Prostaglandin E2 promotes MYCN non-amplified neuroblastoma cell survival via β-catenin stabilization

Amplification of MYCN is the most well-known prognostic marker of neuroblastoma risk classification, but still is only observed in 25% of cases. Recent evidence points to the cyclic adenosine monophosphate (cAMP) elevating ligand prostaglandin E2 (PGE2) and β-catenin as two novel players in neuroblastoma. Here, we aimed to define the potential role of PGE2 and cAMP and its potential interplay with β-catenin, both of which may converge on neuroblastoma cell behaviour. Gain and loss of β-catenin function, PGE2, the adenylyl cyclase activator forskolin and pharmacological inhibition of cyclooxygenase-2 (COX-2) were studied in two human neuroblastoma cell lines without MYCN amplification. Our findings show that PGE2 enhanced cell viability through the EP4 receptor and cAMP elevation, whereas COX-2 inhibitors attenuated cell viability. Interestingly, PGE2 and forskolin promoted glycogen synthase kinase 3β inhibition, β-catenin phosphorylation at the protein kinase A target residue ser675, β-catenin nuclear translocation and TCF-dependent gene transcription. Ectopic expression of a degradation-resistant β-catenin mutant enhances neuroblastoma cell viability and inhibition of β-catenin with XAV939 prevented PGE2-induced cell viability. Finally, we show increased β-catenin expression in human high-risk neuroblastoma tissue without MYCN amplification. Our data indicate that PGE2 enhances neuroblastoma cell viability, a process which may involve cAMP-mediated β-catenin stabilization, and suggest that this pathway is of relevance to high-risk neuroblastoma without MYCN amplification.


Introduction
Neuroblastomas are heterogeneous tumours that vary to a large extent in prognosis and disease outcome and are responsible for 15% of all childhood cancer deaths [1]. Although recent studies have focused on the link between proliferation, differentiation and programmed cell death in neuroblastoma, understanding of the molecu-lar mechanisms driving neuroblastoma heterogeneity is poor. The most extensively defined factor associated with neuroblastoma is the Myc oncoprotein family member MYCN, which is especially amplified in the most aggressive and often metastatic forms of neuroblastoma [2,3]. Although MYCN has important prognostic value, amplification is only observed in about 25% of neuroblastoma cases and it remains largely to be defined what other factors contribute to high-risk neuroblastoma.
Expression of cyclooxygenase-2 (COX-2) and prostaglandin E 2 (PGE 2 ) have been found increased in a variety of malignant tumours, including neuroblastoma [4,5] and pharmacological inhibition of COX-2 has been shown to attenuate cell cycle progression in malignant cells [6][7][8][9]. PGE 2 is produced by a multistep enzymatic process in which the rate-limiting step is mediated by COX enzymes. PGE 2 binds to its membrane bound E-type prostanoid receptors, of which prostanoid receptors type 2 and 4 are known to couple to Ga s and are thereby able to increase intracellular cyclic adenosine monophosphate (cAMP) levels. cAMP is involved in the regulation of diverse cellular processes, including regulation of cytoskeletal dynamics, cellular differentiation, proliferation and programmed cell death in a variety of cells including neural-like cells [10,11].
Of particular interest are recent research lines that focus on molecular interactions between PGE 2 , cAMP and b-catenin. b-catenin contributes to other malignancies such as hepatocellular carcinoma and colorectal carcinoma and its role in paediatric malignancies is well documented [12]. Also, its role in normal physiological development of pluripotent cells from the neural crest has been well-established [13][14][15]. Regarding neuroblastoma, b-catenin expression is increased in MYCN non-amplified neuroblastoma cell lines and b-catenin target gene transcription is increased in neuroblastoma tumours without MYCN amplification [16].
Distinct pools of b-catenin exhibit distinct cellular functions. b-Catenin associates with membrane junctional complexes where it binds to cadherins and a-actin. Free cytosolic b-catenin is rapidly tagged for proteasomal degradation by a multiprotein destruction complex comprised of the kinases glycogen synthase kinase 3b (GSK3b), casein kinase 1 and adaptor proteins like axin2, which is the limiting component in the assembly of this complex [17][18][19]. Stabilized b-catenin translocates to the nucleus, where it activates transcription of TCF/Lef target genes. The result is expression of mitogenic and survival genes including Myc oncogene family members [20] and cyclin D1 [21]. Interestingly, PGE 2 has been shown to enhance b-catenin nuclear localization via dissociation of GSK3b from axin by Ga s [22] and by activating protein kinase A (PKA) [23]. Activated PKA can directly phosphorylate b-catenin at residue ser675 [24] and GSK3b at residue ser9 [10,25,26].
In this paper, we aim to identify the contribution of a molecular link between PGE 2 and b-catenin to cell proliferation and inhibition of apoptosis, independent of MYCN amplification.

Cell culture
Human neuroblastoma cell lines SK-N-AS and SK-N-SH were obtained from ATCC (Manassas, VA, USA). Both cell lines are of epithelial morphology. Cells were maintained in DMEM (1.0 g/l glucose, HEPES) supplemented with 10% v/v heat-inactivated FCS, non-essential amino acids and antibiotics (penicillin 100 U/ml, streptomycin 100 l/ml) in a humidified atmosphere of 5% CO 2 at 37°C. Cells were washed with HBSS (400 mg/l KCl, 60 mg/l KH 2 PO 4 , 8 g/l NaCl, 350 mg/l NaHCO 3 , 50 mg/l Na 2 HPO 4 ÁH 2 O, 1 g/l glucose, pH 7.4), dissociated from the plate with trypsin EDTA and seeded in appropriate cell culture plate format. Cells were serum-deprived for 24 hrs before stimulation. Inhibitors (XAV939, celecoxib and niflumic acid) or antagonists (AH6809 and L-161,982) were added 30 min. prior to stimulation with PGE 2 .

cAMP assay
Experiments were performed in 24-well cell culture plates. When indicated, cells were pre-incubated with niflumic acid for 2 hrs. Cells were incubated in serum-free DMEM supplemented with 0.1 mM 3-Isobutyl-1-methylxanthine for 10 min at 37°C with indicated stimuli. A radioactive competitive binding assay was used to determine cAMP levels, as described earlier [27,28,29,30].

JC-1 mitochondrial membrane polarization assay
Cells were seeded in 24-well plates and incubated with designated stimuli for 24 hrs and mitochondrial membrane polarization was measured using the Mitochondrial Staining Kit (Sigma-Aldrich, Munich, Germany).

Isolation of mRNA and real-time PCR analysis
Total mRNA extraction was performed with the NucleoSpin RNA II Kit (Machery-Nagel, D€ uren, Germany). cDNA was acquired using reverse transcription by AMV Reverse Transcriptase Kit (Promega, Madison, WI, USA). qPCR was performed with the Illumina Eco Personal qPCR System (Westburg, Leusden, The Netherlands). Cycle parameters (30 sec. each): denaturation at 94°C, annealing at 60°C and extension at 72°C. Target genes were normalized to the geometric mean of reference genes GAPDH, SDHA and YWHAZ [29]. Primer sequences are listed in Table 1.

TOP flash assay
TOPflash-or FOPflash-transfected cells were subjected to stimulation in serum-free DMEM for 16 hrs, and luciferase activity was assayed via the Dual Reporter luciferase assay system (Promega).

Immunoblotting
Cells were lysed in ice-cold RIPA buffer supplemented with phosphatase and protease inhibitors (aprotinin, leupeptin, pepstatin A, Na 3 VO 4 , NaF, b-glycerophosphate). Equal amounts of protein were subjected to electrophoresis on polyacrylamide gels and transferred to nitrocellulose membranes. Membranes were blocked with 5% w/v BSA or 5% w/v milk in tris-buffered saline with 0.1% v/v Tween-20. Protein expression was determined by specific primary and horseradish peroxidase (HRP)-conjugated secondary IgGs. Antibodies used are listed in Table 2. Protein expression was visualized by ECL in the Syngene G:BOX HR iChemi gel documentation system (Syngene, Cambridge, UK). Band was quantified by densitometry using ImageJ software.

Study population and samples
Formalin-fixed paraffin-embedded tumour samples from 31 neuroblastomas were analysed. All samples were collected at our own hospital between 1984 and 2012 according to the Research Code of the University Medical Center Groningen (https://www.umcg.nl/en/research/ researchers/general/researchcode/pages/default.aspx) and national ethical and professional guidelines (Code of Conduct; Dutch Federation of Biomedical Scientific Societies; http://www.federa.org). Tumour specimens were subdivided based on risk group assessment in low risk, medium risk and high risk according to Children's Oncology Group risk groups (Table 3).

Fluorescence in situ hybridization
Fluorescence in situ hybridization was performed to exclude MYCNamplified tumours from the study population. Tissue sections were hybridized with Vysis LSI N-MYC (2p24) SpectrumGreen/CEP2 Spec-trumOrange Probe (Abbott Molecular, Green Oaks, IL, USA). The LSI N-MYC (2p24) probe hybridizes to the 2p24 region and contains sequences that flank both 5 0 and 3 0 ends of the MYCN gene. The CEP2 (2p11.1-q11.1) probe hybridizes to alpha satellite sequences specific to chromosome 2. Images were captured using a Leica DMRA2 Fluorescence microscope (Leica Microsystems) equipped with a Leica DC350F camera. The number of spots in 20 adjacent tumour cells was counted

Immunohistochemistry
For determination of b-catenin protein expression, deparaffinized tumour sections were incubated with anti-b-catenin IgG (1:25 in 1% v/v goat serum, 1% v/v human serum) overnight and subsequent with HRPconjugated rabbit antimouse IgG (1:1000; DAKO, Glostrup, Denmark). Staining was performed with 0.5 mg/ml 3,3 0 -diaminobenzidine in 30 mM imidazole containing 0.03% v/v H 2 O 2 and 1 mM EDTA, pH 7.0. As a negative control, the primary antibodies were omitted. For each sample, expression was quantified by measuring the average staining intensity on a scale of 256 channels using ImageJ software in four different random selected regions within the tumour (using haematoxylin and eosin-stained sections). Differences between the groups were tested for significance using the Kruskal-Wallis test. All procedures were performed blind.

Statistics
Data represent means AE SEM, from n separate experiments. Normality and equal variance were evaluated by Shapiro-Wilk test and f-test. Statistical significance of differences was evaluated by Student's t-test and one-way or two-way ANOVA followed by a Tukey multiple comparison test. Non-parametric data were evaluated with the Kruskal-Wallis test followed by Dunn's multiple comparison test. Differences were considered to be statistically significant when P < 0.05.

Results cAMP and PGE 2 enhance neuroblastoma cell viability
To study the effects of cAMP elevation on MYCN non-amplified neuroblastoma cell viability, SK-N-AS cells were incubated for the indicated periods of time with the direct adenylyl cyclase activator forskolin or PGE 2 (Fig. 1A). Forskolin or PGE 2 enhanced time-dependently cellular viability compared to control treated cells. To determine that PGE 2 could act via elevation of cAMP, we measured intracellular cAMP levels. Cells show increase in cAMP levels after incubation with forskolin compared to basal. Similar, 16,16-dimethyl-PGE 2 (in this paper referred to as PGE 2 ) induces cAMP elevation (Fig. 1B).
In addition, expression of cyclin D1, a cell cycle regulatory protein involved in G1/S-phase transition, is increased by both forskolin and PGE 2 (Fig. 1C).
Prostaglandin E 2 mediates its effects through activation of four subtypes of E-type prostanoid receptors, termed EP1-4. Of these receptors, EP2 and EP4 are known to be Ga s -coupled and enhance intracellular cAMP. SK-N-AS neuroblastoma cells express all four receptor subtypes (Fig. 1D). To study which receptor is involved in enhancement of cell viability by PGE 2 , the EP2 and EP4 receptor were antagonized using AH6809 and L-161,982 respectively. AH6809 did not affect PGE 2 -enhanced cell viability. However, L-161,982 completely prevented enhancement of cell viability by PGE 2 (Fig. 1E), demonstrating that the effects of PGE 2 are mediated through the Ga scoupled EP4 receptor.

Cyclooxygenase-2 inhibition attenuates neuroblastoma cell viability
Prostaglandin E 2 is produced by a multistep enzymatic process from arachidonic acid, in which the rate-limiting step is mediated by COX. Expression of COX-2 is increased in many cancers, including neuroblastoma, resulting in high PGE 2 production [31]. As PGE 2 enhances cell cycle progression, inhibition of PGE 2 production by COX-2 inhibition should attenuate cell proliferation. In overnight culture, neuroblastoma cells produce PGE 2 (150 pg/ml) and COX-2 inhibition with niflumic acid decreases PGE 2 (6 pg/ml). Importantly, niflumic acid decreases dose-dependently viability of SK-N-AS cells ( Fig. 2A). To confirm that the niflumic acid data were the result of COX-2 inhibition, another COX-2 inhibitor was used, celecoxib. Similar to niflumic acid, celecoxib treatment resulted in a marked decrease in cell viability (Fig. 2B). To visualize decreased cell survival over a longer time, cells were cultured for 14 days in a colony formation assay in the presence or absence of either niflumic acid (Fig. 2C) or celecoxib (Fig. 2D). Confluency was markedly decreased in the presence of specific COX-2 inhibitors which was more pronounced when higher concentrations of COX-2 inhibitor were used.

Cyclooxygenase-2 inhibition blocks cell cycle progression and activates an apoptotic response
To clarify mechanisms underlying the reduction in cell survival by COX-2 inhibition, we studied cell cycle progression and apoptotic pathway activation. Cell cycle progression was first studied by BrdU incorporation followed by fluorescence-activated cell sorting using cytometry (Fig. 3A). PI was used to measure the overall DNA content of cells. Niflumic acid reduced the amount of BrdU-positive cells and cells with a ploidy between 2n (cells in G0/G1) and 4n (cells in G2/M), indicating a lower fraction of cells in S-phase. In contrast, induction of apoptosis was studied by sorting cells based on membrane expression of annexin V (Fig. 3B). Viable cells are located in the lower left quadrant (double negative), early apoptotic cells are located in the lower right quadrant (annexin V positive, PI negative) and late apoptotic cells are located in the upper right quadrant (double positive). Niflumic acid treatment exerts a higher percentage of cells in late apoptosis, while the percentage of cells in early apoptosis was not affected. To confirm apoptotic pathway activation, we measured aggregation of the JC-1 dye at mitochondrial membranes. A distinctive feature of the early stages apoptosis is the disruption of active mitochondria, which includes changes in the membrane potential. In healthy, polarized membranes, the JC-1 dye spontaneously forms aggregates with red fluorescence (EM 590 nm), while in apoptotic cells, the JC-1 dye remains in its monomeric form (EM 529 nm). Inhibition of COX-2 with niflumic acid results in attenuated mitochondrial membrane polarization (Fig. 3C). Valinomycin served as a positive control.
Amplification and thereby increased expression of N-Myc is the best known marker for neuroblastoma risk profile [1][2][3]. N-Myc is a positive regulator of cell cycle progression and a negative regula-tor of programmed cell death. COX-2 inhibition with niflumic acid decreased expression of N-Myc mRNA (Fig. 3D). To further distinguish between inhibition of cell cycle progression and induction of apoptotic signalling, we looked for expression of cyclin D1. Niflumic acid attenuated expression of cyclin D1, which is consistent with the cell cycle analysis shown in Figure 3A. Finally, niflumic acid-induced cleavage of the apoptosis linked Poly ADP ribose polymerase (PARP) (Fig. 3E).

Rescue of cyclooxygenase-2-dependent cell survival by exogenous prostaglandin E 2
To confirm that the decreased viability of neuroblastoma cells in response to COX-2 inhibition (Fig. 2)   levels, cells were incubated with niflumic acid absence or presence of exogenous PGE 2 . We used a concentration of niflumic acid that results in AE75% viability compared to control-treated cells. Reduced viability resulting from COX-2 inhibition by niflumic acid was completely restored by exogenous PGE 2 in a dose-dependent manner (Fig. 4A). In addition, niflumic acid treatment resulted in a marked decrease in cAMP production, which was rescued by simultaneous addition of exogenous PGE 2 (Fig. 4B).
cAMP enhances b-catenin stability and activates b-catenin target gene transcription b-catenin stability and transcriptional activity are known to determine the balance between cell proliferation and differentiation and are involved in the progression of several cancers [32]. Because several recent reports suggest that PGE 2 can enhance b-catenin stability and transcriptional activity [22,[33][34][35][36], we aimed to identify the effects of forskolin and PGE 2 on b-catenin stability and transcriptional activity. In response to cAMP elevation, PKA is activated. PKA can phosphorylate b-catenin and GSK3b (see introduction). b-catenin not phosphorylated by GSK3b is stable and translocates to the nucleus. In this   Neuroblastoma cells stimulated with forskolin or PGE 2 were analysed for protein levels of phosphorylated b-catenin and GSK3b. Forskolin and PGE 2 induced phosphorylation of b-catenin (ser675), p-GSK3b (ser9) and concomitant higher levels of active, unphosphorylated, b-catenin (Fig. 5A). It has been reported that phosphorylation of b-catenin by PKA (ser675) enhances b-catenin stability and its nuclear translocation [24,37,38]. b-Catenin localization was determined by analysis of cytosolic (C) and nuclear fractions (N) (Fig. 5B). Phosphorylation of b-catenin (ser675) was increased in both nuclear and cytosolic fractions by cAMP and PGE 2 . Levels of active b-catenin were increased in nuclear fractions by cAMP and   To visualize localization of p-b-catenin (ser675), neuroblastoma cells were stimulated with forskolin or PGE 2 after which p-b-catenin (ser675) was visualized by immunofluorescence (Fig. 5C). Both forskolin and PGE 2 transiently increased levels of p-b-catenin (ser675) and induced accumulation of p-b-catenin (ser675) in (peri)nuclear regions (white arrows, Fig. 5C), which was followed by a rapid decline. To assess b-catenin/TCF target gene transcription in response to forskolin or PGE 2 , neuroblastoma cells were transfected with a TCF luciferase reporter gene construct (TOPflash) containing multiple TCF-binding sites or a construct with mutated TCF-binding sites (FOPflash). Either forskolin or PGE 2 induced TCF reporter gene expression, as measured by luciferase activity in cell lysates (Fig. 5D).

b-Catenin is involved in neuroblastoma cell survival
To study if the effects of forskolin and PGE 2 on cell survival are mediated via b-catenin-dependent gene transcription in a non-Wnt liganddependent manner, neuroblastoma cells were transfected with the stable active mutant of b-catenin (b-catenin S33Y ). Transfection with b-catenin S33Y increased protein levels of b-catenin (Fig. 6A). Functionally, b-catenin S33Y increased TCF reporter gene activity (Fig. 6B). In addition, active b-catenin could enhance cellular viability and cell cycle progression. Overexpression of b-catenin S33Y enhanced cell viability compared to cells expressing the empty vector (Fig. 6D).
Conversely, we addressed whether b-catenin is required for PGE 2induced neuroblastoma cell viability. To study this, we lowered active (non-phosphorylated) b-catenin by enhancing the stability of axin2, the limiting factor of the APC/GSK3b/CK1 destruction complex that targets b-catenin for proteasomal degradation, with the tankyrase inhibitor XAV939 [39]. Levels of active b-catenin were reduced by XAV939 treatment, while total levels of b-catenin remained unchanged (Fig. 6C). Importantly, while XAV939 prevented PGE 2induced cell viability, it did not affect neuroblastoma cell viability at a basal level (Fig. 6E). Thus, our data indicate that the effects of PGE 2 on neuroblastoma involve active b-catenin.

SK-N-SH human neuroblastoma cells
Cancer cell lines are often characterized by cell line-specific (genetic) abrogations that do not accurately reflect the behaviour of other cancer cell lines originating from the same disease. We therefore repeated key experiments in a different human neuroblastoma cell line without MYCN amplification, SK-N-SH. Similar to SK-N-AS, forskolin and PGE 2 enhanced SK-N-SH neuroblastoma cell viability, while antagonism of the EP4 receptor with L-161,982 prevented PGE 2 -enhanced cell viability which was not affected by antagonism of the EP2 receptor using AH6809 (Fig. 7A and data not shown). In addition, overexpression of b-catenin S33Y enhanced cell viability (Figs 6D and 7B). Specific COX-2 inhibition with niflumic acid or celecoxib (data not shown) attenuated cell viability in a dose-dependent manner (Figs 2A and 7C). SK-N-SH cells were more sensitive towards COX-2 inhibition compared to SK-N-AS. The PGE 2 -induced enhanced cell viability was prevented by attenuating active b-catenin with the tankyrase inhibitor XAV939 (Figs 6E and 7D). Thus, SK-N-SH and SK-N-AS human neuroblastoma cell lines behave similarly in response to manipulation of PGE 2 or b-catenin signalling properties.
High-risk neuroblastoma without amplification of MYCN exhibit high b-catenin expression Finally, we examined a potential correlation between b-catenin expression in neuroblastoma tumour sections with neuroblastoma risk profile. We investigated tumour specimens that were collected and stored for diagnosis for expression of b-catenin protein by immunohistochemistry. Characteristics of the investigated risk group populations are listed in Table 4. We were only interested in cases without MYCN amplification as it has previously been reported that b-catenin expression is increased in neuroblastoma tumours and cell lines without amplification of MYCN [16]. Fluorescence in situ hybridization was performed to identify amplification status of neuroblastoma specimens. We regarded tumour specimens that scored positive for MYCN amplification as a separate group of tumours. b-Catenin expression was determined by scoring the average staining intensity of the tumour. b-Catenin was found predominantly at the periphery of the cell body and within the neuropil, areas of the tumour that are comprised of non-myelinated axons, dendrites and synaptic dense regions. We did not observe bcatenin-positive nuclei. When high-risk tumours with MYCN amplification were compared to high-risk tumours without MYCN amplification, higher levels of b-catenin were found in high-risk tumours without MYCN amplification. Importantly, we observed higher levels of b-catenin in tumour specimens from high-risk tumours compared to expression in low-and medium-risk cases (P = 0.042) (Fig. 8C).

Discussion
The development of neural crest-derived tissues is under tight control of molecular mechanisms that regulate proliferation, differentiation and programmed cell death. When these core molecular pathways become deregulated, malignancies such as neuroblastoma can develop. The most extensively defined factor contributing to neuroblastoma pathogenesis is amplification of the Myc oncoprotein family member MYCN. Amplification of MYCN is only observed in about one-fourth of neuroblastomas [1,3]. In this study, we focused on the contribution of PGE 2 and b-catenin to progression of neuroblastomas without MYCN amplification. Our findings in MYCN non-amplified neuroblastoma cell lines show that PGE 2 enhances cell viability through the EP4 receptor, whereas this was attenuated by inhibition of PGE 2 with specific COX-2 inhibitors. Elevating cAMP using forskolin mimicked the effects of PGE 2 on cell viability. PGE 2 and cAMP promoted GSK3b inhibition, phosphorylation of b-catenin at the PKA target residue ser675, b-catenin nuclear translocation and TCF-dependent gene transcription. Furthermore, we show that expression of a degradation-resistant b-catenin mutant enhances neuroblastoma cell viability and that inhibition of b-catenin with XAV939 prevents PGE 2 -induced viability. b-Catenin expression in tumour sections was increased in high-risk neuroblastoma without MYCN amplification compared to expression in low-or medium-risk sections. These data suggest that autocrine PGE 2 enhances neuroblastoma cell viability via a mechanism involving cAMP-mediated b-catenin stabilization, and suggest that this pathway is of relevance to high-risk neuroblastoma without MYCN amplification.

221
Increased PGE 2 levels and concomitant up-regulation of COX-2 are frequently observed in malignant tissues of epithelial origin [5,40]. Studies have revealed that inhibition of COX enzymes have potential in anticancer therapy [40]. Regarding neuroblastoma, studies indicate that COX-2 and PGE 2 are involved in progression of neuroblastoma [4,5], although it is largely unknown which downstream molecular pathways are involved. PGE 2 functions by binding to its membrane bound E-type prostanoid receptors which are all expressed in neuroblastoma cells. Of the four receptor subtypes, EP2 and EP4 are Ga s -coupled and stimulate adenylyl cyclase activation and thus intracellular cAMP levels [41]. Indeed, we found that PGE 2 incubation elevated cellular cAMP levels in neuroblastoma. Conversely, inhibiting endogenous PGE 2 synthesis with COX-2 inhibitors decreased cAMP levels. The rise in cAMP levels, either through receptor-mediated adenylyl cyclase activation by PGE 2 or via direct activation of adenylyl cyclase by forskolin, increased cell viability of neuroblastoma cells. Further, we demonstrate that enhancement of cell viability by PGE 2 is mediated through the EP4 receptor as antagonism of the EP4 receptor using L-161,982 completely abolished the enhancement of cell viability by PGE 2 , whereas antagonism of the EP2 receptor using AH6809 did not affect this. In earlier studies, Rasmuson et al. demonstrated that exogenous PGE 2 enhanced viability of MYCN-amplified SK-N-BE(2) neuroblastoma cells, suggesting that PGE 2 is involved in neuroblastoma cell viability independent of amplification of MYCN [4]. Inhibition of COX enzymes has been shown to inhibit neuroblastoma cell survival in vitro [7,42] and to inhibit tumour growth in vivo [5]. Indeed, COX-2 inhibitors attenuated cell viability in neuroblastoma cells. In other studies, inhibition of COX-2 by celecoxib or diclofenac changed mitochondrial membrane potential and subsequent activation of caspase-9-and caspase-3-dependent apoptosis [7,43], while no activation of caspase-8 or BID cleavage were observed, suggesting involvement of the intrinsic apoptotic pathway. In this study, we observed altered mitochondrial membrane potential after treatment with COX-2 inhibitors. In addition, we found that COX-2 inhibition induces apoptosis in neuroblastoma cells. Furthermore, we observed attenuated cell cycle progression after COX-2 inhibition resulting from a cell cycle block between G1 and S-phase. This cell cycle block was further confirmed by decreased expression of cell cycle regulatory  protein cyclin D1. Although it seems evident that the effects of COX-2 inhibition are mediated by decreased PGE 2 production, there are several reports that indicate that accumulation of arachidonic acid, the precursor for PGE 2 synthesis, is also involved. For instance, combined treatment with arachidonic acid and COX inhibitors results in a synergistic effect on cell survival [44]. However in this study, we show that COX-2 inhibition with niflumic acid or celecoxib complemented with exogenous PGE 2 , rescues neuroblastoma cells from decreased viability as a result of COX-2 inhibition. This is also reflected on intracellular cAMP levels, which are decreased by niflumic acid, but restored by simultaneous addition of exogenous PGE 2 .

A B
Of particular interest are emerging research lines that focus on molecular interactions between PGE 2 and b-catenin. In 2002, Fujino et al. showed that PGE 2 stimulates TCF/Lef promoter activity through EP2 and EP4 receptor signalling [33]. Downstream of EP2 and EP4, an increase in cAMP levels activates PKA. Phosphorylation of GSK3b at ser9 by PKA inhibits its kinase activity and thereby enhances b-catenin stability, nuclear translocation and TCF/Lef-dependent gene transcription [23,32,33]. It was, however, unknown if this molecular interaction occurs in neuroblastoma. Stimulation of neuroblastoma cells with forskolin or PGE 2 increased p-GSK3b (ser9) and elevated levels of unphosphorylated b-catenin at the GSK3b residues. In addition, we found increased nuclear localization of active b-catenin. These findings are in agreement with studies in different malignant cell models. In nonsmall cell lung cancer and colorectal cancer, PGE 2 enhances b-catenin nuclear localization and inhibits GSK3b by phosphorylation which is dependent on cAMP and kinase activity of PKA [34,36]. Apart from inhibiting GSK3b, PKA has been shown to directly activate b-catenin by phosphorylation at residue ser675 [24,37,38]. Indeed, we observed increased p-b-catenin (ser675) in response to PGE 2 or forskolin. Subsequently, we observed a transient increase in p-b-catenin (ser675) at the (peri)nuclear regions. Importantly, we also found increased TCF-dependent gene transcription, which confirms earlier studies in HEK cells in which PKA-dependent p-b-catenin (ser675) resulted in increased TCF luciferase activity [24,37]. More recently, increased TCF-dependent gene transcription downstream of PGE 2 and subsequent PKA activity was found in in vitro and in vivo models for colorectal cancer [23,34]. Conversely, inhibition of PGE 2 synthesis using the COX-2 inhibitor celecoxib has been found to attenuate b-catenin stability, nuclear translocation and TCF luciferase activity in colorectal cancer [34,35], non-small cell lung cancer [36], and osteosarcoma [45], with concomitant lower expression of cyclin D1 and cell survival.
The contribution of b-catenin and TCF-dependent gene transcription in colorectal carcinoma has been well-established. In neuroblastoma, it has been shown that b-catenin expression is increased in MYCN non-amplified neuroblastoma cell lines [16]. The result of b-catenin transcriptional activity is expression of mitogenic and survival genes. In this study, we demonstrate increased TCF-dependent gene transcription and enhanced cellular viability in neuroblastoma cells transfected with a b-catenin mutant that is untargetable for proteasomal degradation (b-catenin S33Y ). Importantly, the enhanced cell viability resulting from PGE 2 was completely prevented by the tankyrase inhibitor XAV939. XAV939 prevents tankyrase-mediated degradation of axin2, thereby stabilizing axin2 which is the concentration-limiting component of the destruction complex that sequesters b-catenin and GSK3b. Thereby, XAV939 functions as an inhibitor of b-catenin activity [39]. Thus, b-catenin is, at least to some extent, responsible for PGE 2 -induced enhanced cellular viability. Recently, XAV939 has been shown to decrease neuroblastoma cell survival by inhibiting b-catenin transcriptional activity and decreasing expression of cyclin D1 in both MYCN-amplified and non-amplified cell lines [46]. Moreover, knockdown of b-catenin using RNA interference has been shown to inhibit cell cycle progression and induce apoptosis in neuroblastoma cells [47]. This, together with our data indicates that b-catenin plays a crucial role in neuroblastoma cell survival.
Interestingly, investigation of b-catenin expression in neuroblastoma tumour sections without MYCN amplification revealed particularly high expression in high-risk tumours. This further strengthens our hypothesis that b-catenin plays an important role in neuroblastoma, not only in vitro cell models but also in a clinically more relevant context. Other recent studies have also found increased expression of b-catenin in neuroblastoma, which was involved in maintaining a neuroblast-like phenotype of neuroblastoma cells and conferring resistance to chemotherapeutic agents [48,49]. However, these studies did not distinguish between MYCN-amplified and non-amplified tumours. We are aware that based on our study, we cannot draw strong conclusions on the functional role of increased b-catenin in high-risk neuroblastoma tumours as expression of b-catenin was found predominantly at the cell periphery and the neuropil, where it exerts a different function than in the nucleus. However, during epithelial-mesenchymal transition, which is an early step in the metastatic cascade in cancer [50], b-catenin is liberated from the junctional complexes where it is sequestered. In such an event in the presence of a b-catenin stabilizing stimulus such as PGE 2 , b-catenin translocates to the nucleus and activates TCF/Lef-dependent transcriptional programmes, thereby enhancing survival and metastatic potential [51].
The promising therapeutic potential of COX-2 inhibitors has resulted in multiple compounds that have reached the clinical market. These compounds were, however, rejected as a consequence of concerns regarding drug safety [40]. Recent insight in the systemic effects of COX-2 inhibition and its metabolites, questions the notion that the safety issues are the result of COX-2 inhibition and not COX-1 [52,53]. We believe that with better insights in the molecular events that occur downstream of COX inhibition, more specific chemopreventive compounds might be developed that exploit the therapeutic potential of specific COX-2 inhibition, but circumvent the safety issues.
Taken together, our results demonstrate that PGE 2 is an important factor being involved in the regulation of neuroblastoma cell viability. cAMP activates PKA and thereby stabilizes b-catenin either directly or indirectly by inhibiting GSK3b, resulting in increased b-catenin activity and neuroblastoma cell viability. The increased expression of b-catenin in high-risk neuroblastoma, without MYCN amplification, suggests that b-catenin could play a major role in neuroblastoma. Thus, strategies that target either PGE 2 production by the tumour (COX-2 inhibitors) or target the interaction between PGE 2 and b-catenin warrant further investigation.