation also displayed a 60% reduction in VEGFR2 phosphorylation in heparin lyase-treated cells when compared with mock-treated controls (Fig. 3E).

eNOS is an important regulator of VEGF-induced hyperpermeability that lies downstream of VEGFR2 and its activation is associated with autophosphorylation of VEGFR2 at Tyr-1175 (40–42). Heparan lyase treatment of mBMEC attenuated eNOS phosphorylation at Ser-1177 by VEGF₁₆₅ and VEGF₁₂₁ (Fig. 4). Src kinase is another downstream regulator of VEGF-induced hyperpermeability. Blockade of Src has been shown to reduce tissue injury following cerebral ischemia and myocardial infarction (4, 5). However, attempts to measure Src activation (phosphorylation at Tyr-418) in cultured mBMEC did not meet with success due to the high basal level of phosphorylation in these cells (data not shown).

FIGURE 4.

Loss of heparan sulfate attenuates eNOS phosphorylation in response to both VEGF₁₆₅ and VEGF₁₂₁. Immunoblot analysis of eNOS phosphorylation in mBMEC cells after treatment with heparan lyases I, II, and III (Hep I–III) for 15 min and stimulation with 20 ng/ml VEGF₁₆₅ or 50 ng/ml VEGF₁₂₁ was performed. Representative immunoblots are shown. The bar graphs represent the normalized relative signal intensity. n = 3. Error bars indicate S.E. pS1177, phospho-Ser-1177.

Pharmacological Inhibition of Heparan Sulfate Inhibits Vascular Hyperpermeability

The above findings suggested that antagonizing the action of endothelial heparan sulfate might provide an effective method to reduce vascular hyperpermeability. To explore this possibility, we first examined the effect of protamine, a highly positively charged protein commonly used to neutralize anticoagulant heparin (43). The addition of protamine to mBMEC caused a dose-dependent suppression of VEGFR2 phosphorylation in response to VEGF₁₆₅ (Fig. 5A, left). Inhibition was achieved at a concentration as low as 0.1 μm, and with near complete inhibition occurring at 1 μm. Protamine also inhibited VEGF₁₂₁-induced VEGFR2 phosphorylation, although inhibition required higher concentration (Fig. 5A, right). We also tested the effect of surfen, a recently described low molecular mass (372 Da) heparan sulfate antagonist (average molecular mass of protamine = 5,000 Da) (28). Surfen showed a dose-dependent inhibition of VEGFR2 phosphorylation induced by VEGF₁₆₅ in mBMEC (Fig. 5B, left), with complete inhibition occurring at 10 μm. Like protamine, surfen was able to inhibit VEGF₁₂₁-induced VEGFR2 phosphorylation but required higher concentration for unknown reasons (Fig. 5B, right).

FIGURE 5.

Protamine and surfen inhibit VEGFR2 phosphorylation in vitro and dermal vascular permeability in vivo. A, immunoblot analysis of VEGFR2 phosphorylation in mBMEC stimulated with VEGF₁₆₅ (50 ng/ml) or VEGF₁₂₁ (100 ng/ml) for 5 min. Cells were treated with protamine at the indicated concentration prior to application of VEGF. The immunoblots are representative of three similar experiments. B, immunoblot analysis of VEGFR2 phosphorylation was performed as in A, except that cells were treated with surfen. C, mice were injected intradermally with 200 ng of VEGF₁₆₅ alone (filled squares) or with 12.5 μg of protamine (open squares), and the amount of Evan's Blue leakage was measured; n = 9. D, mice were injected intradermally with 200 ng of VEGF₁₆₅ alone (filled squares) or with 54 μm surfen (open squares); n = 9. E, mice were injected with 200 ng of VEGF₁₂₁ alone (filled squares) or with 54 μm surfen (open squares); n = 11. Experimental and control treatments of the same mouse in C–E are linked with a line. PBS with 0.1% BSA was used as negative control, which showed an average leakage of 0.9 ± 0.1 μg (C), 0.8 ± 0.1 μg (D), and 0.7 ± 0.1 μg (E). pY1175, phospho-Tyr-1175.

We next tested the ability of protamine and surfen to block VEGF-induced leakage in the skin. Protamine attenuated VEGF₁₆₅-induced Evan's Blue modestly (22% reduction, Fig. 5C). The lack of potency of protamine in this assay might have been caused by poor tissue penetration as we observed a highly focal inhibition of dye leakage at a site injected with a mixture of protamine and VEGF. Consistent with this idea, intradermal injection of surfen reversed VEGF-induced leakage more effectively than protamine (57%; 1.9 ± 0.2 μg of Evan's Blue with VEGF₁₆₅ versus 1.2 ± 0.2 μg with VEGF₁₆₅ plus surfen, Fig. 5D). A similar effect of surfen was also observed when VEGF₁₂₁ was tested (50% reduction; 1.3 ± 0.1 μg of Evan's Blue versus 1 ± 0.1 μg with surfen (Fig. 5E). Thus, pharmacological inhibition of heparan sulfate-protein interactions in wild-type animals mimics the effect of altering the biosynthesis of heparan sulfate genetically.

VEGF₁₆₅ and VEGF₁₂₁ Induce a Transient Increase of Heparan Sulfate-VEGFR2 Complexes

To study the mechanism by which heparan sulfate facilitates VEGF-induced VEGFR2 signaling and hyperpermeability, we examined the formation of heparan sulfate-VEGFR2 complexes in situ by proximity ligation assay in isolated mBMEC using antibodies to heparan sulfate (mAb 10E4) and the extracellular domain of VEGFR2. The specificity of the two antibodies was verified by performing proximity ligation using mBMEC treated with heparan lyases and by using mBMEC isolated from Vegfr2^+/− mice (supplemental Fig. S2). Interestingly, heparan sulfate-VEGFR2 complexes were readily detected under basal condition in the absence of VEGF, which may reflect the range of interaction assessed by the assay (up to 30 nm) (44) (Fig. 6A). The apparent number of complexes underwent a rapid increase in response to both VEGF₁₂₁ and VEGF₁₆₅ stimulation (70 and 60%, respectively). The response was rapid (2 min was the earliest feasible time point), and then the number of complexes decreased to the basal level by 15 min (Fig. 6B). The formation of new complexes depended on heparan sulfate based on the lack of induction of complexes by either VEGF₁₂₁ or VEGF₁₆₅ in mBMEC isolated from Ndst1-deficient mice (Fig. 6C). A 47% decrease in the number of heparan sulfate-VEGFR2 complexes was noted under basal conditions (compare basal conditions for mutant and wild-type cells in Fig. 6C), which was expected because the mutant heparan sulfate contains fewer mAb 10E4 epitopes (N-sulfated glucosamine residues) (29). Although one might argue that the induction of complexes of VEGFR2 with heparan sulfate by VEGF₁₆₅ reflects a bridging activity of the ligand, VEGF₁₂₁ does not bind to heparan sulfate yet induces a transient increase in receptor complexes as well (Fig. 1C).

FIGURE 6.

Heparan sulfate-VEGFR2 complexes show a transient increase when stimulated by VEGF₁₆₅ or VEGF₁₂₁. A, representative images were obtained by proximity ligation assay. Cells were fixed and incubated with antibodies to heparan sulfate (mAb 10E4) and to the extracellular domain of mouse VEGFR2 (goat IgG) followed by proximity ligation assay reagents. Each red dot indicates an interaction between heparan sulfate and VEGFR2. Nuclei are shown in blue. The Control panel represents cells incubated with 10E4 and nonspecific goat IgG. B, mBMEC were stimulated with 50 ng/ml VEGF₁₂₁ (striped bars) or 20 ng/ml VEGF₁₆₅ (white bars) for the indicated time. The data are presented as red dots (Duolink signals)/cell. As a negative control, cells were incubated with mAb 10E4 and nonspecific goat IgG, which gave rise to a background of 3.9 ± 0.3 Duolink signals/cell. The error bar represents S.E., n = 4 separate experiments. C, mBMEC derived from Ndst1^f/fTie2Cre⁻ (WT) and Ndst1^f/fTie2Cre⁺ (labeled as Ndst1^−/−) mice were stimulated with VEGF as indicated and analyzed by proximity ligation assay. The error bars represent S.E., n = 3 separate experiments. All images were taken with a DeltaVision deconvolution microscope (200× magnification), deconvolved with SoftWoRx software and processed with NIH Image J.

DISCUSSION

In this report, we provide two lines of evidence demonstrating that VEGF-induced vascular hyperpermeability depends on endothelial heparan sulfate. First, undersulfation induced by tissue-specific inactivation of the sulfotransferase, Ndst1, attenuated VEGF-induced dermal and tumor vascular hyperpermeability in vivo (Figs. 1 and 2). Secondly, pharmacological antagonism of heparan sulfate-protein interactions diminished VEGF-induced hyperpermeability in the dermis and blocked VEGFR phosphorylation in isolated endothelial cells (Fig. 5). We also showed that VEGF₁₆₅, as well as VEGF₁₂₁, which does not bind to heparan sulfate, induced transient association of VEGFR2 with heparan sulfate (Fig. 6).

All VEGF isoforms bind to heparan sulfate with the exception of VEGF₁₂₁, a splice variant lacking exons 6 and 7, which encode one or more heparin-binding domains (13). The interaction of VEGF₁₆₅ with VEGF receptors depends on cell surface heparan sulfate because treatment with heparin lyase severely reduces binding (27, 45). In contrast, VEGF₁₂₁ is thought to act independently of heparan sulfate because cross-linking of VEGF₁₂₁ to VEGFR2 and phosphorylation of VEGFR2 were unaffected by heparin lyase or exogenous heparin in human umbilical vein endothelial cells (22, 27). The data presented here challenge this view. Heparin lyase treatment of mBMEC prevented VEGF₁₂₁-induced VEGFR2 phosphorylation and signaling in vitro (Fig. 3E), and genetic alteration of heparan sulfate in vivo reduced VEGF₁₂₁-induced vascular permeability (Fig. 1). Possibly, differences in the source of cells used in the earlier studies (human umbilical vein endothelial cells) versus microvascular endothelia (mBMEC) used here might explain the discrepancy. Microvascular endothelial cells may be more susceptible to leak from hydrostatic and/or pathologic hyperpermeability stress.

Proximity ligation assays of primary mBMEC provided direct evidence that both VEGF₁₆₅ and VEGF₁₂₁ induced an interaction between heparan sulfate and VEGFR2. VEGF₁₆₅ might accomplish this interaction by acting as bridging molecule, binding to both heparan sulfate and VEGFR2. Consistent with this idea, the heparin-binding domain in VEGF₁₆₅ is physically distinct from the region that interacts with the receptor and thus can function independently (46). However, the mechanism by which VEGF₁₂₁ facilitates heparan sulfate-VEGFR2 interaction must be different due to the lack of a heparin-binding domain in the ligand. The extracellular domain of VEGFR2 can bind to heparin with weak affinity (47), perhaps through the sequence QDRKTKKRHC located between the sixth and seventh Ig-like domains (48). Because all VEGF isoforms are secreted as covalent dimers, with the receptor-binding domains located on opposite sides of the dimer interface, VEGF can facilitate dimerization and autophosphorylation of the receptor (49). Thus, VEGF₁₂₁ might enhance the interaction of heparan sulfate with VEGFR2 by an increase in avidity due to the receptor dimer binding to an extended oligosaccharide sequence in heparan sulfate (50). VEGFR1 also contains a potential heparin-binding domain in its fourth Ig-like extracellular loop, suggesting that a common mode of ligand-induced interaction with heparan sulfate might exist (15, 51).

Heparan sulfate typically occurs covalently linked to a proteoglycan core protein (52). Cells express several membrane and extracellular HSPGs, including glycosylphosphatidylinositol-anchored proteoglycans (glypicans) and transmembrane proteoglycans (e.g. syndecans). Gengrinovitch et al. (53) have suggested glypican-1 as a VEGF₁₆₅-binding proteoglycan, but endothelial cells express other membrane proteoglycans as well (54). Studies are underway to identify the functional HSPG that interacts with VEGFR2 and VEGF₁₆₅ and whether the interaction with heparan sulfate depends on the arrangement of sulfate residues in the chain, e.g., by examining how hyperpermeability is affected in tissue-specific mutants altered in uronyl 2-O-sulfation (55) and glucosaminyl 6-O-sulfation (56). Systemic mutants altered in the expression of specific core proteins are also available, which should aid in the identification of the active proteoglycan(s) in this system (57).

Tissue-damaging edema resulting from ischemia is associated with vascular hyperpermeability induced by expression of VEGF (2, 4). Despite the long term beneficial effects of VEGF in promoting angiogenesis within ischemic tissue, the acute phase edema induced by VEGF aggravates tissue injury (8). Indeed, several studies that employed agents that inhibited VEGF receptor phosphorylation (58) and/or phosphorylation at the downstream steps in the signal transduction cascade (4, 5, 39) improved outcome. The current study suggests that targeting of hyperpermeability through heparan sulfate has the potential to be a novel approach to alleviate tissue injury caused by edema and may synergize with other VEGFR2 pathway inhibitors. Although the induction of edema following endothelial barrier disruption represents an important aspect of disease progression, the inflammatory response contributes significantly to the outcome of ischemic tissue injury as well (59, 60). Previous studies of heparan sulfate-deficient mice have suggested that endothelial heparan sulfate plays a role in neutrophil trafficking and chemokine presentation (29). Thus, antagonizing heparan sulfate interactions provides a generally applicable approach for treating tissue injury and the unintended deleterious sequelae of edema and inflammation.