Cell Survival Signalling through PPARδ and Arachidonic Acid Metabolites in Neuroblastoma

Retinoic acid (RA) has paradoxical effects on cancer cells: promoting cell death, differentiation and cell cycle arrest, or cell survival and proliferation. Arachidonic acid (AA) release occurs in response to RA treatment and, therefore, AA and its downstream metabolites may be involved in cell survival signalling. To test this, we inhibited phospholipase A2-mediated AA release, cyclooxygenases and lipoxygenases with small-molecule inhibitors to determine if this would sensitise cells to cell death after RA treatment. The data suggest that, in response to RA, phospholipase A2-mediated release of AA and subsequent metabolism by lipoxygenases is important for cell survival. Evidence from gene expression reporter assays and PPARδ knockdown suggests that lipoxygenase metabolites activate PPARδ. The involvement of PPARδ in cell survival is supported by results of experiments with the PPARδ inhibitor GSK0660 and siRNA-mediated knockdown. Quantitative reverse transcriptase PCR studies demonstrated that inhibition of 5-lipoxygenase after RA treatment resulted in a strong up-regulation of mRNA for PPARδ2, a putative inhibitory PPARδ isoform. Over-expression of PPARδ2 using a tetracycline-inducible system in neuroblastoma cells reduced proliferation and induced cell death. These data provide evidence linking lipoxygenases and PPARδ in a cell survival-signalling mechanism and suggest new drug-development targets for malignant and hyper-proliferative diseases.


Introduction
Retinoic acid (RA) is a biologically-active vitamin A metabolite used in the treatment of neuroblastoma and acute promyelocytic leukaemia [1]. In vitro, RA induces growth arrest, down-regulation of MYCN expression [2] and differentiation in neuroblastoma cells [3]. Paradoxically, RA can promote increased proliferation and cell survival in certain cell types [4], [5]. Like other anticancer agents such as cisplatin and tamoxifen, RA induces arachidonic acid (AA) release in cancer cells [6][7][8][9], and this may promote cell survival under conditions of cell stress. Furthermore, celecoxib, a non-steroidal anti-inflammatory drug and cyclooxygenase (COX2) inhibitor which inhibits the metabolism of AA, potentiates the effects of both RA and cytotoxic drugs in neuroblastoma cells [10][11][12].
RA has been reported to activate Peroxisome Proliferator-Activated Receptor (PPAR) d, a ligand-activated transcription factor controlling cell growth and proliferation and important for cell survival [13]. RA is thought to be transported into the nucleus by cellular retinoic acid binding proteins (CRABP) or fatty acid binding protein 5 (FABP5) and it has been proposed that CRABP2 mediates RA transfer to RA receptors (RAR) to promote differentiation or apoptosis, whereas FABP5 mediates RA transfer to PPARd heterodimers promoting cell survival [14]. Evidence for the direct activation of PPARd by RA is controversial, with later studies suggesting that RA does not directly bind to PPARd or activate PPAR target genes [15][16][17]. Nevertheless, there may well be interactions between RAR and PPARd signalling pathways in development; for example, it has recently been suggested that neural differentiation is regulated by an RAR-mediated commitment phase followed by the promotion of differentiation via a PPARd-mediated up-regulation of PDK1 [18]. The role of PPARd in cell signalling is likely to be complex; five different mRNA isoforms of PPARd have been described, with PPARd1 and PPARd2 being the most abundantly expressed in human tissues; although PPARd2 has been suggested to represent an inhibitory isoform, a translational product has yet to be identified [18].
Given the activity of celecoxib in inducing cell death in combination with RA, it is possible that AA metabolites are important in promoting cell survival and may interact with RARand/or PPARd-mediated signalling. To test this hypothesis and elucidate the mechanism of interaction between RA and celecoxib, we investigated the effect of inhibiting AA release, cyclooxygenases and lipooxygenases on the survival of neuroblastoma cells after RA treatment. The data suggest that 5-lipoxygenase (5-LO) inhibition sensitises neuroblastoma cells to apoptosis and that celecoxib promotes RA-induced neuroblastoma cell death in vitro through the inhibition of 5-LO. Further experiments to clarify the potential role of 5-LO suggest that the 5-LO product 5-oxo-eicosatetraenoic acid (5-oxo-ETE) mediates cell survival through PPARd.

Flow Cytometry
Adherent and non-adherent cells were fixed in 70% ethanol, 30% PBS, then spun down and resuspended in PBS containing 40 mg/ml propidium iodide and 100 mg/ml RNAse A. Fluorescence from excitation at 488 nm with a 15-mW argon laser was monitored at 585621 nm using a FACScan flow cytometer with CellQuest software (Becton Dickinson, San Jose, CA). Events were counted using a doublet-discriminator parameter to exclude aggregates and a threshold of forward and side scatter was used to exclude debris. Cell-cycle and cell-death analysis was carried out on the same samples using WinMDI (version 2.8, TSRI, La Jolla, CA).

PPARd siRNA
10 nM PPARd SMART-pool siRNA (Dharmacon, Lafayette, CO) was complexed with lipofectamine in OptiMEM (Invitrogen), according to the manufacturer's instructions. Control siRNAs were a universal scrambled negative control (SCR) (Qiagen, Ontario, Canada). Cells were transfected with siRNA for 24 h prior to treatment with ATRA. Levels of apoptosis were higher in SH-SY5Y cells treated with siRNA (SCR or PPARd), therefore, for assessing the effects of knockdown on ATRA-induced apoptosis the data were normalised to the zero-ATRA control for each siRNA.

PPARd Reporter Assay
Cells were transfected with 667 ng of PPAR response element (PPRE) construct, negative or positive control plasmids from the PPRE cignal dual luciferase reporter assay kit (SAbiosciences, Frederick, MD) using lipofectamine. 24 h after transfection, cells were treated with drug combinations for 2 h. Cells were harvested and transferred to a 96-well plate; 50 ml of Stop and Glow solution (Promega, Madison, WI) was added to each well and luminescence measured on a LB96V Microplate Luminometer (EG & G Berthold, Bad Wildbad, Germany).

Establishment of a Stable Inducible PPARd2 Overexpression System in SH-SY5Y Cells
Human PPARd, mRNA transcript variant 2 cDNA in pOTB7 (IOH5357, Invitrogen) was subcloned into BamH1 and Xho1 sites of pcDNA4/TO (Invitrogen). The construct was transfected into SH-SY5Y tet12 cells expressing the TET repressor [22]. After  Table 1 with additional viability data for NGP and NB89 cells in Table S1. (C) Western blot for cleaved caspase-3, 6 h after treatment with ATRA or ATRA with 22 mM celecoxib, size markers from ladder are shown; the full blot is shown in Figure S3. doi:10.1371/journal.pone.0068859.g001 selection using zeocin (250 mg/ml) and blasticidin (5 mg/ml), a mixed population of cells was used for experiments (SH-SY5Y tet12 PPARd2). Addition of 1 mg/ml doxycycline to the medium induced PPARd2 expression (referred to as PPARd2+ve). Cells grown without doxycycline are referred to as PPARd2-ve. To confirm results, transient transfections were carried out using the pcDNA4/TO PPARd2 construct and native SH-SY5Y cells.

RNA Extraction and Quantitative Reverse Transcriptase PCR for PPARd
RNA was extracted using the RNeasy mini kit (Qiagen) and DNAse-treated using the DNA-free kit (Qiagen). 1 mg of RNA was reverse transcribed using a high-capacity reverse transcription kit. Levels of PPARd1 and PPARd2 were measured using Taqman assays (PPARd1: Hs00606407_m1 and PPARd2 Hs02516538_s1) and normalised to endogenous GAPDH (Hs99999905_m1). Quantitative reverse transcriptase (qRT-PCR) was carried out using the relative quantification protocol on an ABI Prism 7900HT and analysed using SDS software (Applied Biosystems, Foster City, CA).

XTT Viability and Cell Growth Assays
The XTT assay (Roche Diagnostics, Basel, Switzerland) was used as previously described [23]. The experimental conditions were based on standard calibration curves to establish the appropriate initial plating density. For viability dose-response curves, cells were seeded at 2 x10 4 per well into 96-well microtiter plates (Corning, Amsterdam, Netherlands) and left to adhere overnight; inhibitors and ATRA were diluted in OptiMEM to the appropriate concentrations. For growth curves, cells were plated at 5610 3 per well, and XTT absorbance was measured on 3 consecutive days. Treated cells were incubated for 24 h before XTT solution was added; plates were read at 480 nm 4 h after XTT addition.

Western Blotting
Cells were lysed in buffer containing protease inhibitors (Roche Diagnostic, Mannheim, Germany). Protein concentrations were measured using Bradford reagent (Pierce, Rockford, IL) and equal quantities separated by SDS-PAGE, transferred to nylon membranes (Bio-Rad, Hertfordshire, UK) and probed with antibodies against COX2 (Santa Cruz Biotechnology Inc, Santa Cruz, CA), cleaved caspase-3 (Cell Signalling Technologies, Danvers, MA) and b-actin (Sigma-Aldrich). Anti-mouse IgG conjugated with horseradish-peroxidase (Upstate, Billerica, MA, USA) was used as secondary antibody and enhanced chemiluminescence Western blotting detection reagent (GE Healthcare, Buckinghamshire, UK) was used for detection.

Statistics
Graphs were plotted using Sigmaplot 11.0 (San Jose, CA). Doseresponse curves were fitted using the non-linear mixed-effects model fit in the nlme package [24] in R [25]. For other experiments, univariate ANOVA (with contrasts) was performed in SPSS Version 17 (Chicago, IL), and t-tests (unpaired) and linear mixed-effects models (nlme package [24]) were performed using R. Where relevant, combination indices (CI) were calculated using Calcusyn (Biosoft, Cambridge, UK). CI.1 indicates an antagonistic interaction between two drugs and a CI,1 indicates synergy. All errors are expressed as standard error of the mean (SEM). For clinical data, Wilcoxon test was used to assess differential expression in MYCN-amplified (all stages) versus non-MYCN-amplified (all stages). For Kaplan-Meier survival analysis the best cut off between low and high PPARd1 expression was determined as the 1st quartile. The log-rank test was used to test whether the two survival curves are different.

Inhibition of AA Metabolism Promotes RA-mediated Cell Death
Several studies have shown that RA promotes AA release from nuclear membranes [6], [7]. To determine if AA release is important in survival of cells treated with all-trans RA (ATRA), we inhibited phospholipase A2 (PLA2) using 10 mM AACOCF3 prior to ATRA treatment ( Figure 1A & Table 1); this concentration of AACOCF3 has been used previously to inhibit AA release in neuroblastoma cell lines [26]. Flow-cytometry was used to estimate the proportion of cells within the sub G 1 peak as a measure of cell death in SH-SY5Y cells; the results were confirmed by XTT viability assays in SH-SY5Y, NGP and NB69 cells. The data showed that AACOCF3 facilitated ATRA-dose-dependent cell death in all three neuroblastoma cell lines ( Figure 1A; Table 1 and  Table S1). These results suggest that AA is critical for cell survival after ATRA treatment. AA is a substrate for cycloxygenases and lipoxygenases and is metabolised to prostaglandins or lipoxins, leukotrienes and hydroxy-eicosatetraenoic acids [27], [28]. Therefore, we used small-molecule inhibitors to target cyclooxygenases and lipoxygenases in combination with ATRA. Firstly, we treated neuroblastoma cells with a sub-lethal dose (22 mM) of celecoxib; in combination with ATRA, celecoxib synergistically decreased cell viability in all three cell lines (combination index [CI] = 0.119 for SH-SY5Y cells with 15 mM ATRA; Table 1 & Table S1). Apoptosis data revealed that celecoxib in combination with ATRA increased apoptosis in a dose-dependent manner ( Figure 1B). Elevated levels of apoptosis after 24 h in cells treated with ATRA and celecoxib were confirmed by substantially-increased caspase-3 cleavage relative to ATRA alone after 6 h of treatment ( Figure 1C). Although cleaved caspase-3 increased in SH-SY5Y cells after treatment with 10 mM or 15 mM of ATRA for 6 h, at 10 mM ATRA this was not linked to a detectable increase in apoptosis at 24 h. Thus, caspase-3 cleavage occurred to a greater extent and with lower concentrations of ATRA in the combination treatment, and corresponded to significantly higher apoptosis at 24 h.

5-LO Promotes Cell Survival After RA Treatment
As celecoxib is well known as a COX2 inhibitor, we investigated the role of AA metabolism by COX2 in promoting cell survival after ATRA treatment and concluded that COX2 was not the main cell-survival mediator of ATRA in neuroblastoma cells (Text S1, Figure S1 and Figure S4). Celecoxib also inhibits 5-LO [29] and 5-LO is expressed in neuroblastoma cell lines [26] and tumours [30]. To determine if 5-LO inhibition promotes synergy between celecoxib and ATRA we tested the combination of the 5-LO inhibitor MK886 and ATRA. To become catalytically active, 5-LO binds the 5-LO activating protein FLAP; MK886 also binds FLAP and as a consequence inhibits 5-LO activity [31]. In SH-SY5Y cells, 1 mM MK886 is a sub-lethal dose [22] but apoptosis increased significantly when combined with increasing doses of ATRA ( Figure 2A). MK886 in combination with ATRA effectively reduced survival in all three neuroblastoma cell lines (Table 1 and Table S1). We assessed the contribution of AA metabolism by lipoxygenases 12-LO and 15-LO to cell-survival signalling and concluded that 5-LO inhibition was the most effective strategy for promoting cell death in combination with ATRA (Text S2, Table S2).
5-LO oxidises AA to both LTA4 and 5S-hydroperoxy-6,8,11,14-eicosatetraenoic acid; the latter is rapidly reduced and then oxidised by NADP+-dependent 5-hydroxyeicosanoid dehydrogenase (5HEDH) activity to 5-oxo-ETE [32]. If 5-LO metabolites promote cell survival after ATRA treatment, we would predict that pre-incubation with 5-oxo-ETE or LTA4 would rescue cells from ATRA and 5-LO inhibition. In experiments to test this, 5-oxo-ETE partially rescued SH-SY5Y cells from death induced by 22 mM celecoxib and ATRA or 1 mM MK886 and ATRA ( Figure 2B and C). Viability assays confirmed the apoptosis results: at the maximum dose of ATRA (15 mM), 5-  Table 2). The data show that 5-oxo ETE affected levels of cell survival at high concentrations of ATRA (the lower asymptote), and that the effects of 5-oxo-ETE were dose dependent ( Figure 2D). In contrast, 0.1 mM LTA4 did not rescue SH-SY5Y cells from celecoxib or MK886 in combination with ATRA (Table 2). We interpret these results to suggest that metabolism of AA by 5-LO promotes cell survival in neuroblastoma cells.

PPARd Activation by 5-oxo-ETE is Involved in RAmediated Cell Survival Signalling
Previous studies have suggested that PPARd mediates survival response to ATRA [13], [14]. Assays for PPRE binding activity in nuclear extracts from SH-SY5Y cells demonstrated that PPARd was 4.5-to 6-fold more abundant than PPARa or PPARc ( Figure  S2A). PPRE activation increased significantly in SH-SY5Y cells after treatment with 4 or 8 mM ATRA (2.360.4 and 1.660.01fold respectively; Figure 3A). Importantly, 0.1 mM 5-oxo-ETE increased PPRE activation to similar levels observed with 4 and 8 mM ATRA, suggesting that 5-oxo-ETE is substantially more potent than ATRA ( Figure 3A). Furthermore, treatment with MK886 reduced PPRE activation in response to ATRA ( Figure  S2B). Therefore, these results suggest that PPRE activation in response to ATRA may be an indirect effect resulting from stimulation of PPAR by AA metabolites such as 5-oxo-ETE.
PPARd siRNA knock-down in SH-SY5Y cells was used to determine if PPARd mediates increased PPRE-binding activity after ATRA treatment. Levels of PPARd1, the full-length PPARd transcript, and PPARd2 were measured 24 h after treatment with siRNA (control samples) and 2 h after subsequent treatment with 4 or 10 mM ATRA ( Figure 3B). Before ATRA treatment, PPARd1 mRNA was reduced by 58%; after treatment with ATRA, PPARd1 mRNA levels remained between 40-60% compared to the corresponding scrambled (SCR) siRNA control ( Figure 3B). The specificity of the commercially-supplied PPARd siRNA with respect to the PPARd1 and PPARd2 isoforms is unknown but there was no significant effect of the PPARd siRNA on PPARd2 transcript levels (ANOVA, scrambled versus PPAR siRNA, P = 0.46). There was no significant difference in PPRE reporter-gene activity between the PPARd and SCR siRNAtreated controls ( Figure 3C) but levels of PPRE reporter activity in cells treated with 4 or 10 mM ATRA were significantly lower in the PPARd siRNA-treated cells ( Figure 3C).

PPARd Inhibition Promotes Cell Death after ATRA Treatment
To investigate the role of PPARd in cell survival, PPARd activity was inhibited prior to ATRA treatment using two different approaches: a small-molecule PPARd inhibitor, GSK0660, and siRNA-mediated PPARd knockdown. Apoptosis was significantly increased after cells were treated with a combination of GSK0660 (0.2 mM) and ATRA compared to ATRA alone ( Figure 4A). Viability assays confirmed these results with GSK0660 on its own slightly reducing cell viability to 88.0610.6% of the controls, but substantially decreased viability in combination with increasing doses of ATRA (Table 1). Transfection of SH-SY5Y cells with PPARd or SCR siRNA decreased the overall viability of the cells, and this was reflected in the reduced viability of the controls ( Figure 4B). However, SH-SY5Y cells transfected with PPARd siRNA had reduced viability in response to ATRA, compared to control SCR siRNA-treated cells ( Figure 4B); in the control cells, viability increased in response to ATRA, as is frequently observed (see Figure S1B for example). PPARd knockdown alone had no significant effect on apoptosis, but had a small effect of increasing apoptosis in combination with ATRA doses of 8 mM or more (P = 0.014; Figure 4C).
Thus, PPARd inhibition reduced viability and increased cell death after treatment with ATRA, supporting the hypothesis that PPARd mediates survival in ATRA-treated cells. To test this hypothesis further, we pre-treated SH-SY5Y cells with a PPARd agonist (0.1 mM L165-041), predicting that it would rescue cells from death mediated by a combination of 5-LO inhibition and ATRA. In viability experiments, the PPARd agonist did indeed increase the IC 50 of celecoxib/ATRA and MK886/ATRA combinations, although overall survival at the highest concentrations of ATRA (10-15 mM) was not affected (Table 2). Of the five different PPARd isoforms described, PPARd1 and PPARd2 are the most widely expressed in different tissues; current evidence suggests that PPARd2 is an inhibitory isoform lacking the ligand binding domain, although a translation product has yet to be identified [18]. We measured the expression of both PPARd1 and PPARd2 in neuroblastoma cells ( Figure 5A) and found that 2 h after treatment with ATRA, PPARd1 expression increased significantly with ATRA concentration but there was no significant effect of MK886 or celecoxib on ATRA-induced PPARd1 levels. In contrast, celecoxib or MK886 increased PPARd2 expression in combination with ATRA ( Figure 5A). To determine the effect of PPARd2 on cell survival and apoptosis, we created a tetracycline-inducible PPARd2 system in SH-SY5Y cells (SH-SY5Y tet12 PPARd2). In the absence of doxycycline, PPARd2 expression was higher than in the parental SH-SY5Y tet12 cells but was significantly increased 24 and 48 h after treatment with doxycycline ( Figure 5B). Up-regulation of PPARd2 expression in the SH-SY5Y tet12 PPARd2 cells significantly increased apoptosis after 48 h ( Figure 5C). This was confirmed by cell viability assays over the same time scale ( Figure 5D). Doxycycline did not increase apoptosis or slow growth in the SH-SY5Y tet12 cells ( Figure 5C and D). As an additional test, we transiently transfected the empty vector (pcDNA4/TO) or the pcDNA4/TO PPARd2 construct into SH-SY5Y cells. Using this approach, PPARd2 mRNA expression after 48 h was lower than in the stable inducible system ( Figure 5E) and there was no baseline expression of PPARd2 mRNA in the pcDNA4/TO controls. As with the inducible system, cells transfected with the PPARd2 construct had higher levels of apoptosis ( Figure 5F).

PPARd Expression and Prognostic Significance in Primary Neuroblastoma Tumours
The in vitro PPARd expression, knockdown and agonist/ antagonist studies suggest that PPARd may have a role in the survival of tumour cells in vivo. To address this question, microarray data from a cohort of 251 neuroblastoma primary tumours [33] were analysed. The expression of PPARd1 and ALOX5 were examined across the following groups: stages 1 & 2, stage 3, stage 4, stage 4S and MYCN amplified neuroblastoma. Low PPARd1 or ALOX5 expression was significantly associated with MYCN amplification status ( Figure 6A and B), and low PPARd1 expression was significantly associated with poor survival ( Figure 6C). However, when multivariate analysis was carried out, with: age at diagnosis, MYCN amplification status, stage and PPARd1 expression, the independent contribution of PPARd1 was not significant, suggesting that the prognostic significance of low PPARd1 was due to its strong correlation with MYCN amplification status. The SCR control did not significantly affect PPARd1 expression, compared to the mock-transfected control (white bar; P = 0.566). Across treatments, PPARd siRNA significantly decreased PPARd1 mRNA expression compared to SCR and mock-transfected controls (1-way ANOVA, P,0.001). (C) PPRE reporter activation in siRNA-transfected SH-SY5Y cells in the presence of ATRA (4 mM or 10 mM) or control vehicle (0). PPARd (black bars) or SCR (grey bars) siRNA had no effect on the controls (C; 1-way ANOVA with contrasts, P = 0.85). Reporter activation was significantly lower in the PPARd siRNA-transfected cells after treatment with 4 mM or 10 mM ATRA (1-way ANOVA with contrasts, P,0.001 and P = 0.015, respectively). All error bars are6SEM. doi:10.1371/journal.pone.0068859.g003 Discussion AA has opposing effects on cell survival: inhibition of PLA2 is associated with decreased survival and motility [34], but AA release and its subsequent accumulation after treatment with cytotoxic agents has been suggested to sensitise cells to apoptosis [35]. Our study, alongside others [11], [12], has shown that combining RA with celecoxib synergistically promotes cell death of neuroblastoma cells in vitro. It is likely that inhibition of AA metabolism by cyclooxygenases and lipooxygenases promotes cell death by two mechanisms: firstly by preventing the synthesis of mediators of cell survival and, secondly, by promoting an accumulation of AA [36]. Indeed, combinations of inhibitors of AA metabolism (celecoxib) and DNA-damaging drugs that promote AA release, promotes cell death effectively in vitro and in vivo [10]. It has recently been shown that activated caspase-3 from dying cells promotes AA release via PLA2 in surrounding cells, thus promoting the regrowth of irradiated tumour cells [37], indicating that targeting AA signalling after drug treatment may minimise drug resistance.
We used small-molecule inhibitors targeted against cyclooxygenases and lipoxygenases in combination with ATRA to elucidate whether AA metabolism promotes cell survival in neuroblastoma cells. Previous studies suggest that PGE2 is an important mediator of RA-induced cell survival signalling [38][39][40]. However, later studies, including this study, have shown that exogenous PGE2 does not affect ATRA and celecoxib-induced cell death [11]. Results from a range of experimental approaches support a COX2-independent mechanism for RA mediated cell survival signalling. The FLAP inhibitor MK886, prevents 5-LO activation and consistently reduced cell viability in combination with RA, suggesting that LO metabolites are important mediators of cell survival after RA treatment. This is also consistent with findings that celecoxib is a potent inhibitor of 5-LO [29] and synergistically promotes cell death in combination with RA. In B cells, 5-oxo-ETE is a major product of AA metabolism which accumulates as a result of oxidative stress. 5-oxo-ETE production is promoted through increased AA production and also increasing levels of NADP+, a cofactor for 5-HEDH [41]. Furthermore, 5-oxo-ETE production is associated with cell survival after treatment with cytotoxic agents in prostate cancer cells [42], [43]. In our study, 5oxo-ETE partially rescued cells from a combination of RA and celecoxib or MK886 (summarised in Figure 7), suggesting that lipoxygenase metabolites such as 5-oxo-ETE mediate RA-induced cell-survival signalling. Although previous studies have suggested that ATRA is a potent PPARd ligand [13], subsequent work has not been able to demonstrate binding of ATRA to PPARd, PPREreporter activation by 0.3 mM ATRA or increased ATRAmediated expression of PPARd target genes [15][16][17]. In this study we found that ATRA or 5-oxo-ETE treatment activated PPARdmediated PPRE-reporter activity in neuroblastoma cells. In contrast to earlier work, the concentrations of ATRA (1 mM-  PPARd performs a complex role co-ordinating the activity of other PPARs via ligand-independent transcriptional repression at PPREs [44]. Since we found consistent increases in reporter activity after treatment with ATRA or 5-oxo-ETE, which decreased after 5-LO inhibition, this suggests that, at least in part, cell survival signals rely on ligand activation of PPARd1. Additionally, the putative dominant-negative isoform, PPARd2, was markedly induced at the mRNA level after celecoxib/ATRA treatment. A dominant-negative role for PPARd2 was also supported by the finding that ectopic PPARd2 expression increased apoptosis and reduced proliferation of neuroblastoma cells. In our model, the effects of PPARd2 were observed after 72 h, and this is not consistent with the levels of cell death observed after 24 h with the drug combinations. However, PPARd2 up-regulation may function to prolong the suppression of PPARd1-mediated cell survival and growth signalling. We also speculate that since 5-LO inhibitors induced PPARd2, PPARd2 might be repressed in a ligand-dependent manner by PPARd1.
Understanding the balance of cell-survival and cell-death signalling pathways in vivo will help determine the reasons for treatment failure and optimise therapeutic strategies with RA in neuroblastoma patients. Low PPARd and ALOX5 expression was associated with MYCN amplification in a cohort of 251 primary neuroblastoma tumours. Although low PPARd1 expression was significantly associated with poor prognosis, it was not an independent prognostic indicator due the correlation with MYCN amplification. It is possible that PPARd1 expression is upregulated in tumour cells in response to patient treatment and this remains to be investigated. It should also be noted that MYCN expression is down regulated in neuroblastoma cells after treatment with RA [2], so if PPARd1 expression is repressed directly or indirectly by MYCN, its potential up-regulation in response to RA would act as a compensatory cell-survival and cellgrowth mechanism.
The mechanism linking low PPARd1 and ALOX5 to MYCN amplification status in neuroblastoma needs to be investigated, and the study raises questions over the potential role of PPARd1 in promoting cell survival signalling in neuroblastoma (Figure 7). Clearly, if these ideas are supported by future work, this will lead to new targets for increasing the efficiency of RA, particularly PPARd and the enzymes involved AA metabolism. Celecoxib also promotes death of neuroblastoma cells synergistically in combination with cytotoxic drugs in vitro and in vivo [10], therefore it is important to determine if PPARd signalling is a common mechanism of drug resistance in cancer.
In summary, the results of this study in neuroblastoma cells suggest that high levels of ATRA induce a survival response mediated by the metabolism of AA by lipoxygenases, particularly 5-LO. The experimental data implicate the 5-LO product 5-oxo-ETE as a ligand activating PPARd1 to promote survival. Blocking this signalling pathway at the level of AA release, 5-LO-mediated metabolism or PPARd1 antagonism promoted cell death. Furthermore, ligand-dependent and ligand-independent-negative feedback relationships between PPARd1 and a dominant-negative PPARd isoform, PPARd2 may have a role in the control of survival signalling. It will be important to test these ideas using Low PPARd1 or ALOX5 expression was significantly associated with MYCN amplification status (Wilcoxon test; P,0.001 for both). For Kaplan-Meier analysis of patient survival, a cut-off at the first quartile was used to distinguish between low PPARd1 and high PPARd1 expression: low PPARd1 expression defined in this way was significantly associated with poor survival (Log Rank test, P,0.001). Multivariate analysis with: age at diagnosis, MYCN amplification status, stage and PPARd1 expression showed that PPARd1 expression did not contribute significantly (P = 0.866). Box and whisker plots show the median (horizontal line), 25th and 75 th percentiles (box horizontal boundaries), with the whiskers above and below extending to the most extreme data point.1.5 times the interquartile range from the box. Outliers are shown as circles. doi:10.1371/journal.pone.0068859.g006 other cytotoxic drugs which promote AA release and in other cell types. Finally, the possibility of a more-general role in cell survival for PPARd1 with negative-feedback regulation by a dominantnegative isoform needs further investigation. Figure S1 The role of COX2 in retinoic acid mediated cell death. (A) ATRA cell-survival signalling is proposed to be mediated through PLA2-dependent AA release and metabolism to lipoxygenase products such as 5-oxo-ETE, which activate PPARd1 to promote expression of survival genes. We suggest that the expression of inhibitory PPARd2 transcripts is suppressed when this pathway is activated. (B) When the PLA2/5-LO/PPARd pathway is inhibited, we propose that the expression of genes facilitating cell survival is prevented, a process facilitated by up-regulation of PPARd2 and PPARd2-mediated suppression, allowing the expression of either stressinduced or RAR-induced (or both) death genes to predominate and drive cell death. doi:10.1371/journal.pone.0068859.g007