**1. Introduction**

FSAP (factor VII activating protease) is a serine protease that circulates in plasma as an inactive zymogen. It belongs to the family of proteases that also includes the urokinase-plasminogen activator (uPA), tissue-PA (tPA), as well as hepatocyte growth factor activator (HGFA). Although a number of charged molecules can activate FSAP into the active protease (FSAPa), histones are the only endogenous molecules identified so far that can activate the zymogen form into FSAPa in plasma and in vivo [1]. In situations such as tissue injury [2], apoptosis, or necrosis [3], as well as when neutrophils undergo NETosis [4], the DNase activity in blood [5] is likely to release histones. FSAPa in turn can cleave and degrade histones and decrease their toxicity towards cells [3,4]. A single nucleotide polymorphism (SNP) in the FSAP gene, Marburg I (MI, G534E) is associated with a weak proteolytic activity [6] and an increased risk of carotid stenosis [7], stroke [8], venous thrombosis [9,10], liver fibrosis [11], and thyroid cancer [12]. The relationship to venous thrombosis [13] and thyroid cancer [14] was not replicated in a number of subsequent studies.

This relationship between the loss of FSAP activity and diseases is also replicated in FSAP-deficient *(Habp2-*/*-*) mice*. Habp2-*/*-* mice show no explicit characteristics when maintained under standard pathogen-free laboratory conditions and do not exhibit any developmental abnormalities. These mice have been studied in two di fferent models of vascular remodeling. In the wire-induced injury model

of neointima formation, *Habp2-*/*-* mice formed a bigger neointima than wildtype (WT mice) [15]. In the model of hind limb ischemia, arteriogenesis in the adductor muscle was enhanced in *Habp2-*/*-* mice, whereas neovascularization was unchanged in the gastrocnemius muscle [16]. Thus, the lack of *Habp2* gene in mice promotes a more exacerbated repair response that is related to enhanced inflammation and increased activity of the pericellular proteolysis system [15,16].

The effects of FSAP in relation to human diseases and mouse models is likely to be related to proteolysis of different substrates. Although a number of substrates for FSAP have been identified [17] we will focus here only on pathways that are linked to vascular remodeling. Growth factors are cleaved by FSAP, which in some cases leads to a loss of activity, such as platelet derived growth factor-BB (PDGF-BB) [18]. PDGF-BB cleavage leads to an inhibition of vascular smooth muscle cells (VSMC) migration and proliferation, as well as neointima formation. FSAP inhibits basic fibroblast growth factor (bFGF)-mediated endothelial cell proliferation by binding to and/or slowly degrading the growth factor [19] and can also activate bFGF by releasing it from the matrix [20]. Activation of bone morphogenetic protein (BMP)-2 and the conversion of pro-BMP-2 into the active form of cytokine is also a function of FSAP that leads to differentiation of cells [21]. FSAP also cleaves protease activated receptors (PARs)-1 and -3 and influences vascular permeability in combination with hyaluronic fragments of different molecular weights [22]. PAR-1 was identified as a receptor on astrocytes and neurons that mediate the anti-apoptotic effects of FSAP in the context of stroke [23]. Stimulation of VSMC and endothelial cells by FSAP leads to an increased expression of proinflammatory genes in both cells types. Whereas the effect of FSAP could be clearly ascribed to PAR-1 on VSMC, this was clearly not the case for endothelial cells.

Vascular endothelial growth factor (VEGF) is a key factor for determining endothelial lineage, endothelial cell proliferation and migration, as well as recruitment of pericytes and vessel assembly [24]. It belongs to the cysteine knot family of growth factors that include the four genes of the PDGF family as well as placental growth factor (PLGF). Of the four genes encoding for VEGF, denoted A, B, C, and D, VEGF-A is considered to be the most important for hypoxia-driven angiogenesis and is secreted in multiple forms, such as VEGF121, VEGF165, and VEGF189, by alternative splicing [25]. These isoforms have a common N-terminal region for receptor binding, whereas the C-terminal part that mediates binding to co-receptors such as neuropilin and cell- and matrix-associated proteoglycans (ECM) [26] is progressively longer. This C-terminal region has a cluster of negatively charged amino acids and has cleavage sites for uPA, plasmin, and matrix metalloproteinases [27], which regulate VEGF's association with the matrix and co-receptors and results in a different pattern of neovascularization.

With the knowledge that FSAP can cleave proteins at clusters of basic amino acids [17] and that it cleaves PDGF-BB [18], we hypothesized that the homologous protein VEGF-A is also cleaved, and its activity regulated by FSAP. We performed binding and cleavage studies with purified proteins to show that FSAP can indeed cleave long forms of VEGF in their heparin/neuropilin-binding domain and that, as expected, this disturbs their binding properties. However, no modulation of VEGF165 activity in vitro on cellular functions was observed. An inhibiting effect of FSAP in the in vivo matrigel model of neovascularization model was observed, which supports the notion that this effect of FSAP may operate in vivo where matrix association, sequestration, and release of VEGF are decisive.

#### **2. Material and Methods**

*FSAP preparations:* The isolation of wild type-FSAP as well as the MI-SNP (G534E isoform) from human plasma, along with the preparation of enzymatically inactivated Phe-Pro-Arg-chloromethylketone (PPACK)-FSAP has been described before [18,28]. The buffer for storage of FSAP was 0.2 M arginine, 0.2 M lysine, 5 mM citrate, pH 4.5, and was also used at the appropriate dilution to exclude any influence of the vehicle. Because of the rapid auto-activation of the zymogen form of FSAP into FSAPa, under the experimental conditions used in this study, the term FSAP is synonymous for FSAPa.

*Specific cleavage of VEGF isoforms:* VEGF165 or VEGF121 (R & D Systems, Wiesbaden, Germany) (2 μg/mL) was incubated with FSAP (12 μg/mL) in Tris pH 7.4, 100 mM NaCl, 2 mM CaCl2 for 1 h at 37 ◦C in the absence or presence of heparin (10 μg/mL) or aprotinin (15 μg/mL), and the reaction was stopped with SDS sample buffer. Western blots were performed under non-reducing and reducing conditions (β-mercaptoethanol; 10%, vol/vol) and VEGF was detected with a polyclonal goa<sup>t</sup> antibody from R & D Systems. Cleaved VEGF was subjected to amino terminal sequencing using the automated Edman degradation procedure with an online phenylthiohydantoin derivative analyzer (Applied Biosystems, Darmstadt, Germany).

*Binding interactions between FSAP, VEGFR2, VEGF, and neuropilin:* VEGF121 or VEGF165 was immobilized in a Maxisorp microtiter 96-well plate (Nunc, Roskilde, Denmark) at a concentration of 1 μg/mL (50 mL) overnight at 4 ◦C in 50 mM NaHCO3 buffer, pH 9.6. The plate was blocked with 3% (wt/vol) BSA in Tris pH 7.4, 100 mM NaCl. FSAP (0–2 μg/mL) was added to the wells with 0.3% (wt/vol) BSA for 2 h at 22 ◦C. After extensive washing, bound FSAP was detected with an antibody followed by peroxidase-linked secondary antibody. The binding of ligands to BSA-coated wells was used as a blank in all the experiments and was subtracted to obtain specific binding. Similarly, either neuropilin-1-Fc or VEGFR2-Fc (R & D Systems) were immobilized to study the binding of FSAP or VEGF.

*Cellular assays*: Human umbilical vein endothelial cells (HUVEC) were cultivated in ECBM medium (modified MCDB-151) containing 5% (vol/vol) FCS (Promocell, Heidelberg, Germany) on fibronectin-coated dishes. For regular growth of these cells, the medium was supplemented with amphotericin B (50 ng/mL), gentamicin (50 ng/mL), epidermal growth factor (0.1 ng/mL), and bFGF (1.0 ng/mL), as described by the manufacturer. Growth factors were preincubated with FSAP for 60 min at 37 ◦C before stimulation of serum starved cells, as described previously [18]. DNA synthesis was determined using the BrdU incorporation kit from Roche Diagnostics (Mannheim, Germany). Migration was tested in a Boyden chamber on a collagen type I coated membrane with 8 μm pores. Growth factors were preincubated with FSAP for 60 min at 37 ◦C before cell stimulation. Cells were incubated in medium containing 0.1% (vol/vol) FCS in the upper chamber, whereas the lower chamber received the same medium with different additives, as indicated. After an incubation period of 5 h at 37 ◦C, the upper side of the membrane was scraped to remove all cells. Thereafter, the membrane was fixed, stained, and the optical density of each well was measured to quantify cell migration. Cells were lysed in SDS-sample buffer and applied onto SDS-PAGE followed by western blotting and detection of phosphorylated ERK with a phospho-specific antibody with total ERK as a loading control (both from Cell Signaling Technology, Leiden, The Netherlands).

BAF3-VEGFR2 cells were obtained from Steven Stacker and Marc Achen (Ludwig Cancer Research Institute, Melbourne Branch, Australia) and BAF3-VEGFR1 cells were provided by Kari Alitalo (Ludwig Cancer Research Institute, Helsinki, Finland) and were cultured in RPMI-1640 medium containing murine interleukin (IL)-3 (Strathman Biotech, Hannover, Germany). Cell number was determined by the WST-1 assay (Roche Diagnostics).

*Western blotting analysis*: HUVEC were starved for 4 h in serum-free medium and then stimulated for 15 min with the appropriate agonist. Cells were pre-incubated with inhibitors for 30 min before induction with agonist. The experiments were stopped by adding SDS sample buffer containing 10 mM NaF, 1 mM orthovanadate, and 1 mM pyrophosphate, and the samples were processed for western blotting. SDS-PAGE was performed and proteins were transferred to Hybond nitrocellulose membranes (GE Healthcare, Freiberg, Germany). For analysis of western blotting, ECL prime chemiluminescence (GE Healthcare) was used. Tissue pieces were homogenized in a glass homogenizer in TBS (50 mM Tris, pH 7.4, containing 100 mM NaCl) with 1% (w/v) SDS. After centrifugation, the extracts were frozen at −80 ◦C until further analysis. Densitometric analysis was performed to calculate relative expression using ImageJ (NIH, Bethesda, Maryland, USA).

*Matrigel model of* in vivo *angiogenesis*: The matrigel model was performed essentially as described previously [29] and there were 7–8 mice per group. Growth factor-reduced matrigel (BD Biosciences) was supplemented with heparin (200 μg/mL), VEGF165, and bFGF (200 ng/mL each), FSAP, MI-FSAP, or PPACK-FSAP (12 μg/mL), as well as the appropriate volume of bu ffer control. Without the presence of heparin, the neovascularization response is very weak in this model system. The concentration of each growth factor was halved when used in combination. Matrigel was applied subcutaneously into the right and left underside flank of 8–12 week old female C57/BL6 mice. The mice were sacrificed after 7 days and the matrigel plugs were removed, fixed with formaldehyde, and embedded in para ffin. Then, 7 μm serial sections were cut and mounted on slides, depara ffinized in xylene, and rehydrated through graded ethanol washes. After antigen retrieval with Tris-EDTA bu ffer (pH 9.0), the sections were stained with endothelial-specific lectin called bandeirea simplificifolia-1 (BS-1, FITC labelled) [30] and Cy3 labelled antiα-SMA ( α smooth muscle specific-actin) (Sigma) and DAPI. The number of red/green positive vessels counted per section from three di fferent levels of the matrigel plugs and vascular density was expressed as number of vessels per mm2. Percentage area of antiα SMA, BS-1, and DAPI staining was also quantified using ImageJ. All three parameters essentially gave qualitatively very similar results, and only bona-fide vessel density results are presented here. We also localized von Willebrand factor (vWF), an alternative marker for endothelial cells (rabbit polyclonal, DAKO, Glostrup, Denmark), in some staining experiments. Both methods of endothelial cell quantification gave similar results.

*Hind limb ischemia model and western blotting of VEGF-A:* The model of hind limb ischemia and the resulting changes in arteriogenesis and angiogenesis have been described in detail in our earlier manuscript [16]. Western blotting with anti-VEGF-A and anti-COX IV was performed to adjust for di fferences in protein concentrations. Band density was measured using the image analysis system Quantity One system (Biorad, Munich, Germany). Experiments were performed in four mice per group.

*Statistics and reproducibility:* All biochemical and cellular experiments were performed in 3-5 independent experiments. Statistical significance was tested by using analysis of variance (ANOVA) with Bonferroni post-hoc test using the programme Graphpad Prism Biostatistics Software (Graphpad, San Diego, CA). A *p*-value < 0.05 was considered to be significant.

*Study approval:* All procedures involving experimental animals were approved by the local governmen<sup>t</sup> animal care committee (GI 20/10-No. 66/2012) and complied with the Directive 2010/63/EU of the European Parliament.
