FOXP1 inhibits cell growth and attenuates tumorigenicity of neuroblastoma

Background Segmental genomic copy number alterations, such as loss of 11q or 3p and gain of 17q, are well established markers of poor outcome in neuroblastoma, and have been suggested to comprise tumor suppressor genes or oncogenes, respectively. The gene forkhead box P1 (FOXP1) maps to chromosome 3p14.1, a tumor suppressor locus deleted in many human cancers including neuroblastoma. FoxP1 belongs to a family of winged-helix transcription factors that are involved in processes of cellular proliferation, differentiation and neoplastic transformation. Methods Microarray expression profiles of 476 neuroblastoma specimens were generated and genes differentially expressed between favorable and unfavorable neuroblastoma were identified. FOXP1 expression was correlated to clinical markers and patient outcome. To determine whether hypermethylation is involved in silencing of FOXP1, methylation analysis of the 5′ region of FOXP1 in 47 neuroblastomas was performed. Furthermore, FOXP1 was re-expressed in three neuroblastoma cell lines to study the effect of FOXP1 on growth characteristics of neuroblastoma cells. Results Low expression of FOXP1 is associated with markers of unfavorable prognosis like stage 4, age >18 months and MYCN amplification and unfavorable gene expression-based classification (P < 0.001 each). Moreover, FOXP1 expression predicts patient outcome accurately and independently from well-established prognostic markers. Array-based CGH analysis of 159 neuroblastomas revealed that heterozygous loss of the FOXP1 locus was a rare event (n = 4), but if present, was associated with low FOXP1 expression. By contrast, DNA methylation analysis in 47 neuroblastomas indicated that hypermethylation is not regularly involved in FOXP1 gene silencing. Re-expression of FoxP1 significantly impaired cell proliferation, viability and colony formation in soft agar. Furthermore, induction of FOXP1 expression led to cell cycle arrest and apoptotic cell death of neuroblastoma cells. Conclusions Our results suggest that down-regulation of FOXP1 expression is a common event in high-risk neuroblastoma pathogenesis and may contribute to tumor progression and unfavorable patient outcome. Electronic supplementary material The online version of this article (doi:10.1186/1471-2407-14-840) contains supplementary material, which is available to authorized users.


Background
Neuroblastoma is the most common extracranial solid cancer in childhood and the most common cancer in infancy [1]. The tumor is suggested to originate from immature neuroblasts giving rise to the sympathetic nervous system. The clinical hallmark of neuroblastoma is its heterogeneity, with the likelihood of cure varying widely between distinct patient subgroups [2,3]. Like no other tumor entity, neuroblastoma provides unique features of tumor biology, including subgroups with differentiation into benign ganglioneuroma and a high incidence of complete spontaneous regression in the absence of any or with minimal therapeutic intervention [4]. By contrast, high-risk neuroblastoma patients have a particularly poor 5-year survival rate of <40% despite intensive multimodal therapy compared to an average survival rate of 80% among other pediatric malignancies [5]. The molecular mechanisms underlying these distinct neuroblastoma phenotypes are still poorly understood.
The winged helix transcription factor Forkhead box P1 (FoxP1) is one of four members of the subfamily P of the Fox transcription factor family, which is characterized by a 100-amino acid long, evolutionarily conserved DNAbinding domain called Forkhead or winged-helix domain. Members of the Fox subfamily P share several characteristics that are atypical among Fox proteins: their Forkhead domain is located near the carboxy-terminal region and they contain leucine zipper motifs that promote FoxP homo-and heterodimerization and thus highly selective, tissue-or cell-type specific activity [6]. These factors are widely but not ubiquitously expressed in human tissues and have been implicated in both embryonic development and adult tissue homeostasis by regulating cell growth, proliferation, differentiation, longevity and transformation [7,8]. Targeted deletion of FOXP1 is lethal in embryogenesis, which is mainly due to cardiac defects [9]. Loss of FoxP1 function has been reported in several human malignancies such as endometrial cancer, lung cancer, head and neck cancer, prostate cancer, renal cell carcinoma, ovarian carcinoma, and has been shown to be associated with poor prognosis in breast cancer [8,10,11]. Furthermore, genomic loss at the tumor suppressor locus 3p14.1 including the FOXP1 gene has been described in many cancer types [12] including neuroblastoma [13] and has indicated the presence of a neuroblastomaassociated putative tumor suppressor gene located between 3p14.1 and 3p21.32 [14]. On the other hand, FOXP1 may also act as an oncogene as it is highly expressed in hepatocellular carcinoma and certain B cell malignancies, in which it is frequently targeted by activating chromosome translocations placing it under the transcriptional control of the IGH enhancers [15,16].
To elucidate the role of FoxP1 in neuroblastoma pathogenesis, we examined the expression of FOXP1 in a cohort of 476 primary neuroblastomas using microarrays, and assessed its correlation with prognostic markers and patient outcome. To investigate the mechanisms of FOXP1 downregulation, we determined copy number aberrations of the FOXP1 locus in 159 neuroblastomas by array-CGH as well as DNA methylation patterns of the FOXP1 promoter in 47 tumors using mass spectrometry. Furthermore, we characterized the functional consequences of FOXP1 re-expression on cell proliferation, apoptosis, migration and colony formation in neuroblastoma cell lines.

Gene expression microarray data and patients' characteristics
Single-color gene expression profiles were generated from 476 primary neuroblastoma samples (stage 1, n = 118; stage 2, n = 78; stage 3, n = 71; stage 4, n = 148; stage 4S, n = 61) and 3 neuroblastoma cell lines (IMR-32, CHP-212 and SK-N-BE(2)) using a 44 K oligonucleotide microarray as described elsewhere [17]. All patients were registered in respective clinical trials with written informed consent from the patient and/or a parent/legal guardian. We received ethics approval from the Ethics Commission of the Faculty of Medicine of Cologne University for the clinical trials NB97 (NCT00017225) and NB2004 (NCT00410631) including the molecular assessment of tumor material. MYCN-amplification was observed in 66 tumors, while it was absent in 405 tumors (not determined, n = 5). Patients' age at diagnosis ranged from 0 to 296 months (median age, 13 months). Median follow-up for patients without events was 7.6 years (range, 0.4-19.0 years). Stage was classified according to the International Neuroblastoma Staging System (INSS) [18]. Total RNA of FOXP1-and GFP-expressing IMR-32, CHP-212 and SK-N-BE(2) cells was isolated at 0, 12, 24 and 72 h after transgene induction using Trizol (Life Technologies, Darmstadt, Germany). All raw and normalized microarray data are available as a subset at Gene Expression Omnibus (Accession numbers GSE45480 and GSE62419). To determine global differences in the expression profiles of FoxP1-and GFP-induced IMR-32, CHP-212 and SK-N-BE(2) cells, we compared mean expression levels of each gene between FOXP1-induced and control cells as described previously [19]. Differentially expressed genes, either up-or down-regulated after FOXP1 reexpression were determined by applying a fold-change cutoff (≥2) and performing an unpaired, two-tailed Student's t-test. The Gene Set Enrichment Analysis (GSEA) software (Broad Institute, Cambridge, MA, USA) was used to identify coordinated changes in a priori defined sets of functionally grouped genes as described previously [20,21]. The neuroblastoma subgroups used for methylation and array-CGH analyses consisted of 47 and 159 tumors, respectively. Risk estimation of corresponding patients was performed according to the International Neuroblastoma Risk Group (INRG) classification system (methylation analysis: high risk, n = 22; intermediate and low risk, n = 25; array-CGH analysis: high risk, n = 51; intermediate and low risk, n = 108) [22].

Analysis of genomic aberrations and promoter methylation of FOXP1
Array-CGH profiles from 159 neuroblastoma tumors were generated using 44 K or 105 K oligonucleotide microarrays as described previously [23,24]. The raw data were analyzed by Agilent Genomics Workbench software (v. 7.0; Agilent Technologies, Santa Clara, CA, USA, 2012) [23]. Array-CGH data are available as a subset at Gene Expression Omnibus (Accession: GSE45480).
FOXP1 promoter methylation analysis was performed by Sequenom Inc. (Hamburg, Germany) as described elsewhere [19,25]. Genomic DNA from 47 primary neuroblastoma specimens (high FOXP1 expression, n = 23; low FOXP1 expression, n = 24, as defined by the cutoff value for dichotomization of FOXP1 expression) and the neuroblastoma cell line IMR-32 was used to sequence three selected DNA regions covering 45 CpG units downstream of the FOXP1 start site (Table 1). PCR Primers were designed by using Methprimer (www. urogene.org/methprimer/, Additional file 1: Table S1).

Data analysis and statistics
Statistical analysis was performed using SPSS software version 20.0 (IBM). Levels of FOXP1 mRNA determined by microarray analysis were compared in patient groups defined by MYCN status (normal vs. amplified), age (<18 months vs. >18 months), tumor stage (stage 1-3 vs. 4S vs. stage 4) and gene expression-based classification (favorable vs. unfavorable) according to the published PAM classifier [26]. Two-tailed nonparametric tests (Wilcoxon, Mann-Whitney U, and Kruskal-Wallis test) were used where appropriate. Kaplan-Meier estimates for OS and EFS were calculated and compared by log-rank test. Recurrence, progression and death from disease were considered as events. The cutoff value for dichotomization of FOXP1 expression was estimated by maximally selected log-rank statistics [27]. Multivariate analysis was performed for EFS and OS using Cox's proportional hazards regression models. The factors FOXP1 (low vs. high as reference), INSS stage (4 vs. 1, 2, 3, 4S as reference), MYCN (amplified vs. normal as reference) and age at diagnosis (≥18 months vs. <18 months as reference) were fitted into a backward selection. The criterion for inclusion was a likelihood-ratio test p-value less than 0.05 and for exclusion more than 0.10. Quantitative data for functional analyses are shown as means ± SD. Unpaired two-tailed Students t-tests were used where appropriate.

Cell culture
The neuroblastoma cell lines SK-N-BE (2) and CHP-212 were obtained from American Tissue Culture Collection (ATCC, Rockville, MD, USA) and authenticated at the DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). The neuroblastoma cell line IMR-32 was purchased from the DSMZ and PT67 retroviral packaging cells were obtained from Clontech (Mountain View, CA). All parental and inducible cell lines were cultured and maintained as described previously [19]. Doxycycline (Sigma-Aldrich, Deisenhofen, Germany) was used to induce transgene expression (2 μg/ml).

Stable inducible neuroblastoma cell lines
Neuroblastoma cell lines stably expressing either FOXP1 or GFP under the control of the reverse tetracyclinecontrolled transactivator (rtTA) were generated using the RevTet ™ System (Clontech) as described previously [19].

In vitro growth properties assays
The effect of FOXP1 expression on cell viability was assessed using the Trypan Blue dye exclusion test. Neuroblastoma cells were seeded in 48-well plates at a density of 1 × 10 4 cells/well in 200 μl RPMI-1640 (10% Tet-free FCS, 2 μg/ml doxycycline). To ensure continuous supply of nutrients throughout the measurement interval, 400 μl fresh RPMI-1640 (10% Tet-free FCS, 2 μg/ml doxycycline) per well were added on day 4. Cells were harvested at day 6 and evaluated for cell number and viability by Trypan Blue exclusion. The effect of FOXP1 expression on cell proliferation was assessed using the Alamar Blue assay (Life Technologies) according to the manufacturer's instructions. In brief, neuroblastoma cells were seeded in 48-well plates at a density of 6 × 10 3 cells/well in 200 μl RPMI-1640 (10% Tet-free FCS, 2 μg/ml doxycycline). To ensure continuous supply of nutrients throughout the measurement interval, 400 μl fresh RPMI-1640 (10% Tet-free FCS, 2 μg/ml doxycycline) per well were added on day 4. At day 6, 60 μl of Alamar Blue solution were added to each well, plates were incubated for 6 h and absorbance was measured at 570 nm against a reference wavelength of 595 nm using a plate reader (Multiscan Ascent plate reader; Thermo Labsystems, Helsinki, Finland). The relative reduction of Alamar Blue was calculated as described by the manufacturer. Cell cycle distribution was assessed by FACS analysis as described previously [29].

Analysis of apoptosis
Annexin-V-binding analysis was carried out using APC-coupled Annexin-V and 7-AAD according to the manufacturer's protocol (BD Biosciences, Heidelberg, Germany). Apoptotic cells were detected by FACS (FACS Canto; BD Biosciences) using the DIVA software (BD Biosciences). TUNEL analysis was performed 3 days after induction of FOXP1 using the In Situ Cell Death Detection Kit according to the manufacturer's protocol (Roche Diagnostics, Mannheim, Germany) as described previously [19].

Colony formation assay
A soft agar assay was used to assess the effect of FOXP1 on colony-forming of neuroblastoma cells as described previously [19]. After 15 days of incubation at 37°C, colonies were stained with 0.005% (w/v) crystal violet (Life Technologies) and colonies larger than 100 μm diameter were counted.

Migration analysis
To assess the effect of FOXP1 on neuroblastoma cell motility, two different assays were used: a transwell (Boyden chamber) assay and a wound healing assay. First, a Boyden chamber transwell migration assay was performed according to the manufacturer's protocol. In brief, 500 μl RPMI-1640 (20% Tet-free FCS, 2 μg/ml doxycycline) were added in each lower chamber of a 24well transwell plate (No. 3421, 5.0 μm pore size, Costar, Corning Costar, Rochester, NY), and 300 μl cell suspension (0.5 × 10 6 cells/ml) of IMR-32 and SK-N-BE(2) FOXP1 or GFP transgenic cells in RPMI-1640 (FCS-free, 2 μg/ml doxycycline) were seeded in the upper chamber. Cultures were maintained for 6 days, then non-motile cells at the top of the filter were removed and the cells on the lower surface of the membrane were fixed with methanol and stained with DAPI. The number of cells that had migrated to the lower surface of the membrane was counted in four random fields at 10 × 10 magnification under a light microscope. As a second method to determine the effect of FOXP1 re-expression on migration of neuroblastoma cells, CHP-212 and SK-N-BE(2) cells were analyzed in a wound healing assay using IBIDI culture inserts (No. 80209, IBIDI GmbH, Martinsried, Germany) according to the manufacturer's protocol. An IBIDI culture insert consists of two reservoirs separated by a 500 mm thick wall. In brief, an IBIDI culture insert was placed into one well of the 24 well plate and slightly pressed on the top to ensure tight adhesion. Equal numbers of FOXP1 or GFP transgenic cells (70 μl per reservoir; 1.0 × 10 5 cells/ml) in RPMI-1640 (10% Tet-free FCS) were added into the inserts. After 24 hours, the insert was gently removed, creating a gap of 500 μm. The well was filled with RPMI-1640 (10% Tet-free FCS, 2 μg/ml doxycycline) and the migration was documented at 10 × 10 magnification under a light microscope.
All assays were conducted in triplicate and reproduced in two independent experiments.

Results
Low FOXP1 transcript levels are correlated with markers of poor outcome in neuroblastoma To investigate whether FOXP1 expression is associated with prognostic markers of poor outcome in neuroblastoma, its expression levels were evaluated in a cohort of 476 neuroblastoma microarray profiles reflecting the whole spectrum of the disease [17]. We observed similar transcript levels in localized and stage 4S tumors, while expression levels were significantly lower in stage 4 neuroblastoma (Figure 1a; P < 0.001). FOXP1 levels were substantially decreased in tumors of patients older than 18 months at diagnosis (Figure 1b; P < 0.001) as well as in tumors with MYCN amplification (Figure 1c; P < 0.001). Furthermore, low FOXP1 transcript levels significantly correlated with unfavorable gene expressionbased classification (Figure 1d; P < 0.001) according to a prognostic multigene classifier that we have defined previously [26]. This predictive signature consists of 144 genes, but does not include FOXP1. Taken together, these results demonstrate that downregulation of FOXP1 is associated with unfavorable prognostic markers in neuroblastoma.
Low FOXP1 expression is associated with adverse outcome of neuroblastoma patients We next determined whether FOXP1 expression is associated with neuroblastoma patient event-free survival (EFS) and overall survival (OS). We noticed that low FOXP1 expression was associated with poor EFS (Figure 1e significant independent prognostic marker for both EFS and OS ( Genetic and epigenetic characterization of the FOXP1 locus in neuroblastoma We next aimed to evaluate whether deregulated FOXP1 expression is associated with genetic or epigenetic aberrations of the FOXP1 locus. First, we examined the methylation status of three DNA regions covering 45 CpG units downstream of the FOXP1 start site in 47 primary neuroblastoma samples and the neuroblastoma cell line IMR-32, which expresses FOXP1 at low levels endogenously. Unsupervised oneway hierarchical clustering revealed a low degree of methylation in general, and a largely homogeneous methylation pattern in tumors of both subgroups (low expression of FOXP1 vs. high expression of FOXP1 as defined in Methods) and IMR-32 cells (Figure 2). These results suggest that aberrant DNA methylation is not regularly involved in FOXP1 gene silencing in neuroblastoma.
Next, we analyzed array-comparative genomic hybridization (array-CGH) profiles of 159 neuroblastoma samples and compared the results with corresponding microarray gene expression data of the same tumors. In four samples, segmental loss of chromosome 3p14.1 comprising the FOXP1 locus was detected, suggesting that deletion of FOXP1 is not a frequent genetic event in neuroblastoma. However, loss of FOXP1 was accompanied by low FOXP1 expression levels in all four cases (Figure 3), indicating that low FOXP1 expression may be result from a gene dosage effect in a subset of neuroblastoma.

Doxycycline-inducible expression of FOXP1 in neuroblastoma cell lines
To examine FOXP1 functions in neuroblastoma pathogenesis, we generated three stable, inducible FOXP1 transgenic cell lines (IMR-32, CHP-212 and SK-N-BE(2)) using the retroviral RevTet-On ™ System (Clontech). 48 h after induction of transgene expression, polyclonal FoxP1-expressing neuroblastoma cells were compared with polyclonal GFP-expressing cells due to promoter leakage of the pRevTRE Vector System (Figure 4a). Recombinant FoxP1 protein levels were in the range of physiological levels observed in neuroblastoma patients with high FoxP1 expression (Figure 4b). Notably, Western blot hybridization analysis revealed that FOXP1 transcript levels correlate well with FoxP1 protein levels (Figure 4b), which may indicate that measurement of FOXP1 mRNA is predictive of its functional activity in neuroblastoma.

FOXP1 expression inhibits neuroblastoma cell growth
To investigate whether FOXP1 expression is involved in the regulation of growth properties in neuroblastoma cells, FOXP1-expressing IMR-32, CHP-212 and SK-N-BE(2) cells were compared with GFP-expressing controls. Cell proliferation was assayed for 6 days using Trypan Blue dye exclusion tests and Alamar Blue assays. On day 6, the number of viable cells was significantly decreased in all FOXP1expressing cell lines as compared to GFP-expressing controls (Figure 5a and b; P < 0.01).
We next examined whether apoptosis may contribute to the inhibition of neuroblastoma cell growth after FOXP1 re-expression. First, DNA fragmentation was assessed using the terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assay in IMR-32, CHP-212 and SK-N-BE(2) cells. In IMR-32 and CHP-212, FOXP1-expressing cells showed a significantly higher fraction of TUNEL-positive cells as compared to GFP-expressing controls, while no difference in the frequency of apoptotic events was observed in SK-N-BE(2) cells ( Figure 5c). Second, we analyzed the externalization of phosphatidylserine by flow cytometry using Annexin-V staining in CHP-212 cells. In accordance with the results of the TUNEL analysis, we observed a significant increase of the Annexin-V-binding fraction 5 days after FOXP1 induction in comparison to control cells (Figure 5d). To assess whether reduced proliferation upon FOXP1 re-expression might be a consequence of impaired cell cycle progression, cell cycle distribution was determined by fluorescence-activated cell sorting (FACS) on days 2 and 7 after induction of FOXP1 expression. We observed  an increased fraction of cells in the G 0 /G 1 phase at the expense of the S-phase population in FOXP1-expressing cells in all three cell lines (Figure 5e). Taken together, these findings indicate that FoxP1 affects growth properties in human neuroblastoma cells by both cell cycle regulation and induction of apoptosis.

FOXP1 attenuates the malignant phenotype of neuroblastoma cells
To further assess the effect of FOXP1 expression on tumorigenicity, we compared anchorage-independent growth rates and migration capabilities of FOXP1-induced with GFP-induced control cells. Fifteen days after transgene induction, FOXP1-expressing IMR-32 and SK-N-BE(2) cells showed a significantly reduced colonyforming ability in soft agar to about 10 and 20% of GFPexpressing control cells, respectively (Figure 6a; P < 0.001 each). In the Boyden chamber assay, the number of migrated cells in the lower chamber was significantly smaller in the FOXP1 expressing IMR-32 and SK-N-BE (2) cells as compared to the GFP-expressing controls (Figure 6b; 30% and 50%, respectively; P < 0.001 each). Consistent with these results, gap closure was delayed in a b c Figure 7  the FOXP1-induced CHP-212 and SK-N-BE(2) cells as compared to GFP-expressing controls (Figure 6c).
Together, our observations demonstrate that reconstitution of FOXP1 expression suppresses the malignant phenotype of neuroblastoma cells.

FoxP1 modulates the expression patterns of apoptosis-, migration-and differentiation-related genes
To assess the molecular consequences of   To identify downstream targets and signaling pathways regulated by FOXP1, we analyzed differentially expressed genes after FOXP1 re-expression using microarray gene expression analysis (Additional file 2: Table S2). FOXP1 strongly up-regulated the expression of the homeobox C9 gene (HOXC9) and genes involved in neuronal differentiation pathways such as DBH, MAP2, RET, TH, DNER, NEFL and NPY in IMR-32 ( Figure 9a) and CHP-212 cells (Figure 9b). Induction of HOXC9 or genes associated with neuronal differentiation was absent in the p53-mutated background of SK-N-BE(2) cells. By contrast, we found a strong (21.2-fold) induction of TNFSF10 (TRAIL) in SK-N-BE(2) cells immediately after FOXP1 induction (Figure 7c), indicating an alternative pathway of FOXP1-induced growth inhibition in SK-N-BE(2) cells.

Discussion
Altered patterns of gene expression arising from inappropriate regulation or function of key transcription factors cause many human diseases, including cancer. Forkhead box proteins are attracting increasing attention as tissue type dependent regulators and critical mediators of cellular processes including proliferation, migration, differentiation, cell cycle arrest or cell death. Dysregulation of Fox protein expression has been implicated in a broad spectrum of human disorders, including immunological dysfunction, infertility, speech/language disorders, and malignancies [30,31]. FoxP1 is a transcriptional repressor that plays an important role in the development of the brain and lung in mammals and has been described to have diverse and opposing functions in different types of cancer [32]. Results of recent studies indicate that loss of FoxP1 activity promotes solid tumor proliferation, e.g. in endometrial cancer, prostate cancer, renal cell carcinoma, and breast cancer, suggesting a tumor-suppressive role in these malignancies [16,33]. In contrast, high expression of FOXP1 was shown to correlate with poor outcome in certain types of B cell lymphomas [34][35][36]. While several Fox family members have been shown to play critical roles in neuroblastoma biology [37][38][39][40], the role of FoxP1 has not yet been investigated in this context.
Here, we demonstrate that low FOXP1 expression is associated with unfavorable prognostic markers and poor patient outcome in neuroblastoma. The prognostic relevance of low FOXP1 expression was validated by multivariate analysis for OS and EFS, showing that FOXP1 expression predicts neuroblastoma patient outcome independently from well-established prognostic markers. Similar results have recently been published for prostate cancer and non-small cell lung cancer, in which an inverse correlation of FOXP1 expression with increased malignancy grade and reduced patient survival was observed [33]. Together, these findings may suggest a tumor suppressive role of FoxP1 in neuroblastoma. In line with this hypothesis, restoration of FOXP1 expression strongly decreased growth rates and tumorigenic characteristics of all three neuroblastoma cell lines analyzed in vitro. These results are consistent with recent studies of other solid tumors, in which ectopic FoxP1 exhibited strong tumor suppressor activity in prostate cancer and glioma cells [33,41]. Although the results of the migration analyses may in part be explained by a reduction in cell number rather than a change of cell migration ability, the analysis of mRNA levels confirmed down-regulation of cell motilityassociated genes such as MAPK12 [42], PLCG2 [43] and NCF4 [44] in all three FOXP1 transgenic neuroblastoma cell lines analyzed. Considering that the cell lines used in this study differ in several molecular and biological characteristics like genetic alterations, doubling time, morphology and motility, our results may suggest that diminished FOXP1 expression might act as a growthpromoting factor in neuroblastoma in general. We also found that induction of FoxP1 delayed cell cycle progression and up-regulated pro-apoptotic genes such as DIABLO [45], CDC42 [46] and DAPK1 [47], indicating the induction of the intrinsic pathway of apoptosis. In addition, we found that HOXC9 is up-regulated, a well described key regulator of neuroblastoma differentiation which links the intrinsic pathway of apoptosis, cell-cycle exit and neuronal differentiation [48]. In line with previously published analyses of HOXC9 in neuroblastoma cells, we found a significant up-regulation of genes involved in neuronal differentiation pathways such as DNER, NEFL [19], DBH, TH [48], RET [49], MAP2 [50] and NPY [51]. The combined effects of intrinsic apoptosis and induction of differentiation have been described to strongly inhibit of proliferation and reduce tumorigenicity of neuroblastoma cells previously, which is well in line with the effects FOXP1 re-expression shown here. Together, these results indicate that FoxP1 may mediate cellular growth restriction in neuroblastoma by cell cycle control, programmed cell death and induction of differentiation. Interestingly, recombinant FoxP1 failed to induce apoptosis in the p53 mutant cell line SK-N-BE(2), suggesting a p53 dependent mechanism of FoxP1 mediated cell death. In accordance with this finding, we did not observe an induction of HOXC9 or other pro-apoptotic genes in this model cell line. FoxP1 and p53 signaling have been described to be connected in B-cell lymphoma, in which the oncogenic action of FoxP1 is repressed by p53-induced miR-34a [52]. These findings emphasize the need for further studies to elucidate the interaction of these factors and their signaling partners in a tissue-specific context. Of note, SK-N-BE(2) cells showed a strong (21.2-fold) induction of TNFSF10, the tumor necrosis factor-related apoptosisinducing ligand (TRAIL), immediately after FOXP1 induction (Figure 7c). This gene has been described to inhibit cell proliferation in several other tumor entities by regulating autophagy and extrinsic apoptosis, indicating an alternative pathway of FOXP1-induced growth inhibition in SK-N-BE(2) cells.
We also aimed to identify the molecular basis underlying the differential expression of FOXP1 observed in different neuroblastoma subgroups. Loss of chromosome 3p is a genomic aberration frequently found in unfavorable neuroblastoma, suggesting that one or multiple tumor suppressor genes are targeted by this genetic alteration [13,[53][54][55]. We found low FOXP1 expression in four tumor samples with segmental 3p14.1 loss containing the FOXP1 locus, pointing towards haploinsufficiency in these tumors. However, deletions of the FOXP1 locus appear to occur rarely in neuroblastoma, suggesting that other factors contribute to differential FOXP1 expression in this malignancy. Analysis of DNA methylation patterns and FOXP1 expression levels did not reveal any correlation, implicating that differential methylation is not a common cause of FOXP1 repression in neuroblastoma. In addition, somatic mutations of the coding sequence of FOXP1 have not been identified in recent reports on the mutational spectrum of neuroblastoma using next generation sequencing [56,57]. We therefore speculate that additional molecular mechanisms may mediate FOXP1 gene silencing in neuroblastoma. Aberrant expression of FOXP1 in various pathologic conditions has been reported to be caused by a variety of molecular mechanisms, including dysregulation of upstream signaling pathways such as FoxO proteins or components of the estrogen-ER signaling pathway [58][59][60], differential splicing [61,62], viral integration events [63], differential miRNA activity [64,65] and translocations affecting upstream regulatory regions of the FOXP1 gene [66]. Further investigations are required to determine the contribution of these mechanisms to altered FOXP1 expression in neuroblastoma.

Conclusions
In summary, our study indicates that aggressive human neuroblastomas show reduced FOXP1 expression, a finding that is compatible with the previously observed loss of FoxP1 in several types of solid tumors. FOXP1 expression levels are inversely correlated with proliferative activity and tumorigenicity in vitro, supporting its role as a transcriptional regulator with critical tumor suppressor functions in solid tumor biology. Although the mechanisms that regulate FOXP1 expression in neuroblastoma remain unclear, low expression of FOXP1 may add predictive value on top of established risk stratification tools in neuroblastoma. Further analysis of this transcription factor may contribute to the understanding of the molecular processes underlying tumor regression and progression in neuroblastoma.