Effects of protein and gene transfer of the angiopoietin-1 fibrinogen-like receptor-binding domain on endothelial and vessel organization.

The vessel-stabilizing effect of angiopoietin-1 (Ang1)/Tie2 receptor signaling is a potential target for pro-angiogenic therapies as well as anti-angiogenic inhibition of tumor growth. We explored the endothelial and vascular specific activities of the Ang1 monomer, i.e. dissociated from its state as an oligomer. A truncated monomeric Ang1 variant (i.e. DeltaAng1) containing the isolated fibrinogen-like receptor-binding domain of Ang1 was created and recombinantly produced in insect cells. DeltaAng1 ligated the Tie2 receptor without triggering its phosphorylation. Moreover, monomeric DeltaAng1 was observed to bind alpha(5)beta(1) integrin with similar affinity compared with Tie2. Unexpectedly, in vitro treatment of endothelial cells with DeltaAng1 showed some of the known effects of full-length Ang1, including inhibition of basal endothelial cell permeability and stimulation of cell adhesion as well as activation of MAPKs. Local treatment of the microvasculature of the developing chicken chorioallantoic membrane with the DeltaAng1 protein led to profound reduction of the mean vascular length density, thinning of vessels, and reduction of the number of vessel branching points. Similar effects were observed in side-by-side experiments with the recombinant full-length Ang1 protein. These effects of simplification of the vessel branching pattern were confirmed through local gene transfer with lentiviral particles encoding DeltaAng1 or full-length Ang1. Together, our findings suggest a potential use for exogenous Ang1 in reducing rather than increasing vascular density. Furthermore, we show that the isolated receptor-binding domain of Ang1 is capable of mediating some effects of full-length Ang1 independently of Tie2 phosphorylation, possibly through integrin ligation.

The angiopoietin family of morphogens exerts essential functions for vascular and lymphatic growth and remodeling (1). Angiopoietin-1 (Ang1) 1 is sequestered by periendothelial and vascular smooth muscle cells and specifically acts on endothelial cells (ECs) through binding and activation of the cellsurface tyrosine kinase receptor Tie2 (2). Unlike classic angiogens proposed for therapy, such as the vascular endothelial growth factor (VEGF) and fibroblast growth factors, Ang1mediated Tie2 activation fails to stimulate mitogenesis of ECs. Rather, Ang1/Tie2 signaling appears to affect EC activities that ultimately lead to the formation of more mature and stable vessel walls. Specifically, Ang1 was shown to mediate antipermeability, anti-inflammatory, and anti-apoptotic effects (3)(4)(5)(6)(7). Mice transgenically overexpressing Ang1 show increased vascularization and leakage-resistant blood vessels (8,9). Also, systemic adenovirus-mediated Ang1 production protect vessels from acute leaking in adult mice (10).
The Ang1-induced vessel-stabilizing effects appear to depend on tightening of EC cell junctions as well as on increased association of the endothelium with mural cells (i.e. periendothelial and smooth muscle cells) recruited to the vessel wall (5,11). Together, these positive effects suggest Ang1 as candidate agent for therapeutic revascularization approaches that aim at the growth of lasting blood vessels in a more coordinated way. Administration of Ang1 as a singular agent may prevent or reverse blood vessel leakiness, such as in the conditions of diabetic retinopathy: intravitreal injection of recombinant Ang1 protein, i.e. Ang1* (see below), proved successful in protecting rat retinal vasculature from inflammation, EC injury, and blood-retinal barrier breakdown, which are typically associated with this disease (12). Co-administration of Ang1 together with VEGF might help inhibit VEGF-induced vessel permeability and facilitate the growth of new and non-leaky vasculature in chronic wound healing approaches as well as in reperfusion therapy of ischemic heart or limb (13,14). Moreover, studies of tumor growth with human colon or breast cancer cell lines that overexpress Ang1 indicate that Ang1 may inhibit angiogenesis in tumors, possibly by its effect in stabilizing interactions between ECs and pericytes and a concomitant inhibition of initiation of tumor angiogenesis (15)(16)(17).
Ang1 is a 498-amino acid molecule arranged in an N-terminal domain/coiled-coil domain/fibrinogen-like domain order (18). The fibrinogen-like domain mediates ligand activity (19). Dimerization and superclustering of the fibrinogen-like domain, as found in native Ang1, were found to be critical for triggering phosphorylation of the Tie2 receptor. Physiologically, native Ang1 physically associates with components of the extracellular matrix (20). The mechanism of liberation from the extracellular matrix is elusive. Native forms of Ang1 are very difficult to produce: the complex quaternary structure of Ang1 and its stickiness to extracellular matrix components have prevented production at high protein titers. Multiple recent attempts have been directed toward genetically engineering multimeric Ang1 constructs that are less difficult to purify but still retain their ability to stimulate Tie2 phosphorylation. Examples are the receptor-binding domain of Ang1 coupled to the angiopoietin-2 oligomerization domain (Ang1*) (21), constructs that exhibit a tandem arrangement of the Ang1 fibrinogenlike domains fused to the immunoglobulin Fc domain for additional multimerization (18), and constructs that exhibit pentamerization by replacement of the N-terminal domain with the short coiled-coil domain of the cartilage oligomeric matrix protein (22).
In this study, we sought to explore the activities of Ang1 dissociated from its state as a multimer. We studied a new variant (⌬Ang1) that is monomeric and comprises the soluble fibrinogen-like receptor-binding domain of Ang1. We describe unexpected functionalities of ⌬Ang1 in mediating effects in vitro and in vivo that copied those of oligomeric fulllength Ang1.

Recombinant Ang1
Vectors-Human full-length Ang1 was cloned from human aortic smooth muscle cells (kindly provided by Dr. P. Benedikt, University Hospital Zurich). Cloning was performed following standard procedures (23). All plasmids, their features, cloning strategies, and the oligonucleotide primers used for PCRs are given in Supplemental Tables 1 and 2.
Expression and Purification of ⌬Ang1 and Full-length Ang1 in Insect Cells-The⌬Ang1 protein was constructed for use in the insect cell expression vector pFMel (24) by PCR-directed mutagenesis using the full-length cDNA of human Ang1 as template. The ⌬Ang1 protein construct is defined by the following amino acid sequences: an insect cell secretion signal sequence (MKFLVNVALVFMVVYISYIYA) followed by DP (resulting from a BamHI site), the factor XIII transglutaminase substrate sequence NQEQVSPL, a short linker sequence (PVELP), a plasmin-sensitive cleavage motif (LIKMKP) (25), and AS (resulting from an NheI site) followed by the C-terminal fibrinogen-like domain of Ang1 beginning with RDCAD. The final ⌬Ang1 protein contains the additional 23-amino acid motif DPNQEQVSPLPVELP-LIKMKPAS in front of the Ang1 fibrinogen-like domain. The nonglycosylated form of ⌬Ang1 has a theoretical molecular mass of 27,212 Da and an isoelectric point of 7.75. The expression and purification in Sf9 insect cells of the Myc-tagged full-length Ang1 protein have been described previously (26).
Recombinant baculovirus encoding ⌬Ang1 was used to produce soluble ⌬Ang1 protein in the Sf21 insect cell system. All insect cell culture reagents (i.e. Sf21 cells, SF-900 II SFM medium with L-glutamine, antibiotic/antimycotic solution, and Cellfectin) were obtained from Invitrogen and used according to the manufacturer's instructions. The optimal expression parameters (i.e. infection cell density, 2.1 ϫ 10 6 cells/ml; multiplicity of infection of 1; and time of harvest, 72 h postinfection) were elaborated according to the protocol of Weber et al. (27). The insect cell supernatant was clarified by centrifugation and filtration. Subsequent column chromatography steps were performed using an ÄKTA fast performance liquid chromatography system (Amersham Biosciences, Uppsala, Sweden) at 4°C. In the first step, the insect cell supernatant was diluted 1:1 with 20 mM sodium phosphate (pH 6.8) and fractionated on a cation exchange HiPrep SP FF column by gradient elution to 20 mM sodium phosphate (pH 6.8) and 1 M NaCl. Subsequent hydrophobic interaction chromatography was performed by applying ⌬Ang1 supplemented with 0.7 M (NH 4 ) 2 SO 4 to a HiTrap phenyl HP column equilibrated with 50 mM sodium phosphate (pH 6.8) and 1 M (NH 4 ) 2 SO 4 . The protein was eluted with a gradient to 50 mM sodium phosphate (pH 6.8). The ⌬Ang1-containing eluate was then dialyzed overnight against Tris-buffered saline (TBS; 20 mM Tris (pH 7.6) and 150 mM NaCl) using a Spectrapor dialysis membrane with a molecular mass cutoff of 12-14 kDa (Spectrum Laboratories, Inc., Rancho Dominguez, CA). The ⌬Ang1 dialysate was concentrated using 20-ml Vivaspin columns (Vivascience Ltd., Lincoln, United Kingdom) with a 5-kDa molecular mass cutoff and finally polished by size exclusion chromatography using a Superdex 200 HR 10/30 column. The ⌬Ang1 protein was filter-sterilized, aliquoted, and stored at Ϫ80°C. Protein concentration was determined using the BCA assay (Pierce). Purity was verified by SDS-PAGE and Coomassie Blue staining. Identity was confirmed by immunoblotting with Ang1-specific antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA).
Cloning of Lentiviral Vectors of ⌬Ang1 and Full-length Ang1-⌬Ang1 and full-length Ang1 were cloned into the third-generation lentiviral vectors pBM14 and pMF359, respectively, using SbfI and PmeI as restriction sites (28). The lentiviral particles were produced and collected as described previously (28). Lentiviral particle-mediated production of ⌬Ang1 and full-length Ang1 proteins was validated in murine CHO-K1 cell culture experiments using a previously described protocol (28). The Ang1 proteins were resolved by 10% reducing SDS-PAGE and detected by immunoblotting with Ang1-specific antibodies. Control cultures were transduced with lentiviral particles encoding green fluorescent protein (GFP).
Solid-phase Binding Assay-96-Well Maxisorb microtiter plates (Nunc, Naperville, IL) were incubated for 24 h at 4°C with ⌬Ang1 at 2 g/ml in TBS. The plates were washed three times and then blocked for 1 h at room temperature with 1% bovine serum albumin (BSA) in TBS. After three washes with TBS, the plates were incubated for 1 h at room temperature with biotin-labeled soluble Tie2-immunoglobulin chimeric protein (Tie2-Fc) at the concentrations indicated in Fig. 1C in TBS and 0.1% BSA. Nonspecific binding to the plates was assessed by incubating biotin-labeled Tie2-Fc in a 20-fold molar excess of unlabeled Tie2-Fc. After three washes with TBS, alkaline phosphatase-coupled streptavidin was added (1:1000 in TBS), and the plate was incubated for 1 h at room temperature. After three washes with TBS, p-nitrophenyl phosphate was added (10 mM in 50 mM Tris-HCl (pH 9.5), 150 mM NaCl, and 10 mM MgCl 2 ). Absorbance at 405 nm was recorded when a maximal absorbance of 1.0 -1.5 was reached.
⌬Ang1 Pull-down Assay-Binding of soluble ⌬Ang1 to immobilized Tie2-Fc was assessed in a pull-down assay. 10 g of ⌬Ang1 were incubated for 4 h at room temperature with admixtures of 50 g of Tie2-Fc and 25 l of protein A-agarose in 0.5 ml of 0.1% BSA. Control incubations were performed by incubating ⌬Ang1 or Tie2-Fc with protein A-agarose. After three washes with TBS and 0.5% Tween 20, proteins bound to protein A-agarose were subjected to SDS-PAGE and stained with Coomassie Blue.
Preparation of ⌬Ang1-Sepharose-Coupling of the ⌬Ang1 protein to CNBr-activated Sepharose 4B (Sigma) was performed following the standard protocol for affinity chromatography recommended by Amersham Biosciences. 2 mg of ⌬Ang1 in 0.1 M NaHCO 3 (pH 8.3) were coupled to 1 ml of resin for 60 min. Coupling was complete as determined by SDS-PAGE and Coomassie Blue staining of unreacted protein. Control resin (Sepharose alone) was processed identically without the ⌬Ang1 protein. For blocking nonspecific protein-binding sites, ⌬Ang1-Sepharose and control Sepharose were preincubated with 1% BSA in TBS (pH 7.5) before use in affinity binding experiments.
Affinity Binding Experiments with ⌬Ang1-Sepharose-Affinity isolation experiments were performed with lysates prepared from mouse MS1 endothelial cells or human foreskin fibroblasts as indicated in Fig. 3. For affinity precipitation experiments, two confluent T75 flasks of either MS1 cells or fibroblasts were overlaid with ice-cold TBS (1.5 ml/flask), collected by scraping, transferred to 15-ml centrifuge tubes, and pelleted by centrifugation. For protein solubilization, the pelleted cells were lysed in 3 ml of ice-cold detergent buffer (50 mM n-octyl ␤-D-glucopyranoside (AppliChem, Darmstadt, Germany), 10 mM Tris, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 1 mM sodium orthovanadate, and Complete proteinase inhibitor mixture (Roche Diagnostics)) and sonicated for 3 min. The extract was then divided into two Eppendorf tubes and extracted for 30 min in an Eppendorf shaker. The extracts were clarified by centrifugation for 20 min at 14,000 rpm. The resulting supernatants were subjected to a second preclearing step by a 20-min incubation with 20 l of Sepharose alone. Finally, the cleared lysates were incubated with 20 l of ⌬Ang1-Sepharose or control resin for 2 h on a rotating wheel. After incubation, the resins were collected by short centrifugation and washed twice with 1 ml of detergent buffer. Bound proteins were eluted by boiling in 50 l of 2ϫ reducing SDS sample buffer, separated by 10% SDS-PAGE, electrotransferred onto nitrocellulose, and immunoprobed with rabbit anti-Tie2 polyclonal antibodies (sc-324, Santa Cruz Biotechnology, Inc.) at 1:200 dilution or with anti-␣ 5 integrin antibodies (sc-10729, Santa Cruz, Biotechnology, Inc.) at 1:200 dilution. Horseradish peroxidase-conjugated secondary antibodies were obtained from DakoCytomation AG (Zug, Switzerland). Protein signals were visualized by the enhanced chemiluminescence method.
Column affinity chromatography was performed in parallel with ⌬Ang1-Sepharose and Sepharose alone (control). MS1 endothelial cells from seven T75 flasks were used to prepare 3-ml protein lysates using the same buffers and methodology as described above. 0.9-ml volumes of each ⌬Ang1-Sepharose and Sepharose alone were provided in Econo-Columns (Bio-Rad) and equilibrated in detergent buffer before 1.5 ml of protein lysate were added to each column. The mixtures of resin and lysate were incubated for 2 h at 4°C on a rotating wheel. The flow-through was collected, and each resin was washed four times with 1 ml of detergent buffer. Serial elution was performed with a step gradient of sodium chloride added to detergent buffer (0.2-1 M). In each elution step, 0.5 ml of elution buffer was applied to the resin and incubated for 5 min before the eluate was collected. The final elution was performed with 2 mM EDTA in detergent buffer. Aliquots of lysate, wash fractions, and eluates were analyzed by SDS-PAGE and immunoblotting as described above.
Cell Binding and MAPK Phosphorylation Assay-HUVEC attachment assays were performed as described previously (31). Attached cells were fixed with 4% paraformaldehyde in phosphate-buffered saline, followed by crystal violet staining. Phase micrographs of centerfields were taken using the ϫ4 objective of a Zeiss Axiovert 135 microscope equipped with a digital camera. Cells were counted from printed micrographs. MAPK phosphorylation assays of HUVECs plated on tissue culture polystyrene dishes coated with BSA, ⌬Ang1, or fibronectin were performed similarly as described (31). Total lysates derived from bound and unbound cells were separated by 10% SDS-PAGE and immunoblotted with anti-phospho-p44/42 MAPK antibodies (Cell Signaling Technology Inc., Beverly, MA). After stripping, the membranes were reprobed with anti-p44/42 MAPK antibodies.
Permeability Measurements-Permeability experiments were performed similarly as described previously (32). EA.hy926 cells (7,29) were seeded onto microporous polycarbonate membrane inserts of 12well Transwell chambers (0.4-m pore size; Costar, Cambridge, MA) and grown in DMEM and 10% FBS. After reaching confluency, EA.hy926 cells were incubated for 24 h with increasing amounts of ⌬Ang1 that were added to the medium of both the apical and basolateral chambers. Radiolabeled [ 3 H]inulin (42.5 GBq/mmol; Amersham Biosciences, Buckinghamshire, United Kingdom) was added at a final concentration of 4 M to the apical chamber (total volume of 0.5 ml). The appearance of radioactivity in the basolateral compartment (total volume of 1.5 ml) was determined after 1 h by scintillation counting of 20-l aliquots of the basolateral medium. The permeability of [ 3 H]inulin is expressed as a percent of the control value of untreated cells grown under identical conditions.
Preparation of Fibrin Gels-Human fibrinogen (Fluka AG, Buchs, Switzerland) was prepared as described (33). Fibrin gels for experiments on the chicken chorioallantoic membrane (CAM) were prepared by mixing the following components to the indicated final concentrations: 7 mg/ml human fibrinogen, 2 units/ml human factor XIII (kindly provided by Baxter AG, Vienna, Austria), 2 NIH units/ml of human thrombin (Sigma), and 2.5 mM CaCl 2 . Gels were allowed to polymerize for at least 30 min at 37°C in a humidified atmosphere before application to the chicken CAM.
Embryonic Chicken CAM Assay-Grafting experiments were performed on chicken embryos grown by the shell-free culture method (34).
Grafts were formed as disc-shaped fibrin hydrogels (fibrinogen concentration of 7 mg/ml) in a 50-l volume containing 0.5-6 g of ⌬Ang1 or 6 g of Myc-tagged full-length Ang1. The grafts were placed on top of the growing CAM at embryonic day 9. Parallel grafting experiments were performed with plain fibrin gels. On embryonic day 11, microvascular growth and blood flow at and around the fibrin graft were monitored in the living embryo with a Zeiss Axiovert 135 fluorescence microscope. Observations were performed after intravenous injections of 0.1 ml of 2.5% fluorescein isothiocyanate (FITC)-dextran (M r 2,000,000; Sigma) (35).
Transduction of the embryonic CAM with lentiviral particles encoding the gene of ⌬Ang1, full-length Ang1, or green fluorescent protein was performed as described previously (28). Briefly, 50-l lentiviral particles mixed with 0.5 l of CellTracker TM Orange chloromethylbenzoylaminotetramethyl-rhodamine (Molecular Probes, Inc., Eugene, OR) were applied pointwise on top of the growing CAM at embryonic day 9. The pattern of FITC-dextran-perfused vessels at and around the application site was monitored on embryonic day 12 with a Leica fluorescence microscope equipped with an XF114 filter (Omega Optical Inc., Brattleboro, VT) and an LE CCD Optronics video camera (Visitron System, Puchheim, Germany) using ϫ50 and ϫ100 magnifications, respectively.
Morphometric Analysis of the CAM-Grayscale prints from images of the CAM vasculature perfused with FITC-dextran were used for morphometric assessment of two parameters of microvascular growth, i.e."mean vascular length density" and "vascular end point density." For this, the vasculature was skeletonized on grayscale prints and analyzed by the method of grid intersection as described (36,37). CAM morphometric data were analyzed using the General Linear Model for statistical evaluation (38). The answering variable Y (vascular length density or end points) is estimated as follows: Cov. y is the mean value of the intersections or end points of zones A (Ͻ1.6 mm away from the margin of the fibrin graft), B (1.6 -3.1 mm), and C (3.1-4.7 mm) of each CAM, whereas the means of zones D (4.7-6.3 mm) and E (6.3-7.8 mm) give the covariable Cov to eliminate variations introduced by the different sizes of the CAM. is the additive constant, ␣ i (i ϭ 1, 2, 3, 4, 5) is the effects of the different growth factor concentrations, ␤ j (j ϭ 1, 2, 3) is the effects of zones A-C, (␣␤) ij is the effects of the combinations of ␣ i and ␤ j (interactions), ⑀ ijk (k ϭ 1, 2..n ij ) is the errors (⑀ ijk ϭ y ijk Ϫ Y ijk ), n ij is the number of y ij (5 ϫ 3 ϭ 15 combinations), b is the regression coefficient of Cov (⌺␣ i ϭ ⌺␤ j ϭ ⌺(␣␤) ij ϭ ⌺⑀ ijk ϭ 0 and ⌺⑀ ijk 2 ϭ minimum). A complete model including Cov explained the data by 91.2% with p Ͻ 10 Ϫ6 , whereas a model where Cov was omitted explained the data by only 71.4%. For further analysis, the complete model including Cov was used. The significance level was set to 5%, and the confidence interval of the parameters was set to 95% (for interpretations, see Refs. 39 and 40).

⌬Ang1 Is a Soluble Minimal Variant of Ang1 That Binds
Tie2-An Ang1 variant (⌬Ang1) that comprised the fibrinogenlike receptor-binding domain (i.e. residues 284 -498) of human Ang1 was created by recombinant DNA methodology. This ⌬Ang1 construct was designed to contain at its N terminus an additional 23-amino acid sequence tag for prospective application in controlled release schemes using factor XIII-mediated affixation of ⌬Ang1 in engineered fibrin hydrogel matrices (25,33). As shown for a variety of different growth factor proteins, e.g. nerve growth factor (25), bone morphogenetic protein-2 (41) and VEGF (42,43), this N-terminal sequence tag does not compromise the activity of the genetically engineered variant compared with the native sequenced protein.
A baculovirus/Sf21 insect cell system was used to produce the ⌬Ang1 protein. ⌬Ang1 was secreted at high titers into the serum-free insect cell culture supernatant and could be purified by fast performance liquid chromatography using a sequence of cation exchange, hydrophobic interaction, and size exclusion chromatography steps (Fig. 1A). SDS-PAGE and Coomassie Blue staining verified the ⌬Ang1 fraction to be a protein of ϳ30 kDa, in good accordance with a theoretical molecular mass of 27,212 Da for non-glycosylated ⌬Ang1. Typical yields were 2.6 -3.5 mg of ⌬Ang1 protein with an estimated purity of Ն98% from 1 liter of insect cell culture medium. The identity of the ⌬Ang1 protein was validated by immunoblotting with Ang1-specific antibodies (data not shown). The monomeric status of ⌬Ang1 was confirmed by size exclusion chromatography and MALDI-TOF spectrometry and revealed a peak at 28,615 Da (Fig. 1B). The elevation of the experimental molecular mass over the theoretical molecular mass of ⌬Ang1 likely reflects its glycosylation in insect cells. The ⌬Ang1 compound was very stable as a monomer and did not form aggregates even after prolonged storage, as confirmed by a 6-month follow-up with MALDI-TOF spectrometry. Only minute amounts of dimeric ⌬Ang1 were detectable at 57,477 Da. The spectrum shown in Fig. 1B is derived from a ⌬Ang1 preparation that was stored for 3 months at Ϫ80°C.
A direct interaction between ⌬Ang1 and its cognate receptor Tie2 was demonstrated in solid-phase binding assays (Fig. 1C). For this purpose, ⌬Ang1 was adsorbed to tissue culture polystyrene plates and subsequently incubated with increasing doses of soluble biotinylated Tie2-Fc chimeric protein. Binding of Tie2-Fc to surface-coated ⌬Ang1 was measured by colorimetry, revealing specific and saturable binding. Moreover, soluble ⌬Ang1 was found to bind Tie2-Fc complexed with protein A-agarose. The ⌬Ang1 protein captured by Tie2-Fc/protein A-agarose was analyzed by SDS-PAGE with Coomassie Blue staining, revealing that immunoisolation of Tie2-Fc resulted in a concomitant pull-down of ⌬Ang1 (Fig. 1D). Binding of the monomeric ⌬Ang1 compound did not trigger phosphorylation of the Tie2 receptor: examination of immunoprecipitated tyrosine-phosphorylated proteins in lysates of HUVECs or EA.hy926 cells incubated with the ⌬Ang1 construct did not reveal any phosphorylated Tie2 receptor (see figure in supplemental data). These findings are consistent with previous results showing that monomeric forms of the Ang1 fibrinogen-like domain are capable of binding the endothelial Tie2 receptor, but do not stimulate its multimerization and activation, as indicated by tyrosine phosphorylation (18).
In Vitro Effects of ⌬Ang1: Promotion of EC Adhesion and Spreading, MAPK Phosphorylation, and Reduction of Transendothelial Permeability-Recent studies have shown that substrates adsorbed with full-length Ang1 can directly support HUVEC adhesion and spreading, possibly through binding to the ␣ 5 ␤ 1 integrin receptor (31). Our HUVEC culture experiments on surface-adsorbed ⌬Ang1 demonstrated that this celladhesive character was retained in the isolated Ang1 fibrinogen-like domain. Adhesion and spreading of HUVECs cultured on tissue culture wells coated with increasing concentrations of ⌬Ang1 increased in a concentration-dependent manner ( Fig.  2A). Substrates of BSA or heat-denatured FBS served as negative and positive controls, respectively. At the highest ⌬Ang1 coating concentration (i.e. 100 g/ml), adhesion to ⌬Ang1 was comparable with that to heat-denatured FBS. Adhesion mediated by ⌬Ang1 was associated with induction of MAPK phosphorylation, presumably due to signaling downstream of activated integrins (31): immunoblotting of HUVEC lysates with anti-phospho p44/42 MAPK antibodies revealed an increase in phospho-MAPK levels in a ⌬Ang1 concentration-dependent manner. As a reference and positive control, a more profound MAPK phosphorylation signal was found in HUVECs plated on fibronectin (Fig. 2B, FN lane).
A previous study has shown that native Ang1 is capable of lowering the permeability of high molecular mass proteins through confluent layers of the EC line EA.hy926 (7). We tested diffusion of the low molecular mass compound inulin (5 kDa) (32) across EA.hy926 cell layers that were grown on Costar microporous membrane inserts in Transwell chambers and treated with ⌬Ang1 for 24 h (Fig. 2C). Our measurements revealed that incubation with ⌬Ang1 resulted in a significant dose-dependent reduction of [ 3 H]inulin flux across the endothelial monolayer. Doses as low as 0.1 g of ⌬Ang1/ml of culture medium were found to be effective in lowering [ 3 H]inulin diffusion. At the best performing dose of 0.15 g/ml ⌬Ang1, transendothelial permeability was reduced by 55 Ϯ 11%. Collectively, these in vitro results indicate that ⌬Ang1 is capable of inducing some effects on human ECs that were previously associated with full-length Ang1. These include activation of the MAPK pathway and inhibition of EC monolayer permeability. Thus, it appears that some of the signaling effect of Ang1 can proceed independently of Tie2 phosphorylation.
⌬Ang1 Binds ␣ 5 ␤ 1 Integrin with Similar Affinity Compared with Tie2-The cell adhesion studies of Carlson et al. (31) suggest that Ang1 is a ligand of ␣ 5 ␤ 1 integrin; direct binding between Ang1 and ␣ 5 ␤ 1 integrin remained to be demonstrated. Here, we explored whether the ⌬Ang1 monomer is capable of binding ␣ 5 ␤ 1 integrin, in addition to Tie2. To this end, affinity isolation was performed with detergent lysates of ECs or fibroblast cells using the ⌬Ang1 protein coupled to Sepharose as bait (Fig. 3). Lysates of MS1 endothelial cells (which express ␣ 5 ␤ 1 integrin as well as Tie2) or human fibroblasts (which express ␣ 5 ␤ 1 integrin but not Tie2) prepared in n-octyl ␤-Dglucopyranoside-containing lysis buffer were incubated with ⌬Ang1-Sepharose or Sepharose alone (control). Proteins captured by ⌬Ang1-Sepharose were separated by SDS-PAGE and analyzed by immunoblotting with antibodies to ␣ 5 integrin or Tie2, revealing that ␣ 5 ␤ 1 integrin bound specifically to ⌬Ang1-Sepharose (Fig. 3A). Moreover, providing an internal positive reference in this assay, we found that Tie2 from MS1 endothelial cell lysate bound to ⌬Ang1 (Fig. 3B).
⌬Ang1 column affinity chromatography followed by salt gradient elution was performed to examine the relative binding affinities between ␣ 5 ␤ 1 integrin or Tie2 and ⌬Ang1 (Fig. 3,  C-E). For this, detergent lysates of MS1 cells were incubated with ⌬Ang1-Sepharose or Sepharose alone (control). For elu-tion, bound receptors were challenged with a step gradient of 0.2-1 M sodium chloride in detergent buffer and a final elution step with 2 mM EDTA in detergent buffer. Immunoblotting of eluates confirmed the specific binding of ␣ 5 ␤ 1 integrin as well as Tie2 to the ⌬Ang1 column (Fig. 3, D and E), but not to the control column (Fig. 3E). The amounts of ␣ 5 integrin and Tie2 in the eluates peaked at 0.5 and 0.4 M sodium chloride, respectively. Very little protein was contained in eluates derived from elution with sodium chloride concentrations above 0.6 M, and no protein was detectable in eluates with EDTA. Hence, at least under these experimental conditions, it appears that the ⌬Ang1 monomer exhibits similar binding affinity for ␣ 5 ␤ 1 integrin and Tie2.
Collectively, these binding experiments suggest that ␣ 5 ␤ 1 integrin could act as an alternative receptor for ⌬Ang1. As described above, ligation of ␣ 5 ␤ 1 integrin resulted in activities characteristic of integrin activation, including cell adhesion and activation of MAPKs.
Effects of Exogenous Ang1 on the Developing CAM Vasculature-The in vivo effects of exogenous administration of fulllength Ang1 and ⌬Ang1 were examined through protein and gene transfer in the embryonic chicken CAM model of developing angiogenesis. For this, Ang1 was delivered locally as a protein or gene to the developing CAM at embryonic day 9. 50-l fibrin gel matrix grafts (42) admixed with 6 g of fulllength Ang1 or ⌬Ang1 compound were used to locally release Ang1 proteins to the CAM. Control experiments were performed with grafts made of fibrin only. The microvascular growth pattern in the area of the developing CAM adjacent to the graft application site was examined on embryonic day 11 by in vivo fluorescence microscopy following perfusion with FITCdextran to monitor the vessels in the living embryo. Visual inspection of the CAM microvasculature revealed that treatment with the full-length Ang1 protein (Fig. 4C) as well as the ⌬Ang1 protein (Fig. 4B) resulted in the formation of straight, thinned vessel structures that were associated with a striking decrease in the number of branching points of the vessels and their connections with the underlying capillary plexus. In contrast, the vessel pattern remained unchanged in experiments with control grafts made of fibrin alone (Fig. 4A).
Gene transfer experiments corroborated the effects observed when the Ang1 or ⌬Ang1 protein was administered to the CAM surface. We have previously reported third-generation lentiviral expression vector platforms that tested suitable for infection of the CAM with human VEGF (28). Here, we utilized this vector platform to accommodate transgenes corresponding to native human Ang1 or its derivative, ⌬Ang1. Expression of Ang1 and ⌬Ang1 was validated in lentivirus-transduced murine CHO-K1 cells by SDS-PAGE analysis and immunoblotting with Ang1-specific antibodies (Fig. 4G). 50-l volumes of high titer preparations of recombinant lentiviral particles were applied pointwise on top of the growing CAM at embryonic day 9. Vessel pattern formation at the application site was examined on embryonic day 12. Again, a strong effect of simplification of the vessel branching pattern became apparent in experiments with both full-length Ang1 (Fig. 4F) and ⌬Ang1 (Fig. 4E). This effect was completely absent in control experiments with lentiviral particles encoding a control protein that is irrelevant in angiogenesis, GFP (Fig. 4D). Morphologies observed with protein or genetic administration of full-length Ang1 or ⌬Ang1 were qualitatively indistinguishable.
Morphometric Analyses of Exogenous Ang1 Effects on the CAM Vasculature-The effects of Ang1 or ⌬Ang1 protein delivery were carefully quantified by side-by-side experimentation and morphometric analyses using the method of grid intersection (Fig. 5) (42). We employed two parameters of microvascular growth, i.e. mean vascular length density as an indicator of arterial/venous vessel formation and density of vascular end points as a measure of connectivity between arterial/venous feeding vessels and capillaries. Images of the FITC-labeled CAM were acquired from concentric optical zones between 0 and 7.1 mm from the margin of the site of graft application. Administration of the ⌬Ang1 protein significantly reduced both the mean vascular length density and the number of vascular end points/unit area compared with grafts of fibrin alone (Fig.  5, A and B). The effects were strongest in areas directly in FIG. 3. Monomeric ⌬Ang1 binds ␣ 5 ␤ 1 integrin and Tie2 with similar affinity. Cell adhesion studies have suggested that full-length Ang1 is a ligand of ␣ 5 ␤ 1 integrin (31). Affinity precipitation and column affinity chromatography were performed to assess whether such binding could occur between ␣ 5 ␤ 1 integrin and ⌬Ang1 and to evaluate the relative affinity of ␣ 5 ␤ 1 integrin compared with that of Tie2. A and B, ⌬Ang1-Sepharose (⌬Ang1) was incubated with detergent lysates prepared from mouse MS1 endothelial cells (which express both Tie2 and ␣ 5 ␤ 1 integrin) or human fibroblasts (which express ␣ 5 ␤ 1 integrin but not Tie2). Parallel incubations were performed with Sepharose alone (control). After several washes, bound proteins were eluted and analyzed by 10% SDS-PAGE and immunoblotting (IB) with antibodies specific for Tie2 or ␣ 5 integrin. This revealed specific pull-down by ⌬Ang1 of both ␣ 5 ␤ 1 integrin and Tie2 receptors from lysates of MS1 endothelial cells; ␣ 5 ␤ 1 integrin from lysates of human fibroblasts was captured. WCL, whole cell detergent lysates. The positions of ␣ 5 integrin and Tie2 are indicated by arrows. C-E, the relative binding affinity of ␣ 5 ␤ 1 integrin versus Tie2 for ⌬Ang1 was assessed via column affinity chromatography using detergent lysates from MS1 cells. For elution, bound proteins were challenged with a step gradient of increasing sodium chloride concentrations in detergent buffer as indicated (x M NaCl). The protein eluates were then analyzed by SDS-PAGE and immunoblotting with anti-␣ 5 ␤ 1 or anti-Tie2 antibodies. ␣ 5 ␤ 1 integrin and Tie2 could be detected in eluates from ⌬Ang1-Sepharose, but not from Sepharose alone (control). The elution profiles suggest that both receptors bind to monomeric ⌬Ang1 with roughly comparable low affinity. Low binding affinities are typical for interaction of integrins with extracellular matrix ligands. The similar low affinity of Tie2 for monomeric ⌬Ang1 conforms published data showing that, in the absence of oligomeric presentation of the Ang1 receptor-binding domain, the affinity for Tie2 is dramatically reduced (18).
contact with the fibrin graft. Both the mean vascular length density and the vascular end point density were significantly lowered by all tested doses of ⌬Ang1, i.e. 0.5 g/application site (lowest dose) to 6 g (highest dose). The effects on the mean vascular length density and the vascular end point density at 6 g of ⌬Ang1 matched those at 6 g of full-length Ang1 (Fig. 5,  C and D). For all data sets, the General Linear Model for statistical evaluation was used to establish significance as described under "Materials and Methods." Collectively, our results show the effects of exogenous Ang1 administration for simplification of the vessel branching pattern in the developing CAM. Furthermore, a functionality of the isolated receptor-binding domain of Ang1 in inducing the effects of full-length Ang1 is indicated. Morphometric quantification revealed that the actions of full-length Ang1 and ⌬Ang1 in the developing CAM vasculature are quantitatively indistinguishable.

DISCUSSION
Pharmacological and biotechnological issues have presented limits to therapeutic angiogenesis. For example, in the case of VEGF, complications such as vascular leakage may arise from tissue exposure to high VEGF concentrations. Such untoward behavior has driven the development of strategies for co-administration of growth factor agents such as Ang1 (44) and the platelet-derived growth factor (45), which could potentially help grow more functional and persistent blood vessels. In the specific case of Ang1, a challenge has remained as to how to produce the complex multimeric structure of Ang1 in a signaling-competent and clinically relevant form. In this study, through the testing of ⌬Ang1, we identified the effects of the monomeric Tie2 receptor-binding domain of Ang1 that unexpectedly parallel those of native Ang1, such as inhibition of basal transendothelial permeability and reduction of the branching of developing vessel structures. Furthermore, our biochemical data together with published data (31) indicate that ␣ 5 ␤ 1 integrin, besides Tie2, represents an additional endothelial cell-surface receptor of ⌬Ang1.
Functional competence has typically been assigned to native FIG. 4. Exogenous administration of Ang1 through protein and gene transfer in the developing chick CAM leads to simplification of the vessel branching pattern. Side-by-side examination of the effects of full-length Ang1 and ⌬Ang1 delivered locally to the chick CAM. A-C, 50-l fibrin gel matrix grafts were used as vehicles for the local release of Ang1 or ⌬Ang1. Control grafts of fibrin only were prepared. The gels were grafted on top of the growing CAM at embryonic day 9 and cultured for 2 days before the graft areas were optically examined. Microvascular growth was examined by in vivo fluorescence microscopy using perfusion with FITC-dextran of the living embryo. A, shown are the results with fibrin alone. B and C, shown is the microvascular growth response to the fibrin gel matrix formulated with 6 g of ⌬Ang1 or full-length Ang1, respectively. ⌬Ang1 and full-length Ang1 induced a pattern of vessel structures with longer vessels having fewer branch points than the controls. White dotted lines mark the margin of a fibrin graft. Scale bars ϭ 0.6 mm. D-F, gene transfer experiments corroborated these Ang1 effects. Third-generation lentiviral particles (28) were produced to accommodate the gene sequences of ⌬Ang1, wild-type Ang1, and GFP. 50-l volumes of lentiviral particles were administered pointwise to the CAM at embryonic day 9. The microvascular growth pattern at the application site was visualized on embryonic day 12 by fluorescence microscopy in the living embryo after intravenous injection of FITC-dextran. Scale bars ϭ 0.4 mm. D, a regular vessel branching pattern comparable with that of the untreated CAM (not shown) was observed in CAM cultures transduced with control viral particles encoding GFP. E, strong reduction of vessel branching was observed in response to ⌬Ang1. F, a similar reduction of vessel branching resulted in a response to wild-type Ang1. G, lentiviral particle-mediated production of ⌬Ang1 and full-length Ang1 (wild-type (wt)) proteins was validated in murine CHO-K1 cell culture experiments. Ang1 proteins in the cell culture supernatant or cell lysates were resolved by 10% reducing SDS-PAGE and detected by immunoblotting with Ang1-specific antibodies. Control cultures (C) were transduced with lentiviral particles encoding GFP. The molecular masses of ⌬Ang1 (Ϸ30 kDa; *) and full-length Ang1 (Ϸ70 kDa; #) are indicated. The majority of full-length Ang1 remained associated with the cell fraction. The effects on the mean vascular length density (as a cumulative measure of arterial/venous growth) and the vascular end point density (as a measure of connectivity between arterial/venous vessels and the underlying capillary plexus) were morphometrically quantitated. Skeletonized images of the FITC-dextran-labeled CAM vasculature from zones between 0 and 7.1 mm from the margin of the fibrin graft were analyzed by the method of grid intersection (37). A, reduction of the mean vascular length density by 6 g of ⌬Ang1 in fibrin gel matrix grafts (black bars) compared with fibrin alone (white bars). B, profound reduction of the vascular end point density by 6 g of ⌬Ang1 in fibrin gel matrix grafts versus fibrin alone. Black bars, 6 g of ⌬Ang1 formulated in fibrin; white bars, fibrin alone. Values are means Ϯ S.D. (n ϭ 6 for fibrin alone and n ϭ 4 for fibrin with 6 g of ⌬Ang1). *, p Ͻ 0.05. C and D, dose-dependent reduction by ⌬Ang1 of the mean vascular length density and the vascular end point density, respectively, in the vicinity of the fibrin graft. For both parameters, the effects of ⌬Ang1 paralleled the effects of full-length Ang1 (n ϭ 6 for full-length Ang1 (fl) and fibrin alone (0) and n ϭ 4 for 0.5, 2, and 6 g of ⌬Ang1). Statistical significance (*) was established with the General Linear Model as described under "Materials and Methods." Ang1 or recombinant Ang1 designer constructs based on their capacity to signal phosphorylation of the Tie2 receptor expressed in ECs: monomeric or dimeric Ang1 constructs that bind endothelial Tie2 without triggering phosphorylation were considered signaling-inactive. Instead, negative dominant effects on Tie2 were attributed to these variants, similar to those of naturally occurring antagonists such as the related angiopoietin-2 or angiopoietin-3. As discussed by Ward and Dumont (46), the presumed antagonistic functionality with Tie2 may well depend on dose, context, and EC type: for example, at high doses, angiopoietin-2 was found to trigger Tie2 phosphorylation and phosphatidylinositol 3Ј-kinase/Akt downstream signaling, leading to enhanced EC survival (3). Another example of the importance of context is provided by the study of Joussen et al. (12), who reported that in vitro application of Ang1 stimulates Akt phosphorylation, whereas in vivo application of Ang1 reduces Akt phosphorylation.
Some effects of ⌬Ang1 observed in this study strongly support a mechanism of induction of signaling in ECs rather than its negative interference with Tie2 signaling: the best example is that application of ⌬Ang1 resulted in inhibition of the basal permeability of EA.hy926 cell layers to the low molecular mass compound inulin (Fig. 2C). This observation adds to the findings of other laboratories that the full-length Ang1 protein exerts direct effects on the permeability of high molecular mass molecules across EA.hy926 as well as HUVEC monolayers (5). Studies of the mural cell-free vasculature exposed to Ang1 in mouse models of retinopathy further support the idea that the Ang1 protein could exert its anti-permeability effects directly through maintenance of EC integrity (47), possibly by strengthening EC-cell junctions through the recruitment of the cell-cell adhesion protein platelet-endothelial cell adhesion molecule into those junctions (5). Our results with ⌬Ang1 suggest that this effect proceeds independently of Tie2 phosphorylation. It is important to note that other activities of full-length Ang1 were not recapitulated by ⌬Ang1: we were unable to detect promotion of in vitro sprouting angiogenesis of HUVECs by exposure to ⌬Ang1 (48,49), and we did not find any anti-apoptotic effect of ⌬Ang1 on serum-starved HUVEC cultures (3) (data not shown).
Recent studies suggest that integrins, specifically ␣ 5 ␤ 1 , can mediate adhesion and migration on Ang1-coated surfaces (31). Our biochemical data as presented in Fig. 3 demonstrate that ⌬Ang1 can physically interact with ␣ 5 ␤ 1 integrin, in addition to Tie2, on ECs. By comparing cells that express both ␣ 5 ␤ 1 integrin and Tie2 with those that express ␣ 5 ␤ 1 but not Tie2, binding between ⌬Ang1 and ␣ 5 ␤ 1 was found to occur independently of Tie2, indicating direct interaction between ⌬Ang1 and ␣ 5 ␤ 1 . Of note, our examination revealed that binding by ⌬Ang1 of both types of cell-surface receptors was of roughly similar low affinity (Fig. 3, C and D). Such low binding affinities are typical for interactions of integrins with extracellular matrix ligands: dissociation constants ranging from 10 Ϫ7 to 10 Ϫ9 M have been reported (31,50,51). For binding between full-length Ang1 and Tie2, calculations have revealed a dissociation constant of ϳ3 ϫ 10 Ϫ9 M (21), indicative of a high affinity interaction. However, in the absence of oligomeric presentation of the Tie2binding domain as by provided in wild-type Ang1, the affinity of Ang1 for Tie2 is dramatically reduced: the binding studies of Yancopoulos and co-workers (18) have shown that the affinity of a monomeric Ang1 receptor-binding domain for Tie2 is ϳ290fold lower compared with that of the full-length form of Ang1. From our data and the published data, we infer that the affinity of ⌬Ang1 for Tie2 is substantially lower than that of fulllength Ang1.
Our examinations of exogenous Ang1 and ⌬Ang1 adminis-tration through protein or gene transfer in the embryonic CAM model of the developing vasculature revealed surprising similarities between the multimeric and monomeric forms, viz. reduction of the density of arterial/venous vessels and their connectivity with the underlying capillary plexus (Fig. 4). Morphometric analysis of the mean vascular length density and the vascular end point density revealed that the effects of exogenous ⌬Ang1 administration mirrored those of exogenous fulllength Ang1 administration in both quality and quantity (Fig.  5). This surprising similarity in effect between ⌬Ang1 and full-length Ang1 demands discussion. The most obvious predicted effect of ⌬Ang1 would be as an inhibitor, by binding Tie2 without inducing its activation, thereby inhibiting Tie2 activation by endogenous Ang1. Antagonizing mechanisms of ⌬Ang1 may indeed explain the thinned and low branching phenotype of blood vessels that we observed in our CAM experiments. Although such explanation is a possible one for ⌬Ang1, it does not explain the effect of full-length Ang1. The thinned vessel phenotype resulting from CAM exposure to full-length Ang1 contrasts with previous findings showing more and enlarged vessels in experimental systems of Ang1 overexpression in mouse skin (10) as well as of Ang1 administration by injection into developing mouse retinal vasculature (47). The profound reduction of vessel branching in our experimental system may originate from the well appreciated role of Ang1 in stabilizing vessels, which then became less susceptible to sprouting and remodeling. The decrease in vascular density by exogenously administered Ang1 in our experiments could reflect the result from overstabilization of vessels. It is also possible that the phenomenon observed in our CAM experiments could be ascribed to Tie2-independent mechanisms of action, e.g. activation of ␣ 5 ␤ 1 integrin through ligation by monomeric or fulllength Ang1 could facilitate interaction of ECs with extracellular matrix components such that the stability of vessels is enhanced. In any case, our studies in the developing CAM vascular system indicate a yet unrecognized role of Ang1 in reducing rather than increasing vascular density. It will be FIG. 6. Conceptual scheme of shared and exclusive effects of ⌬Ang1 and full-length Ang1. Our studies with ⌬Ang1 suggest that the isolated fibrinogen-like receptor domain of Ang1 is capable of mediating some effects of full-length Ang1 by a yet unknown mechanism of action, perhaps through integrins. The effects of ⌬Ang1 on mediating EC adhesion and spreading and reduction of transendothelial permeability are likely to be mediated beyond Tie2 phosphorylation (Tie2-P). In contrast, administration of ⌬Ang1 to ECs had no significant effect on endothelial migration, sprouting, or survival as reported for oligomeric full-length Ang1. These latter effects appear to be critically dependent on induction of Tie2 tyrosine phosphorylation for subsequent activation of the phosphatidylinositol 3Ј-kinase (PI3'-kinase)/Akt signal (survival) as well as focal adhesion kinase (FAK) (migration and sprouting).
interesting to see whether reduction of vascular branching is observed in other tissues that are treated with Ang1 through protein or gene transfer. The reduction of the angiogenic process through the vessel-stabilizing role of Ang1 presents a potential approach for anti-angiogenic, anti-tumor therapies: a recent study has shown that addition of Ang1 to the tumor microenvironment by Ang1-overexpressing cells significantly inhibits the angiogenic process, resulting in an overall inhibition of tumor growth (15). It remains a subject for future investigation to determine whether Ang1 application at sites of tumor angiogenesis could concomitantly slow tumor growth.
The molecular basis of ⌬Ang1 action on ECs remains speculative. Direct activation of EC integrin adhesion receptors, presumably ␣ 5 ␤ 1 , by ⌬Ang1 could be one mechanism contributing to its effect on cultured ECs and vascular morphology (31). Other types of interactions between ⌬Ang1 and the Tie2 receptor that are independent of Tie2 phosphorylation should be considered as well. For example, recent studies of EphB receptors, a class of receptor tyrosine kinases with prominent functions both in the nervous system and in vascular development, have revealed multiple, unexpected functionalities for this receptor class. In the case of EphB2, a kinase-independent role of the EphB2 extracellular domain in steering axons was elucidated (52). Moreover, molecular links between EphB receptors and ion channels and aquaporins in the cell membrane could provide fluid regulation (53). Similarly, yet unaddressed functional links may be envisioned as well for the Ang1/Tie2 and the Ang1/integrin signaling systems. In the conceptual scheme of Fig. 6, we illustrate shared and exclusive signaling effects of monomeric ⌬Ang1 and oligomeric full-length Ang1.