The GPR 55 agonist, L-α-lysophosphatidylinositol, mediates ovarian carcinoma cell-induced angiogenesis

Cell culture

Human ECFCs were isolated from neonatal cord and peripheral blood and their distinct endothelial phenotypes were verified by flow cytometry as previously described (see Supporting Information Fig. S1) (Hofmann et al., 2009; 2012; Reinisch and Strunk, 2009; Reinisch et al., 2009). HUVECs were obtained from Lonza (Basel, Switzerland). ECFC and HUVECs were grown in endothelial growth medium-2 (EGM-2) (Lonza) containing 2% FBS and 1% penicillin/streptomycin/L-glutamine/heparin (Life Technologies, Carlsbad, CA, USA) and EGM-2 growth factor supplements (composed of bFGF, IGF-2, EGF, VEGF, ascorbic acid, hydrocortisone). Ovarian carcinoma cell lines OVCAR-3 (American Type Culture Collection, Manassas, VA, USA), OVCAR-5 (kindly provided by the Cell Culture Core, Vascular Biology Program, Boston Children's Hospital, Boston, MA, USA) and COV-362 (Sigma Aldrich, St. Louis, MO, USA) were grown in DMEM containing 10% FBS.

Ethics statement

Prior approval was obtained for human cell and tissue sample collection from the Institutional Review Board of the Medical University of Graz (protocols 19-252 ex 07/08, 18-243 ex 06/07, 21.060 ex 09/10). Adult samples were collected after written informed consent from healthy volunteers, and umbilical cord samples after written informed consent by the mother after full-term pregnancies in accordance with the Declaration of Helsinki.

Extraction of lipids

Up to 19 mL of conditioned medium from 6–10 million cells of OVCAR-3, OVCAR-5, DMEM and 14 mL from COV-362 cells were extracted with 40 mL of acidified 2:1 methanol : chloroform and 0.05 N HCl in a 60 mL separator funnel. The bottom layer was collected and dried under a gentle stream of nitrogen in a 20 mL glass vial. The lipid extract was reconstituted in 200 μL chloroform : methanol (2:1 v/v) prior to injection.

LPI measurement

Chemical standards of LPI were obtained from Sigma Aldrich. An LC-MS/MS method was optimized on an Agilent (Agilent Technologies, Santa Clara, CA, USA) 6460 triple-quadrupole mass spectrometer using multiple reaction monitoring (MRM) in negative ion mode. Specifically, the MRM transitions used for LPI were 571.3 – >255.1 m/z for quantification and 571.3 – >152.9 m/z for confirmation. A collision energy of 41 V and a fragmentor setting of 207 V were used to monitor both MRM transitions. The most abundant fragment corresponds to the loss of palmitic acid and the secondary fragment corresponds to the subsequent loss of the inositol, leaving a C₃H₆O₅P⁻ ion at 152.9 m/z. Mass spectrometer parameter settings were gas temperature (350°C), gas flow (10 L·min⁻¹), nebulizer (30 psi), sheath gas temperature (400°C), sheath gas flow (11 L·min⁻¹), capillary voltage (3800 V) and nozzle voltage (500 V). Liquid chromatography conditions with a Dikma-Biobond C4 column 4.6 × 50 mm 5 μm particle size were used for separation. Chromatography method included gradient elution at 0.400 mL·min⁻¹ with solvent 20 mM ammonium carbonate/0.1% ammonium hydroxide as mobile phase A and acetonitrile for mobile phase B. The gradient started at 0% B and progressed to 100% A in 16 min, and then changed back to 0% B over 0.1 min, and re-equilibrated for 3.9 min before the next injection. A 10 μL sample injection was used for all standards and samples. An external standard curve was used to calculate concentrations of LPI in different samples between 0.0025 and 0.25 μM. The lower limit of detection and quantification was determined to be at 2.5 nM with a S/N > 12. A validation of the method was done using 20 nM LPI added to DMEM containing 10% FBS and extracted using the method described, and reconstituted in 200 μL chloroform : methanol (2:1 v/v) prior to injection. The control experiment was done using DMEM containing 10% FBS prepared in the same manner. A linear calibration curve was measured for LPI with an R² of 0.95.

Proliferation assay

ECFCs from three different cord blood donors, HUVECs and peripheral blood ECFC were seeded in 24-well plates (Nalge Nunc, Rochester, NY, USA) in EGM-2 at a density of 3000 cells/cm² and allowed to adhere for 24 h. Subsequently, cells were subjected to growth factor-reduced medium [EBM-2 (Lonza) containing 2% FBS and 1% penicillin/streptomycin/L-glutamine/heparin (Life Technologies) without the addition of EGM-2 growth factor supplements] with or without different concentrations of LPI (Sigma) and/or endocannabinoid receptor antagonists: CID16020046 (Tocris Bioscience, Northpoint, Avonmouth, Bristol, UK) and AM251, SR144528 (both Cayman Chemical Europe, Tallinn, Estonia). A 30 min treatment with 10 μM U0126 (Cell Signaling, Cambridge, UK) was also tested. After 48 h, treated cells were harvested and the cell number was counted by a Casy cell counter (Roche, Mannheim, Germany). Nine independent experiments per group were performed in triplicate.

Matrigel angiogenesis assay

Capillary-like network formation of ECFC, isolated from three different donors, plated on growth factor-reduced Matrigel^® (BD, Biosciences, San Jose, CA, USA) was performed according to the instruction manual included in the purchase of Matrigel. The influence of LPI and different endocannabinoid receptor antagonists was tested in growth factor-reduced medium [EBM-2 (Lonza) containing 2% FBS and 1% penicillin/streptomycin/L-glutamine/heparin (Life Technologies) without the addition of EGM-2 growth factor supplements]. Network formation (14–16 h) was documented with a Nikon SPOT camera on a Nikon microscope (Nikon, Amsterdam, The Netherlands). Branch points were counted after 16 h by ImageJ [National Institutes of Health (NIH), Bethesda, MD, USA]. Nine independent experiments per group were performed in triplicate.

Ovarian cancer cell conditioned media

Ovarian cancer cells OVCAR-3, OVCAR-5 and COV-362 cells were grown in a 10 cm dish with DMEM and 10% FBS until approximately 80% confluent. After washing with PBS, cells were incubated in 10 mL phenol-free DMEM without serum for 24 h. Conditioned medium was collected, centrifuged and immediately used. Non-conditioned phenol red-free DMEM served as negative control.

CAM assay

Chicken eggs were purchased from Charles River Laboratories (Wilmington, MA, USA) and placed in an incubator at 37°C, 40% humidity. On day 3, up to 8 mL of albumin was aspirated from a small hole made at the bottom of the egg and the hole was sealed with candle wax. Then an approximately 2 cm large window was cracked into the rounded part of the upright egg using Dumont tweezers (6) and the egg membrane was completely removed. The window was covered with a cap of sterilized aluminium foil. The eggs were then incubated in a cell culture incubator at 37°C, 40% humidity, 3% CO₂. On day 7, up to 2 mm filter paper patches were punched out of sterilized Whatman-filter papers (Sigma Aldrich) and placed on a sterile surface. Five microlitres of treatment solution, control medium or ovarian cancer cell conditioned medium from OVCAR-3, OVCAR-5 or COV-362 cells , as indicated, were dropped on each filter paper allowing it to dry for 15 min and carefully placed on the developing CAM. Blood vessel development was observed daily and pictures were taken 3 days (day 10 of egg development) after treatment with a stereo microscope. Vessels crossing the outline of the filter paper were analysed using ImageJ (NIH). Six to nine independent experiments per group were performed in triplicate.

Reduction of GPR55 gene expression by siRNA

Transfection of ECFCs with a pool of four validated target-specific GPR55 siRNAs (FlexiTube siRNA; Qiagen, Venlo, The Netherlands) (referred to as siGPR55) or scrambled siRNA (Qiagen) (referred to as sicontrol) was performed using Lipofectamine^® RNAiMAX reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. All experiments were performed 36–48 h after transfection. The efficiency of siRNAs and of an appropriate negative control was determined by Western blot.

Proteome profiler arrays

The Proteome Profiler Human Phospho-Kinase Array Kit (Cat. No: ARY003B) and Human Angiogenesis Array Kit (Cat. No: ARY007) were obtained from R&D Systems (Minneapolis, MN, USA). ECFCs were treated with vehicle or LPI (10 μM) for 15 min or 24 h, respectively, and analysed according to manufacturer's instructions. The average signal (pixel density) of duplicate spots representing each protein was evaluated by ImageJ (NIH). After background subtraction, a twofold increase was considered to be significant.

Western blot analysis

ECFCs were serum starved for 6–24 h and subsequently treated with vehicle, LPI and/or CID366791 for 0–60 min. In order to extract the sturdily membrane-bound GPR55, cells were lysed directly with 1× reducing Laemmli (SDS sample) buffer (Boston BioProducts, Inc., Boston, MA, USA) and precipitated 10 min at 90°C. Otherwise, cells were lysed and Western blots were performed as previously described by us (Hofmann et al., 2012; 2014). Specific proteins were detected using antibodies against GPR55 (Thermo Scientific, Tewksbury, MA, USA) and total or phosphorylated ERK1/2 or p38 (all obtained from Cell Signaling), and compared with housekeeping protein control β-actin (Santa Cruz, Dallas, TX, USA). Pixel intensity was determined using ImageJ (NIH). Six independent experiments per group were performed in triplicate.

Data analysis

‘n’ values refer to the number of individual experiments performed. Data were compared using anova and subsequent Bonferroni post hoc test or two-tailed Student's t-test assuming unequal variances, where applicable. Statistical significance was assumed at P < 0.05. EC₅₀ and IC₅₀ values were calculated out of at least three independent experiments with three to five repeats for each concentration using GraphPad Prism^® 5.0f (GraphPad Software, La Jolla, CA, USA) and expressed with the 95% confidence interval provided in parenthesis.

Ovarian cancer cells produce LPI and mediate angiogenesis through GPR55

Increased serum levels of the GPR55-ligand LPI have been found in patients with high-grade ovarian carcinoma (Xiao et al., 2000; 2001; Xu et al., 2001; Sutphen et al., 2004; Murph et al., 2007; Pineiro et al., 2011; Pineiro and Falasca, 2012). To test our hypothesis, that ovarian cancer cells secrete LPI, and thus promote tumour angiogenesis in vivo via an LPI/GPR55-dependent mechanism; conditioned medium from the human ovarian cancer cell lines OVCAR-3, OVCAR-5 and COV-362 was analysed for its LPI levels and in the CAM angiogenesis model. LC-MS/MS revealed that OVCAR-3, OVCAR-5 and COV-362 cells produced significant but quite different amounts of LPI (Figure 1A). Within 3 days, conditioned medium from OVCAR-3, OVCAR-5 and COV-362 strongly induced angiogenesis in vivo to a similar extent (90–100% increase), compared with unconditioned medium (Figure 1B). Selective inhibition of the LPI receptor GPR55 with CID16020046 (20 μM) effectively blocked ovarian cancer-induced angiogenesis of all tested cell lines (Figure 1B). Together, these results suggest that LPI produced by ovarian cancer cells induces angiogenesis in a GPR55-dependent manner.

LPI regulates angiogenic potential of endothelial cells in vitro and angiogenesis in vivo

The effects of purified LPI on endothelial cell proliferation, network formation and migration were tested in vitro on isolated endothelial colony-forming progenitor cells (ECFCs) derived from three different donors. The isolated human neonatal cord ECFCs showed a distinct endothelial phenotype as shown by expression of typical endothelial cell surface markers (Supporting Information Fig. S1), as previously shown (Hofmann et al., 2009; 2012; Reinisch and Strunk, 2009; Reinisch et al., 2009). LPI stimulated ECFC proliferation in a dose-dependent manner with an EC₅₀ of 2.8 (2.2–3.6) μM (Supporting Information Fig. S2a). Low concentrations of LPI, resembling the endogenous LPI levels secreted by 10⁷ ovarian carcinoma cells (1 nM), were sufficient to stimulate proliferation of ECFCs (Supporting Information Fig. S2a). The maximum proliferative increase (1.55 ± 0.1-fold increase) was measured within 48 h upon applying 10 μM LPI as compared with vehicle controls (Figure 2A and Supporting Information Fig. S2a) and further experiments were performed at this 10 μM concentration. HUVECs showed a similar increase in proliferation as did isolated human adult peripheral blood ECFCs (Supporting Information Fig. S2b). Furthermore, compared with vehicle controls, 10 μM LPI significantly increased ECFC network formation in an in vitro Matrigel assay (Figure 2B) and closure of an endothelial wound in an in vitro scratch assay (Figure 2C).

To investigate whether these stimulatory effects could occur in vivo, we analysed angiogenesis in a CAM assay. For this purpose, we placed a filter paper soaked and then dried with either vehicle or 10 μM LPI on the developing 7-day-old CAM. Within 72 h, 10 μM LPI had significantly increased vessel formation, compared with vehicle control (Figure 2D). Together, these in vitro and in vivo results indicate that LPI is a potent pro-angiogenic factor.

LPI-induced angiogenesis is GPR55 dependent

To identify a pharmacological inhibitor of LPI-mediated pro-angiogenesis, we tested specific antagonists of known LPI receptors such as the CB₁, CB₂ recptors and GPR 55 (Pineiro and Falasca, 2012). The GPR55 antagonist CID16020046 (Kargl et al., 2013) decreased LPI-induced ECFC proliferation in a concentration-dependent manner with an IC₅₀ of 17.9 (17.3–18.5) μM (Supporting Information Fig. S3a). LPI-stimulated ECFC proliferation was most effectively inhibited with a CID16020046 concentration of 20 μM, without affecting basal ECFC proliferation (Figure 3A). In contrast, the LPI-stimulated effect was not significantly inhibited by addition of the CB₁ receptor antagonist/GPR55 agonist (AM251) or by antagonism of CB₂ receptors with SR144528 (Supporting Information Fig. S3b). Furthermore, CID16020046 totally suppressed the LPI-induced network formation (Figure 3B) and endothelial wound healing (Figure 3C), without affecting the basal angiogenic capacity of endothelial cells. To confirm that LPI activity was GPR55 dependent, GPR55 was genetically knocked down with a mix of four validated siRNAs in ECFCs (Figure 3D). In response to LPI, siGPR55-ECFCs showed significantly reduced proliferation as compared with ECFCs transfected with control siRNA (Figure 3E). Simultaneous treatment with the GPR55 inhibitor CID16020046 significantly reduced the LPI-stimulated angiogenesis in the in vivo CAM model (Figure 4). Neither CID16020046 nor silencing of GPR55 significantly affected basal angiogenic activities of ECFCs in vitro nor angiogenesis in the CAM assay in vivo (Figures 3 and 4; Supporting Information Fig. S3). Altogether, these results demonstrate that exogenous LPI stimulates the pro-angiogenic capacity of ECFCs in vitro and angiogenesis in vivo in a specifically GPR55-dependent manner.

LPI/GPR55 stimulates ERK1/2 and p38 activation

A human phospho-kinase array was used to investigate the molecular mechanisms underlying LPI/GPR55-mediated angiogenesis. Of the various phospho-proteins in the array, LPI significantly induced phosphorylation of only ERK1/2 and p38 in ECFCs (Figure 5A and Supporting Information Table S1). Western blot analysis confirmed a time-dependent activation of ERK1/2 and p38 by 10 μM LPI (Figure 5B) but not the potential involvement of CREB (cAMP response element-binding protein) (data not shown). Pharmacological inhibition of GPR55 by 20 μM CID16020046 significantly reduced LPI-induced ERK1/2 and p38 phosphorylation (Figure 5C). To confirm the GPR55-dependent activation of ERK1/2 and p38 by LPI, GPR55 was silenced by siRNA (Figure 5D). Compared with control siRNA, knock-down of GPR55 suppressed the LPI-stimulated ERK1/2 and p38 phosphorylation (Figure 5D). Moreover, ECFCs pretreated with the well-established and highly selective MEK1/MEK2 inhibitor U0126 (10 μM) (Favata et al., 1998), which blocks downstream activation of ERK1/2, eliminated ERK1/2 basal phosphorylation without altering total ERK1/2 amounts (Figure 6A). However, LPI no longer induced ERK1/2 phosphorylation (Figure 6A). Furthermore, ERK1/2 inhibition blocked normal ECFC proliferation, without reducing the initial cell number, and prevented the LPI-induced ECFC proliferation, indicating a crucial role for ERK1/2 during endothelial cell proliferation (Figure 6B). Together, these results suggest that LPI induces GPR55-dependent activation of ERK1/2 and thus leads to increased angiogenesis.

We further investigated whether the observed pro-angiogenic effect of LPI relies on an LPI-induced up-regulation of angiogenesis-related proteins and thereby indirectly leads to an autocrine angiogenic feedback loop with an activation of ERK1/2 and p38. A proteome profiler human angiogenesis array revealed that LPI did not lead to an altered production of any of the 55 tested known angiogenesis-related proteins from ECFCs within 24 h (Figure 6C and Supporting Information Table S2).