Upregulation of MAPK Negative Feedback Regulators and RET in Mutant ALK Neuroblastoma: Implications for Targeted Treatment

This article is featured in Highlights of This Issue, p. 3097

Neuroblastoma is a pediatric malignancy arising from sympatho-adrenergic neural crest progenitor cells of the peripheral nervous system (1). Neuroblastoma originates from a subset of neural progenitor cells that normally differentiate into adrenal chromaffin cells and sympathetic ganglia (2).

Treating neuroblastoma remains a major therapeutic challenge for pediatric oncologists, as overall survival (OS) for patients with aggressive disease remains disappointingly low despite intensified and optimized treatment protocols (1). Previous research work from our laboratory and others has identified activating mutations in the tyrosine kinase domain of the ALK transmembrane receptor tyrosine kinase, which are found in the majority of hereditary neuroblastoma and occur as somatic defects in 7% to 10% of sporadic cases (3–6). During embryonic development, ALK is expressed in the central and peripheral nervous system (7), where it may regulate the interplay between cell proliferation and differentiation of the developing sympatho-adrenal cells of the neural crest (1, 8, 9).

The identification of activated ALK in neuroblastoma as a tractable therapeutic target has raised hope for more successful treatment: several small-molecule ALK inhibitors have recently gone into clinical trials such as a recent phase 1 clinical trial for crizotinib in patients with refractory neuroblastoma (10, 11). Moreover, these small molecules are also emerging as important novel treatment options for other tumor entities with aberrant ALK activity, including a subset of lung cancers (12). Notwithstanding these encouraging findings, intrinsic and acquired resistance almost inevitably occurs when using single compound treatment for receptor tyrosine kinases or other kinases (13–15), warranting the development of higher affinity inhibitors and the exploration of opportunities for rationally designed combinatorial treatment approaches. To achieve this goal, a more comprehensive understanding of the mutant ALK controlled regulatory network is required, including feedforward and feedback loops as well as pathway cross-talk, as these are critical in the adaptive responses of cancer cells leading to therapy resistance.

In this study, we established a robust 77-gene signature marking ALK activity in neuroblastoma cells, using transcriptome profiling. Next, we showed that this signature is preserved in other ALK-driven tumor entities and across ALK^F1174L -driven human and murine neuroblastoma tumors and cell lines. First, we confirm the activation of the PI3K/AKT/mTOR, MAPK/ERK, and MYC/MYCN signaling pathways. Second, we describe important novel specific insights related to mutant ALK signaling: we demonstrate a strong upregulation of MAPK negative feedback regulators and cross-species genomic analysis revealed that mutant ALK drives the expression of the tyrosine kinase RET as well as a set of RET controlled cholinergic markers. We consider these novel findings as important leads for further studies exploring novel therapeutic combination strategies in ALK-mutant neuroblastomas and also offering first insights toward a deeper understanding of the role of ALK signaling in normal sympathetic nervous system development.

Human neuroblastoma cell lines, anaplastic large cell lymphoma (ALCL) cell lines, and non–small cell lung cancer (NSCLC) cell lines (Supplementary Table S1) were cultured in RPMI-1640 medium (Invitrogen) supplemented with FBS (10%), kanamycin (1%), penicillin/streptomycin (1%), l-glutamine (1%), and HEPES (25 mmol/L; Life Technologies). Cells were kept at 37°C in a 5% CO₂/95%O₂ humidified environment.

Details on determining GI₅₀ values are described in the Supplementary Materials and Methods.

Detailed information can be found in Supplementary Materials and Methods. In summary, human wild-type ALK (SK-N-AS, NGP, IMR-32), ALK^R1275Q (CLB-GA, LAN-5, UKF-NB-3), ALK^F1174L (SK-N-SH, Kelly, SMS-KCNR), and ALK-amplified (NB-1) neuroblastoma cell lines were treated in triplicate with 0.3 μmol/L NPV-TAE-684 (Novartis/SelleckChem, further referred to as TAE-684) or DMSO (VWR) for 6 hours, followed by RNA isolation and gene expression profiling (see paragraph: “RNA extraction, RT-qPCR, and gene expression profiling”). CLB-GA was further treated with ALK, MEK, PI3K/mTOR, or RET inhibitors. A gene expression time series for ALK inhibition was performed in CLB-GA cells at 0-, 10-, 30-, 60-, 120-, 240-, and 360-minute time points. For shRNA-mediated knockdown, pGIPZ-ALK shRNAmir and pGIPZ-nonsilencing control shRNAmir vectors were used (Open Biosystems).

Human neuroblastoma SK-N-AS cells were electroporated with pcDNA6/TR (Invitrogen) using the Neon Transfection System (Life Techologies). Single-cell clones were generated using blasticidin (7.5 μg/mL) and limited dilution. Using a TetR antibody (Clonetech), the clone with the highest TetR expression was selected (named SK-N-AS-TR) and used for the transfection with pT-REx-DEST30-ALK, ALK^F1174L, or ALK^R1275Q. After transfection of SK-N-AS-TR with the ALK variants, single-cell clones were raised using geneticin (500 μg/mL) and limited dilution, whereas blasticidin treatment was continued as described above. Clones with moderate expression of the ALK variants were selected for further experiments using RT-qPCR (ALK_fwd: ccatcattttggagaggattgaat; ALK_rev: gaaccccctcagggtcctt) and Western blotting.

Homozygous TH-MYCN transgenic mice (16) were sacrificed at days 7 (n = 4) and 14 (n = 4) of life to harvest sympathetic ganglia containing foci of neuroblast hyperplasia and at week 6 to harvest advanced neuroblastoma tumors (n = 4). In addition, we have dissected the same sympathetic ganglia from wild-type mice at days 7 (n = 4) and 14 (n = 4) and week 6 (n = 4) to control for mRNA expression changes during normal development (17).

Details on RNA extraction and RT-qPCR are described in Supplementary Materials and Methods. Gene expression profiling was performed using Affymetrix HG-U133plus2 arrays or Sureprint G3 human GE or G3 Mouse GE 8 × 60K microarrays (Agilent Technologies). Data were summarized and normalized with the gcRMA and vsn method (for Affy and Agilent data, respectively; ref. 18), in the R statistical programming language using the affy, gcrma, and limma packages. Data can be accessed via ArrayExpress (E-MTAB-3205, E-MTAB-3206, E-MTAB-3207).

Total protein lysates were harvested after washing with ice-cold PBS and total protein isolation was carried out using RIPA lysis buffer, containing Complete Protease Inhibitor Cocktail (Roche Diagnostics) and PhosSTOP Phosphatase Inhibitor (Roche Diagnostics). Antibodies directed against phosphorylated ALK (Y1604), total ALK (C26G7), phosphorylated RET (Y905), total RET (C31B4), secondary anti-rabbit and anti-mouse antibodies were obtained from Cell Signaling. Antibodies directed against the loading control proteins β-actin and α-tubulin were obtained from Sigma-Aldrich.

Validation of the ALK signature was performed using signature score analysis in published datasets of neuroblastoma tumors (E-MTAB-161; ref. 19; Oberthuer dataset), neuroblastoma mice tumors (GSE32386; ref. 20), and other ALK-driven tumor entities, including NSCLC (GSE25118; ref. 21) and ALCL (GSE14879, ref. 22; GSE6184, ref. 23). In addition, signature score analysis was performed in a partly published dataset of mRNA expression data of 283 neuroblastoma tumor samples (NRC dataset; refs. 24, 25). In addition, GSE42762 (26) was used to check RET levels upon FOXO3a activation and PI3K/AKT inhibition.

An ALK signaling signature was established using differential expression analysis with the rank product (RP) algorithm. Genes that were significantly (P < 0.05) up- or downregulated after TAE-684 treatment in at least 3 cell lines and that showed a (log)fold-change of at least 1.2 after shRNA treatment were included in the ALK signature (Fig. 1A). MAPK/ERK, PI3K/AKT/mTOR, and RET signature genes were established by differential expression analysis (Limma) of CLB-GA cells treated with the respective inhibitor versus DMSO treatment.

Signature score analysis was performed using a rank-scoring algorithm as described in Fredlund and colleagues. (27). In short, for each sample, expression values were transformed to ranks (a rank of 1 matching with the lowest expressing gene). Next, rank scores for the signature genes were summed for each sample generating a signature score. Correlation of the score with survival was tested using Kaplan–Meier plots and log-rank analysis by grouping the samples in 4 quartiles (R-survival package). Comparison of signature scores or expression between groups of samples was done using the Mann–Whitney test (R-base package). Signature score correlation analysis was performed using Pearson correlation analysis (R-base package).

The probe ID lists of the ALK signature were submitted to DAVID for gene ontology (GO) analysis (28) and to the Connectivity Map database (29) for identification of drugs with similar transcriptional responses as ALK inhibition. Gene set enrichment analysis (30, 31) was performed using the C2 geneset catalogue (genetic and chemical perturbations). This analysis was performed on each cell line separately using the fold changes of TAE-684 versus DMSO treatment and the shALK versus scrambled control transduction. Genesets that were overrepresented in at least 3 cell lines after both ALK inhibitor treatment and shRNA knockdown were withheld. Cross-species genomic analysis was performed by looking for the overlap of the ALK signature genes and the list of differentially expressed genes (RP analysis) for MYCN versus ALK^F1174L-driven or double transgenic tumors.

The transcriptional consequences of constitutive oncogenic ALK signaling were determined through ALK perturbation experiments in 10 neuroblastoma cell lines with either of the 2 hotspot mutations ALK^F1174 and ALK^R1275, high-level ALK amplification (ALK^amp), or wild-type ALK [ALK^wt; protein levels of phospho-ALK (pALK) and ALK are provided in Supplementary Figure S3]. We performed pharmacologic inhibition with ALK inhibitor TAE-684 and controlled for off-target effects by shRNA treatment directed against ALK. Gene expression profiling was performed following 6 hours of TAE-684 treatment or 2 days after shRNA transduction (mRNA and protein expression levels of ALK are provided in Supplementary Figure S4). Importantly, TAE-684 treatment resulted in a marked transcriptome perturbation in all ALK-activated cell lines, whereas no significant effects on the transcriptome were observed in the SK-N-AS cell line with undetectable ALK protein expression. Overall, transcriptional responses in shALK treated cell lines showed similar direction of responses in the majority of up- or downregulated genes albeit with lower fold changes, probably due to incomplete knockdown of ALK (Supplementary Table S5).

Next, we established a gene expression signature recapitulating constitutive ALK signaling in neuroblastoma. Using differential expression analysis of the above described transcriptome data (of both TAE-684–treated cell lines and shALK-treated cells to control for off-target effects), we generated an ALK signature, consisting of 32 downregulated (49 probe IDs) and 45 upregulated (61 probe IDs) genes (see Materials and Methods and Fig. 1A for details). The expression of these signature genes is visualized in Fig. 1B for CLB-GA upon ALK inhibition and for the 10 cell lines in Supplementary Figure S6 for TAE-684 ALK inhibition. The differential expression pattern for these 77 genes was highly reminiscent in the CLB-GA cell line treated with 3 additional ALK small-molecule inhibitors currently under evaluation in clinical trials: crizotinib, X-396, and LDK-378 (10, 32), further supporting the validity of the established 77-gene ALK signature (Fig. 1B). In addition, significant correlation was found between the signature scores for TAE-684 versus the 3three other compounds in 2 independent dataset as shown in Fig. 1C and D. About 44%, 49%, and 66% of the 77-signature genes were shown to be significantly differentially expressed in CLB-GA upon crizotinib, X-396, and LDK-378 treatment, respectively (with significant overlap according to Fisher's exact test, data not shown).

Next, we investigated whether the ALK signature also marks de novo constitutive activation of the mutant ALK receptor. To this end, we generated and performed transcriptome profiling of the SK-N-AS cell lines with tetracycline-inducible overexpression of ALK^F1174L, ALK^R1275Q and ALK^wt. Whereas overexpression of all ALK isoforms resulted in robust accumulation of the protein upon tetracycline induction, only ALK^F1174L and ALK^R1275Q showed constitutive Y1604 phosphorylation (Supplementary Figure S7AS). Indeed, it has been shown previously that even upon overexpression, wild-type ALK displays significantly reduced (to absent) kinase activity compared with the mutant ALK isoforms and is inefficient in transforming NIH3T3 cells (4). Consistent with these data, de novo mutant ALK activation in the SK-N-AS cell line showed to be very well represented by an increased ALK signature score (i.e., a summary score of the expression of the 77-gene ALK signature) when compared with cells overexpressing wild-type ALK, thus confirming the robustness of the developed gene signature list (Fig. 2A; Supplementary Figure S7BS and S7CS).

To validate how well the cell line–based ALK downstream transcriptional profile recapitulates ALK oncogenic activity, we evaluated ALK signature scores in 2 independent gene expression datasets consisting of respectively 252 and 283 neuroblastoma tumors (19, 24, 25), as well as data of 2 in vivo neuroblastoma model systems (20). In human primary tumors with ALK amplification or ALK mutations, ALK signature scores were significantly higher than ALK wild-type tumors (Fig. 2B and D). Moreover, increased ALK signature scores correlate to poor OS (Fig. 2C and E) and event-free survival (EFS; Supplementary Figure S8AS and S8BS) also in tumors with wild-type ALK, pointing at a possible role of ALK signaling in wild-type ALK tumors. In addition, we could show that DBH/ALK^F1174L-driven murine neuroblastoma tumors have significantly higher ALK signature scores than TH-MYCN–driven murine neuroblastoma tumors and normal adrenal tissue (published in ref. 20).

Most interestingly, we could confirm the validity of the ALK signature in compound-treated tumor samples. Indeed, crizotinib-treated TH-MYCN;KI-Alk^R1297Q tumors have reduced ALK signature scores compared with untreated tumors (Fig. 2F).

Next, we analyzed publically available gene expression datasets from other ALK-driven tumor entities (ALKomas). Indeed, we identified high activity scores in primary ALK-rearranged ALCL (Fig. 3A; ref. 22). Furthermore, activity scores also decreased in NSCLC xenografts and in ALCL cell lines following ALK abrogation (either through tyrosine kinase inhibitors or ALK shRNA; refs. 21, 23; Fig. 3B and C). This observation suggests a significant overlap in activated downstream signaling pathways in different ALKoma entities.

Gene set enrichment analysis (GSEA) on ALK inhibitor–treated cell lines yielded enriched gene sets linked to EGFR, PI3K/AKT/mTOR, and MYC/MYCN signaling pathways (Supplementary Table S9) in keeping with previous reports (3–6, 20, 33, 34). Subsequent Gene Ontology (GO) analysis applied to the list of 77 ALK-regulated genes showed that MAPK signaling genes were enriched among the ALK upregulated genes, whereas genes driving cell-cycle arrest and apoptosis were represented in the downregulated gene set (Supplementary Table S10).

To further examine which genes are under control of the MAPK and PI3K/AKT/mTOR signaling branches activated by mutant ALK, we measured the transcriptional effects of trametinib (MEK inhibitor) and BEZ-235 (a dual PI3K/mTOR inhibitor) in the ALK^R1275Q-mutant CLB-GA cell line. Using this approach, we identified 19 and 42 ALK signature genes, which were differentially expressed following trametinib or BEZ-235 treatment, respectively (Supplementary Figure S11). A prominent role for mutant ALK-driven PI3K/AKT/mTOR signaling also emerged from Connectivity Map analysis, yielding LY-294002, sirolimus, and wortmannin as top ranked inhibitors targeting the PI3K/AKT/mTOR pathway (Supplementary Table S12).

Time-series analysis (10, 30, 60, 120, 240, and 360 minutes) of gene expression profiles after exposure of the ALK^R1275Q-positive CLB-GA cells to TAE-684 showed a gradual decrease in ALK signature score starting from 2 hours toward near extinction 6 hours after treatment (Fig. 4A). This is also represented in the time-dependent modulation of transcription levels of ALK-regulated genes (Fig. 4B). Furthermore, GSEA reveals that the overall transcriptional response following ALK inhibition is obvious from 2 hours at which point MAPK pathway–driven gene expression shows the most prominent decrease (Supplementary Table S13). Indeed, 6 of these genes, DUSP4, DUSP5, DUSP6, SPRY2, SPRY4, and MAFF, mark the earliest transcriptional responses with a strong decrease 2 hours after ALK inhibition (Supplementary Figure S14; Supplementary Fig. S4BS, red arrows). Interestingly, these genes are upregulated in SK-N-AS cell lines upon regulable mutant ALK overexpression, indicating that this negative feedback loop is readily installed subsequent to ALK-driven pathway activation (Fig. 4C). These genes are known negative regulators of growth factor signaling, controlling transcription-dependent feedback attenuation. This observation is reminiscent to the effects of MAPK feedback inhibition as described in BRAF^V600E-mutant melanoma (35), but has so far not been observed in mutant receptor tyrosine kinase signaling. This observation prompted us to test MEK inhibitor (trametanib) in a panel of neuroblastoma cell lines. However, responses to the compound were very modest (Supplementary Figure S16).

Next, a stringent cross-species genomics analyses for differentially expressed genes in MYCN- versus ALK^F1174L-driven murine neuroblastoma tumors (16, 20), yielded 7 genes expressed significantly higher in ALK^F1174L;DBH-iCre–positive neuroblastoma tumors. In addition to 3 known MAPK-regulated genes (SPRY4, DUSP6, ETV5), also 4 neuronal markers were identified: RET, ENC1, VGF, and VIP. Interestingly, additional comparison between the 77-gene ALK signature and the differentially expressed genes between TH-MYCN;ALK^F1174L;DBH-iCre–driven versus TH-MYCN–driven murine tumors, further reduced this list to RET, ETV5, VGF, and VIP. Importantly, RET and VIP are part of a gene regulatory network determining noradrenergic and cholinergic sympathetic subtypes during neuronal development (36). Moreover, VGF is expressed in the developing adrenal gland and it is transcriptionally regulated by RET in PC12 cells (37, 38).

Further investigation of additional adrenergic and cholinergic markers in the mice tumors showed higher expression levels for both cholinergic and adrenergic marker genes in double transgenic TH-MYCN;ALK^F1174L;DBH-iCre tumors compared with those from TH-MYCN mice (Fig. 5A). This was confirmed in an independent set of tumors obtained from TH-MYCN;TH-ALK^F1174L mice (Fig. 5B; refs. 20, 33). Moreover, in a dataset of primary human neuroblastoma tumors, we observed that the expression of the cholinergic genes is significantly higher in human primary neuroblastoma with aberrant ALK activation (Fig. 5C; ref. 19).

Furthermore, we also observed low expression of both cholinergic and adrenergic markers in TH-MYCN tumors. Interestingly, in preneoplastic lesions isolated from TH-MYCN mice (17), dramatic downregulation of both the adrenergic and cholinergic markers was observed in full-blown tumors as compared with early hyperplastic lesions in sympathetic ganglia of transgenic mice, whereas expression levels remained unchanged during development of wild-type mice ganglia (Fig. 5D).

Collectively, these data point to a very distinct cholinergic/adrenergic phenotype in MYCN- versus MYCN;ALK-driven neuroblastomas.

An important driver of the sympathetic neuronal markers of the cholinergic lineage is the RET gene. Comparison of RET expression levels in ALK-mutated and wild-type neuroblastoma tumor samples showed a significant higher expression in ALK-mutant versus ALK wild-type samples in 2 independent datasets of primary neuroblastoma tumor samples (Fig. 5E and F).

In addition, RET expression levels showed strong correlation with the 77-gene ALK signature scores in 2 independent patient sample datasets, providing strong support for ALK regulation of RET transcription (Fig. 6D and E).

Importantly, we also confirmed RET total protein expression in 8 of 12 ALK wild-type neuroblastoma cell lines (66.7%) and in 10 of 12 neuroblastoma cell lines with ALK mutation (83%), with moderate-to-strong RET phosphorylation (Tyr905) observed in 7 of 12 ALK-mutant cell lines (58%) and in 1 ALK wild-type cell line (8%; Fig. 5G). Taken together, these data show more pRET-positive mutant ALK cell lines and also more pronounced pRET levels in ALK-mutant cell lines versus ALK wild-type cell lines (Fischer exact test: P = 2.72e-2).

Given the above findings, we sought for further mechanistic evidence for regulation of RET mRNA levels by mutant ALK. To this end, we determined RET expression levels in NIH3T3 cells transformed by mutant ALK^F1174L. These cells were previously reported by Chen and colleagues (4) as a cellular model that demonstrates the transforming capacity of the ALK^F1174L mutation. Using this model system, we observed a robust 2-fold upregulation of RET mRNA levels in NIH3T3^ALKF1174L cells compared with the parental NIH3T3 cell line (Fig. 5H).

Given the previously described regulation of FOXO3a by PI3K/AKT signaling (26) and the observation that the related FOXO1 controls RET expression in mouse spermatogonial stem cells (39), we tested the possibility that FOXO3a could regulate RET under the control of mutant ALK. Pharmacologic inhibition of mutant ALK using LDK-378 clearly leads to the expected loss of FOXO3a phosphorylation (leading to activation of FOXO3a; Supplementary Figure S15B). Next, we investigated published data on overexpression of FOXO3a in neuroblastoma cells (26) and observed strong repression of RET mRNA expression following activation of FOXO3a in combination with PI3K/AKT inhibition (Supplementary Figure S15A). Taken together, these data convincingly show that RET expression is under direct control of mutant ALK.

Next, we explored the effect of RET inhibition for the clinically approved drug vandetanib (RET knockdown in Fig. 6A) in 5 neuroblastoma cell lines (4 ALK-mutant cell lines with varying RET phosphorylation levels and 1 ALK wild-type cell line with RET phosphorylation). All mutant ALK cell lines responded to vandetanib, with strongest effects observed for CLB-GA cells, whereas the ALK wild-type cell line SK-N-BE2C showed no significant response to the compound (Fig. 6B).

Finally, to determine on a more global scale whether RET-driven signaling partially recapitulates mutant ALK signaling, we performed gene expression analysis of CLB-GA cells following vandetanib treatment. Strikingly, the 77-ALK signature genes identified following TAE-684 treatment also showed to be regulated in the same direction following vandetanib treatment, strongly suggesting that RET contributes to ALK-driven neuroblastoma tumors (Fig. 6C). This was further illustrated by the strong correlation observed between vandetanib and ALK signature scores in 2 independent tumor datasets (Fig. 6D and E).

Detailed insights into ALK signaling and its possible interference with other signaling pathways are of utmost importance when introducing small-molecule inhibitors in the clinic, as this may yield improved tools for measuring and predicting therapy response. Also, a deeper understanding of signaling cascades driven by genetic changes and their complex intertwined compensatory regulatory programs activated during tumorigenesis will be a prerequisite for identification of druggable downstream targets for novel combination therapies.

Here, we describe the repertoire of ALK-driven transcriptional alterations in neuroblastoma cells. Using a combined approach of pharmacologic inhibition and shRNA knockdown of ALK, we identified a signature of 32 up- and 45 downregulated genes (ALK signature) and validated this signature in an in vitro cellular model for regulable ALK^F1174L as well as human and murine neuroblastomas. Of further interest, also other ALK-driven tumor entities displayed higher ALK signature scores, thus underscoring the robustness of the developed gene list and revealing a significant overlap in activated downstream signaling pathways in different ALKoma entities. Finally, pathway dissection of the mutant ALK-driven transcriptome was in accordance with activation of the MAPK/ERK, PI3K/AKT/mTOR and MYC/MYCN signaling as described (3–6, 20, 33, 34, 40). Connectivity Map analysis as well as comparison of the ALK gene signature and the transcriptionally regulated genes upon PI3K/mTOR inhibition pointed at a major role of the PI3K/AKT/mTOR pathway downstream of ALK in neuroblastoma. These findings are in concordance with the observations of Berry and colleagues. (33) who showed a synergistic effect of crizotinib and Torin2 treatment of double-transgenic mouse tumors. Both observations further emphasize the possible role of PI3K/mTOR inhibitors in new combinatorial treatment schemes for ALK-positive neuroblastoma tumors.

Importantly, our study uncovered 2 major novel insights, that is, (i) the observation of upregulation of MAPK feedback inhibitors, including DUSPs, SPRYs, and MAFF, and (ii) regulation of RET and cholinergic/adrenergic neuronal differentiation genes through mutant ALK. Of notice, despite the fact that DUSP genes function as negative regulators of MAPKs, their gain of expression has often been correlated with cancer progression, drug resistance, and poor prognosis (41). A similar consitutively activated MAPK feedback loop has been observed in BRAF^V600E-mutant melanoma cells (35). The exact role of this aberrant feedback loop is poorly understood but could reflect the cellular response to cell-cycle arrest and oncogene-induced senescence, keeping in mind that these can be triggered by sustained overactivation of ERK and presence of BRAF^V600E mutations (41). In this scenario, elevated expression of DUSPs and other negative feedback regulators might dampen the primary cellular response to sustained MAPK activation, allowing mutant cells to proliferate. For BRAF^V600E-mutant melanoma, it has been shown that MEK inhibition can counteract the effects from relief of negative feedback loop components as shown in recent clinical trials (42). Therefore, we tested the effects of MEK inhibition in vitro on selected neuroblastoma cell lines but, overall, effects of MEK inhibition were modest.

Feedback loop mechanisms present in normal cells to control for unwanted or temporarily interruption (or activation) of signaling may negatively impact upon molecular treatment and ultimately lead to therapy resistance. In this context, recent work from the Rosen team showed that BRAF inhibitors in BRAF^V600E-mutant melanoma cells leads to derepression of RAS signaling due to downregulation of the negative feedback components as evidenced by a pERK rebounce (43, 44). Moreover and even more worrysome, this same mechanism can also unleash the activity of other receptor tyrosine kinase receptors, further aggravating the unwanted effects of the initial pathway inhibition. Recent work has pointed to increased occurrence of RAS mutations in relapsed neuroblastomas (45). Such mutations may invoke stronger effects on MAPK pathway activation and render cells more sensitive to MEK inhibition. Taken together, the present new insights into downstream signaling of mutant ALK in neuroblastoma cells, points at potential risks of abrogating intrinsic mutant ALK–induced negative feedback regulation of MAPK signaling which may render these cells more sensitive to subsequent accumulation of upstream RAS pathway mutations leading to therapy resistance. Also, our findings are in keeping with our initial finding of predominant AKT signaling in neuroblastoma (24) and more recent articles (33, 46, 47).

A second important novel finding from our study was that mutant ALK controls expression of the RET tyrosine kinase receptor. We first reported a possible role for RET in ALK mutant neuroblastomas first based on cross-species comparative genomic analysis on MYCN versus ALK^F1174L and MYCN/ALK^F1174L-driven mouse tumors and neuroblastoma cell lines (48). Further evidence was provided for this regulation, including the finding that mutant ALK likely controls RET expression levels through FOXO3a signaling. The FOXO3a forkhead transcription factor was previously shown as a key target of the PI3K/AKT pathway in neuroblastoma and essential for their survival (26). Here, we showed that mutant ALK signaling blocks FOXO3a phosphorylation through PI3K/AKT and that FOXO3a regulates RET expression, thus strongly indicating the existence of a mutant ALK–PI3K/AKT–FOXO3a–RET signaling axis in neuroblastoma cells.

This finding has several important consequences. First, this provides novel insights into the regulatory networks controlling early differentiation of sympathetic neuronal progenitors, in particular in relation to ALK and RET. RET regulates cholinergic properties in mouse sympathetic neurons and participates in cross-regulatory interactions which segregate cells toward adrenergic or cholinergic fate in the developing sympathetic ganglia (36, 49). In addition to RET, several genes implicated in cholinergic fate acquisition were identified as being part of the ALK signature, including VGF and VIP. Early neural crest RET/TRKC-positive progenitors committed to sympathetic fate express a hybrid noradrenergic/cholinergic phenotype marked by adrenergic markers TH, DBH, VMAT2, and cholinergic markers ChAT, VAChT, RET, NT3, TRKC, and PRPH expression and are marked by high proliferation levels (36). Shortly hereafter at E15.5, VAChT-positive neurons segregate with RET, PRPH, VIP, and SST and develop toward cholinergic neurons, whereas TRKA expression is initiated in VMAT2-positive neurons that will differentiate toward DBH/TH-positive noradrenergic neurons (50).

Second, the discovery of an ALK-RET regulatory pathway paves the way for novel experiments to deepen our understanding of MYCN and mutant ALK-driven tumor formation. Indeed, analysis of both cholinergic and adrenergic neuronal differentiation genes showed dramatic downregulation during MYCN-driven tumor formation, whereas these marker genes show relatively higher expression levels in MYCN/ALK^F1174L double transgenic mice tumors. This could indicate a possible mutant ALK-driven attenuation of the steep decline of the expression levels of these genes but requires further investigation in hyperplastic preneoplastic lesions from these double transgenic mice tumors. During tumor formation in TH-MYCN mice, we speculate that overexpression of MYCN imposes a very immature phenotype with steeply decreased levels of both adrenergic and cholinergic differentiation markers. We hypothesize, on the basis of our current findings, that mutant ALK may drive the cells slightly further along the differentiation path with installment of the biphenotypic RET/TRKC-positive proliferative progenitor cell type (see Supplementary Figure S17). Possibly, this sustained ALK signaling and upregulation of RET may positively impact on the survival of the cells during tumor initiation, thereby counteracting the initially observed apoptotic loss of most of the tumor precursor cells in early stages of tumor formation (51). Further investigations probing deeper into the mechanisms of the interrelationship between MYCN, ALK, and RET in normal neuron diversification are needed to understand how perturbation of this early gene regulatory network governing the adrenergic/cholinergic switch may impact on neuroblastoma tumor formation.

Third, this study should trigger further work investigating novel therapeutic angles based on our observations. In particular, our data suggest possible co-regulation between mutant ALK and RET (52) and show vandetanib sensitivity in ALK-mutant neuroblastoma cell lines. Moreover, a recent study showed impaired tumor growth in vivo in both MYCN/KI-Alk^R1279Q and MYCN/KI-Alk^F1178L mice upon inhibition of RET by vandetanib (48). Further dissection of the role of RET signaling in neuroblastoma and its impact on therapeutic targeting of the mutant ALK receptor with small molecules is therefore warranted.

In conclusion, we provide the first in-depth analysis of the mutant ALK-driven transcriptome in neuroblastoma cells offering an important resource for future studies toward designing and assessing novel therapies for ALK-driven tumors. In the present study, we were able (i) to establish a mutant ALK activity score that is also recapitulated in other ALKomas, (ii) to uncover a MAPK negative feedback loop that may potentially play an important role in molecular rewiring of mutant ALK neuroblastoma cells following prolonged exposure to inhibitors, and (iii) to show that ALK regulates RET and propose RET as a bona fide target for molecular treatment of neuroblastoma. Taken together, these novel findings should fuel further studies toward understanding the complex interrelationship between ALK and RET signaling in the early steps of sympathetic nervous system and tumor development and provide a basis for further exploration of novel therapeutic strategies, in particular in relation to the interplay between ALK and RET signaling.

J. Gibbons and C. Liang hold ownership interest (including patents) in Xcovery. No potential conflicts of interest were disclosed by the other authors.

Conception and design: I. Lambertz, C. Kumps, S. De Brouwer, J. Gibbons, G. Laureys, T. Van Maerken, J. Cools, J. Vialard, F. Speleman, K. De Preter

Development of methodology: I. Lambertz, C. Kumps, S. Claeys, D.R. Carter, M. De Mariano, S. De Brouwer, C. Liang, D.C. Trujillo, F. Speleman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Lambertz, C. Kumps, S. Lindner, A. Beckers, D.R. Carter, A. Cazes, B.B. Cheung, O. Delattre, I. Janoueix-Lerosey, G. Laureys, G.M. Marchall, M. Porcu, N. Van Roy, A. Van Goethem, P. Zabrocki, J. Cools, J.H. Schulte, F. Speleman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Lambertz, C. Kumps, S. Claeys, A. Beckers, A. De Bondt, I. Janoueix-Lerosey, G.M. Marchall, M. Porcu, D.C. Trujillo, I. Van Den Wyngaert, P. Zabrocki, J. Cools, F. Speleman, K. De Preter

Writing, review, and/or revision of the manuscript: I. Lambertz, C. Kumps, S. Claeys, A. Beckers, S. De Brouwer, G.M. Marchall, D.C. Trujillo, T. Van Maerken, J. Cools, J.H. Schulte, J. Vialard, F. Speleman, K. De Preter

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Lambertz, S. Claeys, M. De Mariano, J. Takita, D.C. Trujillo, I. Van Den Wyngaert, F. Speleman, K. De Preter

Study supervision: F. Speleman, K. De Preter

Other (figures): E. Janssens

The authors thank Jeroen Schacht, Fanny De Vloed, Els De Smet, Justine Nuytens, Lies Vantomme, and Shalina Baute for their outstanding technical support.

This work was supported by NIH grants R01CA102074 and R01CA134878 (D. Danielpour) and the Case Comprehensive Cancer Center P30 CA43703 (for Cytometry and athymic mouse cores). R.S. Wahdan-Alaswad was supported, in part, by a predoctoral fellowship from Research Oncology Training Grant 5T32CA059366-15 (2009) and a National Research Service Award Individual Fellowship Application 1F31CA142311-01 (2010–2011).

C. Kumps was indebted to the Institute for the Promotion of Innovation by Science and Technology (IWT-Vlaanderen; http://www.iwt.be/) for a predoctoral fellowship (grant number 081373); K. De Preter, I. Lambertz, S. Claeys, and T. Van Maerken are supported by the Fund for Scientific Research Flanders (FWO; http://www.fwo.be/). This work was supported by the FWO (grant number: G.0198.08), the Belgian program of Interuniversity Poles of Attraction (IUAP; http://www.belspo.be/belspo/iap/index_en.stm), initiated by the Belgian State, Prime Minister's Office, Science Policy Programming, by the GOA (http://www.ugent.be/nl/onderzoek/financiering/bof/GOA; grant number 01G01910), the FOD (http://www.health.belgium.be/eportal; grant number: NKP_29_014). The research leading to these results has received funding from the European Union's Seventh Framework Program (FP7/2010-2015) under grant agreement n°259348.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.