VEGF₁₂₁b, a new member of the VEGF_xxxb family of VEGF-A splice isoforms, inhibits neovascularisation and tumour growth in vivo

Abstract

Background:

The key mediator of new vessel formation in cancer and other diseases is VEGF-A. VEGF-A exists as alternatively spliced isoforms - the pro-angiogenic VEGF_xxx family generated by exon 8 proximal splicing, and a sister family, termed VEGF_xxxb, exemplified by VEGF₁₆₅b, generated by distal splicing of exon 8. However, it is unknown whether this anti-angiogenic property of VEGF₁₆₅b is a general property of the VEGF_xxxb family of isoforms.

Methods:

The mRNA and protein expression of VEGF₁₂₁b was studied in human tissue. The effect of VEGF₁₂₁b was analysed by saturation binding to VEGF receptors, endothelial migration, apoptosis, xenograft tumour growth, pre-retinal neovascularisation and imaging of biodistribution in tumour-bearing mice with radioactive VEGF₁₂₁b.

Results:

The existence of VEGF₁₂₁b was confirmed in normal human tissues. VEGF₁₂₁b binds both VEGF receptors with similar affinity as other VEGF isoforms, but inhibits endothelial cell migration and is cytoprotective to endothelial cells through VEGFR-2 activation. Administration of VEGF₁₂₁b normalised retinal vasculature by reducing both angiogenesis and ischaemia. VEGF₁₂₁b reduced the growth of xenografted human colon tumours in association with reduced microvascular density, and an intravenous bolus of VEGF₁₂₁b is taken up into colon tumour xenografts.

Conclusion:

Here we identify a second member of the family, VEGF₁₂₁b, with similar properties to those of VEGF₁₆₅b, and underline the importance of the six amino acids of exon 8b in the anti-angiogenic activity of the VEGF_xxxb isoforms.

The advent of anti-neovascular therapies in cancer and eye disease has prompted a burgeoning interest in the mechanisms behind the initiation, development and refinement of new vessel formation. The growth of new vessels from those already established is termed angiogenesis and although a complex process, involving over 50 growth factors, cytokines, receptors and enzymes, is dominated, in most situations, by the 38 kDa cysteine-knot protein vascular endothelial growth factor (VEGF). Following the initial identification of VEGF (Senger et al, 1983) many splice isoforms were subsequently identified, eventually forming a family of mature peptides – the VEGF_xxx family (see Figure 1) (Houck et al, 1991). The individual family members have different expression patterns (e.g., VEGF₁₈₃ expression is mostly restricted to the eye (Jingjing et al, 1999), VEGF₁₄₅ in the reproductive system (Charnock-Jones et al, 1993)); unique characteristics with respect to heparin and receptor-binding properties (resulting from inclusion/exclusion of particular exon segments) and are named by amino acid size. Despite this heterogeneity they are all considered pro-angiogenic, pro-permeability vasodilators. The evidence for these properties for individual splice isoform recombinant proteins in vivo, however, varies from conclusive (VEGF₁₆₅) (Leung et al, 1989), through mixed (VEGF₁₂₁) (Keyt et al, 1996; Zhang et al, 2000) and scanty (VEGF₁₄₅) (Poltorak et al, 1997) to absent (VEGF₂₀₆).

Figure 1.

Alternative splicing of VEGF-A generating VEGF_xxx and VEGF_xxxb isoforms. (A) VEGF-A contains eight exons with receptor binding found in exons 3 and 4, heparin and neuropilin binding in exons 6 and 7 and 8a. (B) Alternative splicing of the C-terminal end using proximal splice site generating VEGF_xxx isoforms or distal splice site generating VEGF_xxxb isoforms. This splicing leads to an alterative last six amino acids (CDKPRR or SLTRKD). (C) The C-terminal splicing leads to the possibility of two sister families of VEGF-A isoforms; VEGF_xxx and VEGF_xxxb, differing only in the C-terminal.

In addition, it has now become evident that even the presence of six splice isoforms is an over-simplification. The pro-angiogenic potential of conventional VEGF isoforms is thought to result from the C-terminus (Keyt et al, 1996), more specifically the six amino acids of exon 8a – CDKPRR (Cebe Suarez et al, 2006; Cebe-Suarez et al, 2008). A sister family of splice isoforms has now been identified (see Figure 1) in which the pro-angiogenic domain (exon 8a) is replaced by the six amino acids coded for by exon 8b – SLTRKD. These exon 8b isoforms (VEGF_xxxb) are exemplified by the only isoform so far extensively investigated, VEGF₁₆₅b. VEGF₁₆₅b has radically different properties to that of the widely studied VEGF₁₆₅. We, and others have shown that VEGF₁₆₅b is not pro-angiogenic in vivo (Woolard et al, 2004) and specifically and actively inhibits VEGF₁₆₅-mediated endothelial cell proliferation and migration in vitro (Bates et al, 2002; Rennel et al, 2008a, 2008b), vasodilatation ex vivo (Bates et al, 2002) and in vivo angiogenesis assays such as rabbit corneal eye pocket, chicken chorioallantoic membrane, mesenteric and Matrigel implants (Woolard et al, 2004; Cebe Suarez et al, 2006; Kawamura et al, 2008). Furthermore, VEGF₁₆₅b not only inhibits physiological angiogenesis in developing mammary tissue in transgenic animals over-expressing VEGF₁₆₅b (Qiu et al, 2008) but also pathological angiogenesis in models of the proliferative ocular lesions of diabetes (retinopathy of pre-maturity) (Konopatskaya et al, 2006) and age-related macular degeneration (laser-induced choroidal neovascularisation, Hua et al, unpublished) and in six different tumour models in vivo (Woolard et al, 2004; Varey et al, 2008; Rennel et al, 2008a, 2008b). The above anti-angiogenic properties occur when delivered as VEGF₁₆₅b-transfected cells (Woolard et al, 2004; Varey et al, 2008; Rennel et al, 2008a), VEGF₁₆₅b-adenoviral constructs (Woolard et al, 2004) or recombinant human protein (Konopatskaya et al, 2006; Rennel et al, 2008b).

Although a significant amount of evidence on the expression and function of the VEGF₁₆₅b isoform has already been published, there is very little evidence on other VEGF_xxxb family members, and their existence and activity is still uncertain. This study, therefore, investigates the properties of a second member of the VEGF_xxxb family – the peptide VEGF₁₂₁b, addressing the hypothesis that VEGF₁₂₁b will also show anti-angiogenic properties. We present descriptive expression data and functional data on cell migration, receptor binding and in vivo tumour and ocular angiogenesis in vivo.

Materials and methods

Tissue samples

Paired colon samples were from partial colon resection for carcinoma. Samples were obtained by taking biopsies of the fresh specimen from a non-necrotic central portion of the tumour and from a peripheral part of the macroscopically normal colon (n=15). The mean patient age was 71.5 (58–80) years, with 62% males and Duke's staging as follows: 6.7% A, 46.7% B and 46.7% C. Informed consent for obtaining the tissue was gained in compliance with local ethics committee approval.

PCR on cDNA from human total RNA

Total RNA was extracted from human tissue using TRIZOL (Invitrogen, Carlsbad, CA, USA) (Chomczynski, 1993) and treated with DNase (Promega, Madison, WI, USA). Complementary DNA was made using oligo dT primers or 3′UTR primer (GCTCTAGAAGCAGGTGAGAGTAAGCG) as described earlier (Bates et al, 2002). PCR was used to identify presence of VEGF₁₂₁b, VEGF₁₆₅ and VEGF₁₆₅b using specific primers; VEGF₁₆₅/VEGF₁₆₅b exon 7 GGCAGCTTGAGTTAAACGAACG, 3′UTR ATGGATCCGTATCAGTCTTTCCTGG, VEGF₁₂₁b exon 5/8b GAAAAATCTCTCACCAGGAAA, 3′UTR II CTGGATTAAGGACTGTTCTG, VEGF₁₂₁ exon 5/8a GAAAAATGTGACAAGCCGAG 3′UTR II CTGGATTAAGGACTGTTCTG, GAPDH forward TGATGACATCAAGAAGGTGGTGAAG, reverse TCCTTGGAGGCCATGTGGGCCAT and PCR products were separated and visualised using 4% agarose/ethidium bromide gel. Putative VEGF₁₂₁b PCR products were purified using QIAquick gel extraction (Qiagen, Crawley, UK) and inserted into pGEM vector and confirmed by sequencing of plasmid.

Production of recombinant protein

Recombinant human VEGF₁₂₁b (VEGF₁₂₁b) and VEGF₁₆₅b were produced in Chinese hamster ovary cells as glycosylated dimers by Cancer Research Technologies (London, UK) as described earlier (Rennel et al, 2008b). VEGF₁₆₅ and VEGF₁₂₁ were from R&D Systems, Minneapolis, MN, USA.

In vitro assays with human umbilical vein endothelial cells

Human umbilical vein endothelial cells (HUVEC) were extracted from umbilical cords from caesarean sections (St Michael's Hospital, Bristol, UK). HUVEC migration was performed as described earlier (Rennel et al, 2008b). Briefly, HUVECs were serum starved for 6 h and 100 000 serum-starved cells were plated into collagen-coated 8 μm inserts (Millipore, Billerica, MA, USA) placed in 24-well plates with 500 μl of chemoattractant in M200 medium with 0.1% v/v FCS and 0.2% w/v BSA and incubated overnight at 37°C to allow for migration. Migrating cells were stained in Mayer's haematoxylin. Migrating cells were counted (10 random fields per insert) and expressed as percentage of migrating cells in relation to the total number of seeded cells. Experiments were performed in triplicates. A peptide (Ac-SLTRKD) was generated and increasing concentrations (1 nM–100 μM) peptide was combined with or without 1 nM VEGF₁₆₅ to analyse impact on HUVEC migration.

Cell cytotoxicity was performed using a lactate dehydrogenase assay (Promega); 13 000 cells were seeded in cell culture 96-well plates in triplicates or quadruplicates in 100 μl of full growth media (FGM) for 24 h. Recombinant VEGF protein was added in 100 μl media with 0.1% FCS for 48 h. Conditioned media and lysed cells were spun at 250 g for 15 min at 4°C followed by addition of kit components, incubated for 30 min and read at 490 nm. Receptor and kinase inhibitors were added at indicated concentrations with or without 1.4 nM VEGF₁₂₁b.

For direct counting, HUVECs were seeded in 24-well plates in triplicates in FGM 24 h followed by 48 h exposure to 1.4 nM growth factors or FGM. Cells were trypsinised and counted using a haemocytometer.

In vivo tumour model

LS174t human colon carcinoma cell lines (ECACC, Salisbury, UK) (Yuan et al, 1996; Lee et al, 2000) were grown in Minimum Essential Eagle’s Medium supplemented with 10% v/v FCS, 2 mM L-glutamine and 1% non-essential amino acids (all from Sigma Aldrich, Gillingham, UK). Cells were transfected with 1 μg/well of pcDNA₃ empty vector or pcDNA₃-VEGF₁₂₁b, using lipofectamine plus (Invitrogen) as per manufacturer's instructions and selected using 500 μg ml⁻¹ geneticin (Sigma Aldrich). Conditioned media and cell lysate were analysed by western blotting to confirm expression levels of VEGF isoforms. Two million cells were injected subcutaneously into the lumbar region of nude mice (six per group). Xenotransplanted tumours were measured by calliper every 2–3 days and tumour volume was calculated according to (length × width × [length+width]/2). Mice were culled by cervical dislocation and tumours were removed when the first tumour reached 16 mm in any direction all mice were killed in that group. Injections, measurements and analysis were all carried out with the investigators blinded to group.

Tumour vessel density was counted in 10 random fields in haematoxylin and eosin stained 6 μm sections frozen sections from tumours. Vessel presence was confirmed by staining by blocking in 5% v/v goat Ig for 30 min, 20 μg ml⁻¹ biotinylated isolectin B4 (L-2140 Sigma-Aldrich) or 1 μg ml⁻¹ Flk-1 (sc-6251, Santa Cruz Biotechnology, Santa Cruz, CA, USA) overnight, 2 μg ml⁻¹ anti-mouse biotin antibody (Vector Laboratories, Burlingame, CA, USA) for 1 h followed by avidin biotinylated enzyme complex (Vector Laboratories) for 30 min followed by DAB substrate (Vector Laboratories). Sections were examined using a Nikon Eclipse E400 microscope and photos were captured using Nikon Eclipse Net software. To visualise apoptosis, tumour sections were stained with cleaved caspase 3 antibody (1 : 400 dilution, CatNo9661 Cell Signaling Technology, Danvers, MA, USA) and positive staining areas were circled and compared to total tumour area using Image J software.

Cell proliferation by flow cytometry analysis and direct counting

Trypsinised LS174t cells were fixed in 70% ethanol for 30 min at 4 °C followed by 100 μg ml⁻¹ ribonuclease (Sigma Aldrich) for 5 min. Propidium iodine (PI, 50 μg ml⁻¹) was added to the cells and analysed by flow cytometer.

For healthy, apoptotic and necrotic cell determination a PI-annexin V-FITC kit was used (Bender MedSystems, Vienna, Austria). Cell culture media was centrifuged at 1500 r.p.m. to collect non-adherent cells, added to trypsined cells and diluted in binding buffer to approximately 300 000 cells/ml. Cells were stained with 25 μl annexin V-FITC for 10 min, washed, stained with 20 μg ml⁻¹ PI and analysed by flow cytometer.

For direct counting, 135 000 LS174t cells/24 well was seeded out in triplicates in media supplemented with 0.1% FCS in the presence or absence of 1 nM VEGF₁₂₁b. Cells were trypsined and counted using a haemocytometer after 24 and 48 h.

In vivo imaging of ¹²⁵I-VEGF₁₆₅b biodistribution

Nude mice were injected with LS174t tumours on the right hindleg. ¹²⁵I-VEGF₁₂₁b was generated using Iodogen tubes (Pierce Biotechnology, Cramlington, UK) and purified with NAP-10 columns (GE Healthcare, Chalfont St Giles, UK). Analysis by thin layer chromatography showed >95% purity (¹²⁵I-VEGF₁₂₁b/total ¹²⁵I). Approximalty 3.2 MBq (70 μg protein) was injected into the tail vein when tumours were >10 mm in diameter. Anaesthesia was maintained by 2% halothane during X-ray and scanning (NanoSPECT/CT, Bioscan, Washington, DC, USA), after 40, 70, 120, 240 or 1440 min. For biodistribution studies, 0.10 MBq (3 μg) ¹²⁵I-VEGF₁₂₁b was injected into the tail vein and mice were culled at 120 or 240 min. Organs and tissues of interest were excised and assessed using a gamma counter (LKB Wallac 1282 Compugamma CS, Wallac, Perkin Elmer, MA, USA). Uptake was expressed as % injected dose/g tissue.

In vitro saturation binding of ¹²⁵I-VEGF_xxxb to Fc-VEGFR-1

One μg ml⁻¹ human VEGFR-1-Fc chimaera or VEGFR-2-Fc (R&D Systems) was bound to Immulon II HB Flat 96-well plates (Thermo Labsystems, Basingstoke, UK) overnight followed by blocking with 3% w/v BSA for 2 h and binding of ¹²⁵I-rhVEGF₁₂₁b or ¹²⁵I-rhVEGF₁₆₅b for 4 h in 1% w/v BSA in triplicates. Plates were washed three times with PBS+0.05% v/v Tween in between each step. To detach the complex 10% w/v SDS was added and counted in a gamma counter. Data were calculated as c.p.m. and expressed as % binding compared with maximum binding concentration.

SDS–PAGE and immunoblotting of tissue and cell lysate

Tissue lysates from transfected cells or human colon tissue were extracted using RIPA buffer (50 mM Tris, 150 mM NaCl, 1% v/v NP-40, 0.25% w/v Na-deoxycholate, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 μg ml⁻¹ of each of aprotinin, pepstatin and leupeptin) and homogenised using a polytron and spun at 12 000 r.p.m. for 10 min at 4°C. Lysates were separated on 15% SDS–PAGE, transferred to polyvinylidene fluoride membranes, blocked in 10% w/v dry milk, incubated with a mouse monoclonal anti-human VEGF_xxxb antibody (2 μg ml⁻¹, R&D Systems) for 1 h followed by horse radish peroxidase conjugated anti-mouse antibody (1/5000, Pierce Biotechnology) for 1 h. Membranes were developed using supersensitive West Femto Maximum Sensitive Substrate (Pierce Biotechnology). Densitometry analysis of western blots was performed using Image J. VEGF₁₂₁b is detected by immunoblotting using anti-VEGF antibodies (A-20 Santa Cruz or AF-293 R&D Systems) and the commercial VEGF_xxxb antibody (A56/1, R&D Systems, cat no. MAB3045). Commercial ELISA for VEGF₁₆₅b (R&D Systems) and panVEGF (Duoset R&D Systems) can pick up VEGF₁₂₁b but panVEGF ELISA underestimates the VEGF₁₂₁b levels (and VEGF₁₆₅b) by approximately 40% (Varey et al, 2008).

Oxygen-induced retinopathy mouse model

The oxygen-induced retinopathy (OIR) model was performed as described earlier (Smith et al, 1994; Konopatskaya et al, 2006) with minor modifications. Neonatal C57/Bl6 mice and nursing CD1 dams were exposed to 75% oxygen between P7 and P12. Return to room air induced hypoxia in the ischaemic areas. On P13, mice received either VEGF₁₂₁b (1 or 10 ng) or Hank's buffered solution in 1 μl intraocular injections using a Nanofil syringe fitted with a 35 gauge needle (WPI, Sarasota, FL, USA) into the left eye under isoflurane anaesthesia. On P17, both eyes were dissected, fixed in 4% paraformaldehyde for 4 h at 4°C and retinas were dissected. Retinas were permeabilised in PBS containing 0.5% Triton X-100 and 1% bovine serum albumin, stained with 20 μg ml⁻¹ biotinylated isolectin B4 (Sigma Aldrich) in PBS pH 6.8, 1% Triton-X100, 0.1 mM CaCl₂, 0.1 mM MgCl_2, followed by 20 μg ml⁻¹ ALEXA 488-streptavidin (Molecular Probes, Eugene, OR, USA) and flat mounted in Vectashield (Vector Laboratories). Quantification of neovascular and ischaemic areas were performed in a blinded fashion using Photoshop CS3 along with Image J and expressed as percentage of total retinal area (=normal+ischaemic+neovascular).

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (version 4.0cx). Data are given as means±s.e.m. or means±95 % confidence interval when stated. One-way ANOVA was used to compare migration, cytoprotection, tumour growth and retinal vascularisation. T-test was used to compare tumour growth, apoptosis and microvascular density and paired t-test used for retinal areas.

Results

Expression of VEGF₁₂₁b in human colon tissue and its reduction in colon carcinoma

Earlier, we reported a shift in expression of VEGF_xxxb isoforms in human malignant melanoma (Pritchard-Jones et al, 2007), prostate (Woolard et al, 2004; Rennel et al, 2008a) and colon carcinoma (Varey et al, 2008) leading to less anti-angiogenic VEGF_xxxb splice isoforms in cancer tissue. The ELISA used to quantify the levels of VEGF_xxxb isoforms does not distinguish between the individual members of the VEGF_xxxb isoform family. Earlier RT–PCR used to detect VEGF_xxxb isoforms (using exon 7 and 3′UTR primers) did not detect VEGF₁₂₁b (see Figure 2A) as the forward primer was in exon 7 (spliced out in VEGF₁₂₁b and VEGF₁₂₁). To identify VEGF₁₂₁b, mRNA primers were used that spanned the exon 5/exon 8b boundary. These primers detected VEGF₁₂₁b but not VEGF₁₆₅b or VEGF₁₆₅ (see Figure 2B). VEGF₁₂₁b was detected in normal tissue such as aorta, spleen, placenta and some normal and tumour colon tissue (see Figure 2B), which also expressed VEGF₁₆₅ and VEGF₁₆₅b (see Figure 2A) and retina and vitreous (Figure 2C). Presence of VEGF₁₂₁b was confirmed by sequencing of PCR products.

Figure 2.

VEGF₁₂₁b is expressed in human tissue and the expression is reduced in human colon tumours. (A–C) PCR on cDNA from different human tissues using primers to detect VEGF₁₂₁b, VEGF₁₆₅b and VEGF₁₆₅. VEGF₁₆₅ and VEGF₁₆₅b are detected in colon cDNA both in tumour and adjacent control colon (A). VEGF₁₂₁b is detected in colon tissue, aorta, spleen and placenta using VEGF₁₂₁b-specific primers (B). VEGF₁₂₁, ₁₆₅, ₁₂₁b and ₁₆₅b can be found in the retina and to a lesser extent in the human vitreous (C). (D) Densitometry analysis of VEGF₁₂₁b and VEGF₁₆₅b protein expression in matched human control colon samples and colon tumours analysed by western blot using a VEGF_xxxb-specific antibody (n=14 pairs, P=0.0067 paired t-test).

Measurement of VEGF levels in an earlier study on human colon carcinoma showed that the angiogenic VEGF isoforms were upregulated compared with matched colon tissue whereas the levels of anti-angiogenic VEGF_xxxb isoforms remained unchanged. Western blot on that tissue showed two bands consistent with expression of VEGF₁₂₁b and VEGF₁₆₅b (Varey et al, 2008). Quantification of the western blot analysis of paired human colon tissues from tumour and adjacent normal colon tissue indicates a downregulation of VEGF₁₂₁b in tumour tissue compared to VEGF₁₆₅b (see Figure 2D, P<0.01 paired t-test, n=14 pairs).

VEGF₁₂₁b is secreted, inhibits migration and reduces apoptosis in endothelial cells

To determine whether VEGF₁₂₁b can be secreted by colon cells, VEGF₁₂₁b recombinant cDNA was transfected into colon carcinoma cells. Western blotting of the lysate and the media showed that VEGF₁₂₁b was found pre-dominantly in the media, indicating that it is freely secreted (see Figure 3A), in contrast to cells that were not transfected with the VEGF₁₂₁b plasmid, in which no VEGF₁₂₁b was seen in the media or the lysate. VEGF₁₂₁b binding to VEGFR-1 was analysed by incubating increasing concentrations of radiolabelled VEGF₁₂₁b with immobilised VEGFR-1 (see Figure 3B). The saturation-binding kinetics of VEGF₁₂₁b (see Figure 3B, log₁₀EC₅₀=−9.9±0.04) was similar to that for VEGF₁₆₅b measured at the same time (EC₅₀=−9.65±0.04, Figure 3C).

Figure 3.

VEGF₁₂₁b inhibits migration of endothelial cells and inhibits serum starved induced cell death in endothelial cells through classical VEGFR-induced pathways. (A) VEGF₁₂₁b is found in the media on transfected cancer cells. (B–C) VEGF₁₂₁b has similar saturation binding kinetics to VEGFR-1 as VEGF₁₆₅b. (D–E) VEGF₁₂₁b inhibits VEGF₁₆₅ (D) or VEGF₁₂₁ (E) induced migration of HUVEC. Cells were allowed to migrate through 8 μm pore inserts with increasing concentrations of VEGF₁₂₁b (0–6 nM) or VEGF₁₆₅b (0–2.5 nM) in the presence of 1 nM VEGF₁₆₅ or 2.9 nM VEGF₁₂₁. Data plotted as percentage migration compared to total number of seeded cells (F) VEGF₁₂₁b is cytoprotective to a similar degree as VEGF₁₆₅ in HUVEC. Release of lactate dehydrogenase as a measurement of cell viability was monitored after 48 h of serum starvation in the presence of increasing amounts of VEGF₁₂₁b (0–3.6 nM), 1 nM VEGF₁₆₅ or full growth media (FGM) with serum and growth factors. (G) Similar cytoprotection of VEGF₁₂₁b was observed in serum starved HUVEC as cell number was not reduced by the same amount after 48 h of cell starvation. (H) VEGF₁₂₁b cytoprotection is mediated via VEGFR-1 and -2, P13-K, MEK1/2. Inhibition of VEGFR1/2=200 nM PTK787, VEGFR-2=10 nM ZM323881, P13- K=15 μM LY294002, p38=10 μM SB203580 and MEK1/2=15 μM PD98059. (One-way ANOVA, Dunnett's post hoc test, 0 nM or control vs addition, ^*P<0.05, ^**P<0.01, ^***P<0.001).

We have shown earlier that VEGF₁₆₅b conditioned media (Bates et al, 2002; Rennel et al, 2008a) or recombinant human VEGF₁₆₅b (Woolard et al, 2004; Rennel et al, 2008b) inhibits VEGF₁₆₅-induced migration of endothelial cells. Migration of HUVEC was inhibited by the addition of VEGF₁₂₁b either when stimulated with VEGF₁₆₅ (see Figure 3D, VEGF₁₆₅ vs any treatment P<0.05) or VEGF₁₂₁ (see Figure 3E, VEGF₁₂₁ vs any treatment P<0.05). The level of inhibition was similar to that observed with VEGF₁₆₅b (see Figure 3D) and no additional significant reduction in migration was observed with the combination of VEGF₁₂₁b and VEGF₁₆₅b (see Figure 3D). Addition of up to 100 μM peptide of the last six amino acids had no effect on VEGF₁₆₅-induced HUVEC migration (data not shown) indicating that the 3D structure and or the full length of the VEGF_xxxb protein is needed for inhibition.

VEGF₁₆₅ is known to protect endothelial cells from apoptosis and incubation of HUVECs with recombinant VEGF₁₂₁b was equally cytoprotective (see Figure 3F, no addition vs any treatment P<0.01). This resulted in a small increase in cell number induced by treatment with VEGF₁₂₁b (see Figure 3G). To determine the mechanism of this cytoprotective effect of VEGF₁₂₁b, HUVECs were treated with inhibitors of receptor tyrosine kinases and signalling kinases (see Figure 3H). Inhibition of VEGFR-1 and VEGFR-2 by 100 nM PTK787 resulted in an inhibition of the cytoprotective effect of VEGF₁₂₁b, as did incubation with the VEGFR-2-specific antagonist ZM323881 (10 nM). The cytoprotective effect was blocked by inhibitors of PI3Kinase (LY294002) and MEK1/2 (PD98059), but not by inhibitors of p38 MAPK (SB203580, see Figure 3H).

Over-expression of VEGF₁₂₁b reduces tumour growth through a direct effect on endothelial cells

We have shown earlier that over-expression of VEGF₁₆₅b reduces tumour growth in five different tumour types (Woolard et al, 2004; Varey et al, 2008; Rennel et al, 2008a), but VEGF₁₆₅b does not directly affect proliferation or apoptosis of LS174t cells (Varey et al, 2008). To verify whether VEGF₁₂₁b could inhibit tumour growth as well, LS174t colon carcinoma tumour cells were transfected with plasmid containing VEGF₁₂₁b or cloning vector alone (control). Expression levels were confirmed by immunoblot on lysate and conditioned media showing that VEGF₁₂₁b was secreted and found in the media (see Figure 3A). Cells were injected into the back of nude mice and tumour growth was monitored over time. VEGF₁₂₁b reduced tumour growth compared with empty vector (290±66 mm³ vs 925±268 mm³ after 14 days, P<0.05, unpaired t-test Welch correction, see Figure 4A). VEGF₁₂₁b tumours were smaller and excised tumours were less haemorrhagic than those with empty vector (see Figure 4A inserted images). Immunohistochemical staining was performed to visualise blood vessel distribution (see Figure 4B for representative VEGFR-2 staining). Analysis of the vessel density in 10 random fields in each tumour showed a decreased number of vessels in VEGF₁₂₁b tumours (control vs VEGF₁₂₁b, 3.5±0.6 vs 1.7±0.1, P<0.05 unpaired t-test n=6 tumours per treatment 10 fields analysed per tumour, Figure 4B). VEGFR-2 staining of tumour sections showed expression of VEGFR-2 in the tumour vessels and not in the general tumour mass (see Figure 4B inserted images) indicating an effect of VEGF₁₂₁b on the endothelium rather than an anti-proliferative effect on the tumour cells. Earlier analysis has shown absence of VEGFR-1 and VEGFR-2 expression in LS174t cells (Rennel et al, 2008b). To determine whether VEGF₁₂₁b expression induced apoptosis in the tumour cells, tumours were stained for cleaved caspase 3. There was no change in caspase staining in control vs VEGF₁₂₁b-expressing tumours (see Figure 4C).

Figure 4.

VEGF₁₂₁b reduces tumour growth in nude mice bearing colon carcinoma tumours by reducing tumour vessel ingrowth. (A) LS174t human colon carcinoma cells were transfected with pcDNA₃-VEGF₁₂₁b or empty pcDNA3 plasmid and injected subcutaneously into nude mice and tumour growth was monitored over time. Over-expression of VEGF₁₂₁b resulted in a reduced tumour growth and contained less blood compared with control cells (inserted images). Scale bar=10 mm. (B) Immunohistochemistry staining of tumour sections for VEGFR-2 visualise microvessels (inserted images). Quantification of microvascular density showed significantly fewer blood vessels per unit area than control tumours. Each point represents the mean of 10 random analysed fields and six tumours per treatment were examined (^*P<0.05 unpaired t-test). (C) Cleaved caspase 3 staining of tumour sections (inserted images) showed no significant difference in apoptosis. (D, E) VEGF₁₂₁b has no effect on LS174t colon carcinoma growth in vitro. Over-expression of VEGF₁₂₁b in LS174t colon carcinoma cells had no effect on cell proliferation analysed by FACS (D). Addition of 1 nM VEGF₁₂₁b had no effect of cell number analysed by direct counting of cells over 48 h (E).

Transfected LS174t colon cells were analysed by flow cytometry and VEGF₁₂₁b had no effect on proliferation (see Figure 4D control vs VEGF₁₂₁b, S/G₂-M 21±2.3 vs 27±4.4, P=0.36 unpaired t-test). In addition, addition of recombinant VEGF₁₂₁b had no effect on tumour cell number (see Figure 4F). These data indicate that VEGF₁₂₁b reduced tumour growth by reducing the tumour-driven angiogenesis, rather than a direct effect on the tumour cells themselves.

Biodistribution of intravenous VEGF₁₂₁b

The results above indicate that VEGF₁₂₁b could be a potential therapeutic agent in angiogenic diseases, if the uptake into neovascular tissues was sufficient. To determine how intravenously (i.v.) injected recombinant VEGF₁₂₁b is distributed in vivo, ¹²⁵I-VEGF₁₂₁b was injected i.v. into tumour-bearing mice and imaged using high-resolution single photon emission computed tomography (NanoSPECT/CT). Radiolabelled VEGF₁₂₁b was distributed quickly through the mouse and images acquired at 40 min post injection (p.i.) showed accumulation in organs of metabolism and secretion. Thereafter, the overall signal gradually declined because of excretion/degradation of the radiolabelled protein. Figure 5A shows whole body images at different time p.i. Uptake of radioactivity can be seen in the tumour as well as abdominal tissues such as stomach and kidney. Coronal and sagittal sections also visualised uptake in the tumour at 2 h p.i. (see Figure 5E-F). Biodistribution of tissue and tumour indicated that approximately 2.5% of the total ¹²⁵I is found in the tumour at 2 h and stays at a similar level for up to 24 h (see Figure 5G). Intestine and liver showed high uptake with a maximum at 4 h. No obvious adverse haemodynamic effects were seen in the animals.

Figure 5.

In vivo distribution of ¹²⁵I-VEGF₁₂₁b in tumour-bearing mice. Tumour-bearing mice received an intravenous injection of ¹²⁵I-VEGF₁₂₁b and 3D imaged using NanoSPECT/CT. (A–D) Time course for biodistribution of ¹²⁵I-VEGF₁₂₁b after tail vein injection using transverse sections. (E) Coronal. (F) Para-sagittal through the centre of the tumour. The tumour is circled and arrows indicate different organs. (G) Quantification uptake into different organs and tissues over time. Data expressed as % in tissue relative to the total injected dose, per gram of tissue.

VEGF₁₂₁b rescues retinal vasculature by reducing neovascularisation and ischaemia

We have shown earlier that the anti-angiogenic VEGF₁₆₅b can inhibit retinal neovascularisation in an OIR mouse model when administered as a single intraocular injection (Konopatskaya et al, 2006). In this model retinal vaso-obliteration occurs in the central retina when pups are exposed to high oxygen (from P7 to P12) and when returned to room air the relative hypoxia induces neovascularisation with a peak reached at P17. Retinas were flat mounted and stained with isolectin to visualise vessels with neovascularisation occurring at the edge of the central ischaemic area (see Figure 6A). Blind quantification of the injected and uninjected contralateral retina, showed reduced neovascularisation after 1 or 10 ng VEGF₁₂₁b (see Figure 6B-C, uninjected control vs 1 or 10 ng VEGF₁₂₁b P<0.05 or control injection vs 1 or 10 ng VEGF₁₂₁b P<0.05). Injection of 10 ng VEGF₁₂₁b reduced the ischaemic area, and increased the area of normal vasculature (see Figure 6D-E, control injection vs 10 ng VEGF₁₂₁b P<0.05 for ischaemic and normal vessel growth). The overall result was an increase in the proportion of normal vasculature in these retinae (see Figure 6E). Injection of control solution had no effect compared with contralateral eye (see Figure 6B uninjected control vs control injected eye, P>0.05, paired t-test).

Figure 6.

VEGF₁₂₁b reduces hypoxia-induced retinal neovascularisation after single intraocular injection. The oxygen-induced retinopathy mouse model leads to neovascularisation induced by hypoxia. (A) Flat mounted mouse retinas with vessels visualised by isolectin staining. Ischaemic and neovascularised areas are circled (i, ischaemic, n, neovascular and r, normal retinal vessels). Normal retinal vessels (vii) are slender compared to neovascular vessels (viii). (B) Quantification of neovascularisation, ischaemic and normal vessel growth compared to matched uninjected control eye (paired t-test, ^*P<0.05, ^**P<0.01). (C–E). Distribution of ischaemic, neovascular and normal vessel growth in injected retinas relative to uninjected eye showed that VEGF₁₂₁b reduced neovascularisation (C) and the ischaemic areas (D) leading to an increase in the revascularisation (E). (One-way ANOVA, Bonferroni post hoc test, ΔP<0.05, ΔΔP<0.01 compared to vehicle injected=0 ng).

Discussion

The existence of a second family of VEGF isoforms with overtly different properties from that of the conventional VEGF species required a radical re-evaluation of VEGF biology, not only academically – thousands of manuscripts have attempted to quantify ‘VEGF’ in body fluids or tissues using primers, probes or antibodies that would not distinguish these contrasting families; but also in terms of the pathogenesis of disease and the potential clinical importance of VEGF_xxx/VEGF_xxxb balance, because these families derive from the same gene. VEGF₁₆₅b is widely expressed (Bates et al, 2002, 2006; Cui et al, 2004; Perrin et al, 2005; Pritchard-Jones et al, 2007; Varey et al, 2008) and may form the pre-dominant isoform in many tissues (Perrin et al, 2005; Varey et al, 2008). The physiological role of these isoforms is only just now being elucidated in normal tissues, and it seems that they act as a brake on excess angiogenesis during conditions of controlled growth or VEGF_xxx expression. Examples include development of the ovary, where the VEGF_xxxb levels are maintained to prevent angiogenesis into the developing germ bud and so prevent formation of a testicular artery (VEGF_xxxb antibodies are angiogenic in this model, Artac et al, in press), in the virgin mammary fat pad, where VEGF_xxx levels are prevented from inducing angiogenesis until lacteal formation, when VEGF_xxxb levels fall (and prevention of this fall results in reduced milk production) (Qiu et al, 2008), and in the eye where a relative drop in VEGF_xxxb levels is accompanied with angiogenesis in the retina in diabetes (Perrin et al, 2005). Increased VEGF_xxxb levels also seem to be required for normal pregnancy, as a failure to upregulate VEGF_xxxb in first trimester pregnancies is associated with subsequent maternal–fetal pathology (Bills et al, 2009).

The development of an angiogenic milieu within a stable and non-angiogenic tissue, may depend to some degree on changes in gene (over) transcription or suppression but also on changes in mRNA splicing control. The control of splicing between the two VEGF isoform families has begun to be elucidated, and it has been shown that multiple stimulatory signals can differentially affect the choice of splice site in the terminal exon, including hypoxia (Varey et al, 2008), IGF, TNF and PDGF (Nowak et al, 2008), which all favour proximal splice site selection in exon 8, and hence angiogenic isoforms, and the latter three growth factors suppress distal splicing, whereas hypoxia seems not to affect it. TGF-β, on the other hand, seems to favour distally spliced isoforms (Nowak et al, 2008). This seems to be through mechanisms involving splice factors such as ASF/SF2 and SRp55. Additionally, alteration of splicing has been seen in human and animal models of disease including in Denys Drash Syndrome (Schumacher et al, 2007), kidney, prostate and colon cancer (Bates et al, 2002; Woolard et al, 2004; Varey et al, 2008; Rennel et al, 2008a), malignant melanoma (Pritchard-Jones et al, 2007) where the PSS is favoured and animal models of glaucoma where the DSS is favoured (Ergorul et al, 2008). However, there is very little known about regulation of splice site choice in the terminal exon from the exon 5 splice donor site, but the earlier studies on hypoxia indicate that this is unlikely to be a regulatory factor for VEGF₁₂₁b.

The most studied VEGF_xxxb family member – VEGF₁₆₅b – has been clearly shown not to be angiogenic. Other researchers have shown that the six amino acids of exon 8b are required for the anti-angiogenic activity of VEGF₁₆₅b because artificially generated VEGF₁₅₉ lacking both exons 8a and b, which is not angiogenic (Cebe Suarez et al, 2006). Exon 8a seems to be required for heparin and neuropilin-1 binding and full activation of the VEGFR-1 and -2, as it is likely to interfere with the correct folding and 3D structure of the protein, resulting in differential activation of VEGFR-2 tyrosine phosphorylation, and hence downstream signalling (Cebe Suarez et al, 2006; Cebe-Suarez et al, 2008; Kawamura et al, 2008). Here, we present for the first time new data suggesting that the presence of exon 8b in a different VEGF_xxxb family member endows it with similar properties to VEGF₁₆₅b: VEGF₁₂₁b inhibits experimental endothelial cell migration in vitro and new vessel formation in vivo in tumour and non-tumour-related angiogenesis. The VEGF_xxxb family members seem to share similar properties.

We have shown that VEGF₁₂₁b has similar bio-distribution after i.v. injection to VEGF₁₆₅b and a similar proportion of injected bolus reaching a single implanted tumour – with around 2% for VEGF₁₂₁b and 6% for VEGF₁₆₅b (Rennel et al, 2008b). Furthermore, we have shown in this study and in our earlier work (Woolard et al, 2004) that these two VEGF_xxxb family members bind the conventional tyrosine kinase VEGF receptors. VEGF₁₆₅b does not bind neuropilin-1 (Kawamura et al, 2008), and there is evidence that VEGF₁₂₁ also binds neuropilin poorly, suggesting that VEGF₁₂₁b would not bind it either. von Wronski et al (2007) have highlighted the likelihood that neuropilin-1 binds the residues coded for by exon 8a and replacement by 8b results in VEGF isoforms that do not seem to show neuropilin-1 binding (Cebe Suarez et al, 2006).

Although we have shown that VEGF₁₆₅b and VEGF₁₂₁b bind to the conventional VEGF receptors, further work will be required to investigate the mechanistic differences between VEGF₁₂₁b and its corresponding isoform VEGF₁₂₁. Studies into the mechanistic contrast between VEGF₁₆₅b and VEGF₁₆₅ are more advanced with increasing evidence now emerging that VEGF₁₆₅b, although binding to VEGFR-2 with equal affinity as VEGF₁₆₅, is not simply a classical competitive inhibitor. VEGF₁₆₅b does stimulate the phosphorylation of VEGFR-2 but qualitatively in a unique way. Although VEGF₁₆₅b does lead to the phosphorylation of tyrosine residues of the VEGF receptor, tryptic phosphopeptide mapping and the use of phosphosite-specific antibodies have shown that VEGF₁₆₅b is considerably poorer than VEGF₁₆₅ in inducing phosphorylation of the angiogenic positive regulatory site Y1052 (Y1054 in human sequence) in VEGFR-2 (Kawamura et al, 2008). This study also confirmed that the ability of different VEGF isoforms to induce angiogenesis correlated with their abilities to bind the VEGF co-receptor neuropilin-1 (Kawamura et al, 2008). Whether or not VEGF₁₂₁b activates VEGFR-2 in a similar way is yet to be investigated, as are the potential effects of simultaneous VEGF₁₂₁b and VEGF₁₆₅b administration of receptor activity, angiogenesis or tumour growth.

Another significant finding here is that VEGF₁₂₁b seems to act as a survival factor for human endothelial cells. This activity seems to be VEGFR dependent, and uses classical survival pathways such as the PI3 kinase and MEK/ERK pathways. This provides a potential explanation for the surprising finding that the ischaemic region in the retinal neovascular preparations was reduced by VEGF₁₂₁b administration. This is of particular interest as it suggests a vascular normalisation in the retina, leading to the possibility that VEGF₁₂₁b may be of use in ischaemic retinal diseases associated with angiogenesis such as diabetic retinopathy.

In summary, we provide new information about the existence, expression and activity of a second member of the anti-angiogenic VEGF_xxxb family. Our data confirm that VEGF₁₂₁b has similar inhibitory properties on the process of in vivo angiogenesis and the components of angiogenesis in vitro as does VEGF₁₆₅b, supporting the notion that the replacement of the terminal six amino acids with those coded for by exon 8b is of crucial importance in converting the dominant pro-angiogenic growth factor into a positively anti-angiogenic molecule that may have broad therapeutic potential. Further studies will be required to elucidate its mechanism of action of VEGF₁₂₁b and its potential clinical value along with its other family member VEGF₁₆₅b.