Phosphoproteomic analysis of anaplastic lymphoma kinase (ALK) downstream signaling pathways identifies signal transducer and activator of transcription 3 as a functional target of activated ALK in neuroblastoma cells

Activation of the anaplastic lymphoma kinase (ALK) receptor tyrosine kinase is a key oncogenic mechanism in a growing number of tumor types. In the majority of cases, ALK is activated by fusion with a dimerizing partner protein as a result of chromosomal translocation events, most studied in the case of the nucleophosmin–ALK and echinoderm microtubule-associated protein-like 4–ALK oncoproteins. It is now also appreciated that the full-length ALK receptor can be activated by point mutations and by deletions within the extracellular domain, such as those observed in neuroblastoma. Several studies have employed phosphoproteomics approaches to find substrates of ALK fusion proteins. In this study, we used MS-based phosphotyrosine profiling to characterize phosphotyrosine signaling events associated with the full-length ALK receptor. A number of previously identified and novel targets were identified. One of these, signal transducer and activator of transcription 3 (STAT3), has previously been observed to be activated in response to oncogenic ALK signaling, but the significance of this in signaling from the full-length ALK receptor has not been explored further. We show here that activated ALK robustly activates STAT3 on Tyr705 in a number of independent neuroblastoma cell lines. Furthermore, knockdown of STAT3 by RNA interference resulted in a reduction in myelocytomatosis neuroblastom (MYCN) protein levels downstream of ALK signaling. These observations, together with a decreased level of MYCN and inhibition of neuroblastoma cell growth in the presence of STAT3 inhibitors, suggest that activation of STAT3 is important for ALK signaling activity in neuroblastoma.

Activation of the anaplastic lymphoma kinase (ALK) receptor tyrosine kinase is a key oncogenic mechanism in a growing number of tumor types. In the majority of cases, ALK is activated by fusion with a dimerizing partner protein as a result of chromosomal translocation events, most studied in the case of the nucleophosmin-ALK and echinoderm microtubule-associated proteinlike 4-ALK oncoproteins. It is now also appreciated that the full-length ALK receptor can be activated by point mutations and by deletions within the extracellular domain, such as those observed in neuroblastoma. Several studies have employed phosphoproteomics approaches to find substrates of ALK fusion proteins. In this study, we used MS-based phosphotyrosine profiling to characterize phosphotyrosine signaling events associated with the full-length ALK receptor. A number of previously identified and novel targets were identified. One of these, signal transducer and activator of transcription 3 (STAT3), has previously been observed to be activated in response to oncogenic ALK signaling, but the significance of this in signaling from the full-length ALK receptor has not been explored further. We show here that activated ALK robustly activates STAT3 on Tyr705 in a number of independent neuroblastoma cell lines. Furthermore, knockdown of STAT3 by RNA interference resulted in a reduction in myelocytomatosis neuroblastom (MYCN) protein levels downstream of ALK signaling. These observations, together with a decreased level of MYCN and inhibition of neuroblastoma cell growth in the presence of STAT3 inhibitors, suggest that activation of STAT3 is important for ALK signaling activity in neuroblastoma.

Introduction
Many of the tyrosine kinases in the human kinome are implicated in human cancers [1], and provide important targets for cancer treatment, as well as biomarkers for patient stratification. Recently, tyrosine kinase inhibitors targeting anaplastic lymphoma kinase (ALK) have been approved for the treatment of ALKpositive non-small-cell-lung-cancer (NSCLC) [2]. In NSCLC, ALK activation occurs via chromosomal translocation, leading to activation of the kinase domain. However, ALK is also known to be mutated in the context of the full-length receptor tyrosine kinase (RTK), this being most clearly understood in neuroblastoma. Neuroblastoma is a childhood cancer that stems from the sympathetic nervous system [3], most commonly originating in the adrenal glands, but also developing at additional sites in the neck, chest, and abdomen. It is considered to be a disease of developing tissue, as it originates from precursor cells of neural crest tissue that are active during embryonic development [4]. ALK gain-of-function mutations have been described in both familial [5,6] and sporadic [6][7][8][9][10] neuroblastoma. Although there is currently no clinically approved specific treatment for ALK mutations in neuroblastoma, ongoing clinical trials are being conducted to determine the potential usefulness of ALK-targeted therapies for use in future treatment approaches [11]. Today, crizotinib (PF-02341066) is in use as a Food and Drug Administration-approved drug for the treatment of ALK-positive NSCLC [12]. Clinical studies today suggest that anti-ALK inhibitor therapy offers promise in the treatment of ALK-mediated tumors [2,11,13].
One consistent problem with kinase inhibitors and their use in a more personalized therapy approach is the drug-resistant mutations that arise in response to treatment, most commonly within the ATP-binding site of the kinase in question. Originally described in the epidermal growth factor receptor and breakpoint cluster region-c-abl kinase domains [14][15][16][17][18], this has now also been described for echinoderm microtubuleassociated protein-like 4 (EML4)-ALK by Choi et al., and others who have described the appearance of such mutations in the fusion EML4-ALK protein that confers resistance to crizotinib [19][20][21][22]. Phase I/II studies of crizotinib in children with relapsed/refractory tumors involving ALK, including neuroblastoma patients, have been initiated (ClinicalTrials.gov, NCT01182896) [23].
Given the increasing clinical importance of ALK activity, we chose to characterize ALK-mediated tyrosine kinase signaling networks by MS-based phosphoproteomics. In the current study, we conducted a comprehensive profiling of endogenous proteins that are tyrosine-phosphorylated upon expression and activation of wild-type ALK in PC12 cells, a cell line originally from the rat neural crest. This list, which reflects both direct and indirect tyrosine phosphorylation targets as a result of ALK activity within the cell, verifies several recently published results such as fibroblast growth factor receptor substrate 2 (FRS2) [23]. One of the proteins that was most significantly tyrosine-phosphorylated in response to ALK activation in our analysis was signal transducer and activator of transcription 3 (STAT3), which has previously been reported to be phosphorylated by both the fulllength ALK and ALK fusion proteins [24][25][26][27]. Here, we investigated this further, and clarified the importance of STAT3 as a mediator in the initiation of transcription of MYCN.

Phosphotyrosine profiling of PC12 cells expressing ALK
To identify sites with altered phosphorylation associated with ALK RTK activity, we utilized an immunoaffinity-coupled LC-MS/MS approach on PC12 Tet-on-inducible cell lines. After induction, ALK was activated by addition of agonist mAbs that have previously been reported to initiate ALK signaling pathways [10,28]. Comparison with controls expressing unstimulated ALK allowed quantification of 336 phosphorylation sites (207 phosphotyrosine, 78 phosphoserine, and 51 phosphothreonine) derived from 189 different proteins (Table S1). Label-free quantification of ALK-activated versus control cells revealed 101 phosphorylation sites (73 phosphotyrosine, 13 phosphoserine, and 15 phosphothreonine) in 72 proteins, with a more than five-fold increase in corresponding phosphopeptide intensities. In contrast, only 19 phosphorylation sites in 12 proteins were found with decreased phosphorylation following ALK induction. Within the list of mapped phosphorylation sites, we identified 11 phosphotyrosine sites that belong to ALK (Fig. 1A). As expected, all of them showed a significant increase in the level of phosphorylation on activation of ALK with mAb as compared with control cells (Table S1). The identified tyrosine phosphorylation sites in full-length ALK overlap with the predicted ones (PhosphoMotif, Human Protein Reference Database, http://www.hprd.org/PhosphoMotif_finder), and cover, with only one exception, the entire set of mapped phosphotyrosine sites from earlier investigations of EML4-ALK and nucleophosmin (NPM)-ALK [29][30][31].
To highlight potential direct downstream substrates of full-length ALK, a network analysis using known protein-protein interactions and experimentally verified kinase-substrate relationships of the human orthologs of the identified phosphoproteins was performed (Fig. 1B). The network model illustrated two of the identified phosphoproteins with increased phosphorylation as potential direct targets of ALK following ALK induction, namely PTPN11 (SHP-2) and STAT3. Neither of these has previously been shown to interact with full-length ALK. In agreement with a role for these proteins downstream of ALK, we detected approxi-   (Table S1), showed decreased phosphorylation upon abrogation of ALK activity in several neuroblastoma cell lines (Fig. S1).

ALK activates STAT3 in PC12 cells
We subsequently decided to focus in more detail on one of the most prominent hits from the phosphotyrosine proteomics screen, namely STAT3. As observed in the screen, activation of ALK resulted in clear phosphorylation of STAT3 at Tyr705. PC12 cells expressing either doxycycline-inducible wild-type ALK or the ALK F1174S mutant were employed to examine STAT3 activation. As ALK is still regarded as an orphan receptor, the ALK receptor was stimulated with an agonist mAb that has been previously reported to bind and activate ALK in a cell culture model [28,32,33]. Interestingly, whereas long-term stimulation of the wild-type ALK receptor did lead to visible tyrosine phosphorylation of STAT3, this was not prominently observed after 30 min of stimulation, a time point when ALK and ERK1/2 were highly phosphorylated ( Fig To investigate the activation process further, we examined whether an interaction between ALK and STAT3 could occur. We were unable to observe an interaction between doxycycline-induced ALK and endogenous STAT3 in PC12 cells (data not shown). However, an interaction between STAT3 and ALK was observed on immmunoprecipitation of tagged STAT3, when wild-type ALK was transiently cotransfected with doxycycline stimulated with mAb31 --+ 30 min Protein expression was induced with 1 lgÁmL À1 doxycycline, and cells were serum-starved for 24 h prior to stimulation with 1 lgÁmL À1 ALK-activating mAb (mAb31) for 30 min or 24 h. Whole cell lysates were analyzed by SDS/PAGE, and this was followed by immunoblotting with antibodies against p-ALK Y1278 , ALK, p-STAT3 Y705 , and p-ERK. Pan-ERK and pan-STAT3 antibodies were employed as loading controls. (B, C) PC12 cells were transfected with either wild-type ALK or the ALK F1174S mutant together with FLAG-tagged STAT3 prior to stimulation with 1 lgÁmL À1 ALKactivating mAb (mAb46) for 24 h, in the presence or absence of 250 nM crizotinib, as indicated. Lysates were immunoprecipitated (IP) with either antibody against FLAG (M2) (B) or with antibody against ALK (mAb31) (C), and this was followed by immunoblotting for ALK, FLAG, and STAT3, as indicated. WCL, whole cell lysate.
FLAG-tagged STAT3 (Fig. 2B). Upon stimulation of ALK, increased interaction between ALK and STAT3 was observed (Fig. 2B, lane 3). This interaction was abrogated by addition of the ALK inhibitor crizotinib (Fig. 2B, lane 4). Similarly, on cotransfection of activated ALK F1174S with FLAG-tagged STAT3, an interaction was observed that was blocked upon addition of crizotinib prior to immunoprecipitation (Fig. 2B, compare lanes 5 and 6). Similarly, we were able to observe STAT3 in both wild-type and gain-of-function ALK immunoprecipitates from PC12 cells transiently expressing ALK and STAT3 (Fig. 2C). In agreement with our previous observations, the STAT3-ALK interaction was crizotinib-sensitive (Fig. 2C, compare lanes 3 and 5 with lanes 4 and 6). Quantification of independent blots showed clear six-fold to 10-fold increased binding of STAT3 to ALK upon stimulation or when a gain-of-function ALK variant was employed (Fig. S2). These data suggest that, upon activation, STAT3 may be recruited to ALK signaling complexes. However, it is important to note that our attempts to verify this interaction at the level of endogenous proteins in neuroblastoma cell lines have not been successful.

STAT3 is important for MYCN expression in response to ALK activation
Given recent observations that ALK regulates MYCN transcription in neuroblastoma cells and collaborates with MYCN in neuroblastoma pathogenesis [34][35][36][37], we decided to investigate a role for STAT3 in this process. Initially, we employed small interfering RNA (siRNA) targeting STAT3 in a number of neuroblastoma cell lines, including CBL-GE, CBL-BAR, CBL-GA and Kelly cells. These neuroblastoma cell lines are all ALK gain-of-function in nature, containing either an activated ALK mutation (ALK R1275Q , CBL-GA; ALK F1174V , CLB-GE; ALK F1174L , Kelly) or overexpressing an ALK receptor with an extracellular domain deletion (CLB-BAR Dexon 4-12 ), and express different levels of MYCN (Fig. S3) [38][39][40]. Cell lines were transfected with either scrambled siRNA, two independent siRNAs targeting STAT3, or a mock control. In each cell line tested, the scrambled siRNA transfection did not reduce STAT3 levels, which were comparable to those in cells with control transfection without siRNA. However, upon transfection with specific STAT3 siRNA, all cell lines tested showed reduced levels of STAT3 as compared with the scrambled transfection controls ( To further validate these results, we employed a number of STAT3 inhibitors, including FLLL32 and STATTIC, which have previously been shown to inhibit STAT3 activation [41][42][43]. We investigated ALK, STAT3 and MYCN levels in CLB-GE, CLB-BAR, Kelly and CLB-GA neuroblastoma cell lines upon treatment with STAT3 inhibitors (Fig. 4A-D). Treatment with either FLLL32 or STATTIC efficiently abrogated the phosphorylation of STAT3 without affecting general STAT3 levels. Importantly, whereas these inhibitors blocked STAT3 activity, they did not affect the phosphorylation status of ERK or ALK itself ( To investigate further whether STAT3 is involved in ALK-activated initiation of MYCN transcription, we employed an MYCNP-luciferase assay in two independent neuroblastoma cell lines [36]. Cells were transfected with MYCNP-luciferase reporter or control, and treated with either STATTIC or FLLL32. Upon treatment with STAT3 inhibitors, both cell lines showed reduced luciferase activity as compared with untreated cells (Fig. 5A). Additionally, we employed quantitative RT-PCR (qRT-PCR) to confirm a role for STAT3 in ALK regulation of MYCN transcription. As controls, we employed primers amplifying part of the coding sequence of ribosomal protein 29 (RPL29) or ribosomal protein 19 (RPL19) (Fig. 5B, data not shown). Neuroblastoma cell lines treated with STATTIC or FLLL32 for 24 h showed a significant reduction in MYCN mRNA levels in comparison with untreated cells (Fig. 5B). Thus, pharmacological inhibition of STAT3 activity in neuroblastoma cell lines harboring ALK gain-of-function mutations results in reduced transcription of MYCN mRNA.

STAT3 activity is required for growth and viability of neuroblastoma cells
We then investigated whether STAT3 is not only important for initiation of MYCN transcription but also might influence the proliferation of our neuroblastoma cell lines. In this analysis, we employed CLB-GE, CLB-BAR, Kelly and CLB-GA neuroblastoma cell lines, measuring their growth in response to treatment with the STAT3 inhibitor FLLL32 or STAT3 siRNA. In all cell lines examined, reduction of endogenous STAT3 expression levels by siRNA or treatment with the STAT3 inhibitor FLLL32 resulted in a decrease in cell growth as compared with untreated cells (Fig. 6B, D-G) in a similar manner as crizotinib decreased cell growth (Fig. 6A). Furthermore, it was clear that treatment of neuroblastoma cell lines with either FLLL32 or crizotinib reduced the phosphorylation status of STAT3 to a similar degree, without increasing cleavage of poly(ADP-ribose) polymerase (PARP), which was used as a measure of apoptosis (Fig. 6C).

Discussion
It is interesting to compare the results of this study with earlier phosphoproteomics analysis on oncogenic ALK fusion proteins, such as EML4-ALK and NPM-ALK [29][30][31]. In terms of phosphorylation of ALK itself, these studies identified overlapping but not identical sites in the ALK intracellular domain. Activation of full-length ALK results in phosphorylation of Tyr1278, Tyr1282, and Tyr1283, which lie in the Y′ RAS′YY autophosphorylation motif in the activation loop. Although our data do not allow us to comment on the order of phosphorylation within the Y′RAS′YY motif, they are in agreement with previous studies elucidating the mechanism of phosphorylation [44][45][46], and further support the use of antibodies against pTyr1278 as an indicator of ALK activation. Another site commonly used to denote activation of the ALK receptor is Tyr1604, which is situated at the C-terminal tail of the receptor, and has been reported to be important for transformation activity and docking of phospholipase Cc in studies on NPM-ALK [47]. Although antibodies against pTyr1604 are commonly used to assay ALK activity, the importance of Tyr1604 in the interaction of phospholipase Cc with full-length ALK has not been confirmed. The identification in this study reinforces the potential importance of this tyrosine in ALK activation. Of the phosphorylated tyrosines identified outside of the activation loop, Tyr1507 of ALK lies within a consensus Shc-binding site (NPTpY), and has been shown to be critical for interaction of full-length ALK with Shc [23]. These results are consistent with earlier work on NPM-ALK [48]. The remaining tyrosines that are phosphorylated in response to ALK activation identified in this study, namely Tyr1078, Tyr1092, Tyr1096, Tyr1131, Tyr1584, and Tyr1586, have been poorly characterized in the context of the full-length ALK RTK, and determination of their significance requires further analysis. However, work on the oncogenic ALK fusion proteins implicates these residues in binding of IRS-1 (Tyr1096) [48], and SNT (FRS2) (Tyr1096 and Tyr1507) [49]. As well as identifying tyrosine phosphorylation sites in ALK, we observed phosphorylation of Ser1086, Ser1281 and Thr1597 on ALK. Although this finding is intriguing, the significance of these phosphorylation events is unclear. Although the phosphorylation of   RTKs on serine and threonine is well recognized, its role is much less understood than that of tyrosine phosphorylation. However, serine/threonine phosphorylation is known to have important regulatory functions for, for example, fibroblast growth factor receptor, epidermal growth factor receptor [50,51], and Kit [52], to name but a few. Although information on the phosphorylation of ALK itself in response to activation represented an important part of the information obtained with this phosphoproteomics approach, the identification of candidate molecules phosphorylated in response to ALK activation was the primary aim of this investigation. The list of phosphoproteins contains both known and novel ALK signaling components, some of which have been characterized, but the majority of which remain to be investigated for their relevance in both physiological and pathological ALK signaling. Of those known candidates, several examples can be briefly mentioned. One molecule identified in our screen was FRS2, which was phosphorylated on Tyr349 in response to ALK activation. FRS2 has previously been reported to bind full-length ALK, although the binding site is unclear [23]. Similarly, MAPK1which was phosphorylated on Thr183 in this studyhas been shown by several groups to be phosphorylated upon stimulation of the ALK receptor [53].
One of the major tyrosine-phosphorylated proteins identified in this study was STAT3, which was phosphorylated on Tyr705 in response to ALK activation. STAT3 has been reported to be phosphorylated downstream of several of the oncogenic ALK fusion proteins [25,27,53]. To date, although several articles have reported STAT3 phosphorylation downstream of fulllength ALK, the significance of this is unclear [8,9,24,28,[54][55][56]. One study of NPM-ALK has suggested that oncogenic ALK binds and activates STAT3 directly via the first tyrosine (Tyr1278) in the Y′RAS′ YY motif of ALK [45]. The crystal structures of ALK show that unphosphorylated Tyr1278 interacts with Cys1097 in the N-lobe of the kinase, implying that phosphorylation of Tyr1278 should make a binding site for STAT3 available, consistent with this hypothesis. In this study, we saw clear phosphorylation of STAT3 on Tyr705, which is particularly robust in response to oncogenic forms of the full-length ALK receptor, such as ALK F1174S , ALK Dexon 2-3 , and ALK Dexon 4-12 [39,55,56]. Although it was not investigated in this study, it would be of interest to examine the importance of STAT3 phosphorylation in neuroblastoma cells harboring a wild-type nonactive ALK receptor. This hypothesis is supported by inhibition of the NB1 neuroblastoma cells lines with crizotinib, which resulted in a loss of pSTAT3 Y705 , although the significance of STAT3 for cell growth was not discussed [54]. In light of the activation of STAT3 by ALK, we also investigated the interaction of ALK with STAT3. Indeed, we observed an interaction of STAT3 with fulllength ALK when overexpressed; however, this was not detectable at the level of endogenous proteins, suggesting that care should be taken when interpreting these results. It is unclear whether this is because of limitations of the antibodies employed here, or a masking of the epitope by the interaction, or simply a low level of endogenous interaction. Although the precise molecular mechanisms underlying ALK activation of STAT3 are unknown, it is clear from these studies that loss of STAT3 activity, either by RNA interference-mediated knockdown of STAT3 or by addition of STAT3 inhibitors [41][42][43], reduces the growth and viability of four neuroblastoma lines. One key component examined here was the effect of STAT3 activity on MYCN levels. Recent work has highlighted the cooperative roles of ALK and MYCN in neuroblastoma, with ALK impacting on both the level of MYCN transcription and the stability of the MYCN protein itself [34][35][36][37]. Given these reports, our finding here that STAT3 is important for changes in MYCN levels in response to ALK signaling is significant. Indeed, inhibition of STAT3 by addition of inhibitors, or by RNA interference, reduced MYCN levels even in the presence of activated ALK.
In conclusion, our phosphoproteomic analysis identifies a number of phosphorylation sites targeted by activation of the ALK RTK. One prominent target identified in this analysis was STAT3, which we show here is important for the regulation of MYCN downstream of activated ALK. From our investigations, it is also clear that there is a difference between the ability of wild-type ALK to activate STAT3 and that of oncogenically activated forms of ALK. The molecular mechanisms underlying this difference are not known, and further work is required to understand this. This work suggests that STAT3 inhibition may be a viable approach in the regulation of MYCN activity in neuroblastoma cells, and may have potential therapeutic value in the future.

Phosphotyrosine profiling
Tyrosine phosphorylation profiling of PC12 cells expressing human ALK with and without activation of ALK was undertaken by immunoaffinity purification with P-Tyr-100 (Cell Signaling Technology, Danvers, MA, USA), followed by LC-MS/MS. This was performed as previously described [29,57], except that 20 and 0.8 pmol of the stable isotopelabeled phosphotyrosine peptides EH[13C 3 ,15N 1 -A]LLA-pYTLGVK and HTDDEMTGpYV[13C 3 ,15N 1 -A]TR, respectively, were spiked into each cell lysate prior to phosphopeptide purification. Mass spectra of two technical replicates were searched against a rat protein database supplemented with the sequence for human ALK (extracted from UniProtKB/TrEMBL ver. 2013_01; 41 749 reviewed and unreviewed sequences) with MAXQUANT version 1.3.0.5. Standard settings were used in MAXQUANT (false discovery rate of < 1% at the protein, peptide and modification site levels), except that the 'match between runs' function was enabled. A list with high-confidence phosphorylation site identifications was generated by filtering the 'Phospho (STY) sites' output table for Localization probability > 75% and Score difference > 5. Intensity values for phosphorylation sites were normalized between samples by using correction factors based on the intensity values of the phosphotyrosine sites of the heavily labeled peptide standards. Changes in intensity ratios between cells with activated ALK and control cells of ≥ 5 were considered to be significantly altered.

Network analysis
Rat-human orthologs were extracted from ENSEMBL. The protein-protein interactions among proteins of interest were retrieved from the Protein Interaction Network Analysis platform [58], and substrate-kinase relationships were downloaded from the PhosphoSitePlus database [59]. CYTOSCAPE [60] was used for visualization of networks.

Cell proliferation
CLB-GA (0.035 9 10 6 per well), CLB-GE (0.025 9 10 6 per well), CLB-BAR (0.05 9 10 6 per well) and Kelly (0.035 9 10 6 per well) cells were seeded in collagen-coated 48-well plates overnight, and, starting on the following day, treated with FLLL32 (1.5 lM) for 0-5 days. Cell viability was determined at 0-5 days after drug treatment, with 55 lM resazurin (Sigma, Stockholm, Sweden) [61]. After 2 h at 37°C, the amount of metabolized resazurin was analyzed as relative fluorescence with an Infinit200 plate reader (TE-KAN, M€ annedorf, Switzerland). Results are from one of three representative experiments, with each experiment being performed in triplicate. Results are presented as fold relative fluorescence from FLLL32-treated cells versus relative fluorescence from control untreated cells.

Luciferase assay
Cells (2 9 10 5 ) of the neuroblastoma cell lines CLB-GE and CLB-BAR were transfected with empty pGL2 (control) or MYCNP-luciferase, by the use of Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Cells were then serum-starved, and STAT3 was inhibited by employing 2.5 lM FLLL32 (generated in the laboratory of P.-K. Li) or STATTIC (Sigma, St Louis, MO, USA) for 12 h. Each sample within an experiment was analyzed in triplicate, and the experiment was carried out three times. Results are presented as relative luciferase activity, where untreated samples transfected with empty pGL2 vector were set to 1.

Supporting information
Additional supporting information may be found in the online version of this article at the publisher's web site: Table S1. Phosphotyrosine profiling of human ALK in PC12 cells. Fig. S1. Verification of phosphorylated downstream target by active ALK in neuroblastoma cell lines. Fig. S2. Immunoprecipitation of ALK and STAT3. Fig. S3. Endogenous ALK and MYCN expression in the cell lines employed.