Research ResourceCancer

Phosphoproteome and gene expression profiling of ALK inhibition in neuroblastoma cell lines reveals conserved oncogenic pathways

See allHide authors and affiliations

Science Signaling  20 Nov 2018:
Vol. 11, Issue 557, eaar5680
DOI: 10.1126/scisignal.aar5680

Alternatives to ALK in neuroblastoma

Neuroblastoma is a common pediatric solid tumor that is often driven by oncogenic mutations or rearrangements of the gene encoding the tyrosine kinase receptor ALK. In relapsed neuroblastoma, the frequency of ALK mutation is increased, highlighting the importance of understanding ALK signaling in this cancer. Two papers identify alternative targets in ALK-driven neuroblastoma cells. By combining various proteomics analyses with protein-protein interaction networks, Emdal et al. found that IRS2, an adaptor protein in the insulin receptor signaling pathway, linked ALK signaling to neuroblastoma cell survival. Van den Eynden et al. integrated proteomics and gene expression analyses to identify ETS family transcription factors and the MAPK phosphatase DUSP4 as targets of ALK signaling. These papers identify new targets that could be exploited to treat ALK-positive neuroblastoma.


Anaplastic lymphoma kinase (ALK) is a tyrosine kinase receptor that is a clinical target of major interest in cancer. Mutations and rearrangements in ALK trigger the activation of the encoded receptor and its downstream signaling pathways. ALK mutations have been identified in both familial and sporadic neuroblastoma cases as well as in 30 to 40% of relapses, which makes ALK a bona fide target in neuroblastoma therapy. Tyrosine kinase inhibitors (TKIs) that target ALK are currently in clinical use for the treatment of patients with ALK-positive non–small cell lung cancer. However, monotherapy with the ALK inhibitor crizotinib has been less encouraging in neuroblastoma patients with ALK alterations, raising the question of whether combinatorial therapy would be more effective. In this study, we established both phosphoproteomic and gene expression profiles of ALK activity in neuroblastoma cells exposed to first- and third-generation ALK TKIs, to identify the underlying molecular mechanisms and identify relevant biomarkers, signaling networks, and new therapeutic targets. This analysis has unveiled various important leads for novel combinatorial treatment strategies for patients with neuroblastoma and an increased understanding of ALK signaling involved in this disease.


Anaplastic lymphoma kinase (ALK) was originally identified as a fusion protein with nucleophosmin in anaplastic large-cell lymphoma (1, 2). Numerous oncogenic ALK variants have since been identified (3). ALK fusion proteins have been described in many different tumor types, including inflammatory myofibroblastic tumors, non–small cell lung cancer, and diffuse large–B cell lymphomas (4, 5). The full-length ALK receptor is activated by the ligands ALKAL1 (also known as FAM150A or AUG-β) and ALKAL2 (also known as FAM150B or AUG-α) (69). However, in both familial and sporadic neuroblastoma, which is a childhood cancer arising from the sympathetic nervous system, full-length ALK is activated by point mutations, predominantly in the kinase domain (3, 1015). Although activating neuroblastoma mutations are spread throughout the kinase domain, most of the mutations are located in one of three hotspot residues [Phe1174, Arg1275, or Phe1245 (4)] and are observed at increased frequencies in relapsed patients (1618).

ALK-positive neuroblastoma mutants fall into three classes: gain-of-function ligand-independent mutations, ligand-dependent mutations that are not constitutively active and require activation with ALKAL1, ALKAL2, or agonist antibodies, and, lastly, kinase-dead mutations (19, 20). The ability of ALKAL ligands to further activate mutant ALK suggests that dysregulation of the ALKAL ligands may potentially play a role in neuroblastoma (6, 8, 9). ALK is an attractive therapeutic target in neuroblastoma. Various small-molecule ALK tyrosine kinase inhibitors (TKIs) show clinical activity against ALK fusion–positive malignancies in both pediatric and adult patient populations (4, 21). In contrast to the favorable response in pediatric ALK fusion–positive malignancies (22), a phase 1 crizotinib study reported a disappointing response rate of only 1 of 11 ALK-positive patients (23). Additional trials in neuroblastoma are now ongoing to test the efficacy of next-generation ALK TKIs, such as ceritinib and lorlatinib ( Neuroblastoma is a heterogeneous tumor that exhibits many gains and losses of chromosome regions, which may contribute to the complexity of the disease (10, 15, 24, 25).

To identify downstream signaling components regulated by ALK and define feedback loops, cross-talk, and adaptive responses in neuroblastoma cells, we performed phosphoprotein and RNA-sequencing (RNA-seq) analysis in response to treatment with the first- and third-generation ALK TKIs crizotinib and lorlatinib (26, 27). From these data, we generated both phosphoproteomic and gene expression signatures reflecting ALK signaling events in several neuroblastoma cell lines. Phosphoproteomic analysis identified several conserved oncogenic downstream signaling pathways of ALK, similar to those involved in insulin receptor (INSR)/tropomyosin receptor kinase (TRK) and fibroblast growth factor receptor (FGFR) signaling. In addition, signaling events involved in feedback and cross-talk were identified, including modulation of DUSP (dual-specificity phosphatase) family phosphatases. Furthermore, from analysis of the RNA-seq data, several transcription factors were predicted and validated as responsive to ALK signaling, including members of the FOXO (forkhead box O) and ETS (E26 transformation-specific or E-twenty-six) transcription factor families. Together, these results identify ALK signaling targets that increase our understanding of ALK signaling in neuroblastoma.


Mass spectrometry and RNA-seq–based identification of ALK-dependent networks

To investigate differences in gene expression and phosphoproteomic patterns downstream of ALK in neuroblastoma cells, we isolated RNA and phosphoproteins after treating cells with either crizotinib or lorlatinib, which are first- and third-generation ALK inhibitors, respectively (Fig. 1A) (26, 27). Two ALK-dependent neuroblastoma cell lines (CLB-BAR and CLB-GE) were chosen as well as one Ras-driven ALK TKI–insensitive neuroblastoma cell line (SK-N-AS), which was included as a control for off-target effects. The CLB-BAR cell line harbors a deletion of exons 4 to 11 in the region of the ALK locus encoding the extracellular domain and displays relatively high expression of ALK and MYCN proteins due to amplification of the locus. The CLB-GE cell line harbors an ALKPhe1174Val mutation and is amplified at the MYCN/ALK locus. Neither CLB-BAR nor CLB-GE cells have mutations in either p53 or RAS, although they display high NF1 protein levels (28). SK-N-AS cells harbor an activating NRASGln61Arg mutation and reduced levels of NF1, resulting in activation of the RAS-MAPK (mitogen-activated protein kinase) pathway, and have undetectable levels of ALK protein (28, 29).

Fig. 1 ALK inhibition experimental setup.

(A) Three different neuroblastoma cell lines (CLB-BAR, CLB-GE, and SK-N-AS) were cultured in control conditions or in the presence of the ALK inhibitors crizotinib or lorlatinib. Cells were harvested for tyrosine (pTyr) and serine/threonine (pSer and pThr) phosphoproteomic analysis or RNA-seq analysis after 1 and 24 hours, respectively. Pie charts show the number of targeted sites for each drug. n = 1 biological replicate with 2 technical replicates for each cell line and treatment condition. (B) Boxplots show raw ALK phosphorylation signal intensities of all targeted ALK sites in untreated cells or in cells treated with the indicated ALK inhibitor. Phosphorylation signal was not detected in SK-N-AS cells. P values were calculated using Wilcoxon rank sum test. (C) Bar plots show the ALK mRNA normalized counts for all three cell lines in control or ALK inhibitor conditions as indicated, as determined by RNA-seq analysis. n = 1 biological replicate for each cell line and treatment condition.

Before mass spectrometry (MS) analysis, ALK TKI responses were examined in cell lysates by focusing on ALK, extracellular signal–regulated kinase (ERK), and AKT because the activity of these kinases is sensitive to ALK TKI treatment (fig. S1) (3032). We found that cell proliferation was inhibited by crizotinib with a half maximal inhibitory concentration (IC50) of about 240 nM in CLB-GE and 150 nM in CLB-BAR cells and by lorlatinib with an IC50 of 25 nM in CLB-GE and 16 nM in CLB-BAR cells (fig. S2, A and B), confirming our earlier results (28, 30, 31, 33, 34). On the basis of these results, we used 250 nM crizotinib or 30 nM lorlatinib for further experiments. As mentioned above, levels of ALK were undetectable in SK-N-AS control cells, and the phosphorylation of ERK1/2 was similar in untreated and ALK TKI–treated cells (fig. S1). In addition, we performed cell cycle and apoptosis analyses of cells in response to treatment with ALK TKIs. After 24 hours, ALK TKI–treated cells exhibited hallmarks of inhibition of cell cycle progression as well as induction of apoptosis. Treatment of ALK-addicted CLB-BAR and CLB-GE cells with either crizotinib or lorlatinib resulted in reduced numbers of cells in S phase and increased levels of cleaved poly(adenosine diphosphate–ribose) polymerase (PARP), effects not seen in the SK-N-AS control cell line (fig. S3, A and B).

To identify sites with altered phosphorylation associated with ALK activity, we used an immunoaffinity-coupled liquid chromatography–tandem mass spectrometry (LC-MS/MS) approach after treating all three cell lines with either crizotinib or lorlatinib for 1 hour (Fig. 1A). In crizotinib-treated cells, signals from 2583 phosphotyrosine, 5013 phosphoserine, and 2223 phosphothreonine residues derived from 3345 different proteins were identified. Treatment using the next-generation ALK TKI lorlatinib resulted in signals from residues in 2252 proteins (2568 phosphotyrosine, 2562 phosphoserine, and 1103 phosphothreonine residues) (Fig. 1A). In parallel, we performed RNA-seq analysis from the same cell lines at 24 hours after treatment with either crizotinib or lorlatinib and measured expression in 19,291 coding genes (Fig. 1A). In agreement with our immunoblotting analysis (fig. S1), LC-MS/MS analysis revealed decreased tyrosine phosphorylation of ALK (Fig. 1B and data file S1). Analysis of ALK expression using RNA-seq revealed an increase in ALK mRNA expression upon treatment of CLB-BAR cells with either crizotinib or lorlatinib. This effect was not observed in either the CLB-GE or SK-N-AS cell lines (Fig. 1C).

MS analysis on ALK TKI treatment of neuroblastoma cells

MS-based analysis identified a phosphoproteomic signature downstream of ALK signaling that was common to both crizotinib and lorlatinib in CLB-BAR cells (Fig. 2A). This signature consists of 56 proteins with decreased phosphorylation [proteins with at least one site with a log fold change (FC) decrease of 1.5 or higher after treatment with both drugs], including ALK itself, and several adaptor proteins such as FRS2/3, IRS2, SHC1 to 3, and GAB1/2 (data file S1). Other proteins identified in our analysis included kinases such as ERK1/2 (also known as MAPK1/3), maternal embryonic leucine zipper kinase, and AKT family members as signaling components downstream of ALK (Fig. 2A and data file S1). Other kinases, phosphatases, and transcriptional regulators that showed decreased phosphorylation in response to both ALK TKIs in CLB-BAR were ATR and NEK9, TNK2, SHP2 (also known as PTPN11), Rictor, BRD2, NDRG, and members of the ETV, FOXO, and ERF transcription factor families (data file S1). A few proteins (such as ZNF496 and YAP1) displayed increased phosphorylation upon inhibition of ALK with both crizotinib and lorlatinib, whereas a substantial number of proteins showed increased phosphorylation upon treatment with lorlatinib but not crizotinib (such as TCF4, TP53BP1 at multiple sites, UHRF1BP1, EP400, VGF, and IRS2 at specific residues) (data file S1). Crizotinib treatment resulted in increased phosphorylation at specific sites in a few proteins such as IRS2 (at Tyr919), RABEP1, BCL9, MAP1B, actin binding LIM protein 1, and AGAP3 (data file S1). To provide an overall picture of the signaling pathways regulated by ALK from the peptides identified in ALK TKI–treated cells, we used a gene set enrichment analysis (GSEA) on the 56 proteins with decreased phosphorylation. This analysis showed enrichment of proteins involved in the FGFR (21.4% of the 56 proteins with decreased phosphorylation), INSR (17.9%), and nerve growth factor (NGF) (19.6%) signaling pathways (Fig. 2B, fig. S4A, and data file S2). A GSEA on the hyperphosphorylated proteins after treatment with either lorlatinib or crizotinib did not result in any enriched pathways at 10% false discovery rate (FDR), suggesting that these increased phosphorylation responses were rather nonspecific (data file S2).

Fig. 2 Phosphoproteomic analysis after ALK inhibition in CLB-BAR cells.

(A) Correlation between the effects of the ALK inhibitors crizotinib and lorlatinib on protein phosphorylation states. Each dot represents a phosphorylation site. Targeted ALK sites are indicated in blue. Proteins with the largest response to both drugs are indicated with the targeted site in parentheses. Thresholds used to determine differential phosphorylation (−1.5/1.5) are indicated by dashed lines. (B) Reactome GSEA on all 56 proteins that were identified with decreased phosphorylation levels upon ALK inhibition. The 10 most significantly enriched pathways are shown and ranked on the basis of FDR values. GSEA was performed using Fisher’s exact test. Results from the complete analysis are available in data file S2. (C) Schematic overview of ALK showing the tyrosine residues in the ALK intracellular domain that are modulated in response to ALK inhibitor treatment. (D) Effect of ALK inhibition on other receptor tyrosine kinases (RTKs). Median values are indicated by horizontal dashed lines. n = 1 biological replicate with 2 technical replicates for each cell line and treatment condition.

Of the 56 proteins with decreased phosphorylation in CLB-BAR cells, only 23 showed decreased phosphorylation in CLB-GE cells after treatment with either crizotinib or lorlatinib (using the −1.5 log FC threshold; fig. S4D). Apart from ALK, proteins with decreased phosphorylation included SHC3, FRS2, ERK1, GAB1, and GAB2. Other ALK downstream signaling molecules identified in the CLB-BAR cell line, such as RICTOR, were identified with only one ALK TKI in CLB-GE cells (data file S1). GSEA on the phosphoproteomic signature downstream of ALK signaling in CLB-GE cells revealed only a weak enrichment of the pathways mentioned above, which were not significant at 10% FDR (data file S2 and fig. S4B). In the non-ALK–dependent SK-N-AS cells, only one protein (AKT3) showed decreased phosphorylation after treatment with either TKI (fig. S4, C and D, and data file S2).

The previous analysis focused on identifying proteins with decreased phosphorylation based on a response to both crizotinib and lorlatinib, using a log FC cutoff of −1.5. Because several residues were only measured using one drug and not the other, several proteins were not included in the analysis. Some of these proteins showed a pronounced response to TKI treatment in CLB-BAR cells, such as AKT1 (−5.7), AKT2 (−4.6), and DUSP4 (−4.5) for crizotinib and SHC4 (−4.9) for lorlatinib (data file S1). Therefore, we also included those proteins that contained sites with a log FC value below −4. Accordingly, 18 additional proteins were identified as showing decreased phosphorylation in response to ALK TKI treatment, leading to a total number of 74 proteins (fig. S5, A to C). The enrichment for the insulin, FGFR, and NGF pathways became stronger when this more complete set of 74 proteins was used for the GSEA (fig. S5D).

Tyrosine phosphorylation sites have been identified by MS-based analyses in the intracellular domain of ALK, both in ALK fusion proteins and in the full-length protein (3538). Among the phosphorylation sites mapped by MS in response to crizotinib or lorlatinib, we identified all of the previously observed 11 phosphotyrosine sites in the ALK intracellular domain (Fig. 2, A, C, and D, and data file S1). The most pronounced decrease in phosphorylation was observed at sites 1078, 1096, 1278, 1282, and 1283 (Fig. 2A). The latter three sites are located in the activation loop and are important for the activation of full-length ALK (39). We also examined the effect of ALK TKI treatment on other RTKs. We observed that crizotinib treatment resulted in a specific decrease in phosphorylation at tyrosine residues in ALK, whereas lorlatinib treatment was more promiscuous, resulting in decreased phosphorylation of discoidin domain receptor 1 (DDR1), DDR2, and insulin-like growth factor-1 receptor (IGF1R)/INSR among others (Fig. 2D), which occurred in CLB-BAR and CLB-GE cells, but not in the SK-N-AS cell line (data file S1).

To examine whether decreased phosphorylation of RTKs other than ALK had a role in neuroblastoma cell growth, we performed analyses of proliferation in response to targeted inhibition of DDR1, epidermal growth factor receptor (EGFR), and IGF1R/INSR. Neither the selective DDR1 inhibitor DDR1-IN-1 (40) nor the selective EGFR inhibitor afatinib affected the growth of these cell lines (fig. S2, C and D). Given the strong insulin pathway enrichment and the decreased phosphorylation of IGF1R/INSR found with lorlatinib, cell lines were treated with either crizotinib or lorlatinib, linsitinib (an IGF1R/INSR inhibitor) (41), or a combination of both. We observed that treatment with linsitinib reduced growth of all neuroblastoma cells, including the SK-N-AS control cell line, whereas crizotinib and lorlatinib were active only in CLB-BAR and CLB-GE cells (fig. S2, A and B). Furthermore, linsitinib had an additive effect on the proliferation reduction induced by crizotinib or lorlatinib (fig. S2, A and B).

RNA-seq analysis in ALK TKI–treated neuroblastoma cells

To investigate the effects of crizotinib and lorlatinib on neuroblastoma transcriptional responses, we performed RNA-seq at 24 hours after treatment. Both drugs resulted in similar expression responses, with 764 genes that were differentially expressed (302 present at decreased expression levels and 462 at increased expression levels) upon treatment with either drug in CLB-BAR cells (Fig. 3A). Among those with decreased expression levels in response to both ALK TKIs were MYCN, ETV4, VGF, SOX21, DUSP4, DUSP6, KRT32, MISP, and TFPI2 (Fig. 3A and data file S3). Although some of the genes observed at decreased expression levels, such as MYCN, have been previously described as transcriptional targets downstream of ALK (31), many have not been investigated in the context of ALK signaling. In addition to down-regulation, we also observed up-regulation of many genes, including SLC12A3, ABCC8, CNTN6, SEMA5A, and GPC5 (Fig. 3A and data file S3). This response was not yet observed at 1 hour after treatment (fig. S6, A and B) or in the other cell lines investigated (fig. S7). Our gene signature was similar to a previously established 77-gene signature of neuroblastoma cells with constitutive ALK signaling treated with TAE684, an ALK TKI that did not enter clinical trials (Fig. 3B) (42, 43). Thirty-one (41%) of the 75 protein-coding genes derived from this gene signature were identified after treatment with both crizotinib and lorlatinib. Eleven additional genes were identified with only lorlatinib, whereas no additional genes were identified with crizotinib.

Fig. 3 Gene expression analysis 24 hours after ALK inhibition in CLB-BAR cells.

(A) Correlation between the effects of both ALK inhibitors on gene expression. ALK (blue), genes encoding factors involved in ALK signaling (green), and genes with the largest expression changes (black) after treatment with both drugs are labeled. Thresholds used to determine differential expression (−1.5/1.5) are indicated by dashed lines. Pearson correlation coefficient is indicated on the top left of the plot. (B) Overlap of differentially expressed genes with the 77-gene ALK signature as described for TAE684 by Lambertz et al. (43). The Venn diagram shows the overlap with genes differentially expressed upon crizotinib and lorlatinib treatment. The bar plot shows the proportions of this gene signature that are differentially expressed after ALK TKI treatment as indicated. Note that only 75 coding genes from the 77-gene signature dataset were used for analysis (the long noncoding RNAs MALAT1 and NEAT1 were excluded). (C) GSEA on all 764 genes that were differentially expressed upon ALK inhibition using four different transcription factor target databases. The five most significantly enriched transcription factors are shown and ranked on FDR values. Results from the complete analysis are available in data file S4. TRANSFAC, TRANScription FACtor Database; TRRUST, Transcriptional Regulatory Relationships Unraveled by Sentence-based Text mining. (D) Differential expression of MYCN targets as determined by GSEA. Bar plots indicate the proportion of MYCN targets (+) and nontargets (−), derived from different databases, as indicated on top, that were found to be differentially expressed. P values were calculated using Fisher’s exact test. n = 1 biological replicate for each cell line and treatment condition.

To relate the phosphorylation changes to this expression response, we next predicted the transcription factors responsible for the observed expression signature. Therefore, we performed a GSEA using four different transcription factor target databases (Fig. 3C), which identified 52 different transcription factors at 10% FDR, including TCF3, FOXO1, FOXO4, ETV4, MEF2A, and JUN (Fig. 3C and data file S4). Unexpectedly, MYCN, which is involved in ALK signaling and was observed at reduced levels upon treatment with both TKIs, was not significantly enriched (Fig. 3D). Repeating the GSEA using a set of specific MYCN targets identified by Cowling and Cole (44) resulted in a small but significant enrichment of MYCN targets among the differentially expressed genes (17.9% of MYCN targets in the differentially expressed gene set, compared to 5.6% of other genes; Fig. 3D).

Because we observed a twofold increase in ALK expression after treatment with either TKI in CLB-BAR cells, but not in other cell lines, we evaluated the effect of TKI treatment on the expression of other expressed RTKs in all investigated cell lines (fig. S8). In CLB-BAR cells, several RTKs, such as MET and NTRK3, showed increased expression, whereas RET showed decreased expression (fig. S8). This effect was confirmed at the protein level, where both crizotinib and lorlatinib treatment resulted in decreased RET levels (fig. S9). The latter finding agrees with the ability of ALK activation to increase RET expression (43, 45). In the other cell lines, the RTK expression signature was different and expression changes were less pronounced (fig. S8).

Integration of ALK phosphoprotein and transcriptional responses in neuroblastoma cells

To identify the main ALK signaling components in neuroblastoma cells based on their response to the ALK TKIs lorlatinib and crizotinib, we performed an integrative biological network analysis using known protein-protein interactions (PPIs). Therefore, the 74 identified proteins with decreased phosphorylation and 53 transcription factors (including MYCN) were mapped to the Human Protein Reference Database (HPRD) PPI network (Fig. 4). Using this PPI network, we determined the 10 shortest paths between the 74 proteins with decreased phosphorylation and 53 transcription factors and compared the lengths of these paths to those in a random model, starting from both the transcription factors and the proteins with decreased phosphorylation. Several statistically significant connections were identified at an FDR of 5%. The largest number and most prominent connections were retrieved between the proteins with decreased phosphorylation (AKT1, ERK1/2, and SHC1) and the transcription factors FOXO1, FOXO4, CEBPB, JUN, and RELA (fig. S10A). The potential involvement of FOXO1 and FOXO4 was confirmed by comparing the number of proteins with decreased phosphorylation in the second-degree neighborhood of each transcription factor to a random model (fig. S10B and data file S5). Additionally, the latter analysis also identified JUND, ESRRA, FOS, KLF11, and TCF3 (all at 10% FDR) as transcription factors with a potential role in ALK signaling.

Fig. 4 Interaction between the signaling proteins with decreased phosphorylation and transcription factors in the HPRD PPI network.

Seventy-four proteins with decreased phosphorylation (red) and 53 transcription factors (blue) were mapped to the HPRD protein interaction network. For visualization purposes, only direct interactions between the identified proteins are shown, with the exception of MYCN. Genes with increased abundance (red borders) and decreased abundance (blue borders) are also indicated. A protein was considered to show decreased phosphorylation when it contained sites with decreased phosphorylation signals in both crizotinib- and lorlatinib-treated cells (log2 FC threshold of −1.5; 56 proteins) or when it contained nonphosphorylated sites in cells treated with either drug when the corresponding site was not measured for the other drug (in this case, log2 FC threshold of −4; 18 additional proteins).

Validation of ETS family transcription factors as ALK signaling targets

In silico analysis of the ALK TKI response expression signature obtained from the RNA-seq analysis predicted the involvement of ETS family transcription factor activity (data file S4). The ETS transcription factors ERF and ETV3 showed decreased phosphorylation after either TKI treatment (data file S1). Because the ETS family of transcription factors have been implicated in the tumorigenesis of solid tumors at the level of gene rearrangement, amplification, and posttranslational modification (46), we analyzed the relationship between ETV3 and ETV4 expression with neuroblastoma survival data from the Sequencing Quality Control (SEQC) cohort of 498 patients using the R2: genomics analysis and visualization platform ( This analysis showed a correlation of increased expression of ETV3 and ETV4 genes with poor prognosis (Fig. 5, A and B). Furthermore, on the basis of our ALK inhibitor data, an ALK gene expression signature was generated as a proxy of ALK activity to segregate the SEQC cohort into two groups: a group of patients predicted to have less ALK activity and a group predicted to have more ALK activity. As expected, this analysis showed not only that high expression of ALK correlated with poor prognosis but also that prognosis was worsened when levels of either ETV3 or ETV4 were also high (fig. S11, A and B). Therefore, we analyzed ETV3 and ETV4 at the protein level in neuroblastoma, because they have not yet been described as targets of ALK activity in neuroblastoma. The phosphoproteomic data showed that ETV3 phosphorylation depended on ALK activity (Fig. 2A and data file S1), predicting that ETV3 may be phosphorylated in response to ALK activation. To validate ALK-dependent phosphorylation of ETV3, we used the IMR-32 neuroblastoma cell line, which responds to the ALKAL ligands for ALK (6, 7). Although phospho-specific antibodies for ETV3 were not available, we detected a more slowly migrating form of ETV3 using pan-ETV3 antibodies that was observed only in the presence of the ALKAL1 ligand and that disappeared in the presence of crizotinib (Fig. 5C). The more slowly migrating ETV3 band was also observed in untreated CLB-BAR and CLB-GE cells, where ALK is constitutively activated (Fig. 5D and fig. S1), and as would be expected for an ALK-dependent phosphorylation event. In support of the slower-migrating ETV3 band representing a phosphorylated form of the protein, this band was not observed on phosphatase treatment (fig. S12). Upon treatment with either crizotinib or lorlatinib, this slower-migrating ETV3 band was no longer observed (Fig. 5D). In response to ALK inhibition, ETV4 protein abundance was decreased, in agreement with our RNA-seq data that identified ETV4 as a transcriptional target of ALK (Figs. 3A and 5D). Knockdown of either ETV3 or ETV4 by small interfering RNA (siRNA) in CLB-BAR cells reduced proliferation, which correlated with loss of either ETV protein in response to siRNA treatment (Fig. 5, E to H). These results suggest that ALK signaling activity affects ETV3 and ETV4 at the level of phosphorylation for ETV3 and at the level of transcriptional initiation for ETV4. Furthermore, both ETV3 and ETV4 appear to be important for neuroblastoma cell growth, because loss of either transcription factor resulted in reduced cell proliferation (Fig. 5, E and G).

Fig. 5 Oncogenic ALK regulates ETV3 and ETV4 in neuroblastoma.

(A and B) Kaplan-Meier event-free survival curves of 498 patients with neuroblastoma from the SEQC cohort stratified according to ETV3 and ETV4 expression. Patients with higher expression are highlighted in blue, whereas patients with lower expression are highlighted in red. The log-rank test P values are indicated. (C) Immunoblotting of lysates from IMR-32 cells pretreated with crizotinib and stimulated with ALKAL1 for 30 min and 6 hours. Actin served as the loading control. n = 3 biological replicates. (D) Western blotting for the indicated proteins in lysates from ALK-positive neuroblastoma cell lines CLB-BAR and CLB-GE treated with crizotinib or lorlatinib for the indicated times. n = 3 biological replicates. (E) CLB-BAR cells were transfected with siRNAs targeting ETV3. Proliferation relative to day 0 and relative to scrambled control transfected cells (SiC) was analyzed after 3 and 6 days using resazurin assay. Data are means ± SD from three independent experiments. RFU, relative fluorescence units. (F) CLB-BAR cells were transfected with either scrambled control or three independent siRNAs targeting ETV3. Lysates were immunoblotted for ALK, pALK, ETV3, and MYCN, and actin was used as a loading control. n = 3 biological replicates. (G) CLB-BAR cells were transfected with siRNAs targeting ETV4. Proliferation relative to day 0 and relative to scrambled control transfected cells was analyzed after 2, 4, and 6 days using resazurin assay. Data are means ± SD from three independent experiments. (H) CLB-BAR cells were transfected with either scrambled control or three independent siRNAs targeting ETV4. Lysates were immunoblotted for ALK, pALK, ETV4, and MYCN, and actin was used as a loading control. n = 3 biological replicates. Immunoblot quantification bar plots show means ± SD. **P < 0.01, ***P < 0.005, paired two-sided Student’s t test.

DUSP4 as an ALK downstream signaling target

Our phosphoproteomic profiling identified many components of the ALK signaling pathway, including core components of the MAPK pathway such as ERK1/2 and phosphatidylinositol 3-kinase (PI3K) that have previously been implicated in neuroblastoma (3, 47). We observed that the MAPK phosphatase DUSP4 (also known as MKP2) was regulated at the level of both phosphorylation and transcription downstream of ALK and that DUSP6 transcript levels were decreased upon TKI treatment (Fig. 3A and data files S1 and S3). DUSP family proteins can dephosphorylate both phosphotyrosine and phosphoserine/phosphothreonine residues within a substrate and have been implicated in the dephosphorylation and inactivation of MAPKs including ERK, JNK (c-Jun N-terminal kinase), and p38 (48). Because DUSP4 plays a role in neuronal development and proliferation of several cancer types (49, 50), we investigated DUSP4 in more detail downstream of ALK in neuroblastoma cells. The R2: genomics analysis and visualization platform ( revealed that high DUSP4 expression is associated with poor prognosis in patients with neuroblastoma (Fig. 6A). In addition, cross-examination of the SEQC cohort with our generated ALK signature showed that high expression of ALK together with high levels of DUSP4 correlated with poor prognosis (fig. S11C). This was consistent with our observation that DUSP4 expression was greater in neuroblastoma cells with high ALK signaling, and DUSP4 expression decreased in response to ALK inhibition by either crizotinib or lorlatinib. Because DUSP4 was only identified upon treatment with crizotinib (data were missing for lorlatinib; data file S1), we investigated DUSP4 protein abundance in CLB-BAR, CLB-GE, and SK-N-AS neuroblastoma cells treated with lorlatinib. After 30 min of treatment with lorlatinib, DUSP4 phosphorylation was decreased, as evidenced by its faster migration speed (Fig. 6B). The slower-migrating DUSP4 band was no longer observed on phosphatase treatment, suggesting that it represents phosphorylated DUSP4 (fig. S12). In addition to modulation of DUSP4 phosphorylation by ALK, extended treatment with lorlatinib for 6 hours resulted in a loss of DUSP4 protein in both ALK-addicted CLB-BAR and CLB-GE cell lines (Fig. 6B). These results suggest that ALK activation phosphorylates and stabilizes the DUSP4 protein and may promote initiation of transcription or stabilization of DUSP4 transcript, resulting in an overall decrease in detectable DUSP4 protein in ALK-addicted neuroblastoma cells treated with ALK inhibitors (Fig. 6B). In the control SK-N-AS cell line, lorlatinib treatment did not alter DUSP4 protein, either at the total protein level or at the posttranslational modification level (Fig. 6B). High levels of phosphorylated DUSP4 coincided with robust ERK1/2 phosphorylation in ALK-positive neuroblastoma cells, suggesting that DUSP4 cannot efficiently dephosphorylate ERK1/2 under these conditions. Our results predicted that ALK activation may induce DUSP4 transcription. To test this, we stimulated IMR-32 neuroblastoma cells with ALKAL1. This resulted in phosphorylation of ALK and ERK1/2 after 30 min of ligand addition, as well as increased DUSP4 protein abundance. At 6 hours after the addition of ALKAL1, DUSP4 protein abundance continued to increase, and ERK1/2 phosphorylation was reduced (Fig. 6C). These results suggest that in neuroblastoma cells in which ALK activity is stimulated by ligand, the predicted feedback loop of ERK1/2 dephosphorylation by DUSP4 is intact. Together, these results validate our observations from the phosphoproteomic and RNA-seq datasets that identified DUSP4 as a signaling component regulated by ALK.

Fig. 6 Lorlatinib reduces DUSP4 protein levels in an ALK-dependent manner.

(A) Kaplan-Meier event-free survival curves of 498 patients with neuroblastoma from the SEQC cohort stratified according to DUSP4 expression. Patients with higher expression are highlighted in blue, whereas patients with lower expression are highlighted in red. The log-rank test P value is indicated. (B) Immunoblotting of lysates from CLB-BAR, CLB-GE, and SK-N-AS cells treated with lorlatinib for 30 min and 6 hours. n = 3 biological replicates. (C) Immunoblots from IMR-32 cells stimulated with ALKAL1 ligand for the specified times. n = 3 biological replicates. Immunoblot quantification bar plots show means ± SD. *P < 0.05, ***P < 0.005, paired two-sided Student’s t test.

Regulation of the subcellular localization of FOXO transcription factors by ALK

From the expression-based in silico analysis, the activity of various transcription factors was predicted to be modulated downstream of ALK (Fig. 3, C and D, and data file S4). Several FOXO family members that are AKT targets (Fig. 7A), namely, FOXO1 and FOXO4, were among the predicted transcription factors (data file S4). Furthermore, FOXO3 showed decreased phosphorylation upon ALK TKI treatment (data file S1). Therefore, we experimentally validated FOXO3 in response to ALK inhibition. The R2: genomics analysis and visualization platform ( showed that low FOXO3 expression was associated with poor prognosis in patients with neuroblastoma (Fig. 7B), in agreement with a previous analysis (51). As above, application of the ALK signature to the SEQC cohort showed that high ALK expression correlated with poor prognosis, which was worsened when levels of FOXO transcription factors are low (fig. S11D). In stimulated cells, FOXO transcription factors are phosphorylated through the PI3K-AKT pathway, as confirmed by our network analysis for FOXO4 (fig. S10, A and B), leading to binding of 14-3-3 proteins and sequestration in the cytoplasm (51, 52). Low levels of FOXO proteins would be expected to impair the ability of this family of transcription factors to suppress growth. Thus, ALK activation would be predicted to lead to the retention of low levels of 14-3-3-bound FOXO in the cytoplasm, whereas treatment with either crizotinib or lorlatinib should result in nuclear translocation of FOXO in neuroblastoma cells (Fig. 7A). To test this idea, we analyzed the subcellular localization of FOXO3 in CLB-BAR and CLB-GE cells. In untreated CLB-GE and CLB-BAR cells, FOXO3 proteins were localized mostly in the cytoplasm (Fig. 7, C and D). In contrast, after 3 hours of treatment with lorlatinib, nuclear localization of FOXO3 was increased 13.4-fold in CLB-GE cells and 2.3-fold in CLB-BAR cells (Fig. 7, C and D).

Fig. 7 ALK regulates subcellular localization of FOXO3.

(A) Schematic representation of FOXO relocalization. (B) Kaplan-Meier event-free survival curves of 498 patients with neuroblastoma from the SEQC cohort stratified according to FOXO3 with event-free survival probability. Patients with higher expression are highlighted in blue, whereas patients with lower expression are highlighted in red. The log-rank test P value is indicated. (C and D) Immunofluorescence staining for both CLB-BAR and CLB-GE neuroblastoma cell lines for FOXO3a with or without lorlatinib and quantification. n = 3 biological replicates. Immunofluorescence quantification bar plots show means ± SD. *P < 0.05, ***P < 0.005, paired two-sided Student’s t test. Scale bars, 10 μm. DAPI, 4′,6-diamidino-2-phenylindole.


To better understand signaling downstream of ALK, we investigated both the signaling and transcriptional events upon inhibition of ALK. Downstream events were characterized by phosphoprofiling and RNA-seq of ALK-dependent and -independent neuroblastoma cell lines after treatment with first- and third-generation ALK inhibitors, represented by crizotinib and lorlatinib, respectively. In response to treatment, most of the tyrosine sites on the ALK RTK itself showed decreased phosphorylation (for example, at Tyr1096, Tyr1278, Tyr1282, and Tyr1283). Our findings agree with previous MS-based studies that have identified tyrosine phosphorylation sites in the intracellular domain of ALK, both in ALK fusion proteins and in the full-length protein (3537, 53). The importance of the phosphorylation sites Tyr1278, Tyr1282, and Tyr1283 in the activation loop, which are functionally important in regulation of the full-length ALK RTK (39), is also highlighted in this analysis.

In addition to ALK, other RTKs were modulated in response to ALK TKI treatment of neuroblastoma cells. Of these, DDR1 and DDR2 exhibited decreased phosphorylation in the activation loop of the kinase domain in response to lorlatinib. There was no effect in the control SK-N-AS cell line that expresses both DDRs, suggesting that the effect of lorlatinib is through ALK. DDRs are activated upon collagen binding and have been implicated in tumor development (54). The DDR1 inhibitor DDR1-IN-1 (40) did not inhibit growth of neuroblastoma cells, and the mechanisms underlying this observation will require further investigation. In addition to DDRs, IGF1R and INSR were sensitive to lorlatinib. We observed abrogation of neuroblastoma cell growth with the IGFR1/INR RTK inhibitor linsitinib (41), which, together with lorlatinib, further decreased their proliferation. Lorlatinib treatment also resulted in increased phosphorylation of the RET RTK, particularly at residue Tyr1062, which has been reported to bind the FRS2 and GAB adaptor proteins (55, 56). Our RNA-seq and protein analyses are in line with reports showing that ALK activation results in increased RET expression in neuroblastoma mouse models and cell lines (43, 45, 57).

The highest-scoring proteins with decreased phosphorylation upon treatment with the inhibitors included signaling kinases such as ERKs and AKTs, members of the ETV and FOXO transcription factor families, and phosphorylated adaptor proteins such as FRS2 and IRS. We generated a 56-protein signature and an extended 74-protein signature on inclusion of proteins only measurable in either crizotinib or lorlatinib treatment. GSEA of both the 56- and the 74-protein signature highlighted several signaling pathways downstream of ALK related to FGFR, INSR, TRK, NGF, and PI3K signaling, which is expected because ALK is a member of the INSR kinase superfamily (58). ALK signaling components that were prominent phosphorylation targets included adaptor proteins such as IRS, SHC, FRS, and GAB that bind to the ALK receptor signaling complex (data file S1) (3, 47). Inhibitory components of the ALK signaling pathway were also identified, such as the E3 ubiquitin ligase CBL (59, 60). Multiple signaling pathways or components of pathways that are activated by ALK, such as components of the PI3K-AKT, CRKL-C3G, MEKK2/3 (MAPK kinase kinase 2/3)–MEK5-ERK5, JAK (Janus kinase)–STAT (signal transducer and activator of transcription), and MAPK pathways, were also observed in our profiling as a downstream target of ALK upon abrogation of its activity (data files 1 to 3).

We performed validation of several ALK signaling components, either identified directly or predicted from our combined analyses. The ETS family of transcription factors (61, 62) govern critical functions in normal cell homeostasis and, when perturbed, contribute to oncogenesis (63). Activation occurs at both the transcriptional and posttranscriptional levels in several cancer types (64, 65). ETV3 and ETV4 abundance is increased in invasive and metastatic solid tumors of breast, lung, colon, pancreatic, and thyroid cancers (46, 64, 66). ALK regulated ETV3 at the posttranscriptional level, but it regulated ETV4 at the transcriptional level. Future investigation should address whether expression of ETV4 could be used as a prognostic biomarker in neuroblastoma.

ALK signaling affected DUSP4 at both the level of phosphorylation and overall protein level, in agreement with reports that ERK phosphorylates and increases DUSP4 protein stability (6769). The phosphorylation of DUSP4 by ERK1/2 does not affect phosphatase activity (69), which may explain the simultaneous presence of high levels of phosphorylated ERK and DUSP4 in both CLB-BAR and CLB-GE cell lines. High DUSP4 abundance may enable maintenance of a delicate ERK signaling balance. Alternatively, DUSP4 has additional enzymatic-independent cellular functions, such as acting as a scaffold protein in JNK signaling (70). Our results from IMR-32 neuroblastoma cells showing that short-term ALKAL1 stimulation increased DUSP4 levels, which correlated with ERK activity, but that long-term stimulation was associated with decreased ERK signaling and high levels of DUSP4 suggest that DUSP4 might be a phosphatase involved in negative feedback for ALK signaling. Together, these results point to a context-dependent function of DUSP4 because neuroblastoma cell lines with constant ALK/ERK signaling maintain high levels of DUSP4 while simultaneously exhibiting high ERK activity.

Various transcription factors were predicted to be regulated downstream of ALK in our expression-based in silico analysis, of which we investigated FOXO3 in neuroblastoma cells. Subcellular localization analysis suggested that ALK activity promoted the retention of FOXO3 transcription factors in the cytoplasm. This is consistent with ALK activation of the PI3K-AKT pathway, which results in the cytoplasmic accumulation of FOXO3a (71, 72). It will be interesting to investigate and validate other transcription factors predicted in our expression-based in silico analysis as functioning downstream of ALK.

A limitation of the comparison of the phosphoproteomic data generated from cells treated with crizotinib and lorlatinib to define ALK-specific responses is the high number of phosphorylation sites that were only detected using one drug and not the other. These missing data for one drug inevitably resulted in false-negative findings, such that some proteins that showed changes in phosphorylation were not included in the 56-protein phosphoproteomic signature. We compensated for this drawback by including sites that were missing for one drug and showed an extreme decrease in phosphorylation for the other drug (using a more stringent FC cutoff of lower than −4; fig. S5). In this way, we identified 18 additional proteins for a total of 74 proteins that showed decreased phosphorylation at minimally one site upon ALK inhibition. The more statistically significant GSEA results that were found using this extended decreased phosphorylation signature underlined the validity of this approach. To integrate the phosphoproteomic response to the transcriptional response, we predicted the involvement of several transcription factors based on the enrichment of their targets in the differential expression signature. This prediction approach is limited by the potentially opposing responses of certain targets to different transcription factors, which may explain why we found only a weak MYCN signature in the expression response. Alternatively, this unexpectedly weak response might also reflect the timing of the expression analysis (24 hours after treatment).

In conclusion, our combined phosphoproteomic and RNA-seq analysis identifies targets that are regulated in response to ALK signaling activity in neuroblastoma. Although we used two neuroblastoma cell lines that show similar proliferative responses to ALK inhibitors, our analysis suggests that many signaling processes are wired differently, and indeed, detailed genetic analysis shows that they exhibit different chromosomal rearrangement profiles (28). In summary, although neuroblastoma is a complex heterogeneous disease, this in-depth investigation of downstream targets of ALK signaling offers future avenues to pursue to inhibit ALK-driven neuroblastoma.


Cell culture and cell treatments

Neuroblastoma cell lines CLB-BAR (gain of function, Δexon4–11 truncated ALK), CLB-GE (gain of function, ALKPhe1174Val mutation), IMR-32 (wild-type ALK, ligand-dependent activation), and SK-N-AS (wild-type ALK) were cultured in complete media, RPMI 1640 supplemented with 10% fetal bovine serum (FBS) and a mixture of 1% penicillin/streptomycin at 37°C and 5% CO2. To validate ETV3 and DUSP4 as ALK signaling targets, 1 × 106 cells were seeded in six-well plates (VWR, catalog no. 10062-892). The following day, the cells were starved for 24 hours and then treated with lorlatinib (30 nM) or ALKAL1 ligand (1 μg/ml) for the times indicated. Cell lysates were harvested after treatment and analyzed by Western blotting.

Phosphoprofiling and RNA-seq sample preparation

Tyrosine/serine/threonine profiling was performed on CLB-BAR, CLB-GE, and SK-N-AS cell lines that were serum starved for 24 hours before treatment for 60 min with crizotinib (250 nM) or lorlatinib (30 nM). Cells were washed once with ice-cold phosphate-buffered saline (PBS), before scraping in urea lysis buffer [20 mM Hepes (pH 8.0), 9 M urea, 1 mM activated sodium orthovanadate, 2.5 mM sodium pyrophosphate, and 1 mM β-glycerol phosphate]. About 20 mg of whole-cell extract was harvested. Cells were sonicated three times for 30 s each with 1 min cooling on ice between each burst, followed by centrifugation at 4°C for 20 min at 20,000g in a Beckman JA-25.50 rotor. Phosphoprofiling was performed by Cell Signaling Technology using techniques described previously (36, 73). Briefly, samples were trypsinized and immunoprecipitated with P-Tyr100 or P-Ser/Thr antibodies (73), and LC-MS/MS analysis was performed as previously described (73). MS/MS spectra were evaluated using a SEQUEST 3G and the SORCERER 2 platform from Sage-N Research (Milpitas, CA, USA) (74). Raw signal intensities for all sites and cell lines, as reported by Cell Signaling Technology, are available in data file S6 and in the PRIDE (PRoteomics IDEntifications) database (accession no. PXD011187). n = 1 biological replicate with 2 technical replicates for each cell line and treatment condition. For RNA-seq experiments, CLB-BAR, CLB-GE, and SK-N-AS cell lines were treated for 24 hours with crizotinib (250 nM) or lorlatinib (30 nM). Total RNA was isolated using the Promega Total RNA Isolation Kit (Promega), and RNA samples were sent to GATC Biotech for analysis. n = 1 biological replicate for each cell line and treatment condition.

RNA-seq data analysis

RNA-seq reads (average number of 44.8 million reads per sample) were aligned to the human hg19 reference genome (excluding alternative haplotype regions) using TopHat2 (75) (options -p 8 -r 100 --library-type fr-unstranded -G). The average alignment efficiency was 93.8%. Gene annotation was done using GENCODE (v19) (76), and genes were quantified using HTSeq (option intersection-strict) (77). Counts were normalized using DESeq2 (78). Only coding genes (19,291 in total) that showed a minimal expression (defined as genes that contained minimal 10 HTSeq counts) within the analyzed cell line were used for downstream analysis. Response to ALK inhibition was quantified by log2 FC (treated/untreated) values. Genes were considered differentially expressed if their absolute log2 FC values were above 1.5.

Phosphoproteomic data analysis

Similar to the RNA-seq data, phosphorylation responses were quantified using log2 FC (treated/untreated) values. Peptides that did not contain the targeted amino acid in their amino acid sequence were excluded from the analysis. Amino acid site-specific responses were defined as the median response of all measured peptides that contained the targeted amino acid. Proteins were considered differentially phosphorylated if their absolute log2 FC values were above 1.5 (unless stated otherwise). For the analysis of related RTKs, a list of 40 different RTKs was downloaded from the HUGO Gene Nomenclature Committee (79).

GSEA and transcription factor predictions

A GSEA was performed to determine the pathways in which the differentially phosphorylated genes were active and to identify the transcription factors underlying the expression changes. Enrichments were determined using Fisher’s exact test and FDR correction using the Benjamini-Hochberg method (80). Reactome pathway information was downloaded from the Molecular Signatures Database v5.2 (81). Different transcription factor target databases were used for transcription factor analyses: TRANSFAC (both the curated and the predicted version) (82), TRRUST v2 (83), and RegNetwork (84). TRANSFAC was downloaded from Harmonizome (85), TRRUST was downloaded from, and RegNetwork was downloaded from Only transcription factors that were expressed in the cell line under analysis and were annotated in AnimalTFDB2.0 (86) were used for enrichment analysis. Transcription factors were identified at 10% FDR.

PPI integrative analysis

To integrate the phosphoproteomic with the transcriptional responses, all the proteins with decreased phosphorylation and the predicted transcription factors underlying the transcriptional response were mapped to the HPRD (87) PPI network. To identify interactions between proteins with decreased phosphorylation and transcription factors, both a shortest-path and a nearest-neighborhood analysis were used. For the shortest-path analysis, the 10 shortest paths in the PPI network were identified between each pair of predicted transcription factor and protein with decreased phosphorylation. To evaluate which pairs had shorter paths than expected by random chance, these 10 paths were compared to the 10 shortest paths between (i) each transcription factor and a random set of 100 nondephosphorylated proteins and (ii) each protein with decreased phosphorylation and a random set of 100 non-enriched transcription factors, both using one-sided Wilcoxon rank sum test. For each pair, the maximum of both P values was used for further analysis, and FDR correction was done using the Benjamini-Hochberg method. In the nearest-neighborhood analysis, all first- and second-degree interaction partners of the identified transcription factors were evaluated for differential phosphorylation using Fisher’s exact test. Transcription factors were identified at 10% FDR. All network analyses were performed using the R igraph package (88).

Kaplan scan

Kaplan scanning was performed within R2 ( on the “Tumor Neuroblastoma - SEQC - 498 - RPM - seqcnb1” dataset. In short, for each gene or other numerical characteristic, R2 calculates the optimal cutoff expression level dividing the patients into a good and a bad prognosis cohort. Samples within a dataset are sorted according to the expression of the investigated gene and divided into two groups on the basis of a cutoff expression value. All cutoff expression levels and their resulting groups are analyzed for survival, with the provision that minimal group number is 8 (or any other user-defined value) samples. For each cutoff level and grouping, the log-rank [as described in (89)] statistical significance of the projected survival is calculated. The best P value and the corresponding cutoff value are selected. This cutoff level is reported and used to generate a Kaplan-Meier graph. The graph depicts the log-rank statistical significance (“raw P”), as well as a P value corrected for the multiple testing (Bonferroni correction) of cutoff levels for each gene.

ALK mRNA signature as proxy for ALK activity

The genes that were differentially expressed upon compound treatment were separated in more or less abundantly expressed genes. These genes can be considered as a proxy for ALK activity (less abundantly expressed genes upon compound administration would be positively associated with ALK activity, whereas more abundantly expressed genes would be negatively associated with ALK activity). We used these two groups of genes to generate a single score for every patient in a cohort, reflecting how well they fit to these genes (compared to all other patients in the cohort) [as described in (90)]. In short, such a score is composed of the average value of more abundantly expressed genes and the average value of less abundantly expressed genes (while controlling for the number of genes), after every gene is normalized with the z score. These scores can then be used as a measure of ALK activity and can be used in analyses like Kaplan scan to segregate a cohort into two groups. Both methods are provided by the R2 platform (

Antibodies and reagents

Pan-ERK1/2 antibody (1:5000, no. 610123) was purchased from BD Transduction Laboratories (Franklin Lakes, NJ), phospho-ALK (1:1000, Y1604, no. 3341), phospho-AKT (1:5000, Ser473, no. 4060), AKT (1:10,000, no. 9272), p44/42 MAPK (1:5000, ERK1/2, no. 4377), FOXO3a (1:200, no. 2497), DUSP4 (1:1000, no. 5149), MYCN (1:1000, no. 9405), β-actin (1:10,000, no. 4970), α-tubulin (1:10,000, no. 2125), and PARP (1:1000, no. 9532) antibodies were from Cell Signaling Technology. ETV3 (1:10,000, no. 176717) and ETV4 (1:1000, no. H00002118-M01) were from Abcam and Abnova, respectively. Horseradish peroxidase–conjugated secondary antibody, goat anti-mouse immunoglobulin G (IgG), and goat anti-rabbit IgG (1:5000) were purchased from Thermo Fisher Scientific. ALKAL1 stimulation was performed as described previously (6, 91). Crizotinib and lorlatinib were from Selleckchem (Munich, Germany).


Cells were lysed on ice with hypotonic lysis buffer [20 mM tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na3VO4, and leupeptin with protease/phosphatase inhibitor cocktail (1 μg/ml) (Cell Signaling Technology)] for 15 min and then centrifuged for 10 min at 4°C. Proteins were separated on 7.5% bis-acryl-tris gels, transferred to polyvinylidene difluoride membranes (Millipore), blocked in 5% bovine serum albumin (BSA) (phosphoprotein blots) or 5% low-fat milk, and immunoblotted with primary antibodies overnight at 4°C. Secondary antibodies were diluted 1:10,000 and incubated with shaking at room temperature for 1 hour. Enhanced chemiluminescence substrates were used for detection (GE Healthcare), and membranes were scanned using LI-COR Odyssey instrumentation. Immunoblots were quantified by normalizing signal intensities to control values. P values were calculated using paired two-sided Student’s t test, as indicated in the figure legends.

Viability assay

Cell viability was assessed as relative redox metabolic activity using a resazurin-based assay. CLB-BAR and CLB-GE neuroblastoma cells (0.4 × 105) were plated on collagen-coated 48-well plates. Cells were incubated with 55 μM resazurin (Sigma-Aldrich) for 3 hours at 37°C. Metabolized resazurin was analyzed by a plate reader (TECAN, Männedorf, Switzerland) as relative fluorescence.

SiRNA transfection

CLB-BAR and CLB-GE cell lines were transfected with one of two duplex siRNAs targeting ETV3 and ETV4 (Stealth RNAi and Silencer RNAi, Invitrogen) according to the manufacturer’s protocols. Cells transfected with scrambled siRNA (Invitrogen) were used as negative controls.


Neuroblastoma cell lines (CLB-BAR and CLB-GE) were seeded on cover glass (no. 1, 12 mm ɸ) precoated with 0.4% solution of type I bovine collagen solution (Advanced BioMatrix, lot no. 7434) in RPMI 1640 culture medium (supplemented with 10% inactivated FBS and penicillin/streptomycin). After 24 hours, cells were treated with lorlatinib (50 nM) for a further 3 hours. Dimethyl sulfoxide was used in controls. Cells were fixed in 4% formaldehyde for 15 min at room temperature and then rinsed three times in PBS for 5 min each. Membranes were permeabilized with 1% Triton X-100 for 5 min and rinsed three times in PBS with Tween20 (PBST) (with 0.5% BSA) for 5 min each. Samples were blocked in blocking buffer for 60 min (5% BSA in PBS) before application of primary antibody overnight at 4°C. Samples were rinsed three times in 1× PBST (with 0.5% BSA) for 5 min each and incubated with fluorochrome-conjugated secondary antibody diluted in antibody dilution buffer for 1 to 2 hours at room temperature in the dark [Alexa Fluor 488 AffiniPure F(ab′)2 Fragment Goat Anti-Rabbit IgG (H+L) (Jackson ImmunoResearch; 1:200) and DAPI (1:1000)], then rinsed twice with 1× PBST (with 0.5% BSA) for 5 min each, and rinsed once in 1× PBS for 5 min. Specimens were mounted with Fluoromount-G (lot no. I1316-W756, SouthernBiotech).

Cell cycle analysis

CLB-BAR, CLB-GEMO, and SK-N-AS cells were seeded in six-well plates (1 × 106 cells per well) precoated with 0.4% solution of type I bovine collagen solution (Advanced BioMatrix, lot no. 7434). Cells were treated with lorlatinib (30 nM) or crizotinib (250 nM) for either 1 hour or 24 hours. The cells were fixed and treated according to the fixed cell cycle–DAPI assay protocol, and cell cycle was determined using the NucleoCounter NC-3000. Data were normalized to cell cycle–specific controls.

Phosphatase treatment

CLB-BAR cells were seeded on six-well plates precoated with 0.4% solution of type I bovine collagen solution (Advanced BioMatrix, lot no. 7434). Cells were treated with lorlatinib (30 nM) for 30 min or not treated before lysis with radioimmunoprecipitation assay buffer. Lysates were treated identically with one sample treated with 1500 U of λ protein phosphatase (P0753S, New England BioLabs Inc.). 4× Laemmli buffer was added to the lysates before Western blot analysis.


Fig. S1. Immunoblot validation of response to ALK inhibition in neuroblastoma cells.

Fig. S2. RTK inhibition effects on three neuroblastoma cell lines.

Fig. S3. Cell cycle analysis and apoptosis analyses upon treatment with ALK inhibitors.

Fig. S4. Phosphoproteomic analysis after ALK inhibition in different neuroblastoma cell lines.

Fig. S5. Identification of 74 proteins with decreased phosphorylation in the CLB-BAR neuroblastoma cell line.

Fig. S6. ALK RNA-seq expression analysis early (1 hour) after ALK inhibition.

Fig. S7. RNA-seq gene expression analysis 24 hours after ALK inhibition in different neuroblastoma cell lines.

Fig. S8. Changes in expression of genes encoding TRKs upon treatment with ALK inhibitors in different neuroblastoma cell lines.

Fig. S9. Effect of ALK inhibition on RET phosphorylation and expression.

Fig. S10. Integrative analysis of ALK-induced phosphorylation signaling and transcriptional response.

Fig. S11. Kaplan-Meier event-free survival curves of 498 patients with neuroblastoma from the SEQC cohort.

Fig. S12. Phosphatase treatment of ETV3 and DUSP4.

Data File S1. Phosphoproteomic responses 1 hour after ALK TKI treatment.

Data File S2. Reactome GSEA on genes encoding differentially phosphorylated proteins.

Data File S3. Expression responses 24 hours after ALK TKI treatment.

Data File S4. Transcription factor GSEA results on 764 differentially expressed genes in the CLB-BAR cell line.

Data File S5. Transcription factor nearest-neighborhood analysis results in CLB-BAR.

Data File S6. Cell Signaling Technology Phosphoscan raw data.


Acknowledgments: We acknowledge the Centre for Cellular Imaging at the University of Gothenburg and the National Microscopy Infrastructure (NMI) (VR-RFI 2016-00968) for providing assistance in microscopy. Funding: This work was supported by grants from the Swedish Cancer Society (BH CAN15/775, RHP CAN15/391, and EL CAN15/541), the Children’s Cancer Foundation (BH 2015-80, BH 2014-150, and RHP 2015-96), the Swedish Childhood Cancer Fund (JG TJ2016-0088 and PR2016-2011), the Swedish Research Council (RHP 2015-04466, BH 2017-01324, and EL 14-3596), the Swedish Foundation for Strategic Research (RB13-0204), the Göran Gustafsson Foundation (RHP2016), the Knut and Alice Wallenberg Foundation (KAW 2015.0144), and the EU Horizon 2020 Marie Skłodowska-Curie Programme grant 675712. Author contributions: J.V.d.E., A.A., and E.L. carried out the bioinformatics analysis. G.U., D.C.-M., J.S., K.R., and J.G. carried out the cell and biochemical analyses. J.K. executed the R2 analysis. J.V.d.E., R.H.P., and B.H. supervised the project and wrote the manuscript with all other authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner (92) repository with the dataset identifier PXD011187. The RNA-seq data have been deposited (ArrayExpress,; accession no. E-MTAB-6654). All other data required to evaluate the conclusions in the paper are in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article