**3. Results**

*Binding and cleavage of VEGF-A by FSAP:* In solid phase binding assays, FSAP bound to immobilized VEGF165, but not to VEGF121, in the presence of heparin (Figure 1A,B). Under non-reducing conditions, there was no change in the intensity or size of the VEGF165 upon incubation with FSAP (Figure 1C,D); however, under reducing conditions, there was a decrease in VEGF immunoreactivity (Figure 1C,D). Cleavage was enhanced by heparin, which often functions as a co-factor for FSAP activity. Inhibition of FSAP by aprotinin reduced this cleavage. The enzymatically inactive PPACK-FSAP had no e ffect (Figure 1C). Cleavage of VEGF165 was time-dependent (Figure 1D), but VEGF121 was not cleaved at all (Supplementary Figure S1). The cleavage site was identified by amino acid sequencing and was found to be in the neuropilin/heparin-binding domain at position 124/125, which was distinct from the plasmin cleavage site at 110/111 (Figure 2). The cleavage site for plasmin and FSAP in relation to the disulphide bridges [31] is shown in Figure 2. Thus, FSAP cleaves VEGF165 in the heparin/neuropilin-binding region.

70

**Figure 1.** Binding of FSAP (factor VII activating protease) to VEGF (vascular endothelial growth factor) and its specific proteolytic cleavage: (**A**) VEGF121 (open bars) or VEGF165 (grey bars) was immobilized, and the binding of FSAP, in the absence or presence of heparin, was detected with an anti-FSAP antibody (mean ± SD of triplicate wells); (**B**) VEGF165 was immobilized, and the binding of increasing concentrations of FSAP in the absence (open circles) or presence (filled circles) of heparin was determined. Error bars are smaller than the size of the symbols; (**C**) Mixtures of FSAP (or Phe-Pro-Arg-chloromethylketone (PPACK)-FSAP), buffer, heparin, VEGF165, and aprotinin, as indicated, were incubated, and the reaction was analyzed by western blotting with an anti-VEGF antibody under reducing or non-reducing conditions; (**D**) FSAP, VEGF165, and heparin were incubated for the indicated time intervals and the samples were analyzed for VEGF, as above. Densiometric analysis was performed to calculate the ratio of VEGF under reduced and non-reduced conditions.

**Figure 2.** Sequencing of proteolytically cleaved VEGF: VEGF165 or VEGF121 (20 μg/mL) was incubated with FSAP (200 μg/mL) or plasmin (200 μg/mL) in the presence of heparin (10 μg/mL) for 2 h at 37 ◦C. The mixture was separated by SDS-PAGE under reducing conditions and processed for N-terminal sequencing. The FSAP and plasmin cleavage sites, as well as the disulphide bond assignments in the heparin/neuropilin-binding domain [31] sequence of VEGF165 are indicated.

*Interaction of FSAP with neuropilin and VEGFR-2:* Neuropilins are co-receptors that bind to larger VEGF isoforms, such as VEGF165, in a heparin-dependent manner, and regulate the activity of long forms of VEGF. Thus, the binding interactions between FSAP, VEGF165, VEGFR, and neuropilin were investigated. FSAP bound strongly to neuropilin-1-Fc in a heparin-dependent manner, but there was no binding to VEGFR-2-Fc (Figure 3A). There was no difference in the binding of PPACK-FSAP, the MI-SNP of FSAP, or WT-FSAP (Figure 3A,B), indicating that the enzymatic activity of FSAP was not involved in the binding to neuropilin. VEGF165 bound to both neuropilin and VEGFR2 in a heparin-dependent manner (Figure 3C). VEGF165 binding to neuropilin-1-Fc was partially inhibited by FSAP in the absence or presence of heparin, but FSAP had no influence on binding of VEGF165 to VEGFR-2-Fc (Figure 3C,D). Hence, FSAP binds to neuropilin, thereby cleaving VEGF165 that, in turn, partially decreases its interactions with neuropilin but not VEGFR2.

*E*ff*ect of FSAP on proliferation and migration of HUVEC or VEGFR-transfected BAF3 cells:* Because FSAP can cleave VEGF165 and inhibits its binding to neuropilin, the effect of FSAP-treated VEGF165 on the activation of HUVEC was investigated in the absence or presence of heparin. FSAP did not influence bFGF- or VEGF165-induced DNA synthesis or cell migration (Figure 4A,B). Phosphorylation of ERK in HUVEC with bFGF or VEGF pretreated with FSAP was not altered (Figure 5).

**Figure 3.** Interactions between FSAP, neuropilin, VEGFR2, and VEGF165: (**A**) Neuropilin-1-Fc or VEGFR2-Fc was immobilized and the binding of wild type (WT)-FSAP (open bars), Marburg I (MI)-FSAP (G534E-SNP) (grey bars), or PPACK-FSAP (hatched bars) was determined in the absence or presence of heparin; (**B**) To immobilized neuropilin-1-Fc, increasing concentrations of WT-FSAP (circles) or MI-FSAP (squares) in the absence (open symbols) or presence (closed symbols) of heparin (filled circles) was added, and FSAP binding was determined; (**C**) Neuropilin-1-Fc or VEGFR2-Fc was immobilized, and the binding of VEGF165 was determined in the absence or presence of heparin (open bars), buffer (grey bars), or FSAP (hatched bars); (**D**) Neuropilin-1-Fc was immobilized and the binding of VEGF165 was determined in the presence of heparin and increasing concentrations of FSAP, as indicated. Results are shown as absorbance (mean + SD of triplicate wells). Error bars in 2B and 2D are smaller than the size of the symbols. \* *p* < 0.05.

**Figure 4.** Effect of FSAP on proliferation and migration of human umbilical vein endothelial cells (HUVEC): basic fibroblast growth factor (bFGF) (40 ng/mL) and/or VEGF165 (20 ng/mL) in the presence of FSAP (12 μg/mL) (dark bars) or buffer control (dotted bars), as well as heparin (10 μg/mL) was preincubated for 60 min at 37 ◦C, and the mixtures were used to stimulate serum-starved HUVEC. (**A**) DNA synthesis was determined using the BrdU incorporation kit; (**B**) Migration was tested in a Boyden chamber. Sphingosine-1-phosphate (S1P) was used a positive control and its concentration was 200 nM. Data are mean + SD of triplicate wells.

**Figure 5.** Effect of FSAP on ERK phosphorylation in HUVEC: Mixtures of FSAP (12 μg/mL), buffer, heparin (10 μg/mL), VEGF165 (20 ng/mL), and/or (bFGF 50 ng/mL) were preincubated for 1 h at 37 ◦C in serum-free medium and then added to cells for 15 min. Cells extracts were analyzed by Western blotting for phosphorylated ERK. Analysis of total ERK was performed to confirm equal loading of gel with lysates. Relative phospho ERK levels were determined by densiometric analysis.

In order to further characterize this result, we also tested the activation of VEGFR-transfected BAF3 cells that are very sensitive to the effects of VEGF. Even in this very sensitive cellular system, FSAP did not inhibit VEGF165- or VEGF121-induced proliferation of BAF-3 cells transfected with VEGFR1 or VEGFR2 (Figure 6A,B). FSAP did not inhibit the effect of bFGF on HUVEC, which is in accordance with our previous results on VSMC [18], but is in contrast to earlier studies on HUVEC [19,20,32]. Expression of neuropilin-1 was observed on both cell types by flow cytometry (data not shown). Thus, in different cellular test systems, FSAP-mediated cleavage of VEGF165 did not alter its ability to activate cellular functions.

**Figure 6.** Effect of FSAP on VEGF-mediated proliferation of VEGFR expressing BAF3 cells: (**A**) VEGFR2-BAF3 cells were stimulated for 4 days with (dark bars) or without VEGF165 (10 ng/mL) (dotted bars) in the absence or presence of FSAP (12 μg/mL), as well as heparin (10 μg/mL); (**B**) VEGFR1-BAF3 cells were stimulated with VEGF165 (circles) or with VEGF121 (squares) in the absence (filled) or presence (open) of FSAP (12 μg/mL), as well as heparin (10 μg/mL). Cell number was determined by the WST-1 assay. Mean + SD of triplicate wells is shown.

*E*ff*ect of FSAP on growth factor-mediated neo-vascularization in matrigel plugs in vivo:* We then tested the effect of FSAP in a model system where VEGF165 interaction with the matrix is important. This model was based on measuring neovascularization in vivo into matrigel plugs that is essentially an extract of tumor extracellular matrix. VEGF165 or bFGF, alone or in combination in the presence of heparin, stimulated the development of new vessels in matrigel, as determined by immunostaining for endothelial and smooth muscle cell markers (Figure 7A; endothelial-specific BS-1 (green) and α-SMA (α smooth muscle specific-actin) (red)). Quantification of the staining showed that the concomitant presence of FSAP reduced neovascularization induced by growth factors (Figure 7B). Enzymatically inactivated PPACK-FSAP did not inhibit growth factor-mediated neo-vascularization (Figure 7A,B), indicating the importance of the FSAP proteolytic activity for this effect. Similarly, the enzymatically inactive MI-isoform of FSAP did not inhibit VEGF165/bFGF-mediated neo-vascularization (Supplementary Figure S2). The effect of VEGF121-mediated neo-vascularization was not inhibited by FSAP (Supplementary Figure S2). Thus, exogenously applied FSAP could inhibit the effects of VEGF165, bFGF, and their combination on neovascularization in matrigel plugs in vivo. Preliminary experiments showed that the neovascularization into matrigel in response to growth factors, in the absence of endogenous FSAP *(Habp2-*/*-* mice), was similar as in WT mice (data not shown).

**Figure 7.** *Cont*.

**Figure 7.** Effect of FSAP on microvascular density in matrigel plugs in vivo: (**A**) Photomicrographs of matrigel plugs after 7 days and stained for BS-1 (FITC, green), α-SMA (α smooth muscle specific-actin) (Cy3, red), and nuclei (DAPI, blue); (**B**) Microvascular density of plugs was determined, bars are means ± SEM (*n* = 7–8), \* *p* < 0.05. Matrigel was supplemented with heparin and with either buffer (dotted bars), FSAP (black bars), or PPACK-FSAP (striped bars), as well as VEGF165 or bFGF, as indicated.

*VEGF expression in the Habp2-*/*- mice subjected to hind limb ischemia:* We then examined whether the levels of the long forms of VEGF were higher in mice in the absence of endogenous FSAP (*Habp2-*/*-* mice). For this, we used the hind limb ischemia model where angiogenesis is induced in the gastrocnemius muscle after femoral artery ligation [16]. Although, collateral growth in the adductor muscle was enhanced in *Habp2-*/*-* mice, capillary density in the gastrocnemius muscle was not altered at 21 days. Western blotting showed that, at an earlier time point of day 3 after ligation, VEGF-A protein was significantly upregulated in the gastrocnemius muscle of *Habp2-*/*-* mice compared to WT mice, whereas no increase was observed in the adductor muscle (Figure 8). At day 7, after ligation, this effect on VEGF-A protein was no longer evident. The VEGF-A isoform detected was largely the VEGF121 dimer (30 kDa), and longer forms seemed not to be produced in this tissue. It was also possible that longer forms were produced and then cleaved to smaller forms. No change in VEGF-A mRNA was detected at the day 3 time point (Supplementary Figure S3). Thus, it was not possible to establish a causal link between the absence of endogenous FSAP and the increased presence of longer forms of VEGF in this experimental system.

**Figure 8.** Changes in VEGF protein levels in the *Habp2*-/- mice after hind limb ischemia: In the hind limb muscles of WT- and *Habp2-*/*-* mice, VEGF-A protein was detected by western blotting at day 3 in the gastrocnemius muscle (top panel) and adductor muscles (bottom panel). VEGF-A was normalized to the expression of cytochrome C oxidase, as measured by western blotting on stripped blots. Relative VEGF-A levels were quantified by densitometry (means + SEM, *n* = 4, \* *p* < 0.05).
