Ligand-induced internalization selects use of common receptor neuropilin-1 by VEGF₁₆₅ and semaphorin3A

The vascular and nervous systems have a treelike branched structure originating from a center and reaching all tissues. Often vessels and nerves run parallel, suggesting that they are guided by common guidance mechanisms. Indeed, there is evidence that axonal guidance cues are also involved in blood-vessel formation. VEGF, an essential regulator of vascular development, has recently been shown to regulate neuronal development.^1-4 

Class 3 semaphorins, a family of 6 secreted glycoproteins that includes Sema3A-3F, function as axonal chemorepellents for growth cones during neuronal development.^5-7  Their receptors are composed of a ligand-binding chain consisting of neuropilins and a signal-transducing chain consisting of plexins.⁸  Except for Sema3E, which directly binds to plexin-D1,⁹  Sema3A, 3B, 3C, and 3D need to bind to neuropilin-1 (Npn-1) or Npn-2 to signal.^10-14  Class 3 semaphorins use distinct combinations of neuropilins and plexins as their receptor system^13,15-17 ; Sema3A uses Npn-1–plexin-A1, -A2, or -A4.^16,17  Recent studies in mice with targeted deletions of semaphorin signal components have revealed an essential role of semaphorin signaling in normal vascular development.^9,15,18 

The VEGF family of proteins is required for angiogenesis during development and after birth, and contributes to neuronal growth, immune regulation, and hematopoiesis.^3,19-21  Npn-1 serves as a VEGF-A isotype-specific receptor.²²  The VEGF-A gene is organized in 8 exons that generate different isoforms, VEGF₁₂₁, VEGF₁₄₅, VEGF₁₆₅, VEGF₁₈₉, and VEGF₂₀₆, by alternative splicing. Exon 7, which codes for a Npn-1 binding site, is expressed in VEGF₁₆₅, VEGF₁₈₉, and VEGF₂₀₆.²³  Although VEGF receptors are required for VEGF₁₆₅ signaling, Npn-1 acts as a coreceptor for VEGF₁₆₅ that enhances VEGF₁₆₅ binding to VEGF receptor-2 (VEGFR-2/KDR/Flk-1) and VEGF₁₆₅ activity.²⁴  VEGF-B isoforms (VEGF-B₁₆₇ and VEGF-B₁₈₆) and placenta growth factor-2 (PlGF-2) bind to Npn-1 and activate VEGFR-1 (Flt-1).^25-27  The VEGF-like protein from orf virus NZ2 binds to Npn-1 and activates VEGFR-2.²⁸ 

Due to its ability to bind various ligands, Npn-1 appears to serve as a “hub” receptor for different ligands. The multiple domain structure of Npn-1 extracellular domain may explain how Npn-1 interacts with various ligands. Npn-1 has a large extracellular domain of 860 amino acids, which consists of 3 subdomains, a1, a2 (CUB), b1, b2 (coagulation factor V/VIII), and c (MAM) domains. Sema3A binds to Npn-1 via a1a2b1 domains, whereas VEGF₁₆₅ binds via b1b2 domains.²⁹  Heparin and PlGF-2 also bind to the b1b2 domain.²⁷  Consistent with Npn-1 playing multiple roles, Npn-1–deficient mice are embryonically lethal and display severe abnormalities in the nervous and cardiovascular systems.^30,31  Endothelial-cell–specific Npn-1–null mice showed severe vascular defects.³²  In vitro, Npn-1 plays a role in cell-to-cell adhesion.³³ 

Npn-1 mediates distinct functions by choice of ligand and ligand-specific signal transducer; the semaphorin–Npn-1 complex uses plexin to signal, whereas the VEGF–Npn-1 complex uses VEGF receptors. In some cases, Npn-1 can mediate opposite signals. For example, VEGF₁₆₅ promotes endothelial-cell adhesion to extracellular matrix and microvessel outgrowth, but Sema3A inhibits these functions.^18,34  Sema3A evokes repulsive signals for growth cone guidance, whereas VEGF₁₆₅ induces axonal outgrowth.^6,35  It is not clear how Npn-1 prioritizes signals derived from different ligands, particularly when the signals are contrasting. To address this question, we have examined Npn-1 from a perspective of a receptor that is shared by multiple ligands. We show that Sema3A exerts a variety of activities on endothelial cells and that VEGF₁₆₅ inhibits these activities. We also show that VEGF₁₆₅ has a priority for use of Npn-1 over Sema3A due to its superior binding to cells and subsequent induction of Npn-1 internalization, which renders Npn-1 unavailable for Sema3A binding. This mechanism allows Npn-1 to select signals from its multiple ligands.

Recombinant human VEGF₁₆₅, VEGF₁₂₁, VEGF-B₁₆₇, VEGF-C, FGF-2, chimeric human Sema3A/Fc, soluble rat Npn-1/Fc (srNpn-1/Fc), human VEGFR-1/Fc, and human B7-1/Fc were from R&D Systems (Minneapolis, MN). Porcine heparin sodium salt and bovine fibronectin were from Sigma (St Louis, MO).

Primary human umbilical vein endothelial cells (HUVECs) were prepared from umbilical cord and cultured as previously described.³⁶  HUVECs were used between passages 2 and 5. The human stromal-cell line HS-5 (ATCC, Manassas, VA) was maintained in DMEM with 10% fetal bovine serum (FBS); the rat pheochromocytoma-cell line PC12 (ATCC) was maintained on fibronectin-coated dishes in DMEM with 10% FBS and 10% horse serum.

HUVECs were starved by overnight incubation in MEDIUM199 (Cellgro; Mediatech, Herndon, VA) with 1% FBS, and then cultured (7000 cells/well) in 96-well flat-bottom tissue culture plates (Costar, Corning, Corning, NY) in MEDIUM199 with 1% FBS and 2 μg/mL heparin. Cultures were supplemented with Sema3A alone or with VEGF₁₆₅ or VEGF₁₂₁. After 16 hours, Cell Counting Kit-8 (Dojindo, Kumamoto, Japan) was added (10 μL/well) and incubation continued for 2 hours, followed by measurement of absorbance at 450 nm. Subsequently, plates were patted dry, and adherent cells were fixed with 4% wt/vol paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) for 15 minutes and stained with 0.05% crystal violet in 20% ethanol. Adherent cells/field were counted at low-power magnification (× 40) by NIH image analysis. Data were obtained from triplicate wells. HUVECs (4000 cells/well) were cultured in 96-well tissue culture plates in MEDIUM199 with 10% FBS and 20 μg/mL heparin, with Sema3A, or Sema3A plus VEGF₁₆₅ or VEGF₁₂₁ for 3 days. Proliferation was measured by ³H thymidine uptake (0.6 μCi/well [0.022 MBq/well]; New England Nuclear, Boston, MA) during last 20 hours of culture. The results reflect mean cpm/culture.

Four-chamber glass slides (LAB-TEK; Nalge Nunc International, Rochester, NY) were coated with 10 μg/mL poly-l-lysine (Sigma) in PBS. HUVECs (40 000/chamber) were added to the chambers in MEDIUM199 with 1% FBS and 2 μg/mL heparin; Sema3A alone or with VEGF₁₆₅ or VEGF₁₂₁ was added to the cultures. After 30 minutes at 37°C, the medium was replaced with 4% wt/vol paraformaldehyde for 15-minute fixation. After removal of the plastic chambers, the glass slide was washed with PBS (3 ×) and mounted. Four nonoverlapping fields were counted at low-power magnification (× 40) to obtain an average ratio of the number of cells with lamellipodia spreading/the total number of adherent cells (40 to 90 cells/field). A cell was considered positive for spreading if it displayed lamellipodia in at least three fourths of the periphery.

Four-chamber glass slides were coated with 5 μg/mL fibronectin in PBS. HUVECs (15 000/chamber) were grown in HUVEC culture medium for 16 hours, washed with MEDIUM199 with 10% FBS and 2 μg/mL heparin, and incubated in the same medium with Sema3A (2 μg/mL or as indicated). After fixation with 4% wt/vol paraformaldehyde, slides were washed with PBS, and 4 nonoverlapping fields were observed to obtain an average retraction score (30 to 60 cells/field). Retraction scores were assigned blindly as described in the legend to Figure 2D.

The assay was performed as described.³⁶  Cells were starved overnight in MEDIUM199 with 1% FBS. Wells (48-well plates) were coated with 150 μL Matrigel (BD Biosciences, San Jose, CA) at 4°C, and incubated for 30 minutes at 37°C. HUVECs (15 000 cells/well) were incubated with MEDIUM199 with 1% FBS and 2 μg/mL heparin for 18 hours onto the Matrigel-coated wells. Cells were photographed under phase-contrast microscopy at low-power magnification (× 40). Average numbers of branching points were counted from triplicate wells.

Endothelial cells were detached with 5 mM EDTA in PBS, washed with binding buffer (MEDIUM199, 1% FBS, 10 mM HEPES, 2 μg/mL heparin), and incubated at the indicated temperatures with VEGF₁₆₅, VEGF₁₂₁, or FGF-2. Cell-surface staining was performed with FITC-conjugated anti–human CD31 mAb (WM59; BD Pharmingen), PE-conjugated ant–ihuman VEGFR-2 mAb (89106; R&D Systems), or PE-conjugated anti–human BDCA-4 (Npn-1) mAb (AD5-17F6; Miltenyi Biotec, Auburn, CA) at 4°C for 20 minutes. For intracellular staining, cells were permeabilized with BD Cytofix/Cytoperm kit (BD Biosciences). Binding of Sema3A to cells was measured after incubation with 2 μg/mL Sema3A/Fc in binding buffer at 4°C for 60 minutes. After washing, Sema3A/Fc bound to cells was detected with FITC-conjugated F(ab′)₂ goat anti–human IgG Fc (Jackson ImmunoResearch Laboratories, West Grove, PA). As a control, we used human B7-1/Fc. Data were collected using a FACScalibur cytofluorometer (Becton Dickinson, San Jose, CA) and analyzed using CELLQuest software (Becton Dickinson).

Flat-bottom microtiter plates (96 well, Immulon 4HBX; Thermo Labsystems, Franklin, MA) were coated overnight with srNpn-1/Fc chimeric protein (2 μg/mL in PBS). After blocking with Superblock (PIERCE, Rockford, IL) for 30 minutes, VEGF₁₆₅ (25 ng/mL), Sema3A/Fc (2 μg/mL), VEGF₁₆₅ (25 ng/mL) plus srNpn-1 (0.05-3.2 μg/mL), or VEGF₁₆₅ (25 ng/mL) plus sVEGFR-1 (0.05-3.2 μg/mL) in PBS 0.1% Tween20, 1% BSA were added with heparin at various concentrations. After incubation at 25°C for 1 hour, bound VEGF₁₆₅ or Sema3A/Fc was detected with rabbit anti-VEGF antibody (A-20; Santa Cruz Biotechnology, Santa Cruz, CA) or rabbit anti-Sema3A antibody (H-300; Santa Cruz Biotechnology) at 1 μg/mL, followed by HRP-conjugated anti–rabbit IgG antibody (Bio-Rad, Hercules, CA) at 1:5000 in PBS 0.1% Tween20 1% BSA. Reactions were visualized with tetramethoxybenzene peroxidase substrate (Kirkegaard & Perry Laboratories, Gaithersburg, MD), followed by addition of 1 M H₂SO₄, and read at 450 nm.

Endothelial cells on glass chamber slides coated with 5 μg/mL fibronectin were incubated (MEDIUM199 with 10% FBS and 2 μg/mL heparin) with 25 ng/mL VEGF₁₆₅ or 2 μg/mL Sema3A at 37°C for the indicated times. The medium was replaced with 4% wt/vol paraformaldehyde for fixation, cells were washed with PBS, and cell membranes were permeabilized with 0.1% Triton X100 in PBS. Cells were then stained with mouse anti–human BDCA-4 (Npn-1) mAb (AD5-17F6; Miltenyi Biotec) in 1% BSA and 3% goat serum in PBS at 4°C for 16 hours. Cells were washed with PBS and incubated with Alexa 488–conjugated goat anti–mouse IgG antibody (Molecular Probes, Eugene, OR) in PBS with 1% BSA and 3% goat serum at 25°C for 1 hour. Slides were mounted with FluorSave Reagent (Calbiochem, San Diego, CA). Images were visualized using an LSM 510 confocal microscope (Carl Zeiss, Thornwood, NY), and reflect the merging of fluorescent slice and DIC (differential interference contrast) images.

Results are expressed as means plus or minus SD. Student t test was applied to evaluate group differences; a P value of less than .05 was considered significant.

We examined the effects of Sema3A on endothelial-cell attachment, survival, and growth, and tested whether VEGF₁₆₅, a VEGF-A isoform that binds Npn-1,²⁴  modifies these activities of Sema3A. Previously, Sema3A was reported to inhibit endothelial-cell adhesion to extracellular matrix.¹⁸  As shown in Figure 1A, we found that Sema3A also inhibits endothelial-cell adhesion to tissue culture wells, in the absence of protein coating. This inhibition was blocked by VEGF₁₆₅, but not by VEGF₁₂₁, a VEGF-A isoform that lacks the Npn-1 binding domain.²⁴  Since previous studies concluded that attachment is critical to endothelial-cell survival,³⁷  we examined whether reduced endothelial-cell attachment correlated with decreased survival. In the same cultures, we measured first cell survival using a colorimetric assay, and after removal of nonadherent cells, we measured cell attachment (NIH image analysis). We found that Sema3A dose-dependently reduces endothelial-cell viability after 16-hour incubation in culture medium with 1% FBS (Figure 1B line graph). VEGF₁₆₅ (25 ng/mL) alone increased endothelial-cell viability and protected the cells from death occurring in the presence of Sema3A. VEGF₁₂₁ (25 ng/mL) alone also increased endothelial-cell viability comparably with VEGF₁₆₅ but, unlike VEGF₁₆₅, failed to protect the cells from death induced by Sema3A. In these cultures, endothelial-cell survival closely correlated with endothelial-cell attachment (Figure 1B bar graph), suggesting that the nonadherent cells are dead.

We also examined the effects of Sema3A on endothelial-cell proliferation after 3-day culture. As expected (Figure 1C), VEGF₁₆₅ and VEGF₁₂₁ dose-dependently promoted the proliferation of endothelial cells (open circles). Sema3A (1 μg/mL, Figure 1C closed triangles) reduced by 90% and 87% cell proliferation induced by 12.5 and 25 ng/mL VEGF₁₂₁, respectively. By contrast, under the same conditions, Sema3A reduced by only 29% and 13% endothelial-cell proliferation induced by 12.5 and 25 ng/mL VEGF₁₆₅, respectively (Figure 1C). These results demonstrate that Sema3A inhibits endothelial-cell attachment, survival, and proliferation, and that VEGF₁₆₅, but not VEGF₁₂₁, can counter these functions of Sema3A.

We examined in detail cell-attachment inhibition by Sema3A, and tested whether Sema3A inhibits the initial step of endothelial-cell spreading or rather promotes cell retraction and detachment once spreading has occurred. Endothelial cells usually spread their lamellipodia for firm adherence to poly-l-lysine–coated glass surfaces within 30 minutes (Figure 2A left panel). Lamellipodia spreading by endothelial cells is not observed on uncoated glass surfaces (not shown). In the presence of Sema3A (2 μg/mL), endothelial cells attached to the poly-l-lysine–coated surfaces but displayed virtually no spreading of lamellipodia (Figure 2A right panel). This effect was dependent upon Sema3A concentration (Figure 2B). VEGF₁₆₅ (25 ng/mL) significantly reduced this effect of Sema3A, whereas VEGF₁₂₁ (25 ng/mL) was ineffective (Figure 2C).

Next, we examined whether Sema3A affects endothelial cells that have already spread their lamellipodia. In neuronal cells, Sema3A induces axonal growth cone collapse, a process whereby filopodia become shorter and fewer, and lamellipodia shrink back from the sides of the growth cone.³⁸  After 16-hour incubation on fibronectin-coated slides, most endothelial cells display spreading of lamellipodia (Figure 2D lower left panel). At low-power magnification, lamellipodia are not clearly visible on endothelial cells, but their presence is reflected by somewhat undefined cell boundaries (Figure 2D upper left). After incubation with Sema3A for 1 hour, the endothelial cells displayed retracted lamellipodia to a various degree (Figure 2D lower), which was reflected by more sharp cell boundaries (Figure 2D upper right). Similar Sema3A-induced lamellipodia retraction was noted in endothelial cells cultured on poly-l-lysine–coated glass surfaces (data not shown). To quantify lamellipodia retraction, we established a scoring system (Figure 2D lower panels). By blindly scoring at least 120 cells from 4 fields, we obtained an average retraction score. As shown in Figure 2E and 2F, respectively, Sema3A dose-dependently and time-dependently promoted lamellipodia retraction in endothelial cells. This effect reached a plateau after 60-minute incubation. VEGF₁₆₅, but not VEGF₁₂₁, dose-dependently inhibited Sema3A-induced retraction; almost complete inhibition was noted at 25 ng/mL (Figure 2G). VEGF₁₆₅ had no or minimal effect on cell morphology when added alone (not shown). FGF-2, an endothelial-cell mitogenic factor reported to bind to Npn-1,³⁹  minimally affected retraction induced by Sema3A (Figure 2G). These results provide evidence that in endothelial cells Sema3A inhibits lamellipodia spreading and promotes lamellipodia retraction.

Endothelial cells can form cordlike structures when incubated onto Matrigel, a crude extract of extracellular matrix proteins.⁴⁰  We examined the effects of Sema3A on Matrigel-dependent endothelial-cell cord formation. As shown in Figure 2H, Sema3A inhibited formation of the characteristic reticular network. VEGF₁₆₅ reversed this effect of Sema3A, whereas VEGF₁₂₁ did not. Comparison of branching point numbers indicated that Sema3A significantly reduces cord formation by endothelial cells and that VEGF₁₆₅, but not VEGF₁₂₁, reverses this effect (Figure 2I).

To explore the mechanisms by which VEGF₁₆₅ counters the biologic effects of Sema3A on endothelial cells, we examined whether VEGF₁₆₅ reduces the binding of Sema3A to Npn-1, which serves as a common receptor for VEGF₁₆₅ and Sema3A. Since it is known that heparin enhances VEGF₁₆₅ binding to Npn-1,²³  we first examined the effect of heparin on Sema3A binding to Npn-1. Using an ELISA-based assay in which Npn-1 is immobilized onto the well, we found that heparin enhances significantly the binding of Sema3A to Npn-1 as well as the binding of VEGF₁₆₅ to Npn-1 (Figure 3A).

Based on these results, we examined whether VEGF₁₆₅ interferes with the binding of Sema3A to Npn-1, with or without heparin (2 μg/mL). As shown in Figure 3B, VEGF₁₆₅ dose-dependently inhibited the binding of Sema3A to Npn-1. However, even at the highest concentration tested (1600 ng/mL), reduction by VEGF₁₆₅ was less than 50% with or without heparin. In the biologic assays above (Figures 1, 2), 25 ng/mL VEGF₁₆₅ inhibited almost completely the activities of Sema3A (1-2 μg/mL), whereas this concentration of VEGF₁₆₅ did not significantly inhibit the binding of Sema3A (2 μg/mL) to purified Npn-1. These results suggested that VEGF₁₆₅ inhibits Sema3A activity in endothelial cells by mechanisms other than simple binding competition for Npn-1.

In the presence of VEGF₁₆₅, Npn-1 forms a complex with VEGFR-2, which enhances the binding of VEGF₁₆₅ to VEGFR-2.²²  Also, Npn-1 forms a stable complex with plexin-A1, and this complex has a higher affinity for Sema3A than does Npn-1 alone.⁴¹  Since endothelial cells express plexin-A1 and VEGFR-2 in addition to Npn-1,¹⁸  we tested whether VEGF₁₆₅ can compete for Sema3A binding to the surface of endothelial cells. FACS analysis showed that Sema3A/Fc bound to endothelial cells dose dependently, whereas a control B7-1/Fc did not (Figure 4A). When added to cells together with VEGF₁₆₅ at 4°C for 1 hour, Sema3A/Fc (2 μg/mL) binding to cells was reduced by VEGF₁₆₅ dose dependently (Figure 4B top). VEGF₁₆₅ (12.5 ng/mL) reduced Sema3A/Fc binding to cells by approximately 50%. VEGF₁₂₁, VEGF-B₁₆₇, VEGF-C, and FGF-2 (all at 100 ng/mL) failed to reduce Sema3A/Fc binding to endothelial cells (Figure 4B).

We examined Sema3A (2 μg/mL) binding to endothelial cells (1 hour at 4°C) after the cells were preincubated (25°C for 1 hour) with VEGF₁₆₅ (12.5 ng/mL). VEGF₁₆₅ was not removed. Under these conditions, VEGF₁₆₅ at 12.5 ng/mL reduced Sema3A/Fc (2 μg/mL) binding to endothelial cells by approximately 80% (Figure 4C top). These results indicated that preincubation of cells with VEGF₁₆₅ at 25°C is more effective at reducing Sema3A binding than simultaneous addition of VEGF₁₆₅ and Sema3A at 4°C. Preincubation of cells with VEGF₁₂₁ or FGF-2 did not affect the subsequent binding of Sema3A to cells (Figure 4C middle and bottom). These results provided evidence that VEGF₁₆₅, but not VEGF₁₂₁, VEGF-B₁₆₇, VEGF-C, or FGF-2, can compete for the binding of Sema3A to endothelial cells. The observation that preincubation enhanced the ability of VEGF₁₆₅ to inhibit Sema3A binding to endothelial cells suggested a contribution by mechanisms other than cell binding competition.

Npn-1 is expressed on various cell types.^42,43  We examined whether VEGF₁₆₅ can inhibit the binding of Sema3A to human bone marrow stromal HS-5 and rat pheochromocytoma PC12 cells. Preincubation with VEGF₁₆₅, but not VEGF₁₂₁, dose-dependently inhibited the binding of Sema3A to stromal HS-5 (Figure 4D) and rat pheochromocytoma PC12 cells (Figure 4E). These results indicate that VEGF₁₆₅ can inhibit the binding of Sema3A to cells of various lineages.

To clarify the mechanisms by which VEGF₁₆₅ reduces the binding of Sema3A to target cells, we examined endothelial-cell–surface expression of Npn-1. Using FACS analysis, we found that 1-hour preincubation at 25°C with VEGF₁₆₅, but not VEGF₁₂₁, dose-dependently reduced surface Npn-1 (Figure 5A left). At the concentration of 12.5 ng/mL (0.3 nM), VEGF₁₆₅ reduced by approximately 50% Npn-1 levels detected on control cells. In parallel experiments, both VEGF₁₆₅ and VEGF₁₂₁ reduced surface VEGFR-2 (Figure 5A right) but did not affect surface CD31 in these cells (not shown). VEGF₁₆₅ also specifically reduced surface Npn-1 in HS-5 cells (not shown; the effect on rat PC12 cells could not be tested because the antibody fails to recognize rat Npn-1).

Since heparin is required for maximal binding of VEGF₁₆₅ to Npn-1, we examined the effect of heparin on VEGF₁₆₅ (250 ng/mL)–induced reduction of Npn-1 on endothelial cells. We found that heparin enhances VEGF₁₆₅-induced reduction of surface Npn-1, with maximal Npn-1 reduction achieved at a heparin concentration of 80 ng/mL (Figure 5B). We also examined the effect of temperature on VEGF₁₆₅-induced reduction of surface Npn-1. At 4°C, levels of surface Npn-1 reached a plateau after 15 minutes, whereas Npn-1 continued to decrease at 25°C and 37°C (Figure 5C). The reduction of Npn-1 was somewhat more rapid and linear at 37°C than seen at 25°C. When endothelial cells were incubated with VEGF₁₆₅ (25 ng/mL) at 37°C for 30 hours, levels of surface Npn-1 were approximately 75% lower than those measured on endothelial cells cultured under the same conditions with 25 ng/mL VEGF₁₂₁ or FGF-2 (Figure 5D). This time- and temperature-dependent pattern of surface Npn-1 decrease induced by VEGF₁₆₅ suggested either that Npn-1 undergoes structural change, which renders it undetectable under these conditions, or that levels of surface Npn-1 decrease.

We examined whether internalization contributes to the reduction of surface Npn-1 by VEGF₁₆₅. Therefore, we compared the effects of VEGF₁₆₅ on cell-surface and total cell-associated levels of Npn-1. By FACS analysis, VEGF₁₆₅ (25 and 250 ng/mL, 25°C 1 hour) substantially reduced the levels of surface Npn-1 but only slightly reduced the levels of Npn-1 detected after cell permeabilization, which include cell-surface and intracellular Npn-1 (Figure 5E), suggesting that internalization contributed to the reduction of surface Npn-1 levels by VEGF₁₆₅. Using confocal microscopy, we traced Npn-1 distribution in endothelial cells after VEGF₁₆₅ stimulation (Figure 5F). After 5-minute incubation with VEGF₁₆₅ (25 ng/mL, 37°C), Npn-1 was detectable more intensely at the endothelial-cell rim, especially on lamellipodia, whereas after 30 minutes Npn-1 was detectable more intensely in the cytoplasm at the perinuclear region with a dotlike distribution. After 60 minutes, Npn-1 staining was detected almost exclusively at the perinuclear region. VEGF₁₂₁ and FGF-2 did not induce such changes (not shown). Based on these results, we conclude that VEGF₁₆₅ reduces cell-surface Npn-1, at least in part, by promoting its internalization.

The experiments shown in Figures 1, 2, 3, 4, 5 indicated that VEGF₁₆₅ inhibits Sema3A function in endothelial cells (Figures 1, 2) by competing with Sema3A for binding to cell-surface receptors (Figure 4) and inducing Npn-1 internalization (Figure 5). Since VEGF₁₆₅ is ineffective at directly competing with Sema3A for binding to Npn-1 (Figure 3B), we examined the potential contribution of VEGF receptors present on endothelial cells. First, we compared a soluble rat Npn-1 (srNpn-1) and a soluble VEGFR-1 (sVEGFR-1) for their ability to compete with VEGF₁₆₅ binding to immobilized Npn-1. srNpn-1, which is expected to serve as a direct competitor for VEGF₁₆₅, inhibited binding of VEGF₁₆₅ to immobilized Npn-1 (Figure 6A left). By contrast, sVEGFR-1, which interferes with VEGF binding to both VEGFR-1 and VEGFR-2,⁴⁴  minimally affected the binding of VEGF₁₆₅ to immobilized Npn-1, even at the highest VEGFR-1 concentration (3.2 μg/mL) tested (Figure 6A right). Thus, sVEGFR-1 does not directly compete with VEGF₁₆₅ binding to Npn-1.

We then compared the effects of srNpn-1 and sVEGFR-1 on VEGF₁₆₅-induced reduction of cell-surface Npn-1. To detect human endothelial-cell–surface Npn-1, we used the anti–human Npn-1 mAb AD5-17F6. This mAb specifically recognizes human Npn-1 and does not cross-react with rat Npn-1 because excess amounts of srNpn-1 (5 μg/mL) did not significantly alter recognition of human Npn-1 by this antibody (Figure 6B upper left). The species specificity of this antibody to human Npn-1 contrasts with the lack of species specificity of human VEGF₁₆₅ and Sema3A, which can both bind to srNpn-1 (Figure 3A). Taking advantage of this anti–human Npn-1 mAb for detection of surface Npn-1 on human endothelial cells, we found that srNpn-1 dose-dependently blocked VEGF₁₆₅-induced reduction of cell-surface Npn-1 (Figure 6B lower left). These results indicated that a direct interaction of VEGF₁₆₅ with membrane-bound Npn-1 is essential for VEGF₁₆₅ to reduce cell-surface Npn-1. By confocal microscopy, we found that srNpn-1 reduced VEGF₁₆₅-induced internalization of Npn-1, indicating that binding to cell-surface Npn-1 is also essential for VEGF₁₆₅ to promote Npn-1 internalization (Figure 6C).

To examine whether interaction with VEGFR-1 and/or VEGFR-2 is necessary for VEGF₁₆₅ to reduce cell-surface Npn-1 and promote Npn-1 internalization, we used sVEGFR-1, which does not compete for VEGF₁₆₅ binding to Npn-1 (Figure 6A) but neutralizes VEGF bioactivity by competing for VEGF binding to VEGFR-1 and VEGFR-2.⁴⁴  Endothelial cells were incubated with VEGF₁₆₅ (25 ng/mL) plus sVEGFR-1 (0-800 ng/mL). Dose dependently, sVEGFR-1 blocked VEGF₁₆₅-induced reduction of cell-surface Npn-1 (Figure 6B lower right), whereas a control protein (B7-1/Fc) had no effect (Figure 6B upper right). In addition, by confocal microscopy, we found that sVEGFR-1 reduced VEGF₁₆₅-induced internalization of Npn-1 (Figure 6C). These results indicated that VEGF₁₆₅ must interact with endothelial-cell–surface VEGFR-1 and/or VEGFR-2 to effectively reduce cell-surface Npn-1 and promote its internalization, perhaps by bridging VEGFR-1 and/or VEGFR-2 with Npn-1.

Finally, we investigated the effect of sVEGFR-1 on functional assays in which Sema3A promotes endothelial-cell retraction and VEGF₁₆₅ reduces this effect. Dose dependently, sVEGFR-1 reduced VEGF₁₆₅ inhibition of the Sema3A-induced retraction of endothelial cells (Figure 6D). This result is consistent with the results shown in Figure 6B-C that sVEGFR-1 reduces Npn-1 internalization induced by VEGF₁₆₅, thereby enhancing surface Npn-1 available for binding to Sema3A.

It was reported that Sema3A could induce Npn-1 endocytosis in the COS7-cell line, which coexpress Npn-1 and the L1 cell-adhesion molecule.⁴⁵  We examined whether Sema3A can also induce Npn-1 internalization in primary endothelial cells. We found that Sema3A dose-dependently reduced endothelial-cell–surface Npn-1 (Figure 7A), and that approximately 1 μg/mL (4 nM) Sema3A was required to achieve a 50% reduction in Npn-1 levels compared with control. This indicates that approximately 13-fold more Sema3A is required to achieve levels of surface Npn-1 reduction comparable with those induced by VEGF₁₆₅ under the same experimental conditions (Figure 5A). As expected, srNpn-1 blocked reduction of surface Npn-1 induced by Sema3A, whereas sVEGFR-1 was ineffective (data not shown). By using confocal microscopy, we confirmed that Sema3A could induce Npn-1 internalization (Figure 7B). After 5-minute stimulation with Sema3A, Npn-1 accumulated mainly in the cell-to-cell junctional regions, and after 30 minutes Npn-1 accumulated in the cytoplasmic region similar to the pattern noted after stimulation with VEGF₁₆₅. These results demonstrate that Sema3A can induce Npn-1 internalization in endothelial cells.

We show that Sema3A inhibits endothelial-cell lamellipodia formation, attachment, survival, and growth, and that VEGF₁₆₅ counters these activities of Sema3A. We further show that this functional competition between Sema3A and VEGF₁₆₅ is attributable to preferential binding of VEGF₁₆₅ to target cells and internalization of the shared receptor Npn-1. This potency difference at promoting Npn-1 internalization can dictate a priority for Npn-1 use by VEGF₁₆₅ over Sema3A, effectively blocking Sema3A function. Thus, our results demonstrate that control of surface levels of the common receptor is a critical mechanism for functional prioritization among ligands that share a receptor. Sema3A could also reduce surface levels of Npn-1 by promoting its internalization but required a higher concentration than VEGF₁₆₅ at reducing Npn-1 surface levels. The conclusion that Sema3A is less potent than VEGF₁₆₅ may not apply to the various endothelial environments in vivo. Relative effective concentrations of endogenous Sema3A and VEGF₁₆₅ in vivo are unknown and may depend on untested variables, such as levels of plexin-A and VEGF receptors in various cells, the contribution of posttranscriptional modifications, and protein degradation.

We established a requirement for linkage between cell-surface Npn-1 and VEGFR-1 or VEGFR-2 in VEGF₁₆₅-induced Npn-1 internalization. This conclusion comes from the observation that soluble VEGFR-1 extracellular domain, which does not interfere with the binding of VEGF₁₆₅ to Npn-1 (Figure 6A) but competes for VEGF₁₆₅ binding to VEGFR-1 or VEGFR-2,⁴⁴  blocked VEGF₁₆₅-induced down-regulation of surface Npn-1 (Figure 6B). Consistent with this conclusion, VEGF₁₂₁, which does not bind Npn-1 but binds to VEGFRs, did not affect Npn-1 levels. VEGFRs are subject to endocytosis and subsequent ubiquitination by the E3 ubiquitin ligase/multiadaptor protein Cbl.^46,47  It will be important to establish whether Npn-1 is also degraded. Of interest, we observed PEST sequences for protein degradation in the region of amino acids 823 to 840 of human Npn-1.⁴⁸ 

In addition to VEGF₁₆₅, PlGF-2 is known to inhibit Sema3A-induced growth cone collapse in dorsal root ganglion neurons.⁴⁹  PlGF-2 has the potential to link Npn-1 to VEGF receptors, and thus may down-regulate Npn-1 by the same mechanism of VEGF₁₆₅. FGF family members can also interact with Npn-1.³⁹  However, our results show that FGF-2 did not change Npn-1 levels and did not inhibit endothelial-cell retraction induced by Sema3A. Thus, FGF-2 may not bridge Npn-1 to FGF receptor. Also, VEGF₁₆₅ could inhibit other class 3 semaphorin family members that use the common receptor Npn-1. If so, our observations could explain why VEGF₁₆₅, but not VEGF₁₂₁, inhibited Sema3B-induced apoptosis in lung and breast cancers.⁵⁰  Also, since VEGF₁₆₅ can bind to Npn-2⁵¹  in addition to Npn-1, it will be interesting to examine whether VEGF₁₆₅ can induce internalization of Npn-2 and thereby inhibit Npn-2–mediated effects of class 3 semaphorins.

Npn-1 is known to serve as a cell-to-cell adhesion molecule,³³  which facilitates contact between resting T cells and dendritic cells and promotes primary immune responses.⁵²  Npn-1 is expressed in stromal cells, B lymphocytes, monocytes, and erythrocytes.^42,53  Our observation that VEGF₁₆₅ reduces surface levels of Npn-1 may explain some of the functions of VEGF on immune and hematopoietic cells.⁵⁴ 

In essence, receptor components shared by multiple ligands play additional roles besides ligand binding. If the signal-transducing chain is shared, as in the case of gp130 that is shared by different members of the IL-6 family of cytokines, the common signaling chain serves as a convergence mechanism for function.⁵⁵  As a result, distinct cytokines, which first bind to their specific receptor and then the ligand/ligand-specific receptor complexes engage the common signal transducer gp130, display overlapping functions. By contrast, if the ligand-binding chain is shared and the signaling chain differs, ligands with different and even contrasting activities converge on the same ligand-binding receptor. Npn-1 represents an example of a common receptor system, which binds multiple ligands as a hub receptor, but ligand/receptor complexes use specific signal transducers to transduce their respective signal. In this case, the common binding chain serves as a convergence of different ligands onto the same target cell, which provides a means for distinct ligands, even with opposing function, to share the same cell target. In such a receptor system, a mechanism for ligand prioritization appears essential to proper cell function. Here, we describe how a common receptor can serve as a cell membrane gatekeeper by selecting a specific ligand and undergoing internalization.

We thank Ms Yumi Narazaki for blindly scoring retraction assays; Ms Susan Garfield and Mr Stephen Wincovitch for technical assistance with confocal microscopy; Dr Lei Yao for isolation of HUVECs; and Drs Anna Virginia Gulino, Takayuki Nakayama, and Yoshiyasu Aoki for helpful suggestions.