VEGF₁₆₅b, an Inhibitory Vascular Endothelial Growth Factor Splice Variant: Mechanism of Action, In vivo Effect On Angiogenesis and Endogenous Protein Expression

Growth of new blood vessels (angiogenesis) is required for all tumor growth, as well as in nonneoplastic pathologies such as neointimal hyperplasia, arthritis, atherosclerosis, and diabetes (1). Angiogenesis is stimulated by the expression of vascular growth factors, the most commonly expressed of which is vascular endothelial growth factor (VEGF). VEGF mRNA and protein are up-regulated in all known solid tumors as in nearly all other known instances of angiogenesis (2, 3). Multiple isoforms of VEGF are generated by alternative splicing (ref. 4; Fig. 1). The multiple isoforms of conventional vascular endothelial growth factor-A result from differential splicing of pre-mRNA from eight exons resulting in at least six mRNA and (presumed) peptide species identified by the exon composition and amino acid length of the final proteins (ref. 5; Fig. 1). All conventional splice variants studied, including the most common isoform, VEGF₁₆₅, have been shown to be proangiogenic, and they exert this effect by stimulating endothelial cell migration, proliferation, and lumen formation and are potent proangiogenic vasodilators and mediators of increased microvascular permeability (6, 7). These effects are mediated principally by activation of VEGF receptor 2 (VEGFR-2)(KDR/flk1; refs. 3, 8). Endothelial cells also express VEGFR-1; however, the specific role of this receptor in endothelial physiology is less well defined (3, 8).

Of all of the growth factors involved in angiogenesis, VEGF appears to play an irreplaceable role. Transgenic knockout models of only a single gene copy are commensurate with embryonic lethality (9, 10). VEGF-A is highly up-regulated in all cases of pathological angiogenesis yet described (11), and these include the major killers—not only cancer but also cardiovascular disease and diabetes—in the developed world (e.g., refs. 12, 13). It is also required for physiological angiogenesis, for example, in embryonic and placental development. VEGF₁₈₉, VEGF₁₆₅, and VEGF₁₂₁ are commonly overexpressed, but VEGF₁₆₅ appears to predominate quantitatively and functionally in most angiogenic states.

VEGF was originally described as a growth and survival factor for endothelial cells. It is, however, becoming increasingly apparent that VEGF is also an important factor in the function, physiology, and pathology of cancer cells themselves, as well as a wide variety of nonneoplastic cell types, including the migration and growth of neurones (14, 15) and migration of monocytes (16). For instance, it has been shown to act as an autocrine survival factor for breast and prostate cancer cells (17, 18) but also for hematopoietic stem cells (19) and podocytes (20).

Considerable interest therefore results from the realization that VEGF expression is highly regulated at both mRNA and protein levels in tissues that are not normally angiogenic, e.g., cerebellum (21), prostate (22), pancreatic islets (23), and glomeruli (24). While investigating this paradox, we recently discovered mRNA encoding a novel isoform, VEGF₁₆₅b, which did not appear to stimulate endothelial cell proliferation or migration and was down-regulated in renal cell carcinoma (25). This isoform, formed by distal splice site selection in the terminal exon of VEGF, predicts an open reading frame encoding an alternate COOH-terminal sequence but the same number of amino acids in the mature protein (Fig. 1 B). This predicted the translation of a protein of the same length as VEGF₁₆₅, but with a different sequence and hence possibly a different mechanism of action. VEGF₁₆₅b was so named because it also contains 165 amino acids. However, the COOH-terminal six amino acids usually coded for by exon 8 (CDKPRR) would be replaced by six different amino acids (SLTRKD) coded for by 18 bases of mRNA spliced 66 bases downstream of the usual acceptor splice site for exon 8, if this isoform was translated. We initially termed this new open reading frame exon 9, although there is no true intron between the two reading frames, and the alternative splicing should perhaps more correctly be referred to as exon 8 proximal and distal splicing sites. This isoform was identified serendipitously, and the high degree of homology between VEGF₁₆₅b and VEGF₁₆₅ (96%) perhaps explains the elusiveness of this isoform until now. VEGF₁₆₅b and VEGF₁₆₅ would not be distinguished in most nucleic acid and protein assays but would be assumed to be the same product.

It has been previously shown that the COOH terminus of VEGF is necessary for determining mitogenic potency (26). We therefore speculated that the novel COOH terminus would influence function and subsequently showed that conditioned media from cells expressing synthetic, recombinant VEGF₁₆₅b inhibited VEGF₁₆₅-mediated endothelial cell proliferation and migration in vitro and vasodilatation ex vivo (25). The mechanisms by which this occurred, however, and whether VEGF₁₆₅b was a true endogenous inhibitor (i.e., it was produced as an endogenous protein and could block angiogenesis in vivo) were not shown. Moreover, the discovery of the exon 8 distal splicing site predicted but did not demonstrate the existence of an entire family of sister isoforms (VEGF_xxxb), all with inhibitory potential. We have therefore carried out experiments to determine (a) whether VEGF₁₆₅b competitively inhibits VEGF₁₆₅-mediated activity by binding to the same receptor (VEGF-R2) but inhibiting receptor phosphorylation and downstream intracellular signaling, (b) whether exon 8 distal splicing site isoforms are expressed in human tissues and plasma, e.g., VEGF₁₈₉b, VEGF₁₆₅b, and VEGF₁₂₁b, (c) whether VEGF₁₆₅b inhibits VEGF₁₆₅-mediated angiogenesis in vivo; and (d) whether the different family of isoforms have differing effects on tumor growth.

Recombinant Human VEGF₁₆₅b was obtained from R&D Systems (Abingdon, United Kingdom). This was produced in the same manner as commercially available recombinant human VEGF₁₆₅.

Immulon II HB Flat well 96-well plates were incubated with 100 μL of 1 μg·mL^-1 VEGFR-2-Fc Chimera (R&D Systems) overnight, and washed three times with 1× PBS (pH 7.4), supplemented with 0.05% Tween 20 (PBS-T; Sigma-Aldrich, Poole, Dorset, United Kingdom). For cell assays human umbilical vascular endothelial cells were grown to confluency in 96-well plates. The plate was blocked with 800 μL of 3% BSA for 2 hours and again washed twice with PBS-T. A total of 10 ng·mL^{-1 125}I-VEGF₁₆₅ and 100 μCi·mL^-1 was mixed with increasing concentrations of VEGF₁₆₅b or VEGF₁₆₅, and 100 μL added to the ELISA plate. This was incubated for 4 h at room temperature, then gently washed with PBS three times. 10% SDS was then added, and the solution transferred to a gamma counter for assays. Counts were expressed as the percent of the counts of wells incubated with ¹²⁵I-VEGF₁₆₅ alone.

Terminally differentiated conditionally immortalized glomerular visceral epithelial cells (podocytes, a kind gift of Moin A. Saleem, Children’s Academic Renal Unit, Southmead Hospital, University of Bristol, Bristol, United Kingdom) were cultured as described previously (27). Chinese hamster ovary (CHO) cells were transfected with either an empty vector (PEQ176) or VEGFR-2 containing expression vector. Cells were routinely passaged in nutrient mixture DMEM:Ham’s F-12 (CHO) or RPMI (podocytes) with l-glutamine (Invitrogen Corp.) supplemented with 10% fetal calf serum (Invitrogen Corp.), 1% penicillin-streptomycin (Invitrogen Corp.), and 1% of the selection agent, Geneticin (Invitrogen Corp). Transfected CHO cells were incubated with serum-free media for 18 to 24 hours followed by incubation with media or media containing 40 ng·mL^-1 VEGF₁₆₅, 40 ng·mL^-1 VEGF₁₆₅b, or 40 ng·mL^-1 VEGF₁₆₅, and 40 ng·mL^-1 VEGF₁₆₅b for 20 minutes at 37°C. Cells were washed and collected with PBS followed by centrifugation at 3000 rpm for 5 minutes and protein extracted as described below.

Blood was collected from normal healthy volunteers ages 21 to 40. Blood was taken into EDTA vacutainers and centrifuged at 4000 × g for 10 minutes. Plasma was removed from the erythrocytes and stored at −80°C until defrosted and used for ELISA. Prostate chips were obtained from patients undergoing transurethral resection of the prostate for lower urinary tract symptoms. Samples were frozen at −80°C immediately until processed. Patients with benign prostatic hyperplasia and advanced prostate cancer (stage T₃N_xM_0–1; UICC2002) were included. Chips were defrosted on ice and manually sliced with a sterile blade. The mass of each tissue was recorded, and samples were homogenized on ice for 10 minutes and protein extracted as described below. Rat lungs were removed from animals that were humanely killed by cervical dislocation. An equivalent volume of PBS containing either saline, 1 nmol/L VEGF₁₆₅, 1 nmol/L VEGF₁₆₅b, or 1 nmol/L VEGF₁₆₅ and 1 nmol/L VEGF₁₆₅b was added to each sample. The samples were chopped up crudely on ice with a sterile scalpel blade and then incubated for 20 minutes at 37°C in a shaking incubator. After this they were snap-frozen in liquid nitrogen.

Lysis buffer (1 μL/mg tissue; 1 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L Va₃N0₄, 1.0 μg/mL leupeptin, 1 μg/mL aprotonin, and 1 μg/mL pepstatin, in radioimmunoprecipitation assay buffer) was added to cells and prostate tissue and homogenized on ice for 10 minutes, placed at 4°C on an agitating rocker plate for 20 minutes, removed and placed in ice for an additional 60 minutes, and agitated every 10 to 15 minutes.

Samples were centrifuged at 4°C for 15 minutes at 13,000 rpm, and the supernatant collected and stored at −20°C. Protein concentrations were determined by photospectrometry. For rat lung tissue, after thawing, the samples were mixed gently, excess liquid was removed, and protein extracted by adding an equal volume of lysis buffer (140 mmol/L NaCl, 3 mmol/L KCl, 10 mmol/L Na₂HPO₄, 2.7 mmol/L KH₂PO₄, 1% NP40/IGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 1 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L sodium orthovanadate, 20 μg·mL^-1 aprotinin, 10 μg·mL^-1 leupeptin, and 10 μg·mL^-1 pepstatin) and homogenizing on ice for 10 minutes. Homogenates were left at 4°C on a rocker for 1 hour and then spun down in a benchtop centrifuge for 15 minutes at 13,000 rpm at 4°C to separate the protein from cell debris.

Synthetic peptide fragments of the nine amino acid COOH-terminal sequence of VEGF₁₆₅b were coupled to keyhole limpet hemocyanin (Department of Biochemistry, University of Bristol) serving as carrier molecules and were then used to immunize 6 to 8-week-old female BALB/c mice. The animals received s.c. injections of 100 μg of peptide-keyhole limpet hemocyanin conjugates in Freund’s Complete Adjuvant on days 1, 21, and 42 and boosted by i.p injections at 63, 64, and 65 days. Mice were killed humanely the next day and spleens collected. Splenocytes were fused to the NS0 mouse myeloma cell line with polyethylene glycol. Fused cells were cultured in 96-well plates for 2 weeks. Cells from positive wells determined by ELISA screening were serially diluted in 96-well plates and cultured in 10% DMEM and hybridoma-cloning enhancing factor. The same procedure was repeated until 100% positivity from each plate was achieved three consecutive times. Screening was done in Immulon II HB Flat well 96-well plates (Thermo Life Sciences Ltd.), coated with goat antihuman VEGF antibody (0.8 μg/mL in PBS; R&D Systems). After washing with PBS-T, 100 μL of 2 ng/mL VEGF₁₆₅b (from the conditioned medium of transfected cells) or recombinant VEGF₁₆₅ (R&D systems) were added to the wells and incubated for 15 minutes at 37°C with shaking. After washing, 100 μL of conditioned medium from hybridoma cells were added and incubated for 15 minutes at 37°C with shaking. After washing, 100 μL of horseradish peroxidase (HRP)-conjugated goat antimouse immunoglobulins (1:1000 in 1% BSA/PBS; DAKO, Carpinteria, CA) were added and incubated for 15 minutes at 37°C with shaking. After final washing, O-phenylenediamine dihydrochloride substrate (Sigma Chemical Co., St. Louis, MO) was added, and the absorbance at 492 nm was measured using a plate reader. To purify and concentrate the monoclonal antibodies, the selected clones of hybridoma cells were cultured in DMEM (Sigma Chemical Co.) containing 10% bovine IgG-depleted FCS (Hyclone, Logan, UT) with 100 units of penicillin, 100 μg of streptomycin, and 2 mmol/L l-glutamine. Monoclonal antibodies were purified on protein-G Sepharose 4 Fast Flow columns (Amersham Biosciences). The antibodies were concentrated with vivaspin 20 (Vivascience AG, Hannover, Germany) and finally dissolved in PBS.

All protein samples were resuspended in SDS sample buffer, heated at 100°C for 5 minutes, and subsequently resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane. The membranes were blocked with PBS, 10% Marvel, and 0.05% Tween 20 for 1 hour at room temperature, followed by overnight incubation at 4°C with the primary antibody (see below).

Membranes containing recombinant VEGF₁₆₅ and/or VEGF₁₆₅b protein (100 ng of each) and protein samples extracted from cells and transurethral resection of the prostate chips (100 to 150 μg of each) were probed with mouse anti-VEGF₁₆₅b IgG1 described above (Fig. 2 A). The membranes were then probed with HRP-conjugated stabilized goat antimouse IgG for 1 hour at room temperature [1:2000, all dilutions in PBS, 5% dried milk, 0.05% Tween 20 (5% Marvel–PBS-T); Santa Cruz Biotechnology, Santa Cruz, CA]. Immunodetection of recombinant protein was done with a BM Chemiluminescence Blotting Substrate (POD) kit (Roche). Immunodetection of other proteins (e.g., from tissues and cells) was assessed using SuperSignal West Femto Maximum Sensitivity Substrate Kit (Pierce, Rockford, IL). Membranes were stripped with enhanced chemiluminescence stripping buffer and probed with a pan VEGF primary antibody [VEGF-(C-1); Santa Cruz Biotechnology] 1:100 dilution in 5% Marvel–PBS-T for recombinant proteins, and 1:200 dilution for transurethral resection of the prostate chip proteins) overnight at 4°C. Membranes were then incubated with the HRP-conjugated goat antimouse IgG used above (1:2000), and immunodetection was done.

Membranes containing protein extracted from CHO cells transfected with VEGFR-2 (Flk) were probed with mouse anti-phospho-p44/42 (Cell Signaling; 1:500) or rabbit anti-phospho-AKT (1:200 and 1:500; BD Bioscience, PharMingen, San Diego, CA) by overnight incubation at 4°C. After washing in PBS-T, the blots were then incubated for 1 hour with HRP-conjugated goat antimouse or goat antirabbit (1:2000; Pierce) antibody, and visualized as above. Subsequently, membranes were stripped with enhanced chemiluminescence stripping buffer at 50°C for 30 minutes with agitation followed by a visualization step to ensure that all of the immunoreactive bands had been removed. The blots were blocked with 10% Marvel–PBS-T for 1 hour at room temperature on a roller followed by overnight incubation at 4°C on a rocker with mouse anti-p44/42 (1:500; Cell Signaling) or mouse anti-protein kinase Bα/AKT1 (PkB-175, 1:400; Sigma) washed and incubated for 1 hour with HRP-conjugated goat antimouse antibody, as above.

Immunoreactive bands were visualized by chemiluminescence with a compact X4 developer from X-ograph imaging systems. Densitometry, measured using the NIH image software, was used to determine the mean intensity of the immunoreactive bands.

Immunohistochemistry was done on formalin-fixed, paraffin-embedded tissue derived from the normal pole of nephrectomy specimens. Five-micrometer thick sections were cut and mounted onto poly-l-lysine–coated glass slides. Sections were dewaxed, rehydrated, washed in distilled water, and then rinsed in 0.01 mol/L Tris-buffered saline (PBS, pH 7.2). Sections were microwave heated in 0.1 mmol/L Tris-HCl/2 mmol/L EDTA pH buffer (pH 9.0) at 650 watts for two cycles of 8 minutes, washed twice with distilled water for 5 minutes, then treated with 0.015% trypsin (Invitrogen Corp.) diluted in PBS for 15 minutes at 37°C. Sections were washed twice, incubated with 3% hydrogen peroxide solution, washed twice with PBS, blocked with 3% BSA (A4378; Sigma) in PBS and then with 1.5% normal horse serum (S-2000; Vector Lab, Peterborough, United Kingdom) in PBS. Sections were then incubated with 2 μg/mL mouse monoclonal anti-VEGF₁₆₅b IgG or a normal mouse IgG (I8765; Sigma) as a negative control diluted in 1.5% normal horse serum in PBS overnight at +4°C in a humid chamber. Sections were washed twice in 0.05% Tween Tris-buffered saline [PBS/Tween (pH7.2)] for 5 minutes and were treated with both the nonspecific blocking solution, as above, then incubated with biotinylated antimouse-IgG (BA2000; Vector Lab) at a 1:750 dilution in 1.5% normal horse serum in PBS at room temperature for 30 minutes in a humid chamber. Sections were washed twice in PBS/Tween, 5 minutes per wash, and then incubated with Vectastain ABC solution (PK4000; Vector Lab) for 30 minutes at room temperature followed by two additional washes in PBS/Tween. Sections were treated with 3,3′-diaminobenzidine peroxidase substrate solution (SK4100; Vector Lab) until color was visualized. Rinsing in distilled water stopped the reaction. Sections were washed twice in distilled water for 5 minutes, then counterstained in hematoxylin for 5 minutes, washed, dehydrated, cleared in xylene, mounted with DPX, and glass coverslipped. Sections were examined under oil immersion with a ×100 objective on a Nikon Eclipse E-400 microscope, and images were captured using a Coolpix 995 digital camera and a DN-100 digital imaging system (Nikon Instruments, Surrey, United Kingdom).

Fifty to 100 mg of transurethral resection of the prostate tissue collected as above was homogenized in Trizol reagent (Life Technologies, Inc., Rockville, MD) and mRNA extracted by using the method of Chomczyinski and Sacchi (28). Eight microliters of RNA were treated with RNase free DNase (Promega) according to the manufacturer’s guidelines to prevent genomic DNA contamination. mRNA was reverse transcribed using Moloney murine leukemia virus reverse transcriptase and poly-d(T). cDNA was then amplified using intron-spanning primers that detect VEGF₁₆₅b only, even in the presence of 1000× greater concentration of VEGF₁₆₅ mRNA (25). One μmol/L of a primer complementary to exon 4 (5′-GAGATGAGCTTCCTACAGCAC-3′) and to exon 9 and the terminal 5 nucleotides of exon 7, incorporating a HindIII restriction site at the 5′-end (exon9/7, 5′-TTAAGCTTTCAGTCTTTCCTGGTGAGAGATCTGCA-3′), together with 1.2 mmol/L MgCl₂, 200 μmol/L deoxynucleoside triphosphates, and 1 unit of TaqDNA polymerase (Abgene) were used in reactions which were cycled 35 times, denaturing at 94°C for 30 seconds, annealing at 63°C for 30 seconds, and extending at 72°C for 60 seconds. PCR products were run on 3% agarose gels containing 0.5 μg/mL ethidium bromide and visualized under a UV transilluminator. This reaction consistently resulted in one amplicon at ∼220 bp (consistent with VEGF₁₆₅b). A reverse primer complementary to exon 8 (5′-TCACCGCCTCGGCTTGTCACAT-3′) was also used that detects VEGF₁₆₅ but not VEGF₁₆₅b. Finally, cDNA was also amplified with primers designed to pick up all of the VEGF isoforms. One μmol/L of primers complementary to exon 2 (5′-GGAGGGCAGAATCATCACGAAG-3′) and exon 3 (5′-CACACAGGATGGCTTGAAGATG-3′) were used. PCR conditions were as above but annealing at 55°C.

A total of 0.8 μg/mL pan-VEGF capture antibody (Duoset VEGF ELISA DY-293; R&D Systems) diluted in 1× PBS (pH 7.4) was adsorbed onto 96-well sterile plates (655161; Greiner Bio-one Ltd., Gloucester, United Kingdom) overnight at room temperature (100 μL/well). The plates were washed three times with 1× PBS-T before and after blocking with 300 μL/well 1% BSA in PBS for 2 hours at 37°C. One hundred μL of duplicate recombinant human VEGF₁₆₅ or VEGF₁₆₅b standards (R&D Systems) diluted in wash buffer (ranging from 15 pg/mL to 4 ng/mL) and blank (wash buffer) or sample were added to each well (Fig. 2 B). After incubation for 1 hour at 37°C and three washes as before, 100 μL of biotinylated goat antihuman VEGF (0.025 μg/mL in wash solution; R&D Systems), or 100 μL of mouse anti-VEGF₁₆₅b (0.025 μg·mL^-1 in wash solution, as described above) was then added to each well, and plates left for an additional hour at 37°C. After washing, 100 μL of streptavidin-HRP (R&D Systems) at 1:200 dilution or HRP-conjugated stabilized goat antimouse IgG (1:200 dilution; Santa Cruz Biotechnology) in PBS were added, and plates left at room temperature for 20 minutes. The plates were washed an additional three times, and 50 μL/well O-phenylenediamine dihydrochloride solution (Substrate reagent pack DY-999; R&D Systems) added, protected from light, and incubated for 20 minutes at room temperature to allow color development. The reaction was stopped with 50 μL/well 1 mol/L H₂SO₄ (10276; BDH Chemicals; Poole, Dorset, United Kingdom), and absorbance read immediately in an ELISA plate reader (Labsystems Multiskan Bichromatic; Life Sciences International) at 492 nm.

MCF-7 cells were transfected with pcDNA₃-VEGF₁₆₅, pcDNA₃-VEGF₁₆₅b, or pcDNA₃ expression vector. Stable cell lines were selected in growth media containing 500 μg/mL G418. VEGF₁₆₅b lines expressed 17.8 fg/cell/hour, whereas VEGF₁₆₅ and pcDNA₃ lines expressed 4.9 and 0.08 fg/cell/hour, respectively. Corneal assays were done in female New Zealand albino rabbits (Charles River; Calco, Como, Italy). Cells were detached using trypsin and the reaction stopped by the addition of EMEM-10% FCS, followed by pelleting and suspension in EMEM-10% FCS. Cell suspensions containing 2.5 × 10⁵ cells per 5 μL (n = 4 eyes for each group) were implanted into a micropocket (1.5 × 3 mm) incised into the cornea under local anesthesia (30 mg/kg sodium pentothal). Animals were observed every 2 days with a slit-lamp stereomicroscope, without anesthesia. Neovascular responses, presence of hyperemia and edema, and infiltration with inflammatory cells in the cornea were monitored. Positive angiogenic responses were scored according to the product of vessel density and distance from the limbus (in mm), as described previously (29). Briefly, a density value of 1 corresponded to the presence of 0 to 25 vessels, a value of 2 to 25–50 vessels, a value of 3 to 50–75 vessels, a value of 4 to 75–100 vessels, and a score of 5 to >100 vessels. The distance from the limbus was measured using an ocular grid.

Adenovirus-expressing human VEGF₁₆₅b was created with the AdEasy Vector system (Qbiogene, Inc., Nottingham, United Kingdom). Briefly, a reverse transcription-PCR fragment encoding a full-length coding region of VEGF₁₆₅b was cloned into pShuttle-CMV and sequenced in both directions to confirm the correct sequence. pShuttle-CMV-VEGF₁₆₅b was linearized with EcoRI and cotransformed with pAdEasy-1 into Escherichia coli (BJ5183). Recombinant clones were screened with kanamycin resistance and checked by restriction digest. Purified, correct pAd-VEGF₁₆₅b clones were linearized and transfected into HEK293Q cells and viral plaques eluted and used to infect 1 × 10⁵ HEK293A cells

Production of VEGF₁₆₅b was assessed by ELISA. Subsequent passages of virus were generated by infecting HEK293A cells with viral stock of the previous amplification with a multiplicity of infection of ∼5. Virus was purified by double ultra-centrifugation in saturated CsCl solution. (Sorvall Discovery Ultracentrifuge 90SE; Kendro) at 65,000 rpm.

All surgical procedures were done under sterile conditions. Male Wistar rats (200 to 300 g) were anesthetized by inhalation of 5% halothane and maintained on 3% halothane. Body temperature was maintained at 37°C by a thermostatic heating pad and rectal temperature probe. After ventral laparotomy, part of the intestine was exteriorized and a mesenteric panel with few vessels was exposed under an intravital microscope (Leica DMIL). The panel was imaged with either a Nikon Coolpix 800 (Nikon Instruments) or a Leica DC350F (Leica, Bucks, United Kingdom) digital camera, and 25 μL of virus (∼10⁷ plaque forming units) were injected into the nearby fat pad with a Hamilton syringe fitted with a 30-gauge needle. Ad-VEGF₁₆₅b, Ad-VEGF₁₆₅ (30), or adenovirus-expressing enhanced GFP, (a kind gift of James Uney; University of Bristol) was used. In double infection experiments, 10⁷ plaque forming units of each virus were injected into separate, nonoverlapping parts of the fat pad (to prevent co-infection and heterodimer formation). Ten microliters of Monastral blue (0.6%, diluted in saline) were injected into the fat pads on either side of the virus-injected panel. The intestine was replaced, and the animal sutured and recovered. Six days later (day 7), the animal was re-anesthetized with halothane, and a laparotomy was done. The mesentery was exposed, and the virus injected panel located from Monastral blue injection sites and re-imaged.

After imaging, the mesenteric panel was fixed in vivo with 4% paraformaldehyde for 5 minutes, and the rat killed by cervical dislocation. The same mesenteric panel was excised and additionally fixed for 30 minutes. Immediately after fixation, the mesentery was washed with 0.5% Triton X-100 in PBS (0.5% PBX) twice for 1 hour, then blocked with 1% BSA in 0.5% PBX for 1 hour at room temperature. The mesentery was incubated with biotinylated Griffonia simplicifolia lectin IB4 (GS-IB4; Molecular Probes, Cambridge, United Kingdom) 1:250, and mouse monoclonal antibodies to either Ki-67 (NCL-l-Ki-67- MM1, 1:100; Novocastra Lab, Newcastle upon Tyne, United Kingdom) or VEGF₁₆₅b, (8 μg·mL^-1) diluted in the blocking solution overnight at 4°C on a rocker. The mesentery was then washed with 0.5% PBX five times for 1 hour and then incubated with tetramethylrhodamine isothiocyanate-labeled streptavidin (Molecular Probes, Cambridge, United Kingdom) 1:1000, Alexa Fluor 488-labeled goat-antimouse IgG (Molecular Probes) 1:500, and Alexa Fluor 633-labeled phalloidin (Molecular Probes) 1:1000, diluted in blocking solution for 2 hours at room temperature on a rocker. The mesentery was then washed in PBX five times for 1 hour and incubated with Hoechst 33324 (1 μmol/L) for 15 minutes at room temperature on a rocker. The mesentery was mounted as flat as possible with VectaShield (Vector Lab) on a slide and coverslipped. The whole mount mesentery was imaged with a Leica Confocal Microscope (Leica confocal SP2 system; Leica), or Nikon E400 Eclipse epifluorescence microscope (Nikon). Proliferating endothelial cells, sprouts, and microvessels were imaged. VEGF₁₆₅b secretion from the fat pad into the mesentery was checked by staining mesenteries with the anti-VEGF₁₆₅b antibody (Fig. 3).

The vessel area was measured using Openlab software (Improvision, Coventry, United Kingdom). The vessels were selected with the wand tool using a threshold of 32 pixels, and the area of these vessels was recorded. Fractional vessel area (FVA), used to measure the angiogenic effect of VEGF, was calculated from the vessel area as a percentage of the mesenteric area. Percent vessel area increase (%AI) was expressed as the difference between the FVA on day 7 and the FVA on day 1 as a percentage of the FVA on day 1.

After immunofluorescence staining, the whole mount mesentery was used for microvessel measurement. For each mesentery, 8 to 12 views were selected randomly with a 40× objective on a Nikon E400 Eclipse epifluorescence microscope and images were taken with the Nikon Coolpix 800 Digital Camera (Nikon, Surrey, United Kingdom). Openlab software (Improvision, Coventry, United Kingdom) was used to measure the area analyzed and vessel parameters between two adjacent branch points. The total vessels were counted and labeled, and the branch points, proliferating endothelial cells, and sprouts in each image were counted. The diameter and length of each vessel were measured. Branch point density, sprout density, and proliferating endothelial cell density were calculated as the number per unit area within five randomly selected fields of view (×40 objective) containing vessels as described previously (31).

A total of 1 × 10⁶A375 human melanoma cells stably transfected with 2.4 μg of pcDNA3 (control) vector or pcDNA3-VEGF₁₆₅ or pcDNA3-VEGF₁₆₅b (both expressing ∼4 fg.cell^-1·day^-1 VEGF) was s.c. injected into each of six nude mice. To determine the effect of a mixture of populations, 5 × 10⁵ of VEGF₁₆₅b and VEGF₁₆₅-expressing cells were combined in PBS and injected as above. Tumor width and length was measured using Vernier’s calipers, and tumor volume was calculated as the product of length width and the average of the two measurements. The mice were humanely killed once the tumors reached 16 mm in diameter or 28 days, whichever occurred soonest.

One-way ANOVA was used to analyze the difference of parameters between the groups and Student Newmann Keuls post hoc test used to compare the individual groups. P < 0.05 was considered significant.

There have been many splice variants of VEGF described, four of which have been clearly demonstrated to be translated into protein-VEGF₁₈₉, VEGF₁₆₅, VEGF₁₂₁, and large-VEGF (32, 33). These are up-regulated in renal cancer (34). We have shown that VEGF₁₆₅b mRNA was down-regulated but not up-regulated in renal cancer (25). To determine whether VEGF₁₆₅b protein was expressed in normal and pathological tissues, we raised a mouse monoclonal antibody to the predicted nine COOH-terminal amino acids of human VEGF₁₆₅b, the final six of which are specific to VEGF₁₆₅b. This antibody was then used to probe a blot of an SDS-PAGE gel loaded with recombinant VEGF₁₆₅b and VEGF₁₆₅. Fig. 2,A shows that this antibody specifically detects VEGF₁₆₅b and does not detect VEGF₁₆₅. This was true even in the presence of 50-fold greater amounts of VEGF₁₆₅. To determine whether VEGF₁₆₅b protein could be detected in normal human cells, this antibody was used on a Western blot analysis containing protein extracted from terminally differentiated conditionally immortalized human glomerular visceral epithelial cells (podocytes, a kind gift of Moin A. Saleem), which express VEGF₁₆₅b mRNA (35). The protein was then extracted from 10⁶ cells and subjected to SDS-PAGE and Western blotting. The membrane was first probed with the VEGF₁₆₅b antibody and then a commercial pan-VEGF antibody to ensure specificity. Fig. 4,A shows that the antibody detected a number of different proteins, which appear to correlate with expected molecular masses for most isoforms of VEGF, including VEGF₁₂₁b (monomer-band a, dimer-band b), VEGF₁₄₅b (band c), VEGF₁₆₅b (band d), VEGF₁₈₉b (band e), and large VEGFbs (bands at f). The pan-VEGF antibody confirmed that these bands were VEGF isoforms (Fig. 4,A). These isoforms are consistent with VEGF₁₆₅b, VEGF₁₈₉b, VEGF₁₄₅b, and VEGF₁₂₁b. A larger molecular mass isoform, which corresponds to large-VEGF formed by alternate start codon usage (33), was present in both blots. Interestingly, however, the relative intensities of the bands were different for the pan-VEGF antibody compared with the VEGF_xxxb antibody. This suggests that the relative amount of the different splice forms is not the same for VEGF_xxxb isoforms as for the conventional isoforms. Moreover, Western blotting of prostate tissue also demonstrated VEGF₁₆₅b and other isoforms, showing that VEGF₁₆₅b is expressed in human tissue (Fig. 4,B). We therefore speculate that podocytes and other human cells produce a sister family of inhibitory isoforms, including that to large-VEGF. We have described these as the VEGF_xxxb family of isoforms. To determine whether the expression of VEGF_xxxb also occurred in podocytes in human glomeruli, VEGF_xxxb protein expression was detected in sections of nephrectomy specimens by immunohistochemistry. Fig. 4 C shows clear expression of VEGF_xxxb isoforms in the podocytes on the periphery of the glomerulus (arrows)—shown to be the location of VEGF expression by numerous studies (24).

An ELISA was then used to quantify the amount of VEGF_xxxb in 23 normal human plasma and 5 benign and 4 malignant prostate tissues. VEGF_xxxb isoforms were detected in 12 of the normal plasma samples (concentration range, 32 to 452 pg·mL^-1) and all of the normal prostate samples (median, 138 pg·mg^-1 tissue; range, 29 to 396 pg·mg^-1). Interestingly, VEGF_xxxb isoforms were detected at lower levels (median, 86 pg·mg^-1 tissue, 34 to 66 pg·mg^-1) in malignant prostate samples (Fig. 4,D). To determine whether VEGF splicing was altered in prostate tissue, mRNA was extracted from transurethral resection of the prostate chips taken from 17 patients with benign prostatic hyperplasia and 9 with malignant prostate cancer. VEGF₁₆₅b mRNA was found in 82% of the benign (14 of 17) but only 22% (2 of 9) of the malignant samples (P < 0.05, Fisher’s exact test, three of each group shown in Fig. 4 E). VEGF₁₆₅b is therefore an endogenously produced protein and is down-regulated at the mRNA level in prostate tumors.

To determine whether VEGF₁₆₅b could inhibit VEGF₁₆₅-mediated signaling, we measured the ability of VEGF₁₆₅b to compete for binding of its receptors. Human umbilical vein endothelial cells were incubated with a mixture of 20 ng·ml^{-1 125}I-VEGF₁₆₅ and increasing concentrations of either unlabeled recombinant VEGF₁₆₅ (Peprotech, Rocky Hill, NJ) or purified recombinant VEGF₁₆₅b (R&D Systems). VEGF₁₆₅b competed with ¹²⁵I-VEGF for binding to the cells with the same affinity as unlabeled VEGF₁₆₅ (Fig. 5,A). To determine whether this was due to binding to the active signaling receptor VEGFR-2, we measured the ability of VEGF₁₆₅b to compete with 20 ng·ml^{-1 125}I-VEGF₁₆₅ binding to the receptor immobilized on an ELISA plate. VEGF₁₆₅b competed with the same affinity (IC₅₀ = 19.8 ± 1.3 ng/mL) as unlabeled VEGF₁₆₅ (IC₅₀ = 17.3 ± 1.3 ng/mL; Fig. 5 B). These results show that VEGF₁₆₅b can bind to VEGFR-2 with the same affinity as VEGF₁₆₅, i.e., the altered COOH terminus of VEGF₁₆₅b does not affect binding to the receptor.

To determine whether VEGF₁₆₅b could inhibit VEGFR-2 phosphorylation, human microvascular endothelial cells were treated with saline, 40 ng·mL^-1 VEGF₁₆₅, 40 ng·mL^-1 VEGF₁₆₅b, or 40 ng·mL^-1 both isoforms and subjected to immunoprecipitation with anti-phosphotyrosine antibodies and precipitate and supernatant probed for VEGFR-2 protein expression. Fig. 6,A shows that VEGFR-2 was less phosphorylated by VEGF₁₆₅b than VEGF₁₆₅ and was no more phosphorylated than untreated cells. Moreover, addition of both isoforms resulted in less phosphorylation than VEGF₁₆₅ alone or saline (P < 0.05, ANOVA, n = 3). VEGF stimulates angiogenesis and vascular permeability by activating VEGFR-2. Human microvascular endothelial cells express both VEGFR-2 and VEGFR-1, which have been shown to interact. To determine whether VEGF₁₆₅b inhibited VEGF₁₆₅ by inhibiting phosphorylation of VEGFR-2 directly, we measured the effect of VEGF₁₆₅b on VEGFR-2 phosphorylation by incubating CHO cells transfected with VEGFR-2 with saline, VEGF₁₆₅, VEGF₁₆₅b, or both isoforms as above. Cells were then immunoprecipitated with anti-phosphotyrosine antibody and both supernatant and pellet subjected to Western blotting with an antibody to VEGFR-2. Fig. 6 Bi shows that transfected cells have an endogenous level of phosphorylation of VEGFR-2 that is increased by VEGF₁₆₅ and decreased by VEGF₁₆₅b. Moreover, VEGF₁₆₅b incubation inhibited phosphorylation by VEGF₁₆₅.

Phosphorylation of p42/p44 mitogen-activated protein kinase (MAPK) by VEGFR-2 activation has been widely reported and appears to be necessary for alterations in cell behavior such as migration through extracellular matrix (36), apoptosis (37), and endothelial cell proliferation (38), as well as in vivo properties of capillaries such as vascular compliance (39). To show that the VEGF₁₆₅b-mediated inhibition of VEGFR-2 phosphorylation inhibited downstream signaling pathways, CHO cells transfected with VEGFR-2 expression vector were treated with recombinant 40 ng·mL^-1 VEGF₁₆₅b, 40 ng·mL^-1 VEGF₁₆₅, or 40 ng·mL^-1 of both isoforms or neither, the protein extracted, and subjected to SDS-PAGE and Western blotting, and immunoblotted with a phospho-p42/p44 MAPK antibody (Fig. 6,Bii). The blot was then stripped and reprobed with a p42/p44 MAPK antibody. Although treatment with VEGF₁₆₅ resulted in a significant increase in phospho-MAPK blot density, this was not the case for incubation with VEGF₁₆₅b (Fig. 6,Bii). Moreover, treatment with both VEGF₁₆₅ and VEGF₁₆₅b did not result in a significant increase in phosphorylation. Fig. 6,Bii shows the mean density of the phospho-p42/p44 MAPK band relative to total MAPK. Furthermore Fig. 6,Biii shows that, unlike VEGF₁₆₅, VEGF₁₆₅b incubation did not result in phosphorylation of AKT, and incubation with both isoforms inhibited AKT phosphorylation compared with VEGF₁₆₅ treatment alone (P < 0.01, ANOVA; Fig. 6 Biii). This shows that VEGF₁₆₅b can inhibit VEGF₁₆₅-mediated signaling of at least two different pathways stimulated by activation of VEGFR-2.

VEGF₁₆₅ has previously been shown to stimulate both Akt and p44/p42 MAPK phosphorylation in endothelial cells through VEGFR-2 activation (39, 40, 41). To determine whether VEGF₁₆₅b could inhibit VEGF-mediated signaling in endothelial cells, human microvascular endothelial cells were treated with either saline, 1 nmol/L VEGF₁₆₅, 1 nmol/L VEGF₁₆₅b, or 1 nmol/L of each isoform for 20 minutes. Protein extracted from these cells was then subjected to SDS-PAGE and Western blotting and probed with antibodies to phospho-Akt and total Akt or phospho-MAPK and total MAPK (Fig. 7). These studies showed that treatment with VEGF₁₆₅ resulted in greater phosphorylation than no treatment but treatment with VEGF₁₆₅b did not. However, surprisingly VEGF₁₆₅b did not appear to inhibit VEGF₁₆₅-mediated phosphorylation of Akt (Fig. 7,A). Furthermore, VEGF₁₆₅b treatment resulted in a significant increase in phosphorylation of p42/p44 MAPK, very similar to that elicited by VEGF₁₆₅. Moreover, treatment with both isoforms also increased phosphorylation of MAPK to at least as great a level as treatment with either 1 nmol/L VEGF₁₆₅ or VEGF₁₆₅b (Fig. 7 B).

To determine the effect of VEGF₁₆₅b on angiogenesis in vivo, we used two different assays. To determine whether cells transfected with VEGF₁₆₅b could induce angiogenesis, MCF-7 breast cancer cells were transfected with an expression vector (pcDNA3) containing VEGF₁₆₅b and implanted into the cornea of four New Zealand White rabbits (42). Cells transfected with either pcDNA3 or pcDNA3-VEGF₁₆₅ were injected into the contralateral cornea. Fig. 8 A shows examples of the vasculature of the cornea 10 days after implantation of the cells. pcDNA3-transfected cells showed no angiogenesis, whereas VEGF₁₆₅-transfected cells expressing 2 fg·cell^-1·day^-1 VEGF₁₆₅ resulted in significant angiogenesis after 5 days that was marked by 10 days.

VEGF₁₆₅b-transfected cells expressing 12 fg·cell^-1·day^-1 on the other hand did not result in angiogenesis in the cornea (P < 0.001, two-way ANOVA; Fig. 8, A and C). Moreover, when cells transfected with VEGF₁₆₅b were mixed with cells transfected with VEGF₁₆₅ and implanted into the cornea, no angiogenesis was seen (Fig. 8,B), despite significant angiogenesis when VEGF₁₆₅-transfected cells were mixed with pcDNA3-transfected cells and implanted (P < 0.0001, two-way ANOVA; Fig. 8, B and D).

To determine whether VEGF₁₆₅b could inhibit VEGF₁₆₅-mediated angiogenesis in a separate assay in rats, we used a modification of a mesenteric angiogenesis assay (31). Adenoviruses expressing EGFP, VEGF₁₆₅b, or VEGF₁₆₅ were injected into the mesenteric fat pad of an anesthetized rat. In rats that received injections of adenovirus-expressing GFP, a small increase in visualized vascular area was seen (Fig. 9,A, i to ii). When receiving injections of Ad- VEGF₁₆₅, however, extensive and florid vasculature was induced in the adjacent mesenteric connective tissue panels (Fig. 9,A, iii to iv) as shown previously (30). In rats that received injections of Ad-VEGF₁₆₅b, the increase in vasculature seen in the fat pad was no greater (Fig. 9,A, v to vi) than that seen with control virus. Injection of both Ad-VEGF₁₆₅b and Ad-VEGF₁₆₅ did not result in an increase in vascular area (Fig. 9,A, vii to viii). The perfused blood vessel area was significantly increased by Ad-VEGF₁₆₅ relative to adenovirus-expressing enhanced GFP but not by Ad-VEGF₁₆₅b or both Ad-VEGF₁₆₅b and Ad-VEGF₁₆₅ (P < 0.0001, ANOVA; Fig. 9,B). Quadruple fluorescent staining (e.g., Fig. 10,A) of the mesenteries with Griffonia simplicifolia isolectin B4 (stains endothelial cells), Ki-67 (proliferating cells), Hoescht (nuclei), and phalloidin (stains actin filaments in sprouts of endothelial cells undergoing migration) enabled quantification of parameters that require angiogenesis to occur. The density of branch points (Fig. 10,B), sprouts (Fig. 10,C), proliferating endothelial cells (Fig. 10,D), and blood vessels (Fig. 10,E) and the mean vessel length (Fig. 10,F) were measured. VEGF₁₆₅ increased all of these measurements, except mean vessel length, which was decreased as expected if sprouting angiogenesis occurred. Injection of Ad-VEGF₁₆₅b, however, did not result in an increase in perfused vessel area, vascular density, sprouting, branching, or proliferation and did not decrease mean vessel length. Moreover, when both viruses were injected, the effect of VEGF₁₆₅ on all these parameters was significantly inhibited (see Fig. 10 B–F).

A375 melanoma cells, stably transfected with VEGF₁₆₅b s.c. injected into each of six nude mice produced significantly smaller tumors (Fig. 11,B) than cells expressing VEGF₁₆₅ (Fig. 11,A). Moreover, injection of a mixture of the two isoform-expressing cells resulted in delayed tumor growth (Fig. 11,C), and at 20 days, the tumors were significantly smaller than in mice that received injections of VEGF₁₆₅-expressing cells. (Fig. 11,D). Moreover, mixing the tumor cell populations resulted in growth at a slower rate, with an increased doubling time, suggesting that growth was inhibited by VEGF₁₆₅b expression, not simply the number of cells developing into a tumor was different (Fig. 11 E).

Since its discovery in 1989 (43), VEGF has been regarded as the most potent and important angiogenic growth factor in both normal physiology and pathophysiological states (2, 3). This has been due to the large numbers of studies investigating mRNA and protein expression in normal and pathological conditions and showing quite clearly that wherever angiogenesis occurs VEGF is expressed (8). Moreover studies of the regulation of VEGF expression have clearly demonstrated that the driving force for expression of these growth factors is the physiologic requirement for increased vascularity–hypoxia (44). Overexpression of VEGF results in increased angiogenesis and inhibition of VEGF results in inhibition of angiogenesis in normal and pathological states. Anti-VEGF agents have been demonstrated to be effective therapies in cancer (45) and other angiogenic conditions (46). Despite this wealth of information, more than 11,000 scientific publications by the middle of 2004, there is still little known concerning the regulation of the different isoforms of VEGF. The evidence that an alternative splice site exists in the 3′-untranslated region of the VEGF mRNA that results in expression of isoforms with an alternate COOH terminus, which may be inhibitory and was down-regulated in renal carcinoma, suggested that regulation of splicing may be a regulatory mechanism of angiogenesis in cancer (25). However, the presence of RNA sequence from a few individuals does not in itself show that a novel protein species is present nor the mechanisms underlying the inhibitory nature of VEGF₁₆₅b. The results described here show that VEGF₁₆₅b significantly inhibits VEGF₁₆₅-mediated activation of its major signaling receptor, VEGFR-2, and hence inhibits downstream signaling of VEGFR-2, such as MAPK activation, and physiological effects, such as angiogenesis. Surprisingly, VEGF₁₆₅-mediated MAPK signaling in microvascular endothelial cells was the only pathway investigated that was not inhibited, suggesting that VEGF₁₆₅b may be able to signal through VEGF-R1 or neuropilin to modulate VEGF₁₆₅ signaling, as has previously been described for placental growth factor (47). The fact that only the six terminal amino acids are different suggests that it is these residues coded for by exon 8 proximal splicing sites that stimulate receptor signaling involved in these responses. Although one study showed that VEGF₁₂₁ (containing exon 8) was equally incapable as VEGF₁₁₀ in stimulating endothelial proliferation (26), subsequent studies have shown that, in contrast, VEGF₁₂₁ is a highly potent angiogenic growth factor on endothelial cells in culture (48) and in vivo (42). It is therefore quite possible that the six amino acids at the end of the VEGF sequence are required for full stimulation of the receptor, although not for binding to the receptor. It is still not clear whether the six amino acids coded for by exon 8 distal splicing are actively inhibitory or are simply not able to activate the receptor.

VEGF has been the most widely implicated protein in stimulating angiogenesis in cancer and other major health concerns of the developed world (3). That VEGF can be differentially spliced to form an inhibitory isoform, which is endogenously expressed, and is antiangiogenic requires a major reevaluation of our current understanding of how angiogenesis is regulated. Because of the nature of this splice variant—distally splicing into the 3′-untranslated region of the VEGF mRNA, most previously investigated expression studies will not have distinguished VEGF_xxxb from other isoforms. This may explain some of the data that does not show clear relationships between VEGF expression and angiogenesis, such as in melanoma and other angiogenic conditions (49). Moreover, recent studies showing that VEGF-neutralizing antibodies are an effective therapy for some types of cancer, retinopathy, and other angiogenic conditions (45, 46, 50, 51) could be made more effective by targeting the proangiogenic splice variants rather than a pan-VEGF strategy. There appears to be a splicing switch in at least two types of cancer—renal and prostate—but the mechanisms that regulate splicing of VEGF are almost completely unknown. Regulation of splicing from anti- to proangiogenic growth factor isoforms may be a component of the control of development of cancers, progression of heart disease, and other pathological states. It has not escaped our notice that other proteins involved in angiogenesis, such as VEGFR-1, fibroblast growth factor, fibronectin, and collagen (52, 53, 54, 55, 56), also have differentially spliced inhibitory isoforms. A common mechanism may therefore exist for the regulation of these antiangiogenic splicing events.