We previously demonstrated elevated expression of the FP receptor, FGF2 and FGFR1 in endometrial adenocarcinoma 12. Using a neoplastic epithelial cell line stably expressing the FP receptor to the levels observed in endometrial adenocarcinoma (Ishikawa FPS cells), we ascertained a role for FGF2, produced by PGF_2α-FP receptor signalling, on epithelial cell proliferation 12. In addition, we found that FP receptor, FGF2 and FGFR1 co-localised within the vascular endothelial cells in endometrial adenocarcinomas suggesting that PGF_2α may directly and indirectly regulate endothelial cell function 12. To determine if the effects of PGF_2α-FP receptor interaction in endometrial adenocarcinoma cells on endothelial cell function were mediated by FGF2, we used conditioned medium (CM) from Ishikawa FPS cells treated with vehicle or 100 nM PGF_2α for 24 hours. The presence of FGF2 in CM from vehicle (V CM) and PGF_2α-treated (P CM) Ishikawa FPS cells was confirmed by ELISA. Immunoneutralisation of P CM with FGF2 antibody significantly reduced FGF2 concentration in P CM 12.

To assess the effects of the CM on differentiation (network formation) and proliferation, assays were performed using HUVECs as a model system. Treatment of HUVECs with P CM significantly increased endothelial cell network formation (Fig. 1A and 1B, P < 0.05) and proliferation (Fig. 1C, P < 0.05) compared to V CM-treated cells. Treatment of HUVECs with P CM in the presence of the FGF2 receptor 1 (FGFR1) tyrosine kinase inhibitor (SU4984) or FGF2-immunoneutralised CM (FGF2-Ab), significantly reduced endothelial cell network formation (Fig. 1A and 1B; P < 0.05) and cellular proliferation (Fig. 1C; P < 0.05), confirming that these alterations in endothelial cell function were mediated by FGF2 in the P CM signalling through endothelial FGFR1.

Next we investigated the signal transduction pathways mediating the role of FGF2 in the P CM on network formation and proliferation. HUVECs were treated with P CM in the presence of cell signalling inhibitors of extracellular signal-regulated kinase (ERK1/2; PD98059), mammalian target of rapamycin (mTOR; rapamycin) or phosphoinositide-3-kinase (PI3K; wortmannin or LY294002). We found that the P CM-induced network formation was significantly inhibited by PD98059 but not rapamycin, wortmannin or LY294002 (Fig. 2A, P < 0.05). However, endothelial cell proliferation was inhibited by PD98059 and rapamycin but not wortmannin or LY294002 (Fig. 2B, P < 0.05).

We confirmed that endothelial cell proliferation but not network formation was mediated by FGF2-mTOR signalling using recombinant FGF2 protein. Treatment of HUVECs with recombinant FGF2 significantly increased network formation (Fig. 2C; P < 0.05) and proliferation (Fig. 2D; P < 0.05). Co-treatment of cells with recombinant FGF2 protein and rapamycin had no effect on network formation, compared to recombinant FGF2 peptide alone (Fig. 2C; P < 0.05). In contrast, rapamycin treatment significantly inhibited endothelial cell proliferation induced by the recombinant FGF2 protein (Fig. 2D; P < 0.05) confirming that endothelial cell proliferation was mediated by P CM via the FGF-FGFR1-mediated induction of the mTOR pathway.

As ERK1/2 was involved in regulating both P CM induced endothelial cell network formation and proliferation, we investigated the effect of P CM on ERK1/2 phosphorylation. HUVECs were treated with V CM or P CM for 0, 5, 10, 15, 20 and 30 mins (Fig. 3A). Treatment of HUVECs with P CM significantly increased ERK1/2 phosphorylation in a time-dependent manner which was maximal after 10 mins of stimulation, compared to V CM (Fig. 3A; P < 0.05). Co-incubation of HUVECs with P CM in the presence of FGFR1 tyrosine kinase inhibitor (SU4984), c-Src inhibitor (PP2) or ERK1/2 inhibitor (PD98059) significantly reduced the P CM-stimulated phosphorylation of ERK1/2 to basal levels (Fig 3B, P < 0.05). However treatment of HUVECs with P CM in the presence of the PI3K inhibitor LY294002 did not significantly reduce the P CM phosphorylation of ERK1/2 (Fig. 3B). Similarly, co-incubation of HUVECs with P CM and the mTOR inhibitor, rapamycin, had no effect on ERK1/2 phosphorylation (data not shown). Treatment of HUVECs with recombinant FGF2 protein phosphorylated ERK1/2 to the levels observed with P CM (Fig. 3B).

FGF2 has been shown to mediate angiogenesis via COX-2 in an in vivo model using rat sponge implants 17, hence we investigated the effect of P CM on the expression of COX-1 and COX-2 in endothelial cells. HUVECs were treated with V CM or P CM for 1, 2, 3, 4, 6, 16 and 24 hrs. We did not observe an alteration in the expression of COX-1 (Fig. 4A) at any of the time points investigated in response to P CM stimulation. However, we observed a significant increase in endothelial COX-2 expression at 3 hours following treatment with P CM (Fig. 4B; P < 0.05). This increase in COX-2 expression was inhibited by treatment with P CM in the presence of the FGFR1 inhibitor, SU4984 or ERK1/2 inhibitor, PD98059 (Fig. 4C; P < 0.05) indicating that COX-2 expression was regulated via the FGF-2-FGFR1-ERK1/2 signalling pathway.

Cyclooxygenase enzymes are responsible for the catalysis of arachidonic acid to prostaglandins (PG). We next investigated the secretion of endothelial PGE₂ and PGF_2α in response to CM treatment. HUVECs were treated with V CM or P CM for 0, 3, 6, and 16 hrs and the secretion of PGE₂ (Fig. 4D) and PGF_2α (Fig. 4E) was measured by ELISA. There was no significant difference in the levels of endothelial PGE₂ secreted by HUVECs in response to CM at any time point tested (Fig. 4D). In contrast, the amount of endothelial PGF_2α was elevated in HUVECs at 6 hrs following P CM treatment compared to both V CM treatment (6 hrs) and P CM at time 0 hrs (Fig. 4E; P < 0.05). To confirm the involvement of COX in the secretion of PGF_2α, we treated HUVECs for 6 hrs with P CM in the presence of the specific COX-2 inhibitor NS398 or the general COX inhibitor indomethacin. We found that the P CM-induced secretion of PGF_2α was significantly reduced by co-treatment of cells with NS398 (Fig. 4F; P < 0.05). Similarly, co-treatment of cells with indomethacin significantly reduced P CM-induced PGF_2α secretion below the levels observed with P CM alone and P CM with NS398 (Fig. 4F; P < 0.05).

Since the secretion of PGF_2α was elevated in HUVECs following P CM treatment, we investigated the role of the F-prostaglandin receptor (FP) in endothelial cell function. We found a significant elevation in endothelial FP receptor expression after 3 hrs of treatment with P CM (Fig. 5A; P < 0.05). To determine if the regulation of FP receptor was mediated by FGF2 in the P CM, we co-incubated HUVECs with P CM in the absence/presence of inhibitors of FGFR1 tyrosine kinase activity (SU4984) or ERK1/2 (PD98059). Incubation of HUVECs with P CM and SU4984 or PD98059 significantly decreased the levels of FP receptor (Fig. 5B; P < 0.05) indicating that FP receptor expression was regulated by FGF2-FGFR1 interaction via the ERK1/2 pathway.

Next, as PGF_2α secretion and FP receptor expression was increased after P CM treatment, we explored the direct effect of PGF_2α, acting via the FP receptor, on endothelial cell network formation and proliferation. We found that addition of 1 μM exogenous PGF_2α significantly increased endothelial cell network formation (Fig. 5C; P < 0.05). In contrast, incubation of HUVECs with 1 μM PGF_2α had no effect on endothelial cell proliferation (Fig. 5D). To investigate the autocrine/paracrine effect of PGF_2α-FP receptor interaction in P CM-induced endothelial network formation, we used a short hairpin RNA (shRNA) adenoviral construct targeted against the FP receptor (sh478) to knockdown FP receptor expression in HUVECs. HUVECs were infected with scrambled adenovirus (scr) or sh478 for 24 hrs. Efficiency of the FP shRNA in ablating receptor expression was confirmed in the laboratory by quantitative RT-PCR and Western blot analysis (data not shown). Infection of HUVECs with sh478, significantly reduced P CM-induced network formation compared to scrambled control virus (Fig. 5E, P < 0.05). In contrast, infection of HUVECs with FP receptor shRNA adenovirus did not alter P CM-induced endothelial proliferation compared to HUVECs infected with the control scrambled adenovirus (Fig. 5F). Similar data were obtained using a second shRNA targeted to a different region of the FP receptor (sh306; data not shown). To confirm the role of COX-2 and FP receptor in endothelial cell network formation, HUVECs were treated with P CM in the absence or presence of the COX-2 inhibitor (NS398) or FP receptor antagonist (AL8810) (Fig 5G). Addition of NS398 or AL8810 significantly inhibited P CM-induced endothelial cell network formation (Fig 5G, P < 0.05).

The FGF2 antibody recognising the 18 kDa isoform of FGF2 (sc1360) was purchased from Santa Cruz Biotechnology (Autogen-Bioclear, Wiltshire, UK). Arachidonic acid (AA), PGF_2α and indomethacin were purchased from Sigma Chemical Co. (Dorset, UK). PD98059, SU4984, LY294002, PP2, NS398, wortmannin and rapamycin were purchased from Calbiochem (Nottingham, UK). Recombinant FGF2 peptide was purchased from PeproTechEC Ltd. (London, UK). Anti-phospho-p42/44 ERK and anti-p42/p44 ERK were purchased from Cell Signalling Technologies/New England Biolabs (Hertfordshire, UK).

Ishikawa cells stably expressing FP receptor (Ishikawa FPS cells) were cultured in Dulbecco's Modified Eagle's Medium (DMEM, Invitrogen, Paisley, UK) with 10% foetal bovine serum (FBS) and 1% penicillin/streptomycin as described previously 13. Human umbilical vein endothelial cells (HUVECs) (Lonza, Walkersville, USA) were cultured in Endothelial Basal Medium (EBM-2) with 2% FBS and growth supplements (VEGF, FGF, PGDF, IGF, EGF, ascorbic acid, heparin and gentamycin) subsequently referred to as Endothelial Growth Medium (EGM-2) (Lonza, Walkersville, USA). Under experimental conditions HUVECs were incubated with EBM-2 containing 1% fetal bovine serum (FBS) with the addition of ascorbic acid and gentamycin (EBM1%) (Lonza, USA). Cell viability in the presence of chemical compounds used to inhibit specific signal transduction pathways was assessed using the CellTitre96AQueous One Solution Proliferation Reagent (Promega, Southampton, UK).

Conditioned medium (CM) was prepared as described previously 12. Briefly, FPS cells were seeded at a density of 2 × 10⁶ cells and allowed to adhere before serum-starvation for 24 hrs. Thereafter, cells were treated with 20 mls of DMEM containing 8.4 μM indomethacin in the presence of 100 nM PGF_2α or vehicle for 24 hrs to create PGF_2α conditioned medium (P CM) or vehicle conditioned medium (V CM). Conditioned medium from three independent experiments was pooled, aliquoted and stored at -20°C until required. FGF2 was immunoneutralised from the PGF_2α conditioned medium by overnight incubation with 0.5 μg/ml FGF2 antibody. The immune complex was removed by 4 hr incubation with 20 μl of a 50% protein A plus G slurry (Calbiochem). Conditioned medium immunoneutralised with Goat IgG was used as a control. Immunoneutralised CM was aliquoted and stored at -20°C until use. The FGF2 content in the CM before and after immunoneutralisation was confirmed by ELISA as described previously 12.

Network assays were carried out using 12-well Transwell plates (Corning Costar, Cambridge, UK). The upper chambers were coated with 80 μl of growth factor (GF)-reduced Matrigel (BD Biosciences, MA, USA) in the absence/presence of SU4984 (20 μM), PD98059 (50 μM), rapamycin (100 ng/ml), wortmannin (200 nM), LY294002 (50 μM), NS398 (10 μM) or AL8810 (50 μM) and incubated at 37°C for 30 mins to allow thin gel formation. HUVECs were plated onto the gel (2.5 × 10⁴ cells/well) in EBM 1%. In the lower chamber V CM or P CM was added. Transwell plates were incubated at 37°C in a 5% CO₂ atmosphere for 16 hrs. Subsequently, cell networks were fixed with 100% ice cold methanol and stained with haematoxylin. To assess network formation, each well was divided into 5 sections. Hotspots of each section were photographed, 5 photos per well at ×10 magnification, using an inverted microscope and camera (Axiovert 200, Carl Zeiss, Germany). The number of network branches was counted blind. Experiments were repeated at least four times in duplicate. Fold difference was determined by dividing the value obtained from P CM treated cells by the value obtained from V CM treated cells. Data were transformed to percentage increase in network formation with V CM = 100% and are presented as mean ± SEM.

HUVECs were seeded in 96-well plates at 3000 cells/well. Following attachment, cell medium was replaced with EBM1% for 3 hours. Cells were then treated with CM, diluted 1:1 (v/v) with EBM1%, in the absence/presence of SU4984, PD98059, rapamycin, wortmannin, LY294002 or FGF2-immunoneutralised CM. Treatments were replaced three times during the 96 hr incubation. Proliferation was determined using the CellTitre96AQueous One Solution Proliferation Reagent (Promega) as per the manufacturer's instructions. The experiments were repeated three times in quadruplicate. Fold difference was determined by dividing the absorbance obtained by P CM treated cells by the absorbance obtained by V CM treated cells. Data were transformed to percentage increase in proliferation with V CM = 100% and are presented as mean ± SEM.

HUVECs were seeded at 2 × 10⁵ cells per 60 mm diameter dish and left to adhere for 24 hrs before GF-starvation overnight. The next day, HUVECs were treated with P CM or V CM for the time indicated in the figure legend, rinsed with ice-cold phosphate-buffered saline and lysed for 20 mins with protein lysis buffer containing inhibitor cocktail mix as described previously 43. Protein concentration was determined with a protein assay (BioRad, Hercules, CA) and approximately 16 μg of protein was resolved and immunoblotted as previously described 44. Immunoblots were blocked in Odyssey Blocking buffer™ (LI-COR Biosciences, Cambridge, UK) before overnight incubation with primary phospho-p42/44 and p42/44 antibodies (diluted 1:1000 in Odyssey blocking buffer) at 4°C. The following day, blots were washed and incubated with the goat anti-mouse IRDye™ 800 (1:10,000) (Rockland Immunochemicals Inc., Gilbertsville, PA, USA) and goat anti-rabbit Alexafluor 680 (1:5000) (Invitrogen) for 60 minutes at room temperature. Immunoreactive proteins were detected and quantified using the Odyssey infrared imaging system (LI-COR Biosciences). ERK1/2 phosphorylation was calculated by dividing the value obtained from the phosphorylated ERK1/2 channel (700 nm) by the value obtained from total ERK1/2 channel (800 nm) and expressed as fold above vehicle controls. Results are expressed as mean ± SEM from at least four independent experiments.

Taqman RT-PCR was performed as described previously using sequence specific primers and probes designed to span an intron 434546. Briefly, HUVECs were seeded at 5 × 10⁴ cells per well with EBM1% in a 6 well plate. Following overnight serum starvation, cells were treated with V CM or P CM, diluted 1:1 (v/v) with EBM 1% in the absence or presence of SU4984 (20 μM) or PD98059 (50 μM) for the time indicated in the figure legends. RNA was extracted, reverse transcribed and RT-PCR performed using the ABI Prism 7900 as described previously 12. COX-1, COX-2 and FP mRNA were normalized using ribosomal 18S as an internal control. Experiments are representative of at least five independent experiments. Fold difference was determined by dividing the value obtained from P CM treated cells by the value obtained from V CM treated cells and represented as mean ± SEM.

HUVECs were seeded at a density of 5 × 10⁴ cells per 35 mm diameter dish and serum starved overnight. Thereafter cells were treated with V CM or P CM diluted 1:1 (v/v) with EBM 1% containing 3 μg/ml arachidonic acid for the time indicated in the figure legend. After treatment CM was collected and stored at -20°C until required. PGF_2α and PGE₂ secretions in the culture media were assayed by ELISA as described previously 4748. When using chemical inhibitors, cells were preincubated with EBM1% and NS398 (10 μM) or indomethacin (8.4 μM) for 30 minutes and subsequently, cells were treated with V CM or P CM in the absence or presence of inhibitors with the addition of 3 μg/ml arachidonic acid for 6 hrs. Experiments were repeated three times. The data are presented as mean ± SEM.

To generate FP receptor knockdown vectors, oligonucleotides encoding short hairpin transcripts were annealed and individually cloned into the adenovirus shuttle vector pDC316. The start codon of the FP receptor (NM_000959) was used as a reference and labelled as basepair 1. Thus the target sequences corresponded to 478 bp downstream and included a scrambled (scr) negative control: Sh478: 5'-GTGGCCTGGTAATCACTGA; scr: 5'-TTACTCGACGCATGTGCTT. High titre stocks (<10e10 viral plaque forming units/ml, pfu/ml) were prepared using the AdMax system (Microbix Biosystems Inc., Toronto, Ontario, Canada). Briefly, an adenovirus genomic plasmid (pBHGLoxdeltaE1,3Cre) was cotransfected with the shuttle vector into HEK293 cells. After 10 days, plaques were purified and seeded into a T75 flask of HEK293 to generate the first seed. When HEK293 cells showed a cytopathic response; virus was released from the cells by three cycles of freeze thawing. This starter culture was used to produce large bulk preparations of adenovirus. Viral particles were purified using a Vivascience AdenoPack column (Generon House, Eton Wick, UK), buffer exchanged into 8 volumes of 2.5% glycerol, 20 mM Tris-HCl (pH 8) and then concentrated. The viral titre was determined using a modified Adeno-X rapid titre kit (Clontech-Takara Bio Europe, Saint-Germain-en-Laye, France) with 1:1000 rabbit anti-adenovirus Serotype 5 hexon antisera (Labfrontier, Seoul, Korea).

To ablate FP receptor expression, HUVECs were seeded at a density of 3 × 10⁵ cells/25 cm² in EGM-2 and incubated with 100 viruses (either scr or sh478) per cell (MOI) for 24 hrs at 37°C in a 5% CO₂ atmosphere. The next day, the viral infected cells were washed, trypsinised, counted and used in the network formation and proliferation assays.

Statistical significance was assessed on untransformed data with one-way ANOVA and Dunnett's post hoc test using Prism 5.0 (Graph Pad, San Diego, CA). A p-value of less than 0.05 (P < 0.05) was considered to be statistically significant.