Expression of Antisense to Integrin Subunit β3Inhibits Microvascular Endothelial Cell Capillary Tube Formation in Fibrin*

αvβ3antagonists are potent angiogenesis inhibitors, and several different classes of inhibitors have been developed, including monoclonal antibodies, synthetic peptides, and small organic molecules. However, each class of inhibitor works by the same principal, by blocking the binding of ligands to αvβ3. In an effort to develop an αvβ3 inhibitor that down-regulates the actual level of αvβ3, we developed an antisense strategy to inhibit αvβ3 expression in vitro. β3 antisense expressed in endothelial cells specifically down-regulated αvβ3 and inhibited capillary tube formation, with the extent of down-regulation correlating with the extent of tube formation inhibition. This inhibition was matrix-specific, since tube formation was not inhibited in Matrigel. These findings support the notion that αvβ3is required for an essential step of angiogenesis in fibrin, namely capillary tube formation. These results suggest that pseudogenetic inhibition of β3 integrins using antisense techniques may ultimately provide a therapeutic means to inhibit angiogenesis in vivo.

Angiogenesis is the process of blood vessel development from preexisting vessels. In the normal adult, little angiogenesis occurs except as part of wound healing and during certain events in reproduction, including the menstrual cycle and embryonic implantation (1,2). Although aberrant angiogenesis is a hallmark of disorders such as solid tumor growth and metastasis (1,2), diabetic retinopathy (3,4), rheumatoid arthritis (5), atherosclerosis (6), and restenosis following angioplasty (7), pharmacologic stimulation of angiogenesis can promote development of a beneficial collateral circulation around occluded blood vessels, including coronary arteries (8). Therefore, much research is currently under way to identify methodologies to modulate angiogenesis.
Endothelial cell adhesion molecules are attractive targets to inhibit angiogenesis. ␤ 1 integrins (9 -11), integrin ␣ v ␤ 3 , and integrin ␣ v ␤ 5 (12) have all been implicated in angiogenesis, although the role of integrin ␣ v ␤ 3 in angiogenesis has been a subject of controversy. Angiogenic blood vessels in humans and other species express ␣ v ␤ 3 , but normal quiescent vasculature expresses little to no ␣ v ␤ 3 (13). Mouse knock-out studies, however, have shown convincingly that ␣ v ␤ 3 is not required for angiogenesis during embryonic development (14,15), and mutations in the ␤ 3 subunit of individuals with Glanzmann's thrombasthenia cause no apparent problems with angiogenesis (16), although these genetic findings do not rule out the possibility of compensation during development by functionally redundant adhesion molecules. In contrast to these genetic data, compelling pharmacologic data shows that monoclonal antibody and peptide antagonists of ␣ v ␤ 3 are potent inhibitors of angiogenesis in animals and humans (13,(17)(18)(19)(20)(21). Recent findings suggest that some of the actions of ␣ v ␤ 3 may be mediated by its coupling to receptors for growth factors, including platelet-derived growth factor (22) and vascular endothelial growth factor (23). A chimeric derivative (24) of a monoclonal antibody directed against ␣ v ␤ 3 , LM609 (25), has been used in clinical trials as an angiogenesis inhibitor. Synthetic peptide inhibitors of ␣ v ␤ 3 have been tested as angiogenesis inhibitors to prevent retinopathy (3,4).
To date, ␣ v ␤ 3 antagonists have been primarily antibodies (24), peptides (3,4), and small organic molecules (26,27). However, each class of antagonist binds directly to ␣ v ␤ 3 , which has the potential to cause unintended cell signaling (28). The use of RNA antisense as an antiangiogenic strategy has not yet been fully explored. While no drug is entirely selective, antisense has a theoretical advantage over antibodies, inhibitory peptides, or organic molecules in that the drug does not bind directly to the receptor. Furthermore, ␣ v ␤ 3 expression can theoretically be targeted specifically, unlike the pleiotypic effects seen with certain cytokines such as transforming growth factor ␤ or interferon ␥ (29). Integrin subunits such as ␣ 2 and ␤ 1 have been successfully targeted in vitro with RNA antisense directed against different regions of the mRNA (30,31). RNA antisense against other target molecules has been used in clinical trials to treat such diverse conditions as ovarian cancer, Crohn's disease, and retinal damage caused by cytomegalovirus (32). In this paper, we present evidence that the expression of endogenous antisense RNA directed against the integrin ␤ 3 subunit inhibits endothelial cell capillary tube formation in vitro.

EXPERIMENTAL PROCEDURES
Construction of ␤ 3 Sense and Antisense Expression Vectors-Fulllength human ␤ 3 cDNA was used to generate four different ␤ 3 antisense and two sense constructs by PCR. 1 A 130-bp antisense fragment of ␤ 3 from nucleotides 1-130 was generated using forward primer 5Ј-CGC GGA TCC CGC CGC GGG AGG CGG ACG AGA TG-3Ј and reverse primer 5Ј-CGG GGT ACC CTC ACA CCT CGC GTG GTA CAG ATG-3Ј. A 240-bp antisense fragment of ␤ 3 from nucleotides 180 -419 was gen-erated using forward primer 5Ј-CGC GGA TCC GAT GAG GCC CTG  CCT CTG GGC TCA-3Ј and reverse primer 5Ј-CGG GGT ACC CAC  CTG CCG CAC TTG GAT GGA GAA-3Ј. A 240-bp sense fragment of ␤ 3  from nucleotides 180 -419 was generated using forward primer 5Ј-CGG  GGT ACC GAT GAG GCC CTG CCT CTG GGC TCA-3Ј and reverse  primer 5Ј-CGC GGA TCC CAC CTG CCG CAC TTG GAT GGA GAA-3Ј. A 419-bp antisense fragment of ␤ 3 from nucleotides 1-419 was generated using the forward primer described for the 130-bp antisense fragment and the reverse primer described for the 240-bp antisense fragment. A 426-bp antisense fragment of ␤ 3 from nucleotides 1068 -1493 was generated using forward primer 5Ј-CGC GGA TCC CTC ATC CCA GGG ACC ACA GTT GGG-3Ј and reverse primer 5Ј-CGG GGT ACC GCC AGG CCC ACA ACG GCA TAC CCC-3Ј. A 426-bp sense fragment of ␤ 3 from nucleotides 1068 -1493 was generated using forward primer 5Ј-CGG GGT ACC CTC ATC CCA GGG ACC ACA GTT GGG-3Ј and reverse primer 5Ј-CGC GGA TCC GCC AGG CCC ACA ACG GCA TAC CCC-3Ј. Forward primers contained the BamHI restriction enzyme site, GGA TCC, and reverse primers contained the KpnI restriction enzyme site, GGT ACC. DNA fragments were inserted into BamHI and KpnI sites in the multiple cloning site of mammalian expression vector pCEP-4/hygromycin (Invitrogen). Full-length ␤ 3 cDNA was cleaved from pGEM-1 using HindIII and HincII and subcloned into the HindIII and PvuII sites of pCEP-4. ␤ 3 antisense and sense fragments were expressed constituitively from a cytomegalovirus promoter in pCEP-4.
All expression vectors were transfected into HMEC-1 cells using Lipofectin (Life Technologies) as per the manufacturer's instructions, using 12 l of Lipofectin for 2 g of DNA. The DNA and Lipofectin were incubated with HMEC-1 cells in serum-free medium for 16 h, and the medium was then replaced with complete serum-containing medium. After 48 h, 200 g/ml hygromycin B (Life Technologies) was added to the medium to select for stable transfectants. Selective pressure was maintained throughout the culture of the transfectants, and clones were used for only 4 -6 passages not only because of the potential loss of the vector but also because of the potential for gradual loss of the antisense effect, despite stable transfection (34,35).
PCR Analysis-Expression of the ␤ 3 sense and antisense fragments in HMEC-1 cells was confirmed by RT-PCR. Cytosolic RNA was isolated from 5 ϫ 10 6 to 1 ϫ 10 7 cells using an RNeasy RNA Isolation Kit (Qiagen) as per the manufacturer's instructions. RNA yields were quantified by spectrophotometry. Reverse transcription and PCR were performed using forward primer 5Ј-CGG GGT ACC GAT GAG GCC CTG CCT CTG GGC TCA-3Ј and reverse primer 5Ј-CGC GGA TCC CAC CTG CCG CAC TTG GAT GGA GAA-3Ј. Glyceraldehyde-3-phosphate dehydrogenase mRNA was used as an internal positive control using the forward primer 5Ј-TTA GCA CCC CTG GCC AAG G-3Ј and the reverse primer 5Ј-CTT ACT CCT TGG AGG CCA TG-3Ј. RT-PCR was performed in a single tube using Titan One-tube RT-PCR (Roche Molecular Biochemicals) as per the manufacturer's instructions. The final reaction was composed of 1ϫ RT-PCR buffer, 1.5 mM MgCl 2 , 5 mM dithiothreitol, 0.2 mM (each) deoxynucleotide, 10 units of RNAsin (Promega), and the supplied enzyme mix containing avian myeloblastosis virus reverse transcriptase and Expand High Fidelity PCR enzyme. The reverse transcription was performed on a PTC-200 Thermocycler (MJ Research) for 30 min at 55°C followed by inactivation at 94°C for 4 min, initiating the denaturation step for PCR. The following steps were performed for 28 cycles: 94°C for 45 s, annealing at 55°C for 45 s, and extension at 72°C for 2 min (7 min during the final cycle). RNA samples were assessed for genomic contamination by performing the PCR without reverse transcription using Vent polymerase (New England Biolabs). pCEP-4 containing the 240-bp antisense fragment was used as a positive control.
Fibrinogen Purification-Plasminogen-free human fibrinogen (Calbiochem) was further purified by DEAE-column chromatography as described previously (41,42). Peak ␥A/␥A and ␥A/␥Ј fractions were pooled separately, precipitated with 0.234 g/ml ammonium sulfate, and resuspended in 137 mM NaCl, 10 mM Hepes, pH 7.4, 2.7 mM KCl, 1 mM CaCl 2 for storage at Ϫ70°C. ␥A/␥A fibrinogen was used for all experiments. Fibrinogen was dialyzed into PBS and sterile-filtered using a 0.2-m filter for use in capillary tube formation assays.
Three-dimensional Fibrin-based Capillary Tube Formation Assay-A modification of the method of Nehls and Drenckhahn (43) was used to measure endothelial cell capillary tube formation in a three-dimensional fibrin-based matrix. HMEC-1 cells were grown to confluence on Cytodex-3 microcarriers (Amersham Pharmacia Biotech) for 2-3 days in spinner flasks at 37°C with 5% CO 2 . Confluent HMEC-1 cell-coated microcarriers were rinsed three times and resuspended in tube formation assay medium: Dulbecco's modified Eagle's medium, 20% fetal bovine serum, 2 mM L-glutamine, 200 kallikrein-inactivating units/ml aprotinin (Bayer). HMEC-1 cell-coated microcarriers were added to sterile-filtered solutions containing 1.5 mg/ml fibrinogen in PBS with 200 kallikrein-inactivating units/ml aprotinin and 30 ng/ml bFGF (R & D Systems). The concentration of bFGF was the same as that used by Nehls and Drenckhahn (43) for stimulating endothelial cell capillary tube formation. Higher doses of bFGF (up to 3-fold) had no additional effect on capillary tube formation. 50-l aliquots containing ϳ15-20 HMEC-1-coated microcarriers were added to wells of a 96-well plate. Human ␣-thrombin (0.625 NIH units/ml, 3000 NIH units/mg; obtained from Dr. Walter Kisiel, University of New Mexico, Albuquerque, NM) was immediately added and incubated for 30 min to induce fibrin clot formation. Abciximab (obtained from Dr. Mark Kozak, Penn State University, Hershey, PA), LM609, JB1a, or PBS as a control was dissolved in assay medium, and 50 l was added on top of the clot and incubated at 37°C with 5% CO 2 for 1 h. An additional 50 l of tube formation assay medium was then added, and plates were incubated at 37°C with 5% CO 2 for 2-5 days. Cell nuclei were stained for 2 h with 50 g/ml bisbenzimide (Sigma). After inverting the plate, cell sprouts, defined as projections containing a minimum of three nuclei (43), were counted using fluorescence microscopy. The mean number of sprouts per microcarrier was determined in triplicate, and each assay was performed a minimum of three times. Photomicrographs were taken using an Olympus B-Max 50 epifluorescence microscope with a Paultek cooled chargecoupled device camera interfaced with a Scion LG3 framestore board mounted in a Macintosh Centris 650 computer and running Adobe Photoshop.
Matrigel-based Tube Formation Assay-Wells of a 96-well plate were coated with Matrigel as per the manufacturer's instructions (Collaborative Biomedical Products) and incubated at 37°C for 30 min. HMEC-1 cells were detached with 0.25% trypsin, 1 mM EDTA, sedimented by centrifugation for 5 min, and resuspended in cell culture medium. HMEC-1 cells were added to Matrigel-coated wells and incubated at 37°C with 5% CO 2 for 16 h. Photomicrographs were taken with Eastman Kodak Co. T-Max 400 film at ϫ 100 magnification using an Olympus OM-2 camera attached to a Nikon TMS inverted phase-contrast microscope.

RESULTS
␤ 3 Antisense Constructs-To down-regulate ␣ v ␤ 3 expression, the ␤ 3 integrin subunit mRNA was targeted with RNA antisense. The region of a cDNA that yields effective antisense RNA cannot be predicted with certainty; therefore, four different ␤ 3 antisense constructs were tested from different regions of the ␤ 3 cDNA. The sequences selected for antisense targeting were designed to avoid regions of high homology among ␤ subunits (45) (Fig. 1, hatched bars), with the exception of ␤ 3 419. A 130-bp ␤ 3 antisense cDNA fragment (␤ 3 130 antisense) from nucleotides 1-130 preceded the first homologous domain and included one intron/exon boundary and the translational start site; a 240-bp ␤ 3 antisense cDNA fragment (␤ 3 240 antisense) from nucleotides 180 -419 between the first and second homologous domains included two intron/exon boundaries; a 419-bp ␤ 3 antisense cDNA fragment (␤ 3 419 antisense) from nucleotides 1-419 included the same region as ␤ 3 130 antisense, but extended through the first homologous domain and ended before the second domain, and contained three intron/ exon boundaries; and a 426-bp ␤ 3 antisense cDNA fragment (␤ 3 426 antisense) from nucleotides 1068 -1493 between the seventh and eighth homologous domains included two intron/exon boundaries. These sequences were amplified using PCR and inserted in a sense or antisense orientation into the mammalian expression vector pCEP-4/hygromycin.
The HMEC-1 human microvascular endothelial cell line was chosen for transfection with these constructs rather than primary endothelial cells, since it is difficult, if not impossible, to isolate stable transfectants from primary microvascular endothelial cells in culture. Transfectants that are isolated from such primary cells have often undergone phenotypic changes and have a limited lifespan prior to senescence. In contrast, the HMEC-1 cell line is a diploid cell line that retains the characteristics of microvascular endothelial cells during prolonged culture in vitro and does not dedifferentiate (46). ␤ 3 sense and antisense constructs, as well as a full-length ␤ 3 /pCEP-4 construct and the pCEP-4 vector alone were stably transfected into HMEC-1 cells. Eight to ten individual clones were isolated for each construct, since vector copy number can alter gene expres-sion in any particular clone.
␤ 3 240 Antisense Expression Specifically Inhibits ␣ v ␤ 3 Levels-The relative levels of ␣ v ␤ 3 on transfected and nontransfected HMEC-1 cell clones were determined using flow cytometry. ␤ 3 130 antisense and ␤ 3 419 antisense transfectants had reductions in ␣ v ␤ 3 levels of 28 -62% and up to 33%, respectively (data not shown). However, these transfectants also had comparable decreases in ␤ 1 levels that could complicate the analysis of the specific role of ␣ v ␤ 3 and therefore were not characterized further. In contrast, ␤ 3 240 antisense transfectants had reductions in ␣ v ␤ 3 levels of 0 -49% with no concomitant changes in ␤ 1 levels and showed no significant changes in ␣ v ␤ 5 levels (Table I). ␤ 3 426 antisense transfectants had somewhat smaller reductions in ␣ v ␤ 3 levels than the ␤ 3 240 antisense transfectants, 2-29% (Table I), and were therefore not characterized further. HMEC-1 cells transfected with pCEP-4 alone or the ␤ 3 240 fragment in the sense orientation had levels of ␣ v ␤ 3 , ␤ 1 , and ␣ v ␤ 5 that were comparable with nontransfected HMEC-1 cells (Table I). HMEC-1 cells transfected with fulllength ␤ 3 in the sense orientation also showed no significant changes in ␣ v ␤ 3 levels (Table I). Cells transfected with the ␤ 3 240 fragment in the sense or antisense orientation expressed the appropriate sense or antisense RNA, as detected by semiquantitative RT-PCR, whereas these transcripts were undetectable in wild-type HMEC-1 cells (Fig. 2). Although the primers will also amplify the endogenous mRNA for ␤ 3 , bands from the endogenous mRNA were not visible in wild-type HMEC-1 cells under the conditions used to amplify the sense and antisense fragments, consistent with the higher expression levels of the fragments from the viral promoters in the pCEP-4 expression vector.
Western blot analysis of the ␤ 3 antisense transfectants showed similar amounts of ␣ v expressed compared with the

TABLE I
Integrin levels of HMEC-1 transfectants Flow cytometry was conducted using either anti-␣ v ␤ 3 (LM609), anti-␤ 1 (JB1a), or anti-␣ v ␤ 5 (P1F6) as primary antibody. The mean fluorescence intensity of transfected clones was normalized to the mean fluorescence intensity for nontransfected HMEC-1 cells. Experiments were conducted in duplicate (*) or triplicate, and the mean Ϯ S.D. is shown. parental HMEC-1 cells (Fig. 3), whereas ␤ 3 expression was markedly reduced. Scanning densitometry of the ␣ v and ␤ 3 bands in Fig. 3 showed only a 2% difference in ␣ v band intensity, compared with a 40% reduction in ␤ 3 band intensity for the ␤ 3 240 antisense transfectant. These results demonstrate that the ␤ 3 240 antisense transfectant had decreased levels of ␤ 3 expression, but ␣ v expression was not affected. Interestingly, no compensatory increases in ␣ v ␤ 5 levels were seen when ␣ v ␤ 3 levels were decreased in the ␤ 3 240 antisense transfectants. This is in contrast to the compensatory changes seen in individuals with a certain type of Glanzmann thrombasthenia.
In individuals with defects in ␣ IIb that result in decreased ␣ IIb ␤ 3 levels on platelets, a compensatory elevation in ␣ v ␤ 3 levels can occur due to pairing of the excess ␤ 3 subunits with ␣ v (16). However, antisense inhibition of ␤ 3 expression did not cause compensatory increases in ␣ v ␤ 5 levels due to pairing of the excess ␣ v subunits with ␤ 5 in the transfected cells. It is not clear if the excess ␣ v subunits are paired with ␤ 1 or retained intracellularly.

HMEC-1 Cells Form Lumen-containing Capillary-like Structures in a Three-dimensional Fibrin
Matrix-Although the inhibition of ␣ v ␤ 3 expression on ␤ 3 240 antisense transfectants was not complete, the reductions in ␣ v ␤ 3 appeared specific, and the cells were therefore tested for their ability to form capillary tubes. Angiogenesis is a multifactorial process that has been divided into at least six distinct steps: 1) proteolytic digestion of extracellular matrix, 2) migration of endothelial cells, 3) proliferation of these cells, 4) extracellular matrix production, 5) vascular tube formation, and 6) anastomosis of newly formed channels resulting in a patent neovessel (47). Assays have been developed to study each of these steps in vitro, thereby providing a reductionist approach to the complex process of angiogenesis. In particular, endothelial cell tube formation assays have been used to identify critical modulators of this process (48). Because the effects of angiogenesis modulators in two-dimensional tube formation assays do not always correlate with their effects in vivo (47), three-dimensional capillary tube formation assays have been developed. These latter assays appear to more faithfully recapitulate the process of angiogenesis (49).
We modified the well characterized three-dimensional capillary tube formation assay developed by Nehls and Drenckhahn (43) to use HMEC-1 cells that were grown to confluence on microcarriers and embedded in a fibrin matrix in the presence of 30 ng/ml bFGF. HMEC-1 cells were cultured in the threedimensional fibrin matrix for 2-4 days at 37°C. To ensure that this in vitro capillary tube formation assay responded to authentic angiogenic modulators, several parameters were examined. HMEC-1 cells without thrombin added to clot the fibrinogen migrated randomly from the microcarrier beads onto the well and did not form capillary-like structures (data not shown). HMEC-1 cells embedded in the fibrin matrix in the absence of an angiogenic stimulator migrated from the microcarriers into the fibrin in a somewhat more organized fashion but formed few distinguishable capillary-like sprouts (data not shown). In contrast, HMEC-1 cells cultured in the presence of all the assay components formed capillary-like sprouts, similar to those seen when large vessel bovine pulmonary artery endothelial cells were used in this assay (43). HMEC-1 cells migrated from the microcarrier into the fibrin matrix and formed multicellular lumen-containing capillary-like structures (Fig. 4A) similar to those formed with bovine endothelial cells (43). Cell/cell junctions were clearly visible between HMEC-1 cells forming capillary-like structures, and the subcellular structures appeared normal (Fig. 4B). Therefore, HMEC-1 cells appeared to undergo capillary tube formation in this in vitro assay similarly to large vessel bovine pulmonary artery endothelial cells (43).

Antibodies against ␣ v ␤ 3 Inhibit Microvascular Endothelial Cell Capillary Tube Formation in a Fibrin
Matrix-To test that microvascular endothelial cell capillary tube formation was dependent on integrin ␣ v ␤ 3 in this assay, HMEC-1 cells on microcarriers were embedded in a fibrin matrix in the presence of monoclonal antibodies directed against particular integrins. The monoclonal antibodies used were abciximab, a chimeric Fab fragment of the monoclonal antibody 7E3 (50), LM609, directed against ␣ v ␤ 3 (25), and as a negative control, JB1a directed against ␤ 1 integrins (37). Although abciximab was originally developed as an antagonist of platelet ␣ IIb ␤ 3 , recent studies demonstrated that abciximab binds with comparable affinity to ␣ v ␤ 3 (51). After 2-4 days of culture, HMEC-1 nuclei were stained with bisbenzimide and photographed to quantitate the effect on capillary tube formation. Capillary tube formation was quantitated as the number of sprouts per microcarrier bead, with a sprout defined as a minimum of three interconnected cells (43). Control HMEC-1 cells stimulated FIG. 2. PCR analysis of sense and antisense expression in HMEC-1 transfectants. Semiquantitative RT-PCR was performed on RNA extracted from wild-type HMEC-1 cells (lanes 1-3) and HMEC-1 cells transfected with pCEP-4 vectors expressing ␤ 3 240 sense (lanes 4 -6) or antisense (lanes 7-9). Two sets of primers were used: one set designed to amplify the ␤ 3 240 fragment and one set to amplify glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a positive internal control. RT-PCRs were conducted in triplicate, and representative reactions are shown from one set of experiments. Although some variability was noted in expression levels, both the sense and antisensetransfected cells gave rise to PCR products that co-migrated with the PCR product produced directly from the ␤ 3 240 antisense expression vector (lane 10), whereas no product was detected in wild-type HMEC-1 cells (lanes 1-3).  1 and 3) were separated on polyacrylamide gels. The gels were Western blotted to nitrocellulose and probed with either rabbit polyclonal antiserum AB 1930 against ␣ v or AB 1932 against ␤ 3 . Bands were visualized using a chemiluminescent substrate and photographed. Parallel lanes on the nitrocellulose were stained for total protein using Ponceau S (lanes 3 and 4). with bFGF sprouted and migrated into the fibrin matrix (Fig.  5A). In contrast, cells stimulated with bFGF but treated with 5 g/ml abciximab showed significant inhibition of capillary tube formation (Fig. 5B). The inhibition was comparable with that seen when the cells were treated with 5 g/ml of the documented angiogenesis inhibitor LM609 (Fig. 5C). In contrast, 5 g/ml of JB1a had no effect on endothelial capillary tube formation (Fig. 5D). Both abciximab and LM609 inhibited HMEC-1 capillary tube formation in a dose-dependent fashion from 29 to 80% and from 63 to 89%, respectively, with 0.01 to 5 g/ml, while JB1a had no effect (data not shown). LM609 antibody was more effective than abciximab at inhibiting capillary tube formation at concentrations of 0.01-1 g/ml, but at 5 g/ml, abciximab and LM609 inhibited fibrin-based capillary tube formation to a similar extent (data not shown). The differences in dose response may reflect the fact that abciximab is a Fab fragment, whereas LM609 is intact IgG, or they may reflect a difference in antibody affinity for ␣ v ␤ 3 (51).

␤ 3 240 Antisense Expression Inhibits Microvascular Endothelial Cell Capillary Tube Formation in a Fibrin Matrix-␤ 3 240
antisense-transfected HMEC-1 cells were tested in the capillary tube formation assay to examine the effect of reducing ␣ v ␤ 3 levels. Several independent clones were analyzed for each type of transfected HMEC-1 cell. pCEP-4-transfected HMEC-1 cells and ␤ 3 240 sense-transfected HMEC-1 cells (Fig. 6, B and C, respectively; Table II) sprouted to a similar extent as nontransfected HMEC-1 cells (Fig. 6A and Table II). All full-length ␤ 3 sense clones sprouted similarly to nontransfected HMEC-1 cells (data not shown). In contrast, all of the ␤ 3 240 antisense transfectants with reduced ␣ v ␤ 3 levels consistently showed reduced sprouting (Fig. 6D). The only ␤ 3 240 antisense transfectant that showed no change in ␣ v ␤ 3 level, clone 17, also showed no change in sprouting (Table II). The number of sprouts/ microcarrier increased in a dose-dependent manner with the level of ␣ v ␤ 3 in fibrin-based tube formation assays (Fig. 7). Although the ␣ v ␤ 3 level on the ␤ 3 240 antisense transfectants was reduced by only 24 -49%, tube formation in fibrin was inhibited by 21-76%. The maximal inhibition of tube formation seen for the ␤ 3 240 antisense transfectants was comparable with the inhibition seen with maximal doses of abciximab and LM609 (Fig. 5, B and C).
The Matrigel tube formation assay shows that the ␤ 3 antisense effect is matrix-specific. DISCUSSION These results demonstrate that antisense-mediated downregulation of ␣ v ␤ 3 expression in microvascular endothelial cells inhibits capillary tube formation in fibrin. Furthermore, the extent of ␣ v ␤ 3 down-regulation correlates with the extent of tube formation inhibition. This inhibition is matrix-specific, since tube formation is not inhibited in Matrigel. Although these results were obtained using a cell line, this cell line has a normal diploid genetic complement and retains many characteristics of normal human microvascular endothelial cells (33), suggesting that the results obtained may be extrapolated to primary cells. The assay may also prove useful as a rapid screening method for pro-and antiangiogenic agents. In addition, ␣ v ␤ 3 down-regulation in these studies did not involve the use of pharmacologic agents with pleiotypic effects. Moreover, evidence for a true antisense effect was demonstrated in trans- FIG. 5. Capillary tube formation by HMEC-1 cells in fibrin is inhibited by antibodies against ␣ v ␤ 3 . A three-dimensional fibrin-based capillary tube formation assay was performed using bFGF-stimulated HMEC-1 cells. On day 2, the nuclei were stained with bisbenzimide and photographed. A, control HMEC-1 cells showed numerous capillary-like sprouts. HMEC-1 cells incubated with 5 g/ml abciximab (B) or 5 g/ml anti-␣ v ␤ 3 (LM609) (C) showed a significant decrease in the number of sprouts per microcarrier. D, HMEC-1 cells incubated with 5 g/ml anti-␤ 1 (JB1a) showed similar capillary tube formation to control HMEC-1 cells. Bar, 100 m.

TABLE II
Capillary tube formation in fibrin of HMEC-1 transfectants A three-dimensional fibrin-based capillary tube formation assay was performed using bFGF-stimulated HMEC-1 cells. On day 3, nuclei were stained with bisbenzimide, and sprouting of nontransfected and transfected HMEC-1 cells was quantitated. The number of sprouts/microcarrier for transfected clones was normalized to the number of sprouts/ microcarrier for nontransfected HMEC-1 cells. The data shown are one representative of three such experiments, and the mean of triplicate determinations Ϯ S.D. is shown.  7. ␣ v ␤ 3 expression correlates with capillary tube formation in fibrin. Flow cytometry was conducted using primary antibody anti-␣ v ␤ 3 (LM609). A three-dimensional fibrin-based tube formation assay was conducted using bFGF-stimulated HMEC-1 cells. On day 3, nuclei were stained with bisbenzimide, and sprouting of nontransfected and transfected HMEC-1 cells was quantified as the number of sprouts per microcarrier and normalized to the number of sprouts/microcarrier for nontransfected HMEC-1 cells. The data shown are one representative of three such experiments, in which each point represents the mean of triplicate determinations Ϯ S.D. fectants expressing the same cDNA fragment in a sense orientation.
One issue raised by the ability of abciximab to inhibit capillary tube formation is the necessity for careful evaluation of the biological effects of integrin antagonists. Our results raise the possibility that the effectiveness of abciximab in patients undergoing percutaneous coronary interventions may reflect, at least in part, its ability to inhibit angiogenesis. Given that ␣ v ␤ 3 is overexpressed in atherosclerotic plaques (55) and co-localizes with fibrin deposition in occluded arteries (56), it is tempting to speculate that the long term efficacy seen at 3 and 6 months in clinical trials of abciximab (57) was due, at least in part, to inhibition of angiogenesis in atherosclerotic plaques.
These results confirm and extend previous studies that show that ␣ v ␤ 3 plays an essential role in angiogenesis (12), since the data shows that ␣ v ␤ 3 is required for an essential step of angiogenesis in fibrin, capillary tube formation, but is not required for capillary tube formation in all extracellular matrices. The matrix specificity of the antisense inhibition suggests that a critical role of ␣ v ␤ 3 may be in angiogenesis that occurs when endothelial cells are in contact with fibrin, as occurs during wound healing and tumor neovascularization. This may explain the observations made in ␣ v and ␤ 3 knockout mice that angiogenesis still occurs during development (14,15), since angiogenesis under these conditions probably involves different extracellular matrices. Given the essential nature of angiogenesis in development, it seems likely that redundant mechanisms exist or that compensatory changes in other adhesive receptors allow for angiogenesis in the knockout mice. The observation that ␣ v ␤ 3 levels do not influence capillary tube formation in Matrigel is consistent with previous reports that other adhesion molecules, such as ␤ 1 integrins, can be used under different conditions to mediate angiogenesis in other types of extracellular matrix (9 -11).
RNA antisense to integrin ␤ 3 may provide a novel approach to specific ␣ v ␤ 3 antagonism that could be used therapeutically for angiogenesis-dependent pathologies. Li et al. (58) have used full-length ␤ 3 antisense to down-regulate ␣ v ␤ 3 expression in tumor cells and shown that cell motility and basement membrane invasion was reduced significantly. However, the effect of this antisense construct on the expression of related integrins such as ␤ 1 was not characterized. Our results provide proof of the concept that an antisense fragment to integrin ␤ 3 subunit can be used to specifically reduce ␣ v ␤ 3 levels and inhibit microvascular endothelial cell capillary tube formation in fibrin and provide an impetus for further investigations of ␤ 3 antisense as an angiogenesis antagonist. In particular, direct delivery of synthetic oligodeoxynucleotides or virus-mediated infection of endothelial cells (59) to express the antisense RNA endogenously may provide methods to inhibit angiogenesis in vivo. FIG. 8. Capillary tube formation by HMEC-1 cells in Matrigel is inhibited by antibodies against ␤ 1 integrins but not ␤ 3 integrins. Matrigel-based capillary tube formation assays were conducted, and representative fields from one of three experiments are shown at ϫ 100 magnification. Control HMEC-1 cells (A-C) formed capillary-like structures in Matrigel. HMEC-1 cells incubated with 5, 20, or 50 g/ml abciximab (D-F) or 5, 20, or 50 g/ml anti-␣ v ␤ 3 (LM609) (G-I) formed capillary-like structures on Matrigel similar to control HMEC-1 cells, but anti-␤ 1 (JB1a) at 5, 20, or 50 g/ml (J-L) significantly reduced capillary-like structure formation in a dose-dependent manner.