Potent and PPARα-independent anti-proliferative action of the hypolipidemic drug fenofibrate in VEGF-dependent angiosarcomas in vitro

Angiosarcomas are highly aggressive tumors of endothelial origin, which carry a poor prognosis. Fenofibrate is a hypolipidemic drug, which acts by activating the transcription factor PPARα. It has also been widely reported to have ‘anti-cancer’ activity. The current study investigated its effect in a murine VEGF-dependent angiosarcoma cell-line, MS1 VEGF. The study utilised assays to monitor cell proliferation and viability, apoptosis, cell cycle progression, mitochondrial membrane potential, changes in protein expression, and changes in miRNA expression using microarrays. Fenofibrate showed potent anti-proliferative action in MS1 VEGF angiosarcoma cells, without inducing apoptosis. It enriched cells in G2/M cell cycle phase and hyperpolarised mitochondria. Other PPARα activators failed to mimic fenofibrate action. Inhibitors of PPARα and NFκB failed to reverse the inhibitory effect of fenofibrate and their combination with fenofibrate was cytotoxic. Fenofibrate downregulated the expression of key VEGF-effector proteins, including Akt, ERK, Bcl-2 and survivin, and a chemical inhibitor screen discovered relevance of these proteins to cell proliferation. A miRNA microarray revealed that fenofibrate differentially regulated cellular miRNAs with known roles in cancer and angiogenesis. The data raise the possibility that fenofibrate could be useful in angiosarcoma therapy, especially considering its well-established clinical safety and tolerability profile.


Potent suppression of MS1 VEGF angiosarcoma cell proliferation by fenofibrate.
To test the effect of fenofibrate in MS1 VEGF angiosarcoma cells, cells were treated with 50 μM fenofibrate (or 0.1% DMSO) for 48 hours. These experiments revealed a robust decrease in cell number after fenofibrate treatment (~20 ± 5.3% of control) (Fig. 1a,b), without reducing cell viability (Control, 96.8 ± 1.9% vs fenofibrate, 91.40 ± 3.3%) (Fig. 1c). MTS proliferation assays also revealed a robust fenofibrate-induced reduction in MS1 VEGF angiosarcoma cell proliferation (~46.0 ± 2% of control) (Fig. 1d). To assess potency, concentration-response experiments were performed and these revealed relatively potent effects of fenofibrate, with cell proliferation reduced by concentrations ≥ 5 μM (Fig. 1e). Parallel comparative experiments were performed in human umbilical vein endothelial cells (HUVEC). Treatment with 50 μM fenofibrate for 48 hours did not affect HUVEC number or viability ( Fig. 1f,g). However, considering the relatively slow proliferation rate of HUVEC, it was hypothesized that a possible inhibitory effect of fenofibrate may be unmasked by allowing HUVEC to proliferate for a longer duration. Indeed, the data suggested a 3.79 ± 0.14-fold increase in HUVEC cell number when cultured for 5 days. Treatment with 50 μM fenofibrate significantly suppressed this increase (fold increase ~1.39 ± 0.18), without reducing cell viability (Fig. 1h). Collectively, the experiments revealed that fenofibrate exerted potent anti-proliferative action in MS1 VEGF angiosarcoma cells, whereas HUVEC, exposed to 10-fold higher concentrations of fenofibrate were less affected.

Fenofibrate did not induce apoptosis, but enriched cells in the G2/M cell cycle phase and hyperpolarised mitochondria in MS1 VEGF angiosarcoma cells.
To investigate if the anti-proliferative action of fenofibrate in MS1 VEGF angiosarcoma cells was associated with early apoptosis, treated cells were stained with either FITC-conjugated Annexin V (early apoptosis) or propidium iodide (cell death) or both. Flow cytometry analysis revealed no significant increase in Annexin V staining after 48-hour treatment with 50 μM fenofibrate (Control, 0.92 ± 0.3% vs fenofibrate, 0.6 ± 0.2%). In contrast, staurosporine -included as a positive control -induced a robust apoptotic response, evidenced by increased Annexin V staining (13.3 ± 0.2%). There was also a small but significant improvement in cell viability (Control, 86.9 ± 2.6% vs fenofibrate, 96.7 ± 0.7%) and a significant decrease in the percentage of dead cells after fenofibrate treatment (Control, 7.8 ± 2.2% vs fenofibrate, 1.5 ± 0.2%) (Fig 2a-g).
To investigate if fenofibrate modulated mitochondrial membrane potential in MS1 VEGF angiosarcoma cells, experiments were performed using the mitochondrial membrane potential-sensitive dye JC-1, which forms red aggregates upon accumulation in hyperpolarized mitochondria. The data revealed that JC-1 Red fluorescence was enhanced by ~2-fold after 48-hour treatment with 50 μM fenofibrate. In contrast, this signal was virtually abolished in cells treated with the mitochondrial depolarizing agent CCCP, which was included as a positive control (JC-1 Red fluorescence: Control, 31.0 ± 3.0% vs fenofibrate, 61.8 ± 1.9% vs CCCP, 2.5 ± 0.8%) (Fig. 4a-d).

Discussion
Although the effects of fenofibrate in primary endothelial cells have been well-studied, there is little known about its actions in angiosarcomas. Data from the current study illustrate that fenofibrate exerted potent anti-proliferative actions in MS1 VEGF angiosarcoma cells, which were independent of PPARα and NFκB. Fenofibrate neither reduced cell viability nor induced apoptosis, arrested cells in G2/M phase, hyperpolarized mitochondria, and downregulated key VEGF-dependent 'oncoproteins' . An inhibitor screen revealed functional relevance of these oncoproteins to angiosarcoma cell proliferation. In addition, a miRNA microarray screen uncovered robust fenofibrate-induced changes in cellular miRNAs, many of which have known roles in angiogenesis and cell proliferation. and mean data for the exemplified experiment (i-l) (n = 3). Statistical analysis was performed using one-way ANOVA followed by Dunnett's multiple comparisons test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s, not significant.
www.nature.com/scientificreports www.nature.com/scientificreports/ Previous studies have reported that fenofibrate suppresses angiogenesis 31,32 , reduces endothelial tube formation 33 , and suppresses proliferation by inducing a G0/G1 block 29 . It also suppresses angiogenesis in vivo via a PPARα-dependent mechanism but accelerates wound healing in diabetic mice 31,32,34,35 . Concentration-response studies of fenofibrate in MDA-MB-231 cells revealed an ~IC 50 of 16 µM for apoptosis-induction 22 . In contrast, our data revealed an apparent IC 50 of 8 µM for fenofibrate's anti-proliferative action in MS1 VEGF angiosarcoma cells (Fig. 1e), which fits well with the plasma concentrations reported in clinical use 36,37 . In contrast to this study, fenofibrate was reported to depolarize mitochondria and trigger apoptosis in glioblastoma 23,38 . These observations suggest that modulation of mitochondrial membrane potential by fenofibrate is cell-type dependent. Importantly, normal human astrocytes were less sensitive to fenofibrate when compared to glioblastoma cells, although fenofibrate suppressed mitochondrial respiration in both cell-types 23 . This observation suggests that alternative mechanisms potentially mediate the inhibitory effects of fenofibrate in cancer cells. The data presented in this study raise the possibility that fenofibrate-mediated changes in cellular miRNAs and oncoprotein downregulation could play an important role. Comparative histological analysis of primary tumors from mice and humans suggested that human angiosarcomas arise from bone-marrow derived hematopoietic stem cells or early EPC, whereas in mice early EPCs appear to play a major role 7,8 . The clinical relevance of the effect of fenofibrate in angiosarcomas could therefore be supported by testing its effects in cells isolated from such tumors. Furthermore, studies comparing the efficacy and potency of fenofibrate in mouse versus human angiosarcoma cells will also be informative and could potentially reveal species-specific differences in fenofibrate action. Importantly, these studies also suggest involvement of VEGF signaling, which could be relevant because we revealed a suppressive effect of fenofibrate on key VEGF-related oncoproteins.
Structure-activity studies using other PPARα agonists revealed that bezafibrate, WY14643 and fenofibric acid were all ineffective. These data not only revealed the PPARα-independence of fenofibrate action but also how minor changes in structure profoundly affected activity. For example, fenofibrate -but not bezafibrate or WY14643 -protected endothelial cells from apoptosis 39 and WY14643 failed to replicate the inhibitory effect of fenofibrate on mitochondrial respiration 23 . Fenofibrate and fenofibric acid differ only slightly in chemical composition (Fig. 5), but only fenofibrate was effective. This cannot be explained by reduced membrane permeability of fenofibric acid because it is readily-permeable and was shown, for example, to protect epithelial cells against high glucose-induced damage 40 . There are few reports directly comparing the actions of fenofibrate and fenofibric acid, although differential effects on AMPK 28 and 11β-hydroxysteroid dehydrogenase have been reported 41 . The latter www.nature.com/scientificreports www.nature.com/scientificreports/ study also observed that esterification/amidation of the carboxy group was essential for activity, which may be relevant to our study. AMPK activation however is unlikely to explain the anti-proliferative action of fenofibrate because bezafibrate also activates AMPK 42 but was ineffective in MS1 VEGF angiosarcoma cells. When used individually, GW6471 (PPARα antagonist) and PDTC (NFκB antagonist) reduced cell proliferation without affecting cell viability but their combination with fenofibrate was cytotoxic. These data not only support the argument that the effects of fenofibrate in MS1 VEGF cells are independent of signaling via the PPARα and NFκB pathways, but also that, inhibiting both pathways results in cytotoxicity. Reversal of fenofibrate effects by GW6471 or PDTC in cancer cells has been reported 22,33,35,38 but, to our knowledge, their cytotoxicity when combined with fenofibrate is a novel observation. NFκB promotes a Warburg effect in pediatric sarcomas 43 , so it is conceivable that the combination of glycolysis-inhibition (by PDTC) combined with downregulation of 'survival' proteins (by fenofibrate) triggers MS1 VEGF angiosarcoma cell death. Inhibition of glycolysis with GW6471 has also been reported 44 . Cytotoxic effects of a combination of fenofibrate with glycolysis inhibitors have been reported and our lab discovered a similar mechanism for metformin cytotoxicity in MS1 VEGF angiosarcoma cells 17,45 . Further studies will be required to clarify the underlying mechanisms and pharmacological studies supporting PPARα-and NFκB-independence may be complemented by RNA interference (RNAi)-mediated 'knock-down' experiments.
Biochemical investigations revealed that treatment with fenofibrate significantly reduced the expression of oncoproteins like Akt, survivin, Bcl-2 and ERK. A chemical screen of inhibitors targeting these oncoproteins uncovered effects of several chemicals, including Akt1/2 inhibitor, YM155 (survivin), TW37 (Bcl-2), and PD98059. Importantly, these experiments also revealed that fenofibrate was the most effective agent among all tested compounds. Although this may be explained by concentration-dependence in some cases, it is also possible that the high efficacy is due to fenofibrate's ability to down-regulate these key proteins simultaneously. There is little known about the relevance of these molecules to murine angiosarcomas, although they have been investigated in human/canine angiosarcomas. For example, survivin was overexpressed in human angiosarcomas and YM155 inhibited proliferation in human angiosarcoma cells 46 . Constitutive activation of the PI3K-Akt-mTOR signaling pathway was also reported in human angiosarcomas 47 . Furthermore, canine angiosarcomas showed constitutive  4). Statistical analysis was performed using one-way ANOVA followed by Tukey's multiple comparisons test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s, not significant.
www.nature.com/scientificreports www.nature.com/scientificreports/ ERK activation and MEK inhibition reduced in vitro cell viability 48 . Fenofibrate-induced growth arrest in G2/M phase coupled to downregulation of anti-apoptotic proteins Bcl-2 and survivin may also sensitize MS1 VEGF angiosarcoma cells to cytotoxic agents 49 . The data therefore confirm the established oncogenic roles for proteins like Akt, survivin, Bcl-2 and ERK in angiosarcomas and expand the pharmacological utility of specific inhibitors to murine VEGF-dependent angiosarcomas.
The relevance of miRNAs in angiosarcomas and soft tissue sarcomas is starting to be elucidated. For example, miR-497-5p, -378-3p and 483-5 were downregulated in angiosarcoma and targeting of a potassium channel (KCa3.1) by miR-497-5p led to inhibition of cell proliferation and invasion 50 . Sarver et al. profiled the expression of miRNAs in over 20 different sarcomas and reported significant upregulation of chromosome 19 miRNAs in angiosarcomas relative to other sarcomas 51 . Concepcion et al. reported Myc-dependent expression of the miR-17-92 cluster, which may be relevant to angiosarcomas that develop secondary to radiation exposure 52,53 . There is therefore considerable interest in identifying miRNAs with functional and therapeutic relevance to angiosarcomas. In our study, we profiled fenofibrate-induced changes in cellular miRNAs and identified several miRNAs that were differentially expressed. Although little is known about the relevance of these miRNAs to angiosarcomas, there is evidence in the literature suggesting that many of the upregulated miRNAs exert anti-angiogenic and anti-proliferative roles in cancer. These miRNAs include miR122, 140-5p and -20b, which are known to target VEGF [54][55][56] . miRNA-210 was robustly induced by fenofibrate in our study and has been shown to also reduce proliferation and induce G2/M arrest in colorectal cancer cells 57 and its overexpression was associated with improved prognosis 58 . Furthermore, fenofibrate-induced miRNAs known to target Akt, survivin, Bcl-2 and ERK include miR29b, -29a-3p and -122 (Akt) 59 , miR31 (Bcl-2) 60 , miR203 (survivin) 61 and miR20b 62 . Among the miRNAs downregulated by fenofibrate in MS1 VEGF angiosarcoma cells, many are known drivers of cancer cell proliferation, including miR335, -146, -130a, and -135b [63][64][65] . The data therefore raise the intriguing possibility that the anti-proliferative effect of fenofibrate in MS1 VEGF angiosarcoma cells may at least partly be driven by differential changes in cellular miRNAs with pro-or anti-proliferative activity. A direct way to test this hypothesis would be to artificially manipulate the expression of individual or a combination of miRNAs using miRNA mimics or inhibitors to evaluate their role in cell proliferation and understand if this manipulation alters the efficacy and/ or potency of fenofibrate.
In conclusion, we report potent inhibitory effects of the cholesterol-lowering drug fenofibrate in MS1 VEGF angiosarcoma cells, which were independent of PPARα and NFκB. Combined treatment with fenofibrate and a PPARα-or NFκB antagonist led to cytotoxicity. Fenofibrate downregulated the expression of Akt, survivin, ERK and Bcl-2 and a chemical screen uncovered a role for these 'oncoproteins' in cell proliferation and viability. Finally, this study discovered that fenofibrate induces robust changes in cellular miRNAs, many of which potentially regulate angiogenesis and have established roles in cancer. The data therefore establish fenofibrate as a potent inhibitor of VEGF-dependent angiosarcoma cell proliferation and highlight important pharmacological www.nature.com/scientificreports www.nature.com/scientificreports/ differences with its observed effects in other cancer cells in terms of potency, effects on apoptosis and mitochondrial function, PPARα-and NFκB-dependence, and interactions with PPARα-and NFκB pathways. The drug may have potential utility in angiosarcoma therapy.

Materials and Methods
Cell culture. MS1 VEGF angiosarcoma cells were purchased from American Type Culture Collection (ATCC R CRL-2460 TM ). Cells were grown in Dulbecco's Modified Eagle's Medium (DMEM, Cat# 11885, Invitrogen), supplemented with 5% fetal bovine serum (Cat# F2442, Sigma Aldrich) and 1% antibiotics (penicillin/streptomycin, Cat# 15140, LifeTechnologies). The total glucose concentration in the medium was adjusted to 11 mM (equivalent to non-fasting basal blood glucose in non-diabetic mice) using a 10% glucose solution (Cat# G8644, Sigma Aldrich). MS1 VEGF angiosarcoma cells were passaged every 2-3 days and used for experiments up to passage 15. 80-90% confluent cells were washed once with PBS and then incubated with trypsin (Cat# T3924, Sigma Aldrich) for 5 minutes to detach cells. Trypsin was neutralized using cell culture medium, the cells were mixed well by pipetting, transferred to a 15 ml polypropylene tube, and centrifuged at 300 g for 5 minutes. The supernatant was removed and the cells were re-suspended in cell culture medium. Human Umbilical Vein Endothelial Cells (HUVEC) were purchased either from ATCC (PCS 100-010 TM ) or Lonza (Cat# C2519A). HUVECs were grown in Medium 199 (M199, Cat# M4530, Sigma Aldrich) supplemented with 15% fetal bovine serum (Cat# F6178, Sigma Aldrich), 30 μg/ml Endothelial Cell Growth Supplement (ECGS, Cat# 356006, BD Biosciences) and 100 μg/ml heparin (Cat# H4784, Sigma Aldrich). Cells were used for experiments up to passage 6. For HUVEC culture, Detachin (Cat# T100100, Genlantis) was used instead of trypsin and cells were not centrifuged. Both MS1 VEGF angiosarcoma cells and HUVECs were grown in a 37 °C humidified incubator supplied with 5% CO 2 .
Drug treatments. Fenofibrate (Cat# F6020, Sigma Aldrich) was dissolved in 100% anhydrous dimethyl sulfoxide (DMSO) and was prepared fresh on the day of the experiment. The drug was always prepared at 1000x concentration (e.g. 50 mM) and then diluted 1:1000 in cell culture medium to achieve the desired final concentration (e.g. 50 μM). The solution was mixed well by pipetting to aid solubility before addition to the cells. In most experiments, the total duration of treatment with fenofibrate was 48 hours and the treatment medium was replenished after 24 hours. The method of preparation and treatment protocol for the other PPARα agonists were the same as those described for fenofibrate. In experiments involving GW6471 (PPARα antagonist) and PDTC (NFκB antagonist), cells were pre-treated with each antagonist for 1 hour prior to treatment with fenofibrate and . Fenofibrate (50 μM) was used as a positive control in the screen. Statistical analysis was performed using one-way ANOVA followed by Dunnett's multiple comparisons test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s, not significant.
Apoptosis assay. Apoptosis assays were performed using the FITC Annexin V Apoptosis Detection Kit I (Cat# 556547, BD Pharmingen) following the manufacturer's protocol. Briefly, treated cells were centrifuged (300 g, 5 minutes), washed once in cold PBS and then re-suspended in 1x Binding Buffer. The resuspended cells (100 μl) were incubated with either propidium iodide or FITC Annexin V (single-stained) or both (double-stained) for 15 minutes in the dark at room temperature. The final volume of the suspension was adjusted to 500 μl using 1x Binding Buffer and analysis was performed within 1 hour using a BD LSR Fortessa Cell analyzer (BD Biosciences). Unstained and single-stained samples were used as compensation controls during the experiment. Staff providing technical support during this experiment (and other flow cytometry experiments) was blinded to the experimental groups to limit bias during data collection and analysis.
Cell cycle analysis. Treated cells were re-suspended in PBS following trypsinization and centrifugation steps as described above in the cell culture section. Cells were then fixed by adding this suspension drop-wise to ice-cold 100% ethanol and stored overnight at −20 °C. Fixed cells were stained using a solution that contained (in PBS): 50 μg/ml propidium iodide (Cat# P1304MP, LifeTechnologies), 100 μg/ml RNase A (Cat# 12091-021, LifeTechnologies) and 0.1% Triton-X 100 (Cat# A16046, Alfa Aesar). Cells were stained for 1 hour in a 37 °C water www.nature.com/scientificreports www.nature.com/scientificreports/ bath, washed once with PBS and re-suspended in 500 μl PBS for analysis. Unstained cells were used as control during the experiment. JC-1 mitochondrial membrane potential assay. Mitochondrial membrane potential measurements were made using the MitoProbe TM JC-1 Assay Kit for Flow Cytometry (Cat# M34152, Life Technologies). Treated cells were detached using trypsin and the cell suspension was centrifuged at 300 g for 5 minutes. The pellet was re-suspended in warm cell culture medium containing 2 μM JC-1 dye and incubated in a humidified, 5% CO 2 incubator at 37 °C for 30 minutes. 2 ml cell culture medium was then added, followed by centrifugation at 300 g for 5 minutes. The pellet was re-suspended in 400 μl PBS and analyzed immediately on a BD LSR Fortessa Cell analyzer (BD Biosciences). Cells not stained with JC-1 and those stained with JC-1 but treated with the mitochondrial depolarizing agent carbonyl cyanide m-chlorophenyl hydrazine (CCCP) were used as controls during the experiment.
Western blotting. Western blotting experiments were performed using standard protocols. Briefly, cell lysates were prepared in RIPA buffer supplemented with a protease/phosphatase inhibitor cocktail (1:100, Cat# 1861284, ThermoScientific) and EDTA (1:100). 30-40 μg proteins were separated on an SDS-PAGE gel followed by transfer onto a nitrocellulose membrane. Membranes were 'blocked' with 5% bovine serum albumin dissolved in TRIS-buffered saline (TBS) containing 0.1% Tween 20 (wash buffer). Incubation with the primary antibody was overnight at 4 °C on a rocker. Following 3 washes, membranes were incubated for 1 hour at room temperature with the appropriate HRP-conjugated secondary antibodies. Following 3 washes, proteins were detected using the Amersham ECL Prime Western Blotting Detection Reagent (Cat# RPN2232, GE Healthcare Life Sciences) and visualized on a Geliance P600 Gel Documentation System (PerkinElmer, Inc. MA, USA). The band densities of individual proteins were quantified using the Bio-Rad Quantity One software. All antibodies were purchased from Cell Signaling Technology ( Chemical inhibitor screen. The chemical inhibitor screen was performed in either 12-well plates (cell counts/viability) or 96-well plates (MTS proliferation assays). Cells were treated with the inhibitors for a total duration of 48 hours (except TW-37, 24-hour treatment), and the treatment media was replenished after 24 hours. The inhibitors and concentrations used were: LY294002 (PI3K inhibitor, 10 μM), Akt1/2 kinase inhibitor (10 μM), TW-37 (Bcl-2 inhibitor, 1 μM), SU1498 (VEGFR inhibitor, 10 μM), PD98059 (ERK inhibitor, 10 μM), YM155 (survivin inhibitor, 1 μM), temsirolimus (mTOR inhibitor, 1 μM), and SU5402 (FGFR inhibitor, 10 μM). Fenofibrate was used as a positive control in all experiments. The supplier information for each inhibitor can be found in the Materials section.
MicroRNA microarray. Expression profiling of mature microRNAs (miRNA) and changes in response to fenofibrate were analyzed using the Mouse miScript miRNA Cancer PathwayFinder PCR array (Cat# MIMM-102Z, Qiagen). The array profiles 84 miRNAs relevant to cancer and includes positive, negative and normalization controls. Cells were treated with fenofibrate as described above. RNA was extracted from samples pooled from 3 independent experiments using the miRNeasy Mini Kit (Cat# 217004, Qiagen) and 250 ng RNA was reverse-transcribed using the miScript II RT kit (Cat# 218160, Qiagen). 200 μl RNase-free water was added to dilute the cDNA prior to use. The reaction mix for the miRNA PCR array was prepared according to the manufacturer's protocol and contained: 2X QuantiTect SYBR Green PCR Master, 10X miScript Universal Primer, RNase-free water, and template cDNA. 25 μl of this reaction mix was added to each well of the 96-well array plate and the plate centrifuged for 1 minute at 1000 g prior to PCR. The cycling conditions for real-time PCR were: Initial activation Step -15 minutes, 95 °C; 3-step cycling (40 cycles) -Denaturation (15 seconds, 94 °C), Annealing (30 seconds, 55 °C) and Extension (30 seconds, 70 °C). Data were collected and after defining the fluorescence baseline and threshold, analysis was performed using the ΔΔC T method of relative quantification using the online data analysis software available at http://pcrdataanalysis.sabiosciences.com/mirna. SNORD61 and SNORD96A were used as normalization controls and their expression was unchanged after treatment with fenofibrate. Changes in expression of miRNAs in response to fenofibrate treatment were plotted as a Heat Map depicted in Fig. 8.
Data and statistical analysis. Data were analyzed using GraphPad Prism 7.0 and presented as mean ± SEM. Statistical analysis was performed using either a Student's t-test (for 2 groups) or ordinary one-way ANOVA (followed by post-hoc analysis) when more than 2 groups were being compared. A 'p' value of less than 0.05 was considered statistically significant. Statistical analysis was only performed on data generated from independent experiments. Specific details of the statistical tests and experimental 'n' values are indicated in the legend of each figure.