VEGF145, a secreted vascular endothelial growth factor isoform that binds to extracellular matrix.

A vascular endothelial growth factor (VEGF) mRNA species containing exons 1-6 and 8 of the VEGF gene was found to be expressed as a major VEGF mRNA form in several cell lines derived from carcinomas of the female reproductive system. This mRNA is predicted to encode a VEGF form of 145 amino acids (VEGF145). Recombinant VEGF145 induced the proliferation of vascular endothelial cells and promoted angiogenesis in vivo VEGF145 was compared with previously characterized VEGF species with respect to interaction with heparin-like molecules, cellular distribution, VEGF receptor recognition, and extracellular matrix (ECM) binding ability. VEGF145 shares with VEGF165 the ability to bind to the KDR/flk-1 receptor of endothelial cells. It also binds to heparin with an affinity similar to that of VEGF165. However, VEGF145 does not bind to two additional endothelial cell surface receptors that are recognized by VEGF165 but not by VEGF121. VEGF145 is secreted from producing cells as are VEGF121 and VEGF165. However, VEGF121 and VEGF165 do not bind to the ECM produced by corneal endothelial cells, whereas VEGF145 binds efficiently to this ECM. Basic fibroblast growth factor (bFGF)-depleted ECM containing bound VEGF145 induces proliferation of endothelial cells, indicating that the bound VEGF145 is active. The mechanism by which VEGF145 binds to the ECM differs from that of bFGF. Digestion of the ECM by heparinase inhibited the binding of bFGF to the ECM and released prebound bFGF, whereas the binding of VEGF145 was not affected by heparinase digestion. It therefore seems that VEGF145 possesses a unique combination of biological properties distinct from those of previously characterized VEGF species.

The vascular endothelial growth factor (VEGF) 1 isoforms display a limited structural similarity to platelet-derived growth factor and are important regulators of angiogenesis and blood vessel permeability (1)(2)(3). The human VEGF isoforms are generated by alternative splicing from a single gene (4 -6). The domain encoded by exons 1-5 contains information required for the recognition of the known VEGF receptors KDR/flk-1 and flt-1 (7) and is present in all VEGF isoforms. The amino acids encoded by exon 8 are also present in all the VEGF splice variants. The VEGF isoforms are distinguished by the presence or the absence of the peptides encoded by exons 6 and 7 of the VEGF gene. VEGF 121 is 121 amino acids long and lacks both exons. VEGF 165 contains the exon 7-encoded peptide, whereas VEGF 189 contains both exon 6-and exon 7-encoded peptides (6,8,9). VEGF 121 and VEGF 165 promote angiogenesis, cause permeabilization of blood vessels, and induce proliferation of vascular endothelial cells (10 -14). VEGF 189 has not yet been purified, but studies with cells expressing VEGF 189 indicate that it may induce endothelial cell proliferation (8,9). Low levels of a mRNA corresponding in size to a mRNA encoding a putative VEGF variant of 145 amino acids (VEGF 145 ) containing exon 6 but lacking exon 7 were detected previously in reverse PCR experiments, but the protein encoded by this mRNA has not yet been characterized (15,16).
The different VEGF isoforms differ in their heparin binding ability. VEGF 121 does not bind to heparin, whereas VEGF 165 and VEGF 189 do (8,(17)(18)(19). The heparin binding affinity of VEGF 189 was reported to be higher than that of VEGF 165 , suggesting that exon 6 contributes to the heparin binding ability of VEGF 189 (8). VEGF 165 and VEGF 121 are secreted efficiently from producing cells and do not bind efficiently to the ECM produced by CEN4 cells (8,9). In contrast, VEGF 189 is retained on the cell surface and in the ECM, from which it can be released by prolonged incubation with heparin. (9). The peptide encoded by exon 7 also seems to affect the receptor recognition patterns of VEGF isoforms. VEGF 121 recognizes a single VEGF receptor in endothelial cells that was identified as the KDR/flk-1 VEGF receptor (20,21). VEGF 165 also binds to this receptor but recognizes two additional VEGF receptors of unknown structure that are found in endothelial cells and in several transformed cell types (21,22).
We have observed that VEGF 145 is one of the main VEGF isoforms expressed by several cell lines derived from carcinomas of the female reproductive system. We have characterized VEGF 145 and have compared its biological properties with those of other VEGF forms. Recombinant VEGF 145 expressing the exon 6-derived peptide binds to heparin but behaves like VEGF 121 with regard to its receptor recognition ability. In contrast to VEGF 121 and VEGF 165 , it binds to a basement membrane like ECM produced by corneal endothelial cells in a biologically active form. VEGF 145 therefore represents a VEGF form possessing distinct biological characteristics.

EXPERIMENTAL PROCEDURES
Materials-Human recombinant VEGF 165 and VEGF 121 were purified from Sf-9 insect cells, whereas recombinant human bFGF was produced in bacteria as described previously (19,23,24). A rabbit polyclonal antibody directed against VEGF 165 (23) and a mouse monoclonal IgM antibody (M-35) directed against full-length VEGF 165 were produced in our laboratory using standard techniques. Heparinases type I, II, and III were kindly donated by Dr. J. Zimermann (Ibex Technologies, Montreal, Canada). The bicystronic mammalian expression vector MIRB was kindly provided to us by Dr. Craig MacArthur (Washington University, St. Louis, MO) (25). Anti-VEGF monoclonal antibody clone 26503.11 and peroxidase-and alkaline phosphataseconjugated anti-rabbit IgG antibodies were from Sigma. Disuccinimidyl suberate was from Pierce. Sodium alginate was from Fluka. Heparin-Sepharose was purchased from Pharmacia Biotech Inc. 125 I-Sodium was obtained from New England Nuclear. Anti-mouse IgM antibodies conjugated to alkaline phosphatase were from Southern Biotechnology Associates Inc. (Birmingham, AL). Tissue culture plasticware was from Nunc, and 96-well dishes for enzyme-linked immunosorbent assays were bought from Corning. Grace's medium was obtained from Life Technologies, Inc. All other tissue culture reagents were from Biological Industries Inc. (Kibbutz Beth Haemek, Israel).
Identification of VEGF 145 mRNA in Cancer Cells-Total mRNA was prepared from OC-238 human epithelial ovarian carcinoma cells (26). Complementary DNA was synthesized from 300 ng of total RNA using oligo(dT) as a primer and avian myeloblastosis virus reverse transcriptase. PCR amplification was carried out in the presence of a [ 32 P]dCTP tracer (2 Ci in a 100-l reaction volume), 1 mM of each dNTP, 2.5 mM MgCl 2 , and 2.5 units of Taq polymerase. 25 amplification cycles were used, each consisting of a 1-min incubation at 94°C, a 2-min incubation at 65°C, and a 3-min incubation at 72°C. The VEGFspecific oligonucleotides used were GGAGAGATGAGCTTCCTACAG and TCACCGCCTTGGCTTGTCACA, corresponding to amino acids 92-98 and to the six carboxyl-terminal amino acids of VEGF, respectively. A pair of primers from the L19 ribosomal protein were included in the reactions as an internal control. Amplified fragments were resolved in a 6% nondenaturing polyacrylamide gel and were visualized by autoradiography. The band corresponding to the VEGF 145 mRNA (see Fig. 1B) was excised, reamplified, and sequenced.
A similar procedure using different primers was used to detect VEGF 145 mRNA in A431 and HeLa cells. The VEGF-specific primers used here were derived from the 5Ј of the coding region (amino acids [13][14][15][16][17][18][19][20] and from the 3Ј region encompassed by the last 6 amino acids of VEGF including the translation stop codon of the VEGF sequence. Construction of VEGF 145 Encoding Expression Vectors-In order to produce recombinant VEGF 145 , we prepared a VEGF 145 cDNA construct by deleting the oligonucleotides encoded by exon 7 out of the VEGF 189 cDNA. Primers used to amplify exons 1-6 of the VEGF cDNA were the external primer GCTTCCGGCTCGTATGTTGTGTGG, corresponding to a puc118 sequence, and the internal primer ACGCTCCAGGACT-TATACCGGGA, corresponding to a sequence at the 3Ј end of exon 6. Primers used to amplify the 3Ј end of the VEGF cDNA were complementary to the puc118 sequence GGTAACGCCAGGGTTTTCCCAGTC and to the 3Ј end of the exon 6 sequence (underlined) and to the start of exon 8 (CGGTATAAGTCCTGGAGCGTATGTGACAAGCCGAGGCGG-TGA). Following amplification, the PCR products were precipitated, and the products were reamplified using only the puc118-derived external primers. The product was gel purified, subcloned into the PCR-II vector, and sequenced using the sequenase-II kit from U. S. Biochemical Corp. This cDNA was further used for protein expression studies.
Production and Purification of Recombinant VEGF 145 -The VEGF 145 cDNA was subcloned into the BamHI site of the MIRB expression vector (25). Following transfection into BHK-21 cells and selection with 0.6 mg/ml G418, VEGF 145 -expressing cells were identified using anti-VEGF antibodies. The VEGF 145 cDNA was also subcloned into the transfer plasmid pVL-1393 (Invitrogen) downstream from the polyhedrin promoter to yield pVL-1393/v145. This plasmid and baculovirus wild type DNA were co-transfected into Sf9 cells using the calciumphosphate co-precipitation method, and recombinant baculoviruses were isolated as described (23).
VEGF 145 was produced in Sf9 cells as described for VEGF 165 (23). The conditioned medium contained approximately 5 mg of VEGF 145 / liter. The conditioned medium was concentrated by precipitation with 70% ammonium sulfate at 4°C for 12 h. The precipitate was solubilized in 20 mM Tris, pH 7, and 0.1 M NaCl, dialyzed extensively against this buffer at 4°C, and applied to a heparin-Sepharose column. The column was washed with the same buffer containing 0.3 M salt, followed by elution with the same buffer containing 0.8 M NaCl. A small residual amount also eluted at 2 M NaCl. The 0.8 M salt eluant was further purified by reverse phase high pressure liquid chromatography on an Applied Biosystems Brownlee C-8 column using a linear gradient of acetonitrile (20 -80%) containing 0.1% trifluoroacetic acid. VEGF 145 was eluted at 48% acetonitrile. The trifluoroacetic acid in the eluant was neutralized using Tris base, and the acetonitrile was removed using a SpeedVac evaporator at room temperature.
Enzyme-linked Immunosorbent Assays-Increasing concentrations of VEGF in 50 l of coating buffer (20 mM K 2 HPO 4 , 10 mM KH 2 PO 4 , 1 mM EDTA, 0.8% NaCl, pH 7.2) were adsorbed to 96-well dishes for 3 h at 25°C. Free VEGF was aspirated, and the wells were blocked with coating buffer containing 1% bovine serum albumin for 1 h. The wells were extensively washed with wash buffer (10 mM Tris-HCl, pH 7.2, 150 mM NaCl, and 0.1% Tween 20), incubated with the anti-VEGF M-35 monoclonal antibody for 2 h at 25°C, washed again, and incubated for 1 h with an alkaline phosphatase-conjugated secondary antibody. After final washing, the amount of bound antibody was determined using para-nitrophenylphosphate as substrate.
Cell Culture and Production of ECM-coated Dishes-Human umbilical vein-derived endothelial cells (HUVECs) were prepared from umbilical veins and cultured as described previously in M199 medium supplemented with 20% fetal calf serum, vitamins, 1 ng/ml bFGF, and antibiotics (21,27). Proliferation assays using HUVECs were done as described previously (11). Bovine corneal endothelial (BCE) cells were isolated from steer eyes and cultured as described previously (28). ECM-coated dishes were prepared from cells grown in the presence or the absence of 30 mM chlorate as described previously (28,29).
Binding of VEGFs to HUVECs and to ECM-coated Dishes-The binding and the cross-linking of 125 I-VEGF 165 to confluent layers of HUVECs grown in 5-cm dishes in the presence or the absence of various competitors was done essentially as described. VEGF 165 was purified from infected Sf9 cells and iodinated as described (19,21,23,30).
Binding of VEGFs to ECM-coated 96-or 24-well dishes was performed at room temperature. The ECM-coated wells were washed with rinse buffer (10 mM Tris-HCl, pH 7.2, 150 mM NaCl, and 0.1% Tween 20). Nonspecific sites were blocked with binding buffer (20 mM K 2 HPO 4 , 10 mM KH 2 PO 4 , 1 mM EDTA, 0.8% NaCl, 1 mg/ml bovine serum albumin, pH 7.2) for 1 h at room temperature. The binding buffer was aspirated and iodinated, or unlabeled growth factors were incubated with the ECM-coated wells in binding buffer for 2 h at 24°C. Free growth factors were removed by aspiration, and the ECM was washed twice with rinse buffer. In binding experiments, 0.2 N NaOH was added to dissociate bound growth factors, and aliquots were counted in a ␥ counter or neutralized using Tris base and analyzed by SDS-PAGE followed by autoradiography. Alternatively, the wells were further incubated with various anti-VEGF monoclonal antibodies in binding buffer for 2 h and washed. Bound antibody was detected with appropriate secondary antibodies coupled to alkaline-phosphatase, using para-nitrophenylphosphate as substrate. Our M-35 anti-VEGF monoclonal antibody and commercial anti-VEGF monoclonal antibodies produced identical results in these assays.
For biological activity experiments, 15,000 HUVECs were seeded on ECM containing various amounts of adsorbed VEGF in a final volume of 1 ml of growth medium. Cells were trypsinized and counted after 3 days in a coulter counter. All experiments were repeated at least twice with similar results.

VEGF 145 Is Expressed as a Major VEGF Splice Variant in Several Tumorigenic Cell Lines Originating in the Female Re-
productive System-Reverse PCR analysis of mRNA from OC-238 human epithelial ovarian carcinoma cells (Fig. 1B), HeLa cells, and A431 cells (Fig. 1C) detected a VEGF mRNA containing a coding region smaller than that of VEGF 165 but larger than that of VEGF 121 . The size of the coding region corresponded to the expected size of a mRNA encoding a VEGF form containing exons 1-6 and 8 and should lead to the production of a VEGF form containing 145 amino acids (VEGF 145 ) (15). In all these cell lines the VEGF 145 cDNA seemed to be expressed at levels comparable with those of VEGF 165 . The VEGF 145 mRNA was not detected in several other transformed cell lines including C6 glioma cells and U937 cells. Sequence analysis of the PCR product from the OC-238 cells showed that this mRNA was indeed generated by alternative splicing and that it contains exons 6 and 8 but not exon 7 of the VEGF gene.
To study the properties of VEGF 145 , we have expressed the VEGF 145 cDNA in Sf9 insect cells using the baculovirus expression system (23). Most of the VEGF 145 produced by the infected Sf9 cells was found in the conditioned medium as a homodimer of ϳ41 kDa, with small amounts of monomeric VEGF 145 (Fig.  2B). The VEGF 145 dimers dissociated into monomers upon reduction with dithiotreitol ( Fig. 2A). VEGF 145 was partially purified by heparin-Sepharose affinity chromatography. VEGF 145 was eluted from heparin-Sepharose columns using a stepwise salt gradient. Most of the VEGF 145 eluted at 0.6 -0.7 M NaCl, indicating that the heparin binding affinity of VEGF 145 is similar to that of VEGF 165 (data not shown) (17,23). The recombinant VEGF 145 was biologically active and induced the proliferation of HUVECs. The ED 50 of VEGF 145 was 30 ng/ml, whereas VEGF 165 was 6-fold more active than VEGF 145 in this assay (Fig. 3).
VEGF 145 Induces Angiogenesis in Vivo-To determine whether VEGF 145 can induce angiogenesis in vivo, the VEGF 145 cDNA was subcloned into the BamHI site of the mammalian bicystronic expression vector MIRB (25). The MIRB/VEGF 145 plasmid was transfected into BHK-21 cells (31), and stable cell lines producing VEGF 145 were isolated. The VEGF 145 produced by the mammalian cells was biologically active and was secreted into the growth medium. A stable clone producing 0.1 g of VEGF 145 per 10 6 cells was isolated. The VEGF 145 expressing cells were embedded in alginate beads, and the beads were implanted under the skin of BALB/c mice (32). The pellets containing the alginate beads were removed after 4 days and photographed. Clusters of alginate beads containing VEGF 145 expressing cells were dark red with blood, whereas beads containing cells transfected with vector alone had a much lower content of blood (Fig. 4). When examined under higher magnification, pellets containing VEGF 145 producing cells appeared much more vascularized than pellets containing control cells.
VEGF 145 Binds to the KDR/flk-1 Receptor but Not to the Two Smaller VEGF Receptors of HUVECs-125 I-VEGF 165 forms high molecular weight complexes with three types of VEGF receptors following cross-linking to HUVECs (Fig. 5, lane 1), whereas 125 I-VEGF 121 only binds to the larger of these receptors. The common receptor to which both VEGF 121 and VEGF 165 bind is the KDR/flk-1 VEGF receptor (Fig. 5, open arrow) (21). In order to compare the receptor recognition pattern of VEGF 145 with those of VEGF 165 , 125 I-VEGF 165 was bound to HUVECs in the presence of 1 g/ml of heparin and increasing concentrations of VEGF 145 . Bound 125 I-VEGF 165 was subsequently covalently cross-linked to the VEGF receptors. VEGF 145 inhibited the binding of 125 I-VEGF 165 to the KDR/flk-1 receptor of the HUVECs (Fig. 5). This result was verified in a cell-free binding experiment in which VEGF 145 competed with 125 I-VEGF 165 for binding to a soluble fusion protein containing the extracellular domain of the flk-1 receptor (data not shown) (33). In contrast, VEGF 145 did not effectively inhibit the binding of 125 I-VEGF 165 to the two smaller VEGF receptors of the HUVECs (Fig. 5, filled arrow), indicating that the affinity of VEGF 145 toward these two receptors is substantially lower than that of VEGF 165 . This behavior resembles the behavior of VEGF 121 (21) and indicates that the presence of exon 6 is not sufficient to enable efficient binding of VEGF 145 to these two receptors, despite the heparin binding properties that exon 6 confers on VEGF 145 .
VEGF 145 Binds to the ECM Produced by Corneal Endothelial   FIG. 1. A, the structure of the VEGF splice variants. The peptides encoded by the various exons of the human VEGF gene are shown in boxes but are not drawn to scale. The number of amino acids in each of the exon-encoded peptides is shown at the bottom. The exon structure of VEGF 145 is shaded. B, expression of VEGF 145 mRNA in OC-238 human epithelial ovarian carcinoma cells. Total RNA from OC-238 cells was translated into cDNA and amplified by PCR using radioactively labeled nucleotides. PCR products were separated on a polyacrylamide gel as described under "Experimental Procedures." Shown is an autoradiogram of the gel. The amplified species of VEGF and L19 cDNA are indicated. C, expression of VEGF 145 in A431 and HeLa cells. Total RNA from HeLa and A431 cells was translated into cDNA and amplified by PCR using radioactively labeled nucleotides as described under "Experimental Procedures." Plasmids containing the VEGF 121 cDNA, the VEGF 165 cDNA, and the VEGF 145 recombinant cDNA were included in separate PCR reactions using the primers described under "Experimental Procedures." Shown is an autoradiogram of the gel.
Cells-VEGF 189 binds efficiently to the ECM produced by CEN4 cells, whereas VEGF 165 binds to it very weakly (9). The fact that VEGF 189 binds heparin with high affinity led to the suggestion that the interaction of VEGF 189 with the ECM is mediated by heparan sulfate proteoglycans (8,9). The heparin binding affinities of VEGF 145 and VEGF 165 are similar and substantially lower than the heparin binding affinity of VEGF 189 (8). We therefore expected VEGF 145 to bind poorly to ECM. Unexpectedly, experiments in which VEGF 145 was bound to an ECM produced by bovine corneal endothelial cells (28,34) showed that VEGF 145 bound efficiently, whereas the binding of VEGF 165 was marginal (Fig. 6A). In these experiments the binding was monitored with anti-VEGF antibodies, but similar results were obtained when binding to the ECM was assayed directly using 125 I-VEGF 145 (30 ng/ml) or 125 I-VEGF 165 (50 ng/ml) (Fig. 6B, first and third lanes). The binding of 125 I-VEGF 145 to the ECM was substantially but not completely inhibited by 10 g/ml heparin (Fig. 6B, second lane). The 125 I-VEGF 145 used in these experiments contained some impurities (Fig. 6C), but the major iodinated protein that was recovered from the ECM had a mass corresponding to that of 125 I-VEGF 145 (Fig. 6B, first lane). To make sure that 125 I-VEGF 145 binds to the ECM and not to exposed plastic surfaces, the ECM was scraped off and washed by centrifugation, and the amount of adsorbed 125 I-VEGF 145 in the pellet was determined. The ECM contained ϳ70% of the adsorbed 125 I-VEGF 145 . It therefore appears that the presence of the exon 6-derived peptide in VEGF 145 enables efficient binding to the ECM, whereas the exon 7 derived peptide of VEGF 165 does not suffice to confer this ability on VEGF 165 .
VEGF 145 Binds to the ECM Using a Mechanism That Is Not Dependent on ECM-associated Heparan Sulfates-The interaction of bFGF with the ECM is mediated by the heparan sulfate moieties of ECM associated proteoglycans (35). It was of interest to determine if VEGF 145 uses a similar mechanism. When 125 I-VEGF 145 was bound to ECM-coated dishes in the presence of 10 g/ml heparin, the binding was inhibited by ϳ60% (Fig.  7A). Under the same conditions the binding of 125 I-bFGF to the were produced in Sf9 insect cells as described under "Experimental Procedures." Conditioned medium containing recombinant VEGF was collected, and 10-l aliquots were either reduced using 0.1 M dithiothreitol (A) or not reduced (B). Proteins were separated by SDS-PAGE (12% gel) and transferred by electroblotting to nitrocellulose. Filters were blocked for 1 h at room temperature with buffer containing 10 mM Tris-HCl, pH 7, 0.15 M NaCl, and 0.1% Tween 20 (TBST) supplemented with 10% low fat milk. The filters were incubated for 2 h at room temperature with rabbit anti-VEGF polyclonal antibodies in TBST (23), washed three times with TBST, and incubated with anti-rabbit IgG peroxidase-conjugated antibodies for 1 h at room temperature. Bound antibody was visualized using the ECL detection system. ECM was inhibited by 80% (Fig. 7A). The binding of 125 I-VEGF 145 to the ECM was also inhibited by 80% in the presence of 0.8 M salt, indicating that the interaction is probably not hydrophobic (data not shown). These results are compatible with the expected behavior of proteins that bind to the ECM via heparin-like molecules. However, 125 I-VEGF 145 also bound efficiently to an ECM that was digested with heparinase-II (36).
In order to further investigate the mode of interaction of VEGF 145 with the ECM, we measured the heparin-or heparinase-II-induced release of prebound VEGF 145 or bFGF from the ECM. When the ECM-coated wells were incubated for 2 h at 37°C with buffer, only 20% of the bound 125 I-bFGF and 13% of FIG. 7. The effects of heparinase-II and heparin on the binding of 125 I-VEGF 145 and 125 I-bFGF to ECM-coated wells. A, effect of heparin and heparinase on growth factor binding. ECM-coated wells were incubated with or without 0.1 unit/ml heparinase-II in binding buffer for 2 h at 37°C. Subsequently, 125 I-VEGF 145 (40 ng/ml) or 125 I-bFGF (114 ng/ml) was added to the wells in the presence or the absence of 10 g/ml heparin. Following incubation for 3 h at 25°C, the wells were washed, and ECM-associated iodinated growth factors were dissociated by digestion with trypsin for 15 min at 37°C. The amount of bound growth factor was determined using a ␥ counter (100% binding was 15,000 and 25,000 cpm/well for 125 I-VEGF 145 and 125 I-bFGF, respectively). B, effect of heparin and heparinase-II on the release of bound growth factors from the ECM. 125 I-VEGF 145 or 125 I-bFGF were bound to ECM-coated wells as described above. The wells were washed and reincubated in binding buffer alone, with 10 g/ml heparin, or with 0.1 units/ml heparinase-II in a final volume of 50 l. Following 12 h of incubation at 25°C, the integrity of the ECM was verified by microscopy, and 45-l aliquots were taken for counting in a ␥ counter. NaOH was then added to the wells, and the amount of ECM-associated growth factors was determined. The experiment was carried out in parallel to the experiment described in A above. The experiments in A and B were carried out in duplicate, and variation did not exceed 10%. Shown are the mean values. The experiments were repeated four times with similar results. the bound 125 I-VEGF 145 dissociated from the ECM (Fig. 7B). This background release may be attributed in part to a proteolytic activity residing in the ECM (8,39). When 10 g/ml heparin were included in the buffer, only 33% of 125 I-VEGF 145 was released from the matrix, as compared with the release of 78% of the prebound 125 I-bFGF. An even sharper difference was observed when heparinase-II was added to the buffer. The enzyme released 72% of the bound 125 I-bFGF, but only 17% of the bound 125 I-VEGF 145 was released (Fig. 7B). Similar results were obtained when the experiment was performed with unlabeled VEGF 145 , using a commercial monoclonal anti-VEGF antibody to detect VEGF associated with the ECM (data not shown).
To determine the efficiency of the heparinase-II digestion, the ECM was metabolically labeled with [ 35 S]sulfate and subsequently digested with heparinase-II. The digestion released 80 -85% of the labeled sulfate residues (data not shown). To determine whether VEGF 145 can bind to ECM depleted of all types of sulfated glycosaminoglycans, BCE cells were grown in the presence of 30 mM chlorate, an inhibitor of glycosaminoglycan sulfation (29). These ECMs were further digested with a mixture of heparinases I, II, and III (36,40). Neither of these treatments significantly inhibited the binding of VEGF 145 to the ECM, despite a Ͼ95% decrease in the content of ECMassociated sulfate moieties (data not shown).
Because endothelial cells do not proliferate when they are seeded on ECM produced in the presence of chlorate (29), we examined whether VEGF 145 bound to such ECM retains its biological activity. Wells coated with ECM produced in the presence of chlorate were incubated with increasing concentrations of either VEGF 145 or VEGF 165 . The wells were subsequently washed extensively and HUVECs were seeded in the wells. ECM incubated with VEGF 145 induced proliferation of vascular endothelial cells, whereas ECM incubated with VEGF 165 did not (Fig. 8). We therefore conclude that the ECMassociated VEGF 145 is biologically active.

DISCUSSION
Alternative splicing represents an important mechanism for the generation of diversity in growth factors and in their receptors. The alternative splice forms generated from the VEGF gene share angiogenic properties and are active as mitogens for endothelial cells. The VEGF 145 mRNA has been previously detected as a rare VEGF mRNA species in placenta (15,16). We have found expression of the VEGF 145 mRNA in several tumorigenic cell types originating from the female reproductive system at levels comparable with the expression levels of VEGF 165 . Several transformed cell lines from other sources did not express this mRNA, indicating that the expression of VEGF 145 may be more restricted compared with other VEGF forms (6). However, it remains to be seen whether production of VEGF 145 is restricted to the female reproductive system.
To study the properties of VEGF 145 , we have produced recombinant VEGF 145 in mammalian and in insect cells. VEGF 145 was found to induce endothelial cell proliferation and in vivo angiogenesis, in agreement with previous studies that have indicated that these functions are not dependent on the presence of either exon 6 or exon 7 (10) and seem to be associated with the ability to bind to the KDR/flk-1 VEGF receptor (41,42). However, VEGF 145 seemed to be somewhat less active than VEGF 165 , and it is possible therefore that the presence of exon 6 in VEGF 145 can subtly alter the conformation of the protein at the KDR/flk-1 binding site (7). The VEGF 145 protein was secreted into the growth medium of producing cells. VEGF 121 , VEGF 145 , and VEGF 165 are secreted into the medium by producing cells, whereas VEGF 189 and VEGF 206 are sequestered by cell surface heparan sulfates (8). It therefore seems that the simultaneous presence of both exon 6 and exon 7 is required for efficient binding of VEGF to cell surfaces, whereas the presence of each of these exons on its own does not confer this property on VEGF.
Unlike VEGF 121 , VEGF 145 was able to bind to heparin-Sepharose columns, indicating that the exon 6 encoded peptide acts as an independent heparin binding domain. The affinity of VEGF 145 for heparin was similar to that of VEGF 165 , even though the structures of the heparin binding domains of the two isoforms differ. All the VEGF isoforms tested to date bind to the KDR/flk-1 receptor of endothelial cells (21), and VEGF 145 was no exception. However, unlike VEGF 165 , VEGF 121 and VEGF 145 do not bind efficiently to two additional VEGF receptors on HUVECs (21). The fact that VEGF 121 and VEGF 145 are both mitogenic and angiogenic indicates that these two receptors do not play a central role in the angiogenic and mitogenic response of endothelial cells. The biological function and the molecular structure of these two additional receptors is unknown. They appear to be novel VEGF receptors because they are not recognized by VEGF 121 , whereas both KDR/flk-1 and flt-1 are recognized by VEGF 121 . 2 In addition these two receptors are not immunoprecipitated by antibodies directed against the intracellular domains of the known VEGF receptors (21), and they can be detected in tumor cells that do not express detectable levels of KDR/flk-1 or flt-1 mRNA (22). Our results suggests that specific exon 7 sequences that are not present in the exon 6-derived peptide are required for the interaction of VEGF 165 with these two receptors. The heparin binding ability conferred on VEGF 165 by the exon 7 peptide may not play a central role in the recognition of these receptors by VEGF 165 because VEGF 145 and VEGF 165 bind with similar affinities to heparin yet differ in their ability to recognize these two receptors.
VEGF 189 is not found in the conditioned medium of produc-2 G. Neufeld, unpublished results.

FIG. 8. VEGF 145 bound to the ECM produced by BCE cells promotes proliferation of endothelial cells. Wells of 24-well dishes
were coated with an ECM produced by BCE cells cultured in the presence of 30 mM chlorate as described (29). The ECM-coated wells were incubated with increasing concentrations of VEGF 145 (f) or VEGF 165 (Ⅺ) as indicated and washed extensively as described. HU-VECs (15,000 cells/well) were seeded in the ECM-coated wells in growth medium lacking growth factors. Cells were trypsinzed and counted after 3 days. The numbers represent the average number of cells in duplicate wells. The experiment was repeated twice with similar results. Variation within duplicates did not exceed 10%.
ing cells and is sequestered on cell surface heparan sulfates and in the ECM. In view of the strong interaction of VEGF 189 with heparin, it was suggested that VEGF 189 binds tightly to cell surface and ECM localized heparan sulfates, allowing the sequestration of VEGF 189 in the ECM and on cell surfaces. This was supported by experiments that showed that heparin can dissociate bound VEGF 189 from the ECM, although long incubation times and high heparin concentrations were required (9). In contrast, VEGF 165 interacts weakly at best with cell surface heparan sulfates and with the ECM and is released into the medium of producing cells (8,9). We expected VEGF 145 to bind weakly to cell surfaces and to the ECM because VEGF 165 and VEGF 145 appear to bind to heparin with similar affinities. Indeed, we have found that VEGF 145 is secreted as expected into the medium of VEGF 145 producing cells. However, VEGF 145 bound to the ECM produced by corneal endothelial cells much better than either VEGF 165 or VEGF 121 . The lowest VEGF 145 concentration at which binding to the ECM was observed was about an order of magnitude lower than the concentration at which binding of VEGF 165 to the ECM was detected. In addition we have observed that VEGF 145 that is bound to ECM is able to promote proliferation of endothelial cells. These observations prompted us to compare the ECM binding behavior of VEGF 145 with that of bFGF, a growth factor that binds specifically to ECM-associated heparan sulfate moieties (38). Unexpectedly, our observations suggested that VEGF 145 and bFGF do not bind to common binding sites on the BCE cell-derived ECM. Digestion of the ECM with heparinases releases bound bFGF but does not release bound VEGF 145 . It is possible that VEGF 145 binds to a heparan sulfate subpopulation that is not recognized by bFGF or by the heparinases used in the present study. However, VEGF 145 was also able to bind efficiently to an ECM produced in the presence of chlorate, an inhibitor of glycosaminoglycan sulfation (43) and to ECM digested with a mixture of three different heparinases resulting in a greater than 95% depletion of ECM-associated sulfate groups. These experiments therefore indicate that VEGF 145 can bind to ECM components distinct from the heparan sulfate side chains of proteoglycans. A similar observation was reported for transforming growth factor-␤1, a heparin binding protein (44) that binds to the core protein of the ECMassociated chondroitin sulfate/dermatan sulfate proteoglycan decorin (45) rather than to ECM-associated heparan sulfate moieties.
All the splice variants induce angiogenesis in vivo, so why are five VEGF variants produced? Angiogenesis is often initiated under adverse conditions, such as the conditions encountered during wound healing. Many cell types produce several VEGF forms simultaneously (6,46), and it is possible that each form offers advantages in different situations. The simultaneous production of several different VEGF forms may therefore ensure a balanced angiogenic response under diverse circumstances. When the properties of the VEGF variants are examined, it is apparent that the differences in their heparin binding abilities may affect their diffusion from a VEGF producing source to target blood vessels. VEGF 121 does not bind to either heparan sulfates or to the ECM and should therefore diffuse more readily than the heparin and ECM binding VEGF forms. The ECM may serve as a storage depot for the VEGF forms that bind efficiently to the ECM, and these forms may dissociate slowly from the ECM providing prolonged angiogenic stimulation or be released from the ECM as a result of the activity of proteases (8). The balance may tip toward the production of preferred VEGF isoforms under certain conditions (46). Production of VEGF 145 may be such a case because the variety of cell types that produce VEGF 145 appears to be limited com-pared with the range of cell types producing VEGF 121 or VEGF 165 . However, the mechanism that determines what VEGF forms should be produced by a given cell type remains to be elucidated.