Therapeutic Targeting of Non-Oncogene Dependencies In

Published OnlineFirst April 5, 2019; DOI: 10.1158/1078-0432.CCR-18-4117

Translational Cancer Mechanisms and Therapy Clinical Cancer Research Therapeutic Targeting of Non-oncogene Dependencies in High-risk Neuroblastoma Chen-Tsung Huang1, Chiao-Hui Hsieh2,Wen-Chi Lee2,Yen-Lin Liu3,Tsai-Shan Yang4, Wen-Ming Hsu4, Yen-Jen Oyang1, Hsuan-Cheng Huang5, and Hsueh-Fen Juan1,2,6

Abstract

Purpose: Neuroblastoma is a pediatric malignancy of the potentially effective single agents and drug combinations for sympathetic nervous system with diverse clinical behaviors. high-risk neuroblastoma. Genomic amplification of MYCN oncogene has been shown to Results: Among these predictions, we validated in vitro drive neuroblastoma pathogenesis and correlate with aggres- efficacies of some investigational and marketed drugs, sive disease, but the survival rates for those high-risk tumors of which niclosamide, an anthelmintic drug approved carrying no MYCN amplification remain equally dismal. The by the FDA, was further investigated in vivo.Wealso paucity of mutations and molecular heterogeneity has hin- quantified the proteomic changes during niclosamide dered the development of targeted therapies for most treatment to pinpoint nucleoside diphosphate kinase 3 advanced neuroblastomas. We use an alternative method to (NME3) downregulation as a potential mechanism for its identify potential drugs that target nononcogene dependen- antitumor activity. cies in high-risk neuroblastoma. Conclusions: Our results establish a gene expression–based Experimental Design: By using a gene expression–based strategy to interrogate cancer biology and inform drug discov- integrative approach, we identified prognostic signatures and ery and repositioning for high-risk neuroblastoma.

Introduction MYCN-mediated transcriptional program (6, 7), the regulators of MYCN protein stability (8), or the downstream effects of MYCN Neuroblastoma is a childhood cancer of the peripheral sym- amplification (9, 10) have shown clinical promise. In addition to pathetic nervous system. Several clinical and biological variables, MYCN (amplified in 20% of neuroblastomas), several genomic including age at initial diagnosis, stage of disease, and amplifi- alterations, for example, ALK-activating mutation (10%; ref. 11), cation of the MYCN oncogene, are used to stratify patients into ATRX-inactivating mutation or deletion (10%; ref. 12), and TERT neuroblastoma risk groups (1). Although the survival rates from promoter rearrangement (25%; ref. 13, 14), have been described neuroblastoma have been improved substantially in recent dec- in aggressive neuroblastoma, among which ALK is currently the ades, children bearing high-risk tumors, regardless of the presence only tractable oncogene for targeted therapy (15). Relapsed of amplified MYCN, still have poor outcomes (2, 3). Owing to the neuroblastoma, by contrast, was found to harbor more druggable lack of recurrent mutations and heterogeneity of mutational mutations, most of which converged on the activation of the RAS– spectrum in neuroblastoma (4), current treatment approach for MAPK pathway (16, 17). Despite tremendous advances in under- high-risk diseases is largely based on intensive combination standing the cancer genome, new therapeutic approaches to chemotherapy, radiotherapy, stem cell transplant, immunother- effectively treat this heterogeneous, aggressive disease are in high apy, and differentiation therapy (1–3). Although MYCN has a demand. well-established role in neuroblastoma development, pharma- Paralleling the dedicated efforts of researchers to identify driver ceutical targeting of this oncogenic transcription factor remains mutations that confer selective growth advantage, the importance challenging (5). However, alternative strategies that target the of acquired dependencies of cancer cells on the activities of certain nonmutated genes has been increasingly recognized (18, 19). This cancer's addiction to both oncogenes and nononcogenes is 1Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan. 2Institute of Molecular and Cellular Biology, required to sustain the hallmark capabilities and tumorigenic National Taiwan University, Taipei, Taiwan. 3Department of Pediatrics, Taipei state (20). In particular, leveraging the nononcogene addiction for Medical University Hospital, Taipei, Taiwan. 4Department of Surgery, National therapeutic intervention has thus far proved beneficial in cancer Taiwan University Hospital, Taipei, Taiwan. 5Institute of Biomedical Informatics, treatment (18, 19, 21, 22). 6 National Yang-Ming University, Taipei, Taiwan. Department of Life Science, Here, we used a gene expression–based approach to identify the National Taiwan University, Taipei, Taiwan. cancer-related transcriptional signatures and inform potential Note: Supplementary data for this article are available at Clinical Cancer therapeutics for treating high-risk neuroblastoma. This was Research Online (http://clincancerres.aacrjournals.org/). achieved by correlating gene expression signatures between Corresponding Authors: Hsueh-Fen Juan, National Taiwan University, 1, Sec. 4, high-risk neuroblastoma and small-molecule perturbations Roosevelt Rd., Taipei, 106, Taiwan. Phone: 8862-3366-4536; Fax: 8862-2367- (23). In this study, we first performed integrative analysis of the 3374; E-mail: [email protected]; and Hsuan-Cheng Huang, transcriptomes of primary neuroblastoma tumors obtained from [email protected] multiple Gene Expression Omnibus (GEO) datasets. This process doi: 10.1158/1078-0432.CCR-18-4117 led to the identification of gene signatures that were prognostic for 2019 American Association for Cancer Research. patient survival in an independent cohort, especially for children

www.aacrjournals.org OF1

Huang et al.

combined gene expression matrix (11,939 by 1,065) was taken by Translational Relevance the model as input with hyperparameters set as follows: NUM High-risk neuroblastoma has few recurrent somatic muta- (number of topics) ¼ 2; GAMMAPARAM (scale b for the gamma tions and is still associated with poor outcomes despite inten- distribution) ¼ 4; BETA (beta for the Dirichlet distribution) ¼ 10; sive treatment. The lack of druggable oncogenes continues to and number of interactions ¼ 100. After this cross-platform restrict the development of targeted therapies for high-risk normalization, the resultant matrix was then quantile-normalized neuroblastoma. Here, we use an alternative, gene expression– again to ensure that the gene expression distributions of all based approach to identify potential drugs that target the samples were identical. nononcogene dependencies in high-risk neuroblastoma. To assess the quality of data after PLIDA transformation, we Among the top predicted drugs by this approach, our work computed the following measures: (i) sample-wise principal then investigates an FDA-approved anthelmintic drug, niclo- component analysis (Supplementary Fig. S1A); (ii) a Spearman samide, as an effective treatment for neuroblastoma through correlation coefficient (SCC) between a given gene vector the regulation of nucleotide biosynthesis and nucleoside before PLIDA and the same gene vector after PLIDA for each diphosphate kinase 3 (NME3) protein. dataset category (within individual or across all GEO datasets; Supplementary Fig. S1B); (iii) SCCs between a gene and any other genes before and after PLIDA for each dataset category (Supplementary Fig. S1C); and (iv) gene expression values before and after PLIDA for each dataset category, for which a with advanced-stage, MYCN-nonamplified tumors. We used the 1-way ANOVA or KruskalWallis test was applied to determine high-risk neuroblastoma gene signature to predict effective drugs whether the mean or median expression values of a gene were or their combinations and validated some of these predictions in equivalent across all dataset categories, respectively (Supple- vitro. The in vivo efficacy of niclosamide, an anthelmintic drug mentary Fig. S1D). approved by the FDA to treat tapeworm infections, was confirmed in neuroblastoma xenograft models. By further investigating the Clustering patients with high-risk neuroblastoma without neuroblastoma proteome following niclosamide treatment, we MYCN amplification identified downregulation of nucleoside diphosphate kinase 3 We used a gene expression intensity-based similarity met- (NME3), an enzyme involved in the nucleotide biosynthesis, as a ric (26) to compute pairwise similarities among the patients with potential molecular mechanism of the drug's effects. Given the high-risk, MYCN-nonamplified neuroblastoma (HR-nonMNA) rarity of actionable mutations, our data present an alternative for clustering analysis (Supplementary Fig. S2A). This intensity- solution to target-based drug screening in this deadly pediatric based similarity metric has proved superior to other commonly neoplasm. used metrics derived from the Pearson correlation or Euclidean distance in a clustering task of drugs with diverse mechanisms of Materials and Methods action (MoA). In brief, for each HR-nonMNA sample (MYCN-nonamplified, Data source, cross-platform normalization, and quality stage 4, >18 months; n ¼ 156), we subtracted the median measures expression vector of all low-risk (LR) samples (MYCN- Six gene expression datasets from different microarray plat- nonamplified, stage 1 or 2, <18 months; n ¼ 247) from itself to forms containing primary neuroblastomas were obtained from derive a "differential profile." The differential profiles of all GEO with accession numbers listed as follows: GSE45547 HR-nonMNA samples were then applied to our gene expression (Agilent-020382 Human Custom Microarray 44k; n ¼ 649), intensity–based similarity framework to obtain the optimal GSE3446 (Affymetrix Human Genome U133A Array; n ¼ 117), parameter set for the intensity-based similarity metric (query GSE19274 (Illumina human-6 v2.0 expression beadchip; gene set size b and decay factor s; Supplementary Fig. S2B). n ¼ 100), GSE16254, GSE12460, and GSE16237 (Affymetrix Instead of using the F1 score, an external clustering validity Human Genome U133 Plus 2.0 Array; n ¼ 88, 64, and 50, index given a ground-truth answer, we used the silhouette respectively). For each dataset, probe set IDs were mapped to score, an internal clustering validity index based on the statis- gene names using an available R package (hgu133b.db, illumi- tical properties of a clustering, because in this case there is no naHumanv2.db, or hgu133plus2.db) or a GEO platform (GPL) gold standard for these patients with HR-nonMNA. The clus- annotation file (GPL16876 for GSE45547). For each sample, the tering produced by the best silhouette score has proved being median log expression value was taken for each gene mapped to highly correlated with that by the best F1 score in biomedical multiple probes, and no missing value was found across all data analysis (27). samples (negative expression values were replaced by a zero). The genes shared among all datasets were selected and combined Differential expression analysis into a matrix, followed by quantile normalization. This data For each comparative category ("MNA," "HR-MNA," "HR- integration process resulted in an intersection of 11,939 genes nonMNA-subgroup1," or "HR-nonMNA-subgroup2," defined in 1,065 primary neuroblastomas. by age at diagnosis, INSS tumor stage, and MYCN amplifica- For cross-platform data normalization, we used platform-inde- tion), we performed differential gene expression analysis using pendent latent Dirichlet allocation (PLIDA; ref. 24), which uses Significance Analysis of Microarrays (ref. 28; SAM; 2-class the generative probabilistic model latent Dirichlet allocation (25) unpaired MannWhitney U test with 1,000 sample-level per- to learn topic model decomposition from gene expression data- mutations). The differentially expressed (DE) genes (the 90th sets from multiple platforms. The PLIDA model was learned using percentile FDR < 0.001) for each comparative category are the MATLAB code released by Deshwar and Morris (24). The provided in Supplementary Table S1.

OF2 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

Pathway–gene association analysis Genetic perturbation analysis We devised an approach based on gene set enrichment anal- We have generated hundreds of thousands of recurrent pertur- ysis (GSEA; ref. 29) to identify genes that might closely associate bation–transcript regulatory associations among >7,000 chemical with the biology of high-risk neuroblastoma (Supplementary and genetic perturbagens and 12,494 transcripts across 10 cell Fig. S3). We performed GSEA analysis as described previous- types, while demonstrating the robustness of these recurrent ly (29) using the KEGG pathway gene set collection (MSigDB relationships in general against cell-line variability (23). For each v5.0, C2 collection) with default settings (1,000 sample-level comparative category, we combined those recurrent regulatory permutations; the minimum and maximum gene set sizes for associations of genetic perturbation type [corresponding to the consideration were set to 10 and 500, respectively). Pathways are exposure to short hairpin RNAs (shRNA) for 96 hours (sh96)] called statistically significant if FDR q values are <0.25. For each with the DE genes to infer gene regulatory relationships in high- comparative category, we (i) performed "group-wise" enrich- risk neuroblastoma (Supplementary Fig. S4). The inferred gene ment analysis by testing the pathways enriched between the regulatory relationships for each comparative category are pro- source (advanced) and reference patient groups, (ii) performed vided in Supplementary Table S3. "gene-specific" enrichment analysis by testing the pathways enriched among those patients in the source patient group with Generating gene expression signatures the highest 25% and the lowest 25% expression of a given gene For each comparative category, we first scored the genes from for each of the top 500 upregulated and 500 downregulated DE integrated transcriptomic analysis. For differential expression genes, and (iii) computed the proportion of enrichments for analysis, we computed the fold change for an upregulated gene each gene as the ratio of the number of "consistently" enriched or the inverse of fold change for a downregulated gene times the pathways in the gene-specific and group-wise analyses over the square root of the absolute value of the penalized t-statistic number of enriched pathways in the group-wise analysis (Sup- reported by SAM. For pathway–gene association analysis, we plementary Fig. S3A). A pathway is called consistently enriched mapped the proportion of enrichments (0%100%) linearly to in the gene-specific and group-wise enrichment analyses if (i) it [0, 80] and the best FDR q value to [0, 20] as follows: [0.2, 0.25) to has a positive normalized enrichment score (NES) in the group- 2; [0.1, 0.2) to 4; [0.05, 0.1) to 6; [0.01, 0.05) to 10; [0.001, 0.01) wise analysis and has a positive NES for an upregulated gene or a to 14; [0.0001, 0.001) to 18; and [0, 0.0001) to 20. For gene negative NES for a downregulated gene in the gene-specific correlation analysis, we calculated twice the sum of the absolute analysis; or (ii) it has a negative NES in the group-wise analysis value of correlation scores between a given gene and all its first and has a negative NES for an upregulated gene or a positive NES neighbors. For genetic perturbation analysis, we counted the for a downregulated gene in the gene-specificanalysis.The number of directed links between a given gene and its first enriched pathways and the genes that contributed to the con- neighbors weighted by a factor of 0.2. In addition, we also sistent enrichments for each comparative category are summa- incorporated human protein interactome from Menche and col- rized in Supplementary Fig. S3B. leagues (34), which comprised a union of 13,460 proteins with 141,296 physical interactions, with the genes included in our Gene correlation analysis integrated transcriptomic analysis and then counted the number We used high-dimensional undirected graph estimation of physical interactions between a given gene and its first neigh- (HUGE; ref. 30, via R package HUGE version 1.2.7), a Gaussian bors weighted by a factor of 0.1. We note that this scoring scheme graphical model that describes probabilistic relationships puts far more emphasis on pathway–gene association analysis between variables in a multidimensional manner, to infer gene than other components such that a gene with the proportion correlation relationships in high-risk neuroblastoma. For each of enrichment of 100% and the best FDR q value < 0.0001 from comparative category, the top 500 upregulated and 500 down- pathway–gene association analysis has already contributed a regulated DE genes were considered and an estimator of popu- score of 100 to the aggregate score. The aggregate score for each lation inverse covariance matrix (also known as concentration or gene in each comparative category is provided in Supplementary precision matrix) was learned using the graphical lasso algorithm Table S4. These aggregate scores were used to generate the gene (ref. 31; function HUGE with method, "glasso"; nlambda, 100; signatures of high-risk neuroblastoma in Fig. 2A and Supplemen- lambda.min.ratio, 0.1), which is aimed to maximize Gaussian log tary Figs. S6A and S7A (cutoff >50), in which the genes were likelihood with l1 regularizer for a sparse solution (i.e., many ordered by the geometric mean of the ranks of aggregate scores. entries in the matrix will be zero). The parameter l was learned The significant enrichments of the entire MSigDB gene set collec- using stability approach to regularization selection (ref. 32; stARS; tion (v5.0) using hypergeometric test (Benjamini–Hochberg function HUGE.select with criterion, "stars"; stars.thresh, 0.05; (BH)-corrected P < 0.05) for each of gene signatures are provided rep.num, 20), which has been shown to generate reasonable in Supplementary Table S5. sparsity in the graph. The resulting estimator of population inverse covariance matrix was then used to derive the estimates Prognostic significance of the gene expression signatures of population covariance matrix and correlation matrix. The We validated the prognostic value of the generated gene presence of gene correlation relationships was judged by nonzero expression signatures in an independent cohort of 477 patients entries in the estimated population inverse covariance matrix with neuroblastoma (ArrayExpress accession number E-MTAB- and the strength of the relationships was represented by the 179; Agilent Custom Human Neuroblastoma Chip 251496110; corresponding entries in the estimated correlation matrix. The ref. 35). The expression profile (using the processed data from – inferred gene gene correlation relationships for each comparative E-MTAB-179.processed.1.zip) was log2-tranformed (a negative category and their overlap with 1,639 known and likely value was replaced by a zero), and the median value was taken human transcription factors (33) are provided in Supplementary for each gene mapped to multiple probes, generating a 3,834 Table S2. (mapped genes in hg19) by 477 (samples) data matrix, followed

www.aacrjournals.org Clin Cancer Res; 2019 OF3

Huang et al.

by quantile normalization (sample-wise) and then Z score bated at 37 C in humidified atmosphere with 5% CO2, and transformation (gene-wise). For each gene expression signature, routinely passaged when 90%–95% confluent. Cells were nega- we applied partitioning around medoids (PAMs) to the sub- tive for Mycoplasma contamination. DMEM (12800), FBS matrix of the Z-transformed dataset containing only genes in a (A3160601), and trypsin–EDTA (15400054) were purchased signature and determined an optimal dichotomy of patients from Thermo Fisher Scientific. IKK-2-inhibitor-V (S2864), niclo- according to which clustering has the best silhouette from 1,000 samide (S3030), NVP-BEZ235 (S1009), and vinblastine (S4505) random initial assignments of k ¼ 2 medoids. We then selected were ordered from Selleckchem. Pyrvinium pamoate (P0027) was patients according to their metadata (MYCN status, age at obtained from Sigma-Aldrich. Antibodies against NME3 (1:2000, diagnosis, tumor stage, and risk stratification) and estimated GTX108324, GeneTex) and b-actin (1:10,000, MAB1501, Merck) the overall survival (OS), relapse-free survival (RFS), and event- were ordered from commercial vendors. Horseradish peroxidase– free survival (no relapse or progression) curves by the Kaplan– conjugated secondary antibodies were ordered from Abcam. Meier method. The differences of survival curves between Enhanced chemiluminescence reagent was purchased from patient subgroups were compared using the log-rank test (R Bio-Rad. package survival, version 2.38.3). Cell viability assays Expression-based therapeutic discovery MTS assay. A total of 5,000 cells per well were seeded in 96-well To predict effective drug treatments for high-risk neuroblasto- plates for 24 hours followed by drug treatments as indicated. Cells ma, we adapted a gene expression–based method that evaluates were added with 20-mL MTS (G1111, Promega) for 2 hours at the degree of reversal of a disease signature by single agents or their 37 C with 5% CO2 before measurement of the absorbance at 490 combinations through their recurrently regulated transcripts nm (A490) using a ELISA Reader (Bio-Rad). Data are represented across multiple cell types (23). This in silico approach uses small by A490 values with background correction and normalized with molecule–regulated recurring transcripts that more closely repre- the control (DMSO) as 100%. Three technical replicates were sent the unique compound mechanisms and activities to inform performed for each of 3 independent biological replicates effective cancer therapies targeting the nononcogene dependen- (Fig. 3C; Supplementary Fig. S9A). cies manifested in a given disease signature. The algorithm will produce a therapeutic score for each considered single agent or Clonogenic assay. For single-agent experiments (Fig. 3D; Supple- drug combination to provide a rough interpretation of gene mentary Fig. S9B), 4 105 cells per well were seeded in 6-well reversal such that a score of 0.1 corresponds to approximately plates for 24 hours followed by drug treatments as indicated for 72 10% of reversal. A drug combination is considered synergistic if hours. The remaining cells were trypsinized, reseeded at 1,000 the drugs in the combination can reverse a significant fraction of cells per well in new 6-well plates, and allowed to grow for 14 disease genes in a nearly nonoverlapping manner. We have days. Colonies were fixed with 100% methanol, stained with 1% validated the capability of this approach to inform pan-cancer crystal violet, and counted. Three technical replicates were per- synergistic drug interactions, indicating that they are likely to be formed for each of at least 3 independent biological replicates. For effective in a large fraction of patients with cancer. drug-combination experiments (Fig. 4C and D; Supplementary To this end, we selected the high-risk gene signature in this Fig. S10), 1,000 cells per well were seeded in 6-well plates for 24 study (n ¼ 99; Supplementary Fig. S6A) and used the recurrent hours and then treated with drugs as indicated for 12 days during relationships corresponding to small-molecule treatment for 24 which the culture media with or without drugs were replaced every hours as reported previously (23). For each single agent or drug 3 days. Colonies were fixed with 100% methanol, stained with 1% combination, we computed an expression-based therapeutic crystal violet, recorded by a digital scanner, and counted. Relative score (as the extent of reversal of the disease genes) using the growth was derived as the number of colonies in each drug original algorithm, except that the weight of gene g, w(g), was now treatment over that of the control (DMSO), and growth inhibition replaced by w(g) ¼ f(g)path(g), where f(g) ¼þ1or1, if gene g is was computed as one minus relative growth. Two technical upregulated or downregulated in the disease signature, respec- replicates were performed for each of the 2 independent biolog- tively, and path(g) is the mapped score of gene g from pathway– ical replicates. Drug synergy was determined by the delta gene association analysis as described in the "Generating gene score (36), a metric ranging from 1 (antagonism) through 0 expression signatures" subsection. The resulting compounds with (additive effect) to þ1 (synergism), which has been shown to nonzero therapeutic scores are provided in Supplementary Table yield superior accuracy for classifying drug combinations with S6. For combinatorial drug discovery, we only considered the known interaction modes compared with those metrics based on compounds with top 25% therapeutic scores in the list (103 the Loewe additivity or Bliss independence model. The average compounds) for the prediction of drug synergy. The predicted growth inhibition values across biological replicates were used to synergistic drug pairs are provided in Supplementary Table S7. To compute delta scores, using the ZIP function in the R package visualize synergistic drugs, we grouped the drugs by their MoA and synergyfinder (version 1.4.2) with arguments correction ¼ TRUE then ordered the drug groups by the occurrences of MoAs, for and nan.handle ¼ "L4." which drugs within each group were further ordered by their occurrences (Fig. 4A). Apoptosis assay. Atotalof2 106 cells per well were seeded in 10-cm plates for 24 hours and then treated with niclosamide for Cell cultures and chemicals 48 hours. The cells, together with the supernatant, were washed SK-N-DZ and SK-N-BE(2)C cells were obtained from ATCC. with cold PBS twice, stained with FITC–annexin V and propi- SK-N-AS and SK-N-SH cells were provided by Yung-Feng Liao dium iodide (PI; 0.5 mg/mL; 556547, BD Biosciences), and (Institute of Cellular and Organismic Biology, Academia Sinica). incubated for 15 minutes at room temperature. After staining, Cells were grown in DMEM supplemented with 10% FBS, incu- samples were analyzed by flow cytometry (Becton Dickinson).

OF4 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

All experiments were performed independently 3 times Protein extraction, digestion, and stable isotope dimethyl (Fig. 5B). labeling Xenograft tumors were minced in liquid nitrogen using a qRT–PCR assay mortar and pestle. In vivo–powdered samples (n ¼ 2 for each A total of 4 105 cells per well were seeded in 6-well plates condition) and in vitro–cultured samples (2 independent for 24 hours and then treated with niclosamide for 24 hours. biological replicates for each cell line treated with DMSO or Cells were washed with cold PBS twice and lysed with TRIzol niclosamide for 48 hours) were homogenized in lysis buffer Reagent (Invitrogen) and chloroform (Sigma). Total mRNA was containing 50 mmol/L Tris-HCl (Bioman), 1% SDS (BioShop), isolated using Direct-zol RNA miniPrep (Zymo Research) accord- 10% glycerol, and a protease inhibitor (16012540, BioShop) ing to the manufacturer's protocol. Reverse transcription of 1-mg under sonication on ice for 2 minutes (60% amplitude and RNA to cDNA was performed using qRT-PCR Kit (Zymo 0.6 cycle time). The lysates were centrifuged at 16,000 g for Research). The amount of cDNA was quantified by CFX Connect 30 minutes at 4C and protein concentrations of supernatants Real-Time PCR Detection System (Bio-Rad) using iQ SYBR Green were determined by the bicinchoninic acid (BCA) assay. Supermix (Bio-Rad). The following primer sequence pairs were Samples (500 mg) were added with lysis buffer and with triethy- used for analysis: human CCNA2 forward 50-CGCTGGCGGTACT- lammonium bicarbonate (TEAB) to a final concentration of GAAGTC-30 and reverse 50-GAGGAACGGTGACATGCTCAT-30; 50 mmol/L (pH 8.5). Samples were reduced with 5 mmol/L human MCM10 forward 50-GCATGATGGTGTGAAGAGGTTT-30 Tris(2-carboxyethyl) phosphine hydrochloride for 30 minutes and reverse 50-TCCCATTTGTAGAGGCCACAG-30; human at 37C and alkylated with 2 mmol/L S-methyl methanethiosul- ERCC6L forward 50-CAGTTGGTTGGTTCTCCCCA-30 and reverse fonate in the dark for 30 minutes at room temperature. To allow 50-AGGGCCTCCTGGATTTTTCC-30; human KIF20A forward 50- gel polymerization for in-gel digestion, samples were mixed with ACTGCTCTGTCGTCTCTACCT-30 and reverse 50-GGTAACAAGG- 30% acrylamide/bisacrylamide (37.5:1, v/v), 10% (w/v) ammo- GCCTAACCCTC-30; human RUVBL1 forward 50-GGAGGTGAAG- nium persulfate, and tetramethylethylenediamine. The gels AGCACTACGA-30 and reverse 50-ACTATGACGCCACATGCCTC-30; were excised into pieces, washed with 25 mmol/L TEAB, and and human GAPDH forward 50-ACACCCACTCCTCCACCTTTG-30 then 25 mmol/L TEAB/50% (v/v) acetonitrile (ACN), and vor- and reverse 50-GCTGTAGCCAAATTCGTTGTCATAC-30. Data are texed several times. The gel pieces were further dehydrated with DD shown based on the 2 Ct method, with GAPDH used as the 100% ACN and then dried in a centrifugal evaporator. After internal control. Three technical replicates were performed for rehydration with 25 mmol/L TEAB, samples were digested with each of 3 independent biological replicates (Fig. 5A). trypsin at a 1:10 enzyme-to-protein ratio in a 37C water bath overnight. Digested peptides were extracted with 0.1% (v/v) Cell-cycle analysis trifluoroacetic acid (TFA) and then 50% ACN/0.1% TFA for 3 A total of 2 106 cells per well were seeded in 10-cm plates for times, and with 100% ACN. Peptides were desalted on reversed- 24 hours and then treated with niclosamide for 48 hours. Cells phase stage tips (3M). Desalted peptides were dissolved in 400 mL were washed with cold PBS twice, incubated with 10 mg/mL PI of 100 mmol/L TEAB. For stable isotope dimethyl labeling, for 1 hour, and subjected to flow cytometry analysis (Becton peptides were labeled with 16 mL of 4% (v/v) CD2O (for niclo- Dickinson). Three technical replicates were performed for each of samide-treated samples) or CH2O (for control samples) and 3 independent biological replicates (Fig. 5C). added with 16 mL of 0.6 mol/L sodium cyanoborohydride freshly prepared before use. The mixture was agitated in the Animal studies dark for 1 hour at room temperature. The reaction was stopped All experiments using animals were approved by the Insti- by adding 64 mL of 1% (v/v) ammonium hydroxide on ice. tutional Animal Care and Use Committee (IACUC) at Nation- Dimethyl-labeled samples were acidified with 32 mL of 10% al Taiwan University (Taipei, Taiwan). Three- to 4-week-old (v/v) formic acid to a final pH of approximately 3. Differentially male athymic nude (nu/nu) mice (BioLASCO) were used labeled peptides were mixed (1:1), desalted on reversed-phase for animal experiments. Mice were kept under pathogen-free stage tips, and dried in a centrifugal evaporator. conditionsandprovidedwithfoodandwaterad libitum.For To increase sample resolution, peptide samples were separated SK-N-DZ and SK-N-AS xenografts, mice were inoculated sub- by strong cation exchange chromatography. The home-made C8 cutaneously with 5 106 cells in 100 mLof50%Matrigel stage tips were washed with 20-mL methanol, added with buffer B (BD Biosciences) in DMEM. Mice with detectable tumors (0.1% TFA, 80% ACN), and buffer A (0.1% TFA, 5% ACN), showing a maximal diameter of 5 mm or higher were sub- and centrifuged at 1,000 g until all solutions were filtered. The jected to experiments. Tumor volume was calculated using the stage tips were then added with 500 mmol/L ammonium acetate 2 3 formula l w /2 (mm ), where l and w represent tumor (NH4OAc) for 10 minutes followed by centrifugation at 1,000 length and width, respectively. Niclosamide was administered g. Peptide samples were reconstituted in 250 mL buffer A, loaded intraperitoneally at 25 mg/kg daily for 14 days in SK-N-DZ on equilibrated C8 stage tips, and eluted with 0.1% TFA, 15% xenografts and for 28 days in SK-N-AS xenografts. Relative ACN, and an increasing concentration of NH4OAc from 0 to tumor growth was calculated as the ratio of tumor volume at 10, 20, 35, 50, 75, 100, 300, and 500 mmol/L. The eluted fractions the treatment endpoint over that starting at day 0 (Fig. 5D and were combined, acidified immediately with 10% TFA (pH < 3), E). At the treatment endpoint, tumor tissues and major organs desalted on reversed-phase stage tips, and dried in a centrifugal were harvested and stored in liquid nitrogen for histologic evaporator. (Fig. 5F and G; Supplementary Fig. S11) and proteomic analysis (see below). For survival studies, mice were eutha- LC/MS-MS analysis nized with CO2 asphyxia until tumor length reached 15 mm in Peptides were separated with the Dionex Ultimate 3000 diameter (Fig. 5H and I). RSLCnano system (Thermo Fisher Scientific) and analyzed on

www.aacrjournals.org Clin Cancer Res; 2019 OF5

Huang et al.

an LTQ-Orbitrap XL hybrid Mass Spectrometer (Thermo Fisher Western blot analysis Scientific) equipped with a nanoelectrospray ion source (Proxeon Cells were lysed in buffer containing 50 mmol/L Tris-HCl Biosystems). Peptides were dissolved in 10 mL of 0.5% acetic acid (pH 7.4; Bioman), 1% SDS (Bioshop), 10% glycerol, and a and 2% ACN, loaded onto a trap column (100 mm 15 cm) in- protease inhibitor (16012540, Bioshop) immediately added house packed with 3-mm ReproSil-Pur 120 C18-AQ Reverse-Phase before use. After sonication, protein concentrations were deter- Beads (Dr. Maisch HPLC GmbH), and separated at a flow rate of mined by the BCA assay (23225, Thermo Fisher Scientific). Each 500 nL/minute. Solvent A (0.5% acetic acid) and solvent B (0.5% sample with equal protein concentrations was subjected to acetic acid and 80% ACN) were used to separate peptides. The SDS-PAGE and transferred to a polyvinylidene difluoride mem- separation liquid chromatography (LC) gradient increased line- brane. Nonspecific binding was blocked with 5% nonfat milk. arly in the percentage of solvent B from 5% to 40% in 60 minutes, After immunoblotting with the desired primary antibodies, mem- from 40% to 100% in 5 minutes, and at 100% for 10 minutes. The branes were incubated with appropriate secondary antibodies. LTQ-Orbitrap XL system was operated in the positive ion mode Protein bands were detected using the enhanced chemilumines- and in a data-dependent mode to automatically switch between cence reagent (Fig. 6F and G). mass (MS) and tandem mass (MS-MS) acquisition. All MS-MS spectra were produced by the CID method. Full-scan MS spectra were acquired with a resolution of 60,000 at 400 m/z, an auto- Results matic gain control (AGC) target of 1 106, and a mass range from Integrated analysis of the neuroblastoma transcriptome 300 to 1,600 m/z. Ten most intense precursor ions from a survey We used 6 microarray datasets obtained from GEO correspond- scan were selected for MS-MS scan from each duty cycle and ing to an intersection of 11,939 genes in 1,065 primary neuro- detected in the Orbitrap at a AGC target of 1 104 and a blastomas with available clinical information, including MYCN normalized collision energy of 35%. Internal calibration was amplification, tumor stage, age at diagnosis, and risk stratification performed using lock-mass from ambient air (m/z (Fig. 1A and B). To remove cross-platform effects, we used 445.1200025). Dynamic exclusion time was set to 90 seconds. PLIDA (24), a normalization method that builds upon a latent All mass spectra have been deposited to the ProteomeXchange Dirichlet allocation generative model to learn a topic decompo- Consortium with the dataset identifier (PXD010745). sition from multiplatform gene-expression profiles (see Materials and Methods). Several benchmarked quality measures confirmed Proteomic data analysis the ability of PLIDA to eliminate platform-induced systematic Raw MS-MS spectra were searched against the UniProtKB/ differences while preserving platform-specific data structure (Sup- Swiss-Prot human protein database (downloaded on February plementary Fig. S1). 14, 2016) using the MaxQuant (37) software (version 1.5.2.8) Recognizing that patients older than 18 months with high-risk with the following parameters: fixed modification carbamido- stage IV neuroblastoma have a very poor outcome regardless of methyl (C), variable modifications oxidation (M), and trypsin the MYCN amplification status (2, 3), we clustered the subset as the enzyme with up to 2 missed cleavages. The maximum of these patients without MYCN amplification (high-risk, MYCN- mass tolerance for precursor and fragment ions was 3 ppm and nonamplified, HR-nonMNA; n ¼ 156) into 2 patient 0.5 Da, respectively. Peptide pairs were searched with a mul- subgroups, of which 1 (HR-nonMNA-subgroup1; n ¼ 92) tiplicity of 2, with DimethLys0/Nter0 and DimethLys6/Nter6 was marked by aberrant cell-cycle activity and the other (HR- chosen as light and heavy labels, respectively. Data were also nonMNA-subgroup2; n ¼ 64) was enriched for immune signaling searched against a decoy database such that peptide and protein pathways (Supplementary Fig. S2; see Materials and Methods; identifications were accepted at a FDR of 1%, with a minimum ref. 26). peptide length of 6 residues and at least 1 razor and unique To investigate which genes are associated with a poor clinical peptide. Peptide identifications by MS-MS were transferred course in high-risk neuroblastoma, we performed an integrated between LC/MS runs to obtain sufficient information for quan- transcriptomic analysis for 4 comparative categories based on tification using a 2.0-minute window after retention time patient risk groups: (i) "MNA," representing HR-MNA (i.e., alignment. All identified proteins in this study are provided MYCN-amplified, stage 4, >18 months; n ¼ 67) versus HR- in Supplementary Table S8 and comparisons of their overlap nonMNA; (ii) "HR-MNA," representing HR-MNA versus LR are shown in Fig. 6A. Proteins were significantly regulated (i.e., MYCN-nonamplified, stage 1 or 2, <18 months; n ¼ 247); if BH-corrected significance B was <0.05 (Fig. 6B and C). (iii) "HR-nonMNA-subgroup1," representing HR-nonMNA- Comparisons of identified proteins with their corresponding subgroup1 versus LR; and (iv) "HR-nonMNA-subgroup2," repre- transcripts from our integrative analysis are provided in Sup- senting HR-nonMNA-subgroup2 to LR. This integrative analysis plementary Table S9. included differential gene expression (DE; Supplementary Table To characterize the pathways enriched for niclosamide- S1), pathway–gene association analysis (Supplementary Fig. S3), regulated proteins, we performed GSEA (29) on enrichments of gene correlation analysis using HUGE (ref. 30; Supplementary canonical pathways (MSigDB v5.0, C2 collection) using log2 Table S2), and genetic perturbation analysis (Supplementary (heavy over light; H/L) ratios of the identified proteins as inputs Fig. S4; Supplementary Table S3), leading to the identification with 1,000 gene-level permutations. Pathways are considered of gene expression signatures in high-risk neuroblastoma (Fig. 1C; statistically significant if FDR q values are <0.25. All significantly Supplementary Tables S4 and S5; see Materials and Methods). enriched pathways are provided in Supplementary Table S10 and comparisons of their overlap are shown in Fig. 6D and E. The gene Prognostic significance of the gene signatures for high-risk lists relevant to KEGG_PYRIMIDINE_METABOLISM and KEGG_- neuroblastoma PURINE_METABOLISM pathways are provided in Supplementa- To assess the prognostic power of the gene expression signa- ry Table S11. tures identified for high-risk tumors, we performed survival

OF6 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

analysis in an additional cohort of 477 patients with neuroblas- example, WEI_MYCN_TARGETS_WITH_E_BOX; Supplementary toma (35). The gene signature shared among the MNA, HR-MNA, Table S5), substantiating our approach for identifying biologi- and HR-nonMNA-subgroup1 categories (common gene signa- cally relevant gene signatures in neuroblastoma. Compared ture, n ¼ 24; Fig. 2A), which revealed the most significant with the other 3 categories, the HR-nonMNA-subgroup2 signa- enrichment for a gene set characteristic of pluripotent cells ture had a distinct set of immune genes (Supplementary Table S5), (ref. 38; MUELLER_PLURINET, BH-corrected hypergeometric while displaying comparable prognostic value to the HR-non- P ¼ 1.54 10 8; Fig. 2B; Supplementary Table S5), was highly MNA-subgroup1 gene signature in patients with advanced-stage predictive of patient survival, particularly in advanced tumors tumors (Supplementary Fig. S8). Interestingly, patients that were without amplified MYCN (Fig. 2C–E; Supplementary Fig. S5). A stratified as positive for both these 2 signatures tended to have similar prognostic strength was observed for the gene signature poorer outcomes than those positive for only 1 gene signature shared between the HR-MNA and HR-nonMNA-subgroup1 (Supplementary Fig. S8). Together, our data demonstrate the categories (high-risk gene signature, n ¼ 99; Supplementary Fig. ability of these gene signatures to independently prognosticate S6) or that shared between the MNA and HR-MNA categories patients with neuroblastoma. (aggressive MNA gene signature, n ¼ 74; Supplementary Fig. S7), both of which were significantly enriched for cancer-related Gene expression–based drug discovery in high-risk signatures (Supplementary Table S5). Notably, these gene neuroblastoma signatures associated with high-risk disease also revealed signif- We next sought to identify effective drugs for high-risk neuro- icant enrichments for several MYC(N)-regulated gene sets (for blastoma using a gene expression–based approach that compares

AB GSE45547 Mycn Stage Risk Dataset 1,448 Mycn Amp. 4 High GSE45547 621 588 GSE19274 Nonamp. 3 Intermediate or high GSE3446 Age 4,663 1,254 NA 2 Intermediate GSE19274 448 Stage 1 Low or intermediate GSE16254 0 11,939 768 GSE12460 Age 4S Low GSE12460 0 152 GSE16237 Risk 1,431 > 18 mo. NA NA GSE16237 0 162 GSE16254 Dataset < 18 mo. 0 NA GSE3446 C

Category Source (n ) Reference (n )

HR-nonMNA Differential expression Pathway association Gene correlation Genetic perturbation Protein interaction MNA HR-MNA (67) (156)

HR-MNA HR-MNA (67) LR (247) +

HR-nonMNA- HR-nonMNA- LR (247) subgroup1 subgroup1 (92) HR-nonMNA- HR-nonMNA- LR (247) subgroup2 subgroup2 (64)

Analysis

HR-nonMNA- HR-nonMNA- Integrative analysis MNA HR-MNA subgroup1 subgroup2 Differential expression analysis HR-nonMNA-subgroup1 No. upregulated genes 1,815 3,378 2,617 1,659 MNA No. downregulated genes 1,410 2,873 2,346 2,270 65 286 5 Pathway–gene association anlaysis 50 3 Scoring No. relationships 2,982 4,848 6,915 4,314 22 24 0 212

Gene correlation analysis 0 No. relationships 17,289 14,356 16,272 15,986 75 1 00 Genetic perturbation analysis HR-MNA 0 No. relationships 924 2,048 2,567 468

Protein interaction analysis HR-nonMNA-subgroup2 No. relationships 5,389 5,154 4,197 2,717

Figure 1. Integrative transcriptome analysis of high-risk neuroblastoma. A, Clinical variables and risk stratification of primary neuroblastomas (n ¼ 1,065) across 6 cohorts profiled using miscellaneous microarray platforms. B, Comparison of measured genes across different microarray platforms. C, Integrative analysis of the neuroblastoma transcriptome. Four comparative categories based on patient risk groups (top left) were fed into the integrated transcriptome analysis, including differential expression, pathway association, gene correlation, genetic perturbation, and protein interaction (top right). The results from the integrative analysis (bottom left) were used to generate gene-expression signatures for high-risk neuroblastoma (bottom right). MNA, MYCN-amplified; HR-MNA, high-risk MYCN-amplified.

www.aacrjournals.org Clin Cancer Res; 2019 OF7

Huang et al.

A Score rank

First Last RUVBL1 EXO1 CDCA4 GMPS CCNB1 SNRPD1CHEK1 MRPL3 PAICS RNASEH2ACKS2 TEX15 POLE3 AHCY DCTPP1MRPL11 SNRPF CDT1 ABLIM3 GAR1 SERINC1MIS18A CBX7 ATP2B4 HR−nonMNA− 114.8 130 124 106 123.3 107.3 120.7 88.5 64.6 106.8 112.2 98.2 85.4 92.1 64.9 104 103.2 89.5 90.1 100.3 103.5 80.7 96.4 89.4 subgroup1

HR−MNA 95.4 84.9 77.6 77.7 81.5 75.8 71.2 74.8 71.5 71.2 68.6 71.4 71 63.7 63.2 56.9 63.4 64.7 57 55.9 53.8 58.8 52.3 54.1

MNA 90.4 80.3 77.3 80.4 70.7 80.2 65.6 79.9 80.1 70.6 67.3 65.8 76.6 71.6 71.8 64.1 63.4 65.7 65 59.5 56.4 63.2 56.5 56.5

B D High/intermediate−risk, MYCN−nonamplified

| MUELLER_PLURINET | ||| ||| MSigDB Collection 100 ||||| ||| ||| ||||||||| ||||| || |||| SOTIRIOU_BREAST_CANCER_GRADE_1_VS_3_UP | | | |||| || || | Curated ||||| ||| |||||||||| |||| | || ||| ||| ||| | | | || | | | | ||| 80 | BERENJENO_TRANSFORMED_BY_RHOA_UP Computational ||| Signature group | GSE22886_UNSTIM_VS_IL2_STIM_NKCELL_DN | Immunologic 60 || |||| ||| | Positive (n = 61) PUJANA_CHEK2_PCC_NETWORK ||| | | | | 40 Negative (n = 155) MODULE_17 OS (%) KOBAYASHI_EGFR_SIGNALING_24HR_DN 20 − P = 7.09 × 10 6 CAIRO_HEPATOBLASTOMA_CLASSES_UP 0 GSE24634_TREG_VS_TCONV_POST_DAY5_IL4_CONVERSION_UP 050100 150 200 GSE36476_CTRL_VS_TSST_ACT_40H_MEMORY_CD4_TCELL_OLD_DN Time (months) KINSEY_TARGETS_OF_EWSR1_FLII_FUSION_UP RHODES_CANCER_META_SIGNATURE MYCN−nonamplified, > 12 months, stage 4 FEVR_CTNNB1_TARGETS_DN | 100 | | DODD_NASOPHARYNGEAL_CARCINOMA_DN | |||| | WINNEPENNINCKX_MELANOMA_METASTASIS_UP 80 || || | || | Signature group ||| 60 ||| | | | | || | | | | n 02468 | Positive ( = 36) n P ||| || Negative ( = 49) Enrichment (−log10 ) OS (%) 40 ||| | 20 P = 0.0116 0 C 0 50 100 150 200 Relative expression Time (months)

Cluster E Patient cluster Mycn High/intermediate−risk, MYCN−nonamplified 1 Stage ||| |||| | 2 100 ||||||| | | ||| Age || || | ||||||||||| ||||| ||| | || || | | ||||| ||| |||||||||| |||| | | ||| ||| ||| | | | || | | | | Mycn Risk 80 | Amplified || Signature group || Nonamplified AHCY 60 n || Positive ( = 61) NA || || n ATP2B4 40 || ||| || | | | | | Negative ( = 155)

Stage RFS (%) 4 CCNB1 20 −9 3 CDCA4 P = 5.9 × 10 2 0 1 CDT1 0 50 100 150 200 4S CKS2 Time (months) Age EXO1 > 18 months MYCN−nonamplified, > 12 months, stage 4 12−18 months GMPS < 12 months 100 | MRPL3 | || || Risk | PAICS 80 | High Signature group Intermediate or high ||| || POLE3 60 | n Intermediate | | || ||| | | | || | | | | Positive ( = 36) Low or intermediate RNASEH2A 40 Negative (n = 49) Low | RFS (%) | RUVBL1 || | | | | Gene color 20 − SNRPD1 P = 4.64 × 10 4 Upregulated 0 Downregulated SNRPF 0 50 100 150 200 Time (months)

Figure 2. Validation of the gene signature common to high-risk neuroblastoma. A, The gene signature shared among the 3 comparative categories (common gene signature, n ¼ 24). Genes were ordered by the geometric mean of the ranks of prioritization scores. Upregulated and downregulated genes were colored in red

and blue, respectively. B, Top 15 enrichments of the common gene signature for MSigDB collections (log10 BH-corrected hypergeometric P value). C, PAM clustering of 477 additional neuroblastoma samples using the common gene signature into 2 groups with positive and negative associations. D and E, KaplanMeier analysis of OS and RFS of the subsets of patients with neuroblastoma stratiﬁed by the common gene signature in C (log-rank P value).

the patterns of reversal of a disease signature achieved by the small proliferative effects were confirmed in 2 MYCN-amplified SK-N- molecule–regulated transcripts (23). We selected the high-risk DZ and SK-N-BE(2)C and 2 MYCN-nonamplified SK-N-AS and gene signature (Supplementary Fig. S6A) for computational pre- SK-N-SH neuroblastoma cell lines treated with each compound diction. Using this approach, we first identified hundreds of (Fig. 3C and D; Supplementary Fig. S9). compounds whose patterns of perturbation-induced gene expres- The gene expression–based framework also allowed us to sion changes correlated significantly and inversely with the neu- predict synergistic drug combinations for high-risk neuroblas- roblastoma gene signature (Fig. 3A and B; Supplementary Table toma by looking for compound pairs that displayed significant S6). For 4 of the top predictions, IKK-2-inhibitor-V (an IkB kinase and nearly nonoverlapping patterns of gene reversal. From inhibitor), niclosamide (an anthelmintic drug approved by the among the top 103 single agents tested for combination (Sup- FDA), NVP-BEZ235 (PI3K inhibitor), and pyrvinium pamoate plementary Table S6), we identified 660 drug pairs for which (an FDA-approved anthelmintic drug), the cytotoxic and anti- thecombinedpatternsofreversalofthehigh-riskgene

OF8 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

A 0.7 n = 2,214 compounds 0.7 0.6 0.6

0.5 0.5 Top 25 0.4 0.4

0.3 0.3

0.2 0.2 Therapeutic score Therapeutic score

0.1 0.1 Others PKC−i PI3K−i Antimicrotubule IKK−i Others HDAC−i IKK−i Anthelmintic Others PI3K−i JAK−i Prostanoid mod. HDAC−i JAK−i EGFR−i Antibiotic ER stressor Calcineurin inh. mTOR−i Uncoupler Others HDAC−i Anthelmintic Ca2+ ch. blocker 0.0 0.0

Ordered compounds rin rin−a ISOX xin−b O−143 to Rottle feldin−aDG−041 Apicidin K IKK−16 Lasalocid osedostatTGX−115 orinostat NiguldipineT AZD−8055 myr V Niclosamide Bre Malonoben TG−101348O−28−1675 NVP−BEZ235 R Cyclospo Cucurbitacin−i Cytochalasin−b rvinium pamoate phostin−AG−1478 Rhodo IKK−2−inhibitor−V r Py Ty B AHCY ASPM AURKA CBX7 CCNA2 CENPF CENPM CHAF1A DLGAP5 EXO1 EXOSC9 FEN1 HELLS KIAA0101 KIF22 MCM10 MCM3 MCM6 MIS18A OIP5 POLA2 POLE3 RNASEH2A RUVBL1 SERINC1 SNRPB SPC25 TACC3 TPX2 TRIP13 TYMS VRK1 BIRC5 BLM CDC25A CDC45 CDK1 CDKN3 CDT1 CENPA CENPN CENPU DCTPP1 DHFR DSCC1 DTL E2F1 ERCC6L FANCI FAXDC2 FOXM1 GINS2 HJURP HNRNPD KIF15 KIF20A KIF4A MAD2L1 MCM2 MRPL3 NCAPG NEIL3 NUP37 PAICS PKMYT1 POLD2 POLE2 POLQ PTTG1 RFC4 RFWD3 SLIRP SNRPD1 SNRPF TIMELESS TIPIN TMEM33 UBE2C WHSC1 IKK−2−inhibitor−V Rottlerin Compound−transcript Niclosamide recurrent regulation Lasalocid Downregulation Brefeldin−a Upregulation DG−041 Cyclosporin−a Strength of regulation Niguldipine Tosedostat Disease transcript TGX−115 Upregulated NVP−BEZ235 Downregulated AZD−8055 Apicidin Cucurbitacin−i Malonoben KO−143 Tyrphostin−AG−1478 ISOX Cytochalasin−b IKK−16 Rhodomyrtoxin−b TG−101348 RO−28−1675 Vorinostat Pyrvinium pamoate

C SK−N−DZ SK−N−DZ SK−N−DZ SK−N−DZ 160 120 140 140 140 100 120 120 Treatment duration 120 100 100 100 80 80 80 24 h 80 60 60 60 48 h 60 40 72 h 40 40 40 Cell viabliity (%) Cell viabliity 20 (%) Cell viabliity 20 (%) Cell viabliity 20 (%) Cell viabliity 20 0 0 0 0 0 .25 .5 .75 1 2.5 5 7.5 10 0 .25 .5 .75 1 2.5 5 7.5 10 0 .25 .5 .75 1 2.5 5 7.5 10 0 .25 .5 .75 1 2.5 5 7.5 10 IKK−2−inhibitor−V (μmol/L) Niclosamide (μmol/L) NVP−BEZ235 (μmol/L) Pyrvinium pamoate (μmol/L)

SK−N−AS SK−N−AS SK−N−AS SK−N−AS 120 120 100 120 100 100 80 100 80 80 80 60 60 60 60 40 40 40 40 20 Cell viabliity (%) Cell viabliity (%) Cell viabliity Cell viabliity (%) Cell viabliity 20 20 (%) Cell viabliity 20 0 0 0 0 0 .25 .5 .75 1 2.5 5 7.5 10 0 .25 .5 .75 1 2.5 5 7.5 10 0 .25 .5 .75 1 2.5 5 7.5 10 0 .25 .5 .75 1 2.5 5 7.5 10 IKK−2−inhibitor−V (μmol/L) Niclosamide (μmol/L) NVP−BEZ235 (μmol/L) Pyrvinium pamoate (μmol/L) D SK−N−DZ SK−N−DZ SK−N−DZ SK−N−DZ

P= 0.0394 P=0.0136 P=0.129 P=0.00108 280 ● 280 200 ● ● 240 ● 240 240 ● ● 200 160 ● 200 200 ● 160 160 120 ● 160 ● ● ● 120 ● 120 120 ● ● 80 ● ● ● 80 80 80 ● 40 ● 40 ● ● 40 40 ● ● 0 ●●● 0 0 0 ● Number of colonies Number of colonies Number of colonies 01.4(μmol/L) Number of colonies 01(μmol/L) 0 500 (nmol/L) 02.2(μmol/L) IKK−2−inhibitor−V Niclosamide NVP−BEZ235 Pyrvinium pamoate

SK−N−AS SK−N−AS SK−N−AS SK−N−AS

P= 60 0.0305 60 P=0.0132 80 P=0.0153 80 P=0.00451

● ● ● ● ● ● ● 60 ● 60 40 40 ● ● ● ● 40 40

● 20 ● 20 20 ● 20 ● ● ● ● ● ● ● ● 0 0 0 0 ● Number of colonies 02.1(μmol/L) Number of colonies 01.6(μmol/L) Number of colonies 01.4(μmol/L) Number of colonies 01.8(μmol/L) IKK−2−inhibitor−V Niclosamide NVP−BEZ235 Pyrvinium pamoate

Figure 3. Predicted small-molecule treatments for high-risk neuroblastoma. A, Compounds prioritized by their ability to invert the high-risk neuroblastoma signature. The top 25 compounds with their primary MoAs are shown. For full results, see Supplementary Table S6. B, Reversal relationships between the top 25 compounds and the high-risk neuroblastoma signature. C and D, MTS and clonogenic assays of SK-N-DZ and SK-N-AS neuroblastoma cells treated with 4 of the top predicted compounds. For full results, see Supplementary Fig. S9. Error bars, SEM. Three technical replicates were performed for each of at least 3 independent biological replicates. P values were calculated using paired 2-tailed t test.

www.aacrjournals.org Clin Cancer Res; 2019 OF9

Huang et al.

A Combinatorial therapeutic score 0.62 0.62 0.61 0.6 0.6

0.65 0.59 0.58 0.57 0.56 0.56 0.55 0.55 0.55 0.55 0.55 Individual 0.5 therapeutic score Compound MoA TG−101348 0.35 JAK−i Vinblastine 0.19 Antimicrotubule Vincristine 0.11 PI3K−i TGX−115 0.46 Caspase−activator NVP−BEZ235 0.45 HSP−i PAC−1 0.3 CDK−i Tanespimycin 0.14 Proteasome inh. NVP−AUY922 0.24 Bcl−2 inh. Aminopurvalanol−a 0.25 Retinoid Purvalanol−a 0.26 Ca2+ ch. blocker MLN−2238 0.26 mTOR−i Bortezomib 0.25 MEK/ERK−i TW−37 0.33 Anthelmintic CD−437 0.28 Niguldipine 0.49 AZD−8055 0.43 PD−198306 0.32 Pyrvinium pamoate 0.34 B KIF15 ERCC6L EXOSC9 MAD2L1 MCM10 MRPL3 CKS2 OIP5 PKMYT1 PTTG1 RFC4 RNASEH2A SPC25 TIMELESS TRIP13 VRK1 ASPM CDK1 FAXDC2 FOXM1 GAR1 GINS2 HMGB3 HNRNPD KIAA0101 KIF20A KIF22 DHFR DSCC1 DTL E2F1 EXO1 MCM2 MCM3 MIS18A NCAPG NEIL3 CDT1 CENPF CENPM CENPN CENPU CHAF1A POLD2 POLE3 POLQ RFWD3 RUVBL1 SERINC1 SLIRP SNRPB SNRPD1 SNRPF TMEM33 TPX2 TYMS UBE2C AHCY AURKA BIRC5 CBX7 CCNA2 CCNB1 CDC25A CDC45 FANCI FEN1 HJURP KIF4A DCTPP1 DLGAP5 NUP37 CDKN3 CENPA POLE2 BLM MCM6 PAICS POLA2 TACC3 TIPIN TG−101348 Vinblastine Compound−transcript Vincristine recurrent regulation TGX−115 Downregulation NVP−BEZ235 Upregulation PAC−1 Tanespimycin Strength of regulation NVP−AUY922 Aminopurvalanol−a Disease transcript Purvalanol−a Upregulated MLN−2238 Downregulated Bortezomib TW−37 CD−437 Niguldipine AZD−8055 PD−198306 Pyrvinium pamoate

Vinblastine (nmol/L) Vinblastine (nmol/L) CDSK-N-DZ, vinblastine (nmol/L) SK-N-AS, vinblastine (nmol/L) 0 1 5 10 25 0 1 5 10 25 0 0.1 0.5 1 5 00.10.515 0 0 1 22 24 68 0 0 4 25 66 95 50 18 16 35 51 89 1 0115756100 100 19 24 31 60 98 5 19 23 33 81 100 300 40 46 75 81 100 25 1133579 100 750 61 77 95 99 100 50 33 46 49 79 100 NVP−BEZ235 (nmol/L) NVP−BEZ235 (nmol/L) (nmol/L) (nmol/L) SK−N−DZ SK−N−AS

Vinblastine (nmol/L) Vinblastine (nmol/L)

0 1 5 10 25 00.10.515 NVP-BEZ235 NVP-BEZ235 0 0 50 0.03 0.06 0.14 0.28 1 0.13 0.17 0.01 0.01 100 0.02 0.07 0.23 0.31 5 0.09 0.17 0.09 0.01

300 0.14 0.38 0.42 0.27 50 25 5 1 0 25 0.13 0.15 0.12 0.01 750 300 100 50 0 750 0.17 0.36 0.32 0.15 50 0.39 0.26 0.13 0.01 NVP−BEZ235 (nmol/L) NVP−BEZ235 (nmol/L) SK−N−DZ SK−N−AS

Growth inhibition (%) Delta score

≤ 0 100 −1 0 +1

(Additive) (Synergistic) (Antagonistic)

Figure 4. Predicted synergistic drug combinations for high-risk neuroblastoma. A, Top 15 synergistic drug pairs targeting the high-risk neuroblastoma signature. For full results, see Supplementary Table S7. B, Reversal relationships between the top 15 synergistic drugs and the high-risk neuroblastoma signature. C and D, Clonogenic assay of SK-N-DZ and SK-N-AS neuroblastoma cells treated with the combination of NVP-BEZ235 and vinblastine. Drug synergy was calculated using the delta score. Two technical replicates were performed for each of the 2 independent biological replicates. For full results, see Supplementary Fig. S10.

signature were predicted to be synergistic (Fig. 4A and B; vincristine treatments across many cell types (26), indicating Supplementary Table S7). We found that 4 of the top 15 that these drugs belonging to the same transcriptional clusters synergistic drug pairs were the combinations of a PI3K inhibitor may replace each other in a drug combination to produce (TGX-115 or NVP-BEZ235) and an antimicrotubule (vinblas- similar synergistic effects. Overall, these data demonstrate the tine or vincristine; Fig. 4A), of which NVP-BEZ235 and vin- utility of this gene expression–based strategy to inform poten- blastine were tested in vitro to synergistically reduce clonogenic tial therapies for high-risk neuroblastoma. growth in 4 neuroblastoma cell lines (Fig. 4C and D; Supple- mentary Fig. S10). Consistent with this finding, similar patterns In vitro and in vivo effects of niclosamide in neuroblastoma of gene expression changes were observed between TGX-115 Among the compounds identified in our computational and NVP-BEZ235 treatments and between vinblastine and screen, we subsequently investigated the therapeutic efficacy of

OF10 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

A Niclosamide treatment for 24 h 1.2 1.0 SK−N−DZ 0.8 SK−N−BE(2)C 0.6 SK−N−AS 0.4 SK−N−SH 0.2

Relative expression Relative 0.0 CCNA2 MCM10 ERCC6L KIF20A RUVBL1 B C 40 100 Phase Figure 5. 80 30 G /M Therapeutic efﬁcacy of niclosamide in cells (%) 2 + 60 S neuroblastoma. A, Relative expression 20 /PI of the top 5 genes of the high-risk + 40 G1 signature in neuroblastoma cells with 10 20 niclosamide (concentrations as

0 of cells Percentage 0 in Fig. 3D) for 24 hours (n ¼ 3 Annexin independent biological replicates). Niclosamide −+ −+ −+ −+ Niclosamide −+ −+ −+ −+ B and C, Analysis of annexin V/PI double positive cells and cell-cycle changes in neuroblastoma cells treated with niclosamide SK−N−DZ SK−N−AS SK−N−SH SK−N−DZ SK−N−AS SK−N−SH (concentrations as in A) for 48 hours SK−N−BE(2)C SK−N−BE(2)C (n ¼ 3). D and E, Waterfall and box plots of SK-N-DZ and SK-N-AS DESK−N−DZ *** SK−N−AS ** xenografts transplanted 5 5 5 5 ) subcutaneously in mice treated with ) start vehicle or niclosamide (25 mg/kg) for start V / 14 (SK-N-DZ) or 28 (SK-N-AS) days. / V end end 0 0 Vend and Vstart, tumor volumes at the 0 0 V V ( ( 2 end and start of treatment, Vehicle 2 Vehicle respectively. Box plots depict the Niclosamide Niclosamide Log interquartile range (IQR) and whiskers −5 −5 Log −5 −5 represent 1.5 IQR. F and G, Representative hematoxylin and eosin Vehicle Vehicle (H&E) stains of SK-N-DZ and SK-N-AS Niclosamide Niclosamide xenografts collected at the end of treatment in D and E. Scale bars, 100 mm. H and I, Kaplan–Meier survival FGSK-N-DZ SK-N-AS analysis of mice bearing SK-N-DZ and SK-N-AS xenografts and treated as in D and E (n 8 per group). Error bars, SEM. A, One-sample t test. B, D, and E, Unpaired 2-tailed t test. H and I, Log-rank test (, P < 0.05; , P < 0.01; Vehicle Niclosamide Vehicle Niclosamide , P < 0.001; #, P < 1 104).

HISK−N−DZ SK−N−AS 100 Vehicle 100 Vehicle 80 Niclosamide 80 Niclosamide 60 60 40 40 Survival (%) Survival (%) P = 0.00134 20 | 20 P = 0.0181 | 0 0 050100150 0 20406080 Time (days) Time (days)

niclosamide, an anthelmintic drug approved by the FDA for sites (39, 41, 42). In this study, expression of 45% of genes in treating tapeworm infections in human (39, 40). The anthelmin- the high-risk gene signature were predicted to be reversed by tic activity of this drug is linked to mitochondrial uncoupling, niclosamide (Fig. 3B), of which the top 5 scoring genes (CCNA2, which inhibits ATP synthesis by disrupting electron transfer across MCM10, ERCC6L, KIF20A, and RUVBL1; Supplementary Fig. the inner mitochondrial membrane, in drug-sensitive para- S6A) were experimentally conﬁrmed in 4 neuroblastoma cell

www.aacrjournals.org Clin Cancer Res; 2019 OF11

Huang et al.

lines (Fig. 5A). We observed that most of these cells underwent Fig. S3B) and those also enriched using niclosamide-regulated apoptosis after niclosamide treatment (Fig. 5B), with growth transcript abundances (Supplementary Fig. S12D). Importantly, arrest at different phases of the cell cycle (Fig. 5C). NME3 contributed as the top 1 protein to the core enrichment in We next examined the in vivo efficacy of niclosamide in these nucleotide biosynthetic pathways (Supplementary Fig. S13; SK-N-DZ (MYCN-amplified) and SK-N-AS (MYCN-nonampli- Supplementary Table S11), implicating NME3 downregulation as fied) neuroblastoma xenografts. Notably, treatment of these a novel MoA of niclosamide to effectively kill neuroblastoma. This xenografts with niclosamide led to reduced tumor growth and diminished expression of NME3 following niclosamide treatment improved survival with no apparent toxicity (Fig. 5D–I; Supple- was also confirmed by Western blotting (Fig. 6F and G). Together, mentary Fig. S11). These results suggest that niclosamide could be these data corroborate our gene expression–based strategy to potentially repurposed as an effective and relatively safe treatment identify niclosamide as a promising drug targeting the nononco- for neuroblastoma. gene dependencies in high-risk neuroblastoma, while providing a molecular explanation for its efficacy at the levels from transcrip- Proteomic analysis of niclosamide-treated neuroblastoma cells tion to translation. To further elucidate the functional consequences of niclosamide treatment in neuroblastoma, we used a stable isotope dimethyl labeling approach to assess the proteomic changes Discussion during niclosamide treatment in SK-N-DZ and SK-N-AS cells, Complementary to traditional target-based approaches, this both in vitro (cell culture) and in vivo (xenograft). Across 4 study demonstrates the feasibility of using transcriptional profil- conditions tested (2 cell types in 2 models), we identified ing of primary neuroblastomas and small-molecule treatments 2,458 proteins at a 1% FDR, including 91 downregulated and for drug discovery and further investigates the potential of repur- 117 upregulated proteins (BH-corrected significance B < posing the anthelmintic, niclosamide, to treat this disease. Given 0.05; Fig. 6A–C; Supplementary Fig. S12A; Supplementary Table the lack of druggable mutations in high-risk neuroblastoma, we S8). Remarkably, only NME3, an enzyme involved in the bio- sought to exploit nononcogene addiction to guide the drug synthesis of nucleoside triphosphates (43), was downregulated in discovery process through an integrative transcriptome analysis, all conditions (Fig. 6B). Of the other significantly regulated enabling us to identify gene-expression signatures with indepen- proteins, leucine zipper protein 1 (LUZP1) and caveolin 1 (CAV1) dent prognostic value. One of these signatures relevant to high- were upregulated in 3 conditions (Fig. 6C). Thirteen proteins were risk status was compared with the perturbational signatures to downregulated in 2 in vitro conditions, including fatty acid desa- look for drugs or their combinations that are predicted to reverse a turase 2 (FADS2), glutaminase (GLS), and components of significant proportion of the disease signature. the minichromosome maintenance protein (MCM) helicase Our in-depth analysis of the neuroblastoma transcriptome has (MCM2–7; Fig. 6B), whereas several proteins associated revealed many pathways of relevance to the high-risk phenotype, with extracellular matrix structural constituent (for example, including nucleotide biosynthesis, RNA metabolism, DNA repli- COL14A1, COL1A1, COL6A1, COL6A2, COL6A3, LAMA1, cation and repair, and amino acid and one-carbon metabolism LAMB1, and LAMC1) were upregulated in 2 in vivo conditions (Supplementary Fig. S3B). Many genes were found to associate (Fig. 6C). However, we observed marked differences between with these deregulated pathways in a manner that is biologically the transcriptomic and proteomic changes during niclosamide meaningful. For example, the serine pathway genes phosphoglyc- treatment (Supplementary Fig. S12B and S12C; Supplementary erate dehydrogenase (PHGDH), phosphoserine aminotransferase Table S9), consistent with the complexity of mRNA and protein 1(PSAT1), and serine hydroxymethyltransferase 2 (SHMT2), regulation in perturbed systems (44). For example, although the one-carbon metabolism genes methylenetetrahydrofolate þ RUVBL1 (RuvB like AAA ATPase 1) and PAICS (phosphoribosy- dehydrogenase (NADP -dependent) 2, methenyltetrahydrofo- laminoimidazole carboxylase and phosphoribosylaminoimida- late cyclohydrolase (MTHFD2), MTHFD1 and dihydrofolate zolesuccinocarboxamide synthase) are members of the high-risk reductase (DHFR), and the nucleotide biosynthesis genes gene signature (Supplementary Fig. S6A) and were deemed to be guanine monophosphate synthase (GMPS), 5-aminoimidazole- transcriptionally repressed by niclosamide (Fig. 3B), changes in 4-carboxamide ribonucleotide formyltransferase/IMP cyclohy- their protein abundances were not obvious during the drug drolase (ATIC), PAICS, phosphoribosyl pyrophosphate amido- treatment in neuroblastoma cells (Supplementary Table S9). In transferase (PPAT) and thymidylate synthetase (TYMS) were contrast, treatment of neuroblastoma cells with niclosamide successfully captured in MYCN-amplified neuroblastoma led to considerable changes in the expression level of NME3, (Supplementary Figs. S3B, S6A, and S7A; Supplementary Table FADS2, and GLS proteins without affecting their transcript S4), consistent with direct MYC(N) target genes (45–48) and a abundances (Supplementary Table S9). The expression of genes cooperative role in fueling tumor growth (49–51). Moreover, encoding cyclin dependent kinase 1 (CDK1), thymidylate syn- several of these metabolism genes have been reported to be thetase (TYMS), helicase, lymphoid specific (HELLS), ubiquitin overexpressed in many other cancers (such as MTHFD2, GMPS, conjugating enzyme E2 C (UBE2C), and the MCM helicase was SHMT2, TYMS, and PAICS; ref. 52), supporting that cancer found to be regulated by niclosamide at both transcriptional and metabolism serves as an attractive target for therapeutic interven- translational levels (Supplementary Fig. S12C; Supplementary tion. We also note that the spliceosome machinery is upregulated Table S9). in high-risk neuroblastoma, consistent with its crucial role in Pathway analysis revealed that proteins involved in the pyrim- MYC-driven cancers (53, 54). These signature genes thus represent idine and purine biosynthesis were commonly downregulated in a variety of nononcogene dependencies that are reflected at the these neuroblastoma cells treated with niclosamide (Fig. 6D and mRNA level and required for neuroblastoma tumorigenesis. E; Supplementary Table S10), consistent with a dependency of Recently, it has been demonstrated that MYC drives a subset of high-risk neuroblastoma on these 2 pathways (Supplementary high-risk neuroblastoma without amplified MYCN (5%–10%)

OF12 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

through enhancer hijacking and focal enhancer ampliﬁca- probably due to the low percentage of high-risk tumors harboring tion (55). However, we did not detect MYC overexpression in these alternative MYC activation mechanisms or other more the HR-nonMNA-subgroup1 category (Supplementary Table S1), critical factors inﬂuencing the aggressive phenotype in a large

A Identified protein

SK-N-DZ, in vitro 262

13 102 107 SK-N-DZ, in vivo 64 26 310 730 42 223 36 41 SK-N-AS, in vivo 34 52

416 SK-N-AS, in vitro

BCDownregulated protein Upregulated protein

SK-N-DZ, in vitro SK-N-DZ, in vitro 31 15

0 0 0 0 in vivo 0 16 SK-N-DZ, 42 SK-N-DZ, in vivo 0 0 0 13b 1a 2c 1c 0 20d a Figure 6. 0 0 1 1b 7 SK-N-AS, in vivo 8 SK-N-AS, in vivo Proteome analysis of 0 0 1 0 neuroblastoma cells after niclosamide treatment. 21 SK-N-AS, in vitro 28 SK-N-AS, in vitro Comparisons of identified proteins (A), significantly regulated proteins a (B, C; BH-corrected significance NME3 aLUZP1 dANXA1, ANXA2, ANXA6, BGN, b b B < 0.05), and significantly FADS2, G3BP1, GLG1, GLS, MCM2, MCM3, CAV1 CKM, COL14A1, COL1A1, COL6A1, c regulated pathways (D, E; FDR MCM4, MCM6, MCM7, PCNA, PHOX2A, SQLE COL6A2, COL6A3, FABP4, LAMA1, q value < 0.25) among SK-N-DZ POTEE, TSFM LAMB1, LAMC1, LCP1, LGALS1, c and SK-N-AS cells treated with RPL36AL, SF3A2 LGALS3, NID1, OGN, SERPING1 niclosamide in vitro and in vivo. Gene symbols or pathways of DEDownregulated pathway Upregulated pathway interest are shown below each panel. Western blot analysis of SK-N-DZ, in vitro SK-N-DZ, in vitro SK-N-DZ and SK-N-AS cells treated 4 1 with vehicle or niclosamide in vitro (F)orin vivo (G; samples collected 0 0 0 0 b 3 SK-N-DZ, in vivo 24 SK-N-DZ, in vivo from Fig. 5D and E). 0 1 0 0 9 1a 0 0 0 33 0 0 0 4a 2 SK-N-AS, in vivo 8 SK-N-AS, in vivo 1 0 3 1

13 SK-N-AS, in vitro 10 SK-N-AS, in vitro

aKEGG: Pyrimidine metabolism aNABA: Matrisome | Reactome: Integrin bKEGG: Purine metabolism cell surface interactions | Reactome: Hemostasis | Reactome: Platelet activation signaling and aggregation FG

SK-N-DZ SK-N-AS

− 48 − 48 Niclosamide (h) VehicleNiclosamide No. 1 Vehicle No.Niclosamide No. 1 2 Vehicle No. 2Niclosamide No. 1 Vehicle No.Niclosamide No. 1 2 No. 2

NME3 NME3

β-Actin β-Actin In vitro lysate SK-N-DZ xenograft SK-N-AS xenograft

www.aacrjournals.org Clin Cancer Res; 2019 OF13

Huang et al.

fraction of these tumors (like nononcogene dependencies decades to treat intestinal parasites through its mitochondrial revealed in this study). Interestingly, of 8 significantly mutated uncoupling activity (39, 41, 42). Niclosamide has an excellent driver genes that have been reported in neuroblastoma (NRAS, safety profile, with an oral dosage of 2,000 mg/day in adults and PTPN11, ATRX, MYCN, ALK, SHANK2, TERT,andPTPRD; 1,000–1,500 mg/day in children (40). Our results demonstrate a refs. 4, 56), only protein tyrosine phosphatase, receptor type repurposing opportunity for niclosamide in high-risk neuroblas- D(PTPRD) was found in the signature genes of HR-nonMNA- toma. Recent studies have, indeed, shown that niclosamide may subgroup2 category (Supplementary Table S4). This suggests be effective against a broad spectrum of cancers, such as colorectal that the nononcogene dependencies identified by our gene cancer (64) and advanced prostate cancer (65), and moreover expression–based approach are very distinct from the driver may be repurposed to treat type 2 diabetes (66) and Zika virus genes discovered through the genomic approaches in high-risk infection (67), highlighting its potential for use in various disease neuroblastoma. areas. However, despite efforts to identify new therapeutic indica- Prompted by the paucity of somatic mutations in neuro- tions, the exact molecular mechanisms underlying the pleiotropic blastoma, we employed a drug-discovery approach based biological effects of niclosamide still remain elusive. Besides its on nononcogene addiction to identify compounds whose well-known function of mitochondrial uncoupling, niclosamide perturbational gene expression signatures are significantly and has been reported to inhibit STAT3 (68) or Wnt (69) pathway in inversely correlated with the high-risk neuroblastoma signa- cancer cells in vitro. In our proteomic data, we nevertheless ture. The perturbational gene expression signatures inferred observed the downregulation of STAT3 (significance B < 0.05) with a few core cell lines have been shown to sufficiently reflect and the identification of b-catenin (encoded by CTNNB1) only the unique compound mechanisms in independent datasets in niclosamide-treated SK-N-AS cells in vitro (Supplementary while being generalizable to other cell lines, corroborating the Table S9), suggesting that these 2 pathways do not contribute applicability of this approach to drug discovery in neuroblas- to the antineuroblastoma effects of niclosamide. Moreover, toma (23). The increasing need to incorporate nononcogene although we found that transcripts associated with the respi- dependencies in cancer therapy is further supported by the ratory electron transport chain tend to be downregulated by observation that somatic mutations alone are poor predictors niclosamide across many cell types (REACTOME_RESPIRATORY_ of drug sensitivity (57–59). Recently, several studies have ELECTRON_TRANSPORT, FDR q value < 0.25; Supplementary highlighted the potential of using gene expression to target Table S10), the corresponding proteins were downregulated after tumor dependencies. Although it has been shown that reversal niclosamide treatment in neuroblastoma cells only significantly of cancer gene expression might predict drug efficacy (22) and in vitro but not in vivo (Supplementary Table S10). This indicates pharmacologic targeting of mechanistic tumor dependencies that mitochondrial uncoupling may not be a central mechanism can be applied in precision oncology (21), an added advantage for the drug's efficacy observed in neuroblastoma. Instead, of our approach lies in its ability to inform synergistic drug our proteomic analysis confidently revealed downregulation of combinations. However, we should note that global reversion nucleotide biosynthetic pathways in neuroblastoma cells treated of disease signature can only be used to predict drug sensitivity with niclosamide both in vitro and in vivo, with NME3 being the butnotitsmechanism,aswe observed that the top 25 most significantly downregulated protein in these pathways compounds prioritized by this approach induced similar pat- (Fig. 6). terns of gene reversal while representing a diverse range of NME3 was originally identified as an inhibitor of granulocyte compound mechanisms (Fig. 3B). Consistent with the previ- differentiation and an inducer of apoptosis in myeloid cells (43), ous report (22), the extent of disease gene reversal was only but later, it was found to be widely expressed in many cancer highly, but not perfectly, correlated with drug sensitivity, as lines (70). However, its role in cancer is not fully understood. A shown by our finding that the first compound, IKK-2-inhib- recent report has shown that NME3 expression positively corre- itor-V, and the 25th compound, pyrvinium pamoate, induced lates with patient survival in breast or lung cancer but negatively comparably higher cytotoxicity than by the third compound correlates with that in gastric cancer (71). By contrast, our analysis niclosamide or by the 11th compound, NVP-BEZ235 (Fig. 3C did not reveal a dependency of high-risk neuroblastoma tumors and D; Supplementary Fig. S9). Notably, the neuroblastoma on NME3 examined at the mRNA level (Supplementary Table S4), signature used for gene reversal was mainly derived from but we cannot exclude the contribution of this gene product to the pathway–gene association analysis (Supplementary Fig. S3). cancer biology. However, owing to the lack of efficacy targets of This framework was developed to correlate genes with a niclosamide, it is still unable to explain how the direct binding potential biological significance in advanced tumors by com- events could relate to the decreased nucleotide biosynthesis paring disease-related pathways enriched between those and NME3 protein expression in neuroblastoma, as well as to patients with the highest and lowest expression of a given the many other therapeutic effects in different disease settings. gene. Thus, the compounds capable of inverting a large pro- Molecularly, it has been demonstrated that NME3 can enhance portion of the disease signature identified through this frame- NF-kB signaling downstream of Toll-like receptor 5 activa- work are more likely to represent effective treatments. Con- tion (71) and that the direct interaction of NME3 with Tip60 sistently, our analysis identified several compounds that have (also known as lysine acetyltransferase 5, or KAT5) is crucial for been implicated in neuroblastoma therapy, such as the histone DNA repair (72). Further system-level studies of cellular signaling deacetylase inhibitors, vorinostat (60) and panobinostat (61), responses following niclosamide treatment are anticipated to the proteasome inhibitor, bortezomib (62), and many che- uncover clues about the drug's efficacy targets and MoAs in motherapeutic agents including doxorubicin and etoposide different contexts. (ref. 63; Supplementary Table S6). In summary, our findings provide a rationale for neuroblasto- Among the compounds tested in vitro, we proceeded to validate ma drug discovery by targeting nononcogene dependencies the in vivo effects of niclosamide, a drug that has been used for identified through an integrative transcriptomic analysis. The

OF14 Clin Cancer Res; 2019 Clinical Cancer Research

Drugging Non-oncogene Addiction in Neuroblastoma

compound predicted by this approach shows in vitro efficacies in Administrative, technical, or material support (i.e., reporting or organizing neuroblastoma, as further confirmed in vivo for an FDA-approved data, constructing databases): Y.-J. Oyang anthelmintic drug niclosamide. Proteome-wide profiling of Study supervision: H.-C. Huang, H.-F. Juan these neuroblastoma cells after niclosamide treatment identifies nucleotide biosynthesis and NME3 as likely mechanisms for its Acknowledgments antitumor effects. These results suggest an opportunity to deter- This work was supported by the Ministry of Science and Technology (MOST mine the clinical efficacy of niclosamide for treating high-risk 105-2320-B-002-057-MY3 and MOST 106-2320-B-002-053-MY3) and the National Health Research Institutes (NHRI-EX107-10530PI and NHRI- neuroblastoma. EX107-10709BI). We thank the Technology Commons at National Taiwan University College of Life Science (Taipei, Taiwan) for assistance in the FACS Disclosure of Potential Conflicts of Interest analysis. We thank the Mass Spectrometry Laboratory of Tzong Jwo Jang at Fu No potential conflicts of interest were disclosed. Jen Catholic University College of Medicine (New Taipei City, Taiwan) for assistance in the LC/MS-MS analysis. Authors' Contributions Conception and design: C.-T. Huang, H.-C. Huang, H.-F. Juan The costs of publication of this article were defrayed in part by the Development of methodology: C.-T. Huang payment of page charges. This article must therefore be hereby marked Acquisition of data (provided animals, acquired and managed patients, advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate provided facilities, etc.): C.-H. Hsieh, W.-C. Lee, Y.-L. Liu, T.-S. Yang, W.-M. Hsu this fact. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.-T. Huang Writing, review, and/or revision of the manuscript: C.-T. Huang, H.-C. Huang, Received December 17, 2018; revised February 17, 2019; accepted March 28, H.-F. Juan 2019; published first April 5, 2019.

References 1. Matthay KK, Maris JM, Schleiermacher G, Nakagawara A, Mackall CL, 17. Schramm A, Koster J, Assenov Y, Althoff K, Peifer M, Mahlow E, et al. Diller L, et al. Neuroblastoma. Nat Rev Dis Primers 2016;2:16078. Mutational dynamics between primary and relapse neuroblastomas. 2. Maris JM. Recent advances in neuroblastoma. N Engl J Med 2010;362: Nat Genet 2015;47:872–7. 2202–11. 18. Luo J, Solimini NL, Elledge SJ. Principles of cancer therapy: oncogene and 3. Cheung NK, Dyer MA. Neuroblastoma: developmental biology, cancer non-oncogene addiction. Cell 2009;136:823–37. genomics and immunotherapy. Nat Rev Cancer 2013;13:397–411. 19. Nagel R, Semenova EA, Berns A. Drugging the addict: non-oncogene 4. Pugh TJ, Morozova O, Attiyeh EF, Asgharzadeh S, Wei JS, Auclair D, et al. addiction as a target for cancer therapy. EMBO Rep 2016;17:1516–31. The genetic landscape of high-risk neuroblastoma. Nat Genet 2013;45: 20. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 279–84. 2011;144:646–74. 5. Fletcher JI, Ziegler DS, Trahair TN, Marshall GM, Haber M, Norris MD. Too 21. Alvarez MJ, Subramaniam PS, Tang LH, Grunn A, Aburi M, Rieckhof G, many targets, not enough patients: rethinking neuroblastoma clinical et al. A precision oncology approach to the pharmacological targeting of trials. Nat Rev Cancer 2018;18:389–400. mechanistic dependencies in neuroendocrine tumors. Nat Genet 2018;50: 6. Puissant A, Frumm SM, Alexe G, Bassil CF, Qi J, Chanthery YH, et al. 979–89. Targeting MYCN in neuroblastoma by BET bromodomain inhibition. 22. Chen B, Ma L, Paik H, Sirota M, Wei W, Chua MS, et al. Reversal of cancer Cancer Discov 2013;3:308–23. gene expression correlates with drug efficacy and reveals therapeutic targets. 7. Carter DR, Murray J, Cheung BB, Gamble L, Koach J, Tsang J, et al. Nat Commun 2017;8:16022. Therapeutic targeting of the MYC signal by inhibition of histone chaperone 23. Huang CT, Hsieh CH, Chung YH, Oyang YJ, Huang HC, Juan HF. Pertur- FACT in neuroblastoma. Sci Transl Med 2015;7:312ra176. bational gene-expression signatures for combinatorial drug discovery. 8. Gustafson WC, Meyerowitz JG, Nekritz EA, Chen J, Benes C, Charron E, iScience (in press). et al. Drugging MYCN through an allosteric transition in Aurora kinase A. 24. Deshwar AG, Morris Q. PLIDA: cross-platform gene expression nor- Cancer Cell 2014;26:414–27. malization using perturbed topic models. Bioinformatics 2014;30: 9. Rader J, Russell MR, Hart LS, Nakazawa MS, Belcastro LT, Martinez D, et al. 956–61. Dual CDK4/CDK6 inhibition induces cell-cycle arrest and senescence in 25. Blei DM, Ng AY, Jordan MI. Latent Dirichlet allocation. J Mach Learn Res neuroblastoma. Clin Cancer Res 2013;19:6173–82. 2003;3:993–1022. 10. Hogarty MD, Norris MD, Davis K, Liu X, Evageliou NF, Hayes CS, et al. 26. Huang CT, Hsieh CH, Oyang YJ, Huang HC, Juan HF. A large-scale gene ODC1 is a critical determinant of MYCN oncogenesis and a therapeutic expression intensity-based similarity metric for drug repositioning. target in neuroblastoma. Cancer Res 2008;68:9735–45. iScience 2018;7:40–52. 11. Mosse YP, Laudenslager M, Longo L, Cole KA, Wood A, Attiyeh EF, et al. 27. Wiwie C, Baumbach J, Rottger R. Comparing the performance of biomed- Identification of ALK as a major familial neuroblastoma predisposition ical clustering methods. Nat Methods 2015;12:1033–8. gene. Nature 2008;455:930–5. 28. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays 12. Cheung NK, Zhang J, Lu C, Parker M, Bahrami A, Tickoo SK, et al. applied to the ionizing radiation response. Proc Natl Acad Sci U S A Association of age at diagnosis and genetic mutations in patients with 2001;98:5116–21. neuroblastoma. JAMA 2012;307:1062–71. 29. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, 13. Peifer M, Hertwig F, Roels F, Dreidax D, Gartlgruber M, Menon R, et al. et al. Gene set enrichment analysis: a knowledge-based approach for Telomerase activation by genomic rearrangements in high-risk neuroblas- interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A toma. Nature 2015;526:700–4. 2005;102:15545–50. 14. Valentijn LJ, Koster J, Zwijnenburg DA, Hasselt NE, van Sluis P, Volckmann 30. Zhao T, Liu H, Roeder K, Lafferty J, Wasserman L. The HUGE package for R, et al. TERT rearrangements are frequent in neuroblastoma and identify high-dimensional undirected graph estimation in R. J Mach Learn Res aggressive tumors. Nat Genet 2015;47:1411–4. 2012;13:1059–62. 15. Carpenter EL, Mosse YP. Targeting ALK in neuroblastoma–preclinical and 31. Friedman J, Hastie T, Tibshirani R. Sparse inverse covariance estimation clinical advancements. Nat Rev Clin Oncol 2012;9:391–9. with the graphical lasso. Biostatistics 2008;9:432–41. 16. Eleveld TF, Oldridge DA, Bernard V, Koster J, Daage LC, Diskin SJ, et al. 32. Liu H, Roeder K, Wasserman L. Stability approach to regularization Relapsed neuroblastomas show frequent RAS-MAPK pathway mutations. selection (StARS) for high dimensional graphical models. Adv Neural Inf Nat Genet 2015;47:864–71. Process Syst 2010;24:1432–40.

www.aacrjournals.org Clin Cancer Res; 2019 OF15

Huang et al.

33. Lambert SA, Jolma A, Campitelli LF, Das PK, Yin Y, Albu M, et al. The 54. Koh CM, Bezzi M, Low DH, Ang WX, Teo SX, Gay FP, et al. MYC regulates human transcription factors. Cell 2018;172:650–65. the core pre-mRNA splicing machinery as an essential step in lymphoma- 34. Menche J, Sharma A, Kitsak M, Ghiassian SD, Vidal M, Loscalzo J, et al. genesis. Nature 2015;523:96–100. Disease networks. Uncovering disease-disease relationships through the 55. Zimmerman MW, Liu Y, He S, Durbin AD, Abraham BJ, Easton J, et al. MYC incomplete interactome. Science 2015;347:1257601. drives a subset of high-risk pediatric neuroblastomas and is activated 35. Oberthuer A, Juraeva D, Li L, Kahlert Y, Westermann F, Eils R, et al. through mechanisms including enhancer hijacking and focal enhancer Comparison of performance of one-color and two-color gene-expression amplification. Cancer Discov 2018;8:320–35. analyses in predicting clinical endpoints of neuroblastoma patients. 56. Ma X, Liu Y, Liu Y, Alexandrov LB, Edmonson MN, Gawad C, et al. Pan- Pharmacogenomics J 2010;10:258–66. cancer genome and transcriptome analyses of 1,699 paediatric leukaemias 36. Yadav B, Wennerberg K, Aittokallio T, Tang J. Searching for drug synergy in and solid tumours. Nature 2018;555:371–6. complex dose-response landscapes using an interaction potency model. 57. Basu A, Bodycombe NE, Cheah JH, Price EV, Liu K, Schaefer GI, et al. An Comput Struct Biotechnol J 2015;13:504–13. interactive resource to identify cancer genetic and lineage dependencies 37. Cox J, Mann M. MaxQuant enables high peptide identification rates, targeted by small molecules. Cell 2013;154:1151–61. individualized p.p.b.-range mass accuracies and proteome-wide protein 58. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, quantification. Nat Biotechnol 2008;26:1367–72. et al. The Cancer Cell Line Encyclopedia enables predictive modelling of 38. Muller FJ, Laurent LC, Kostka D, Ulitsky I, Williams R, Lu C, et al. Regulatory anticancer drug sensitivity. Nature 2012;483:603–7. networks define phenotypic classes of human stem cell lines. Nature 2008; 59. Garnett MJ, Edelman EJ, Heidorn SJ, Greenman CD, Dastur A, Lau KW, 455:401–5. et al. Systematic identification of genomic markers of drug sensitivity in 39. Sheth UK. Mechanisms of anthelmintic action. Prog Drug Res 1975;19: cancer cells. Nature 2012;483:570–5. 147–57. 60. DuBois SG, Groshen S, Park JR, Haas-Kogan DA, Yang X, Geier E, et al. 40. Andrews P, Thyssen J, Lorke D. The biology and toxicology of mollusci- Phase I study of vorinostat as a radiation sensitizer with 131I-metaiodo- cides, Bayluscide. Pharmacol Ther 1982;19:245–95. benzylguanidine (131I-MIBG) for patients with relapsed or refractory 41. Frayha GJ, Smyth JD, Gobert JG, Savel J. The mechanisms of action of neuroblastoma. Clin Cancer Res 2015;21:2715–21. antiprotozoal and anthelmintic drugs in man. Gen Pharmacol 1997;28: 61. Waldeck K, Cullinane C, Ardley K, Shortt J, Martin B, Tothill RW, et al. Long 273–99. term, continuous exposure to panobinostat induces terminal differentia- 42. Terada H. Uncouplers of oxidative phosphorylation. Environ Health Per- tion and long term survival in the TH-MYCN neuroblastoma mouse spect 1990;87:213–8. model. Int J Cancer 2016;139:194–204. 43. Venturelli D, Martinez R, Melotti P, Casella I, Peschle C, Cucco C, et al. 62. Brignole C, Marimpietri D, Pastorino F, Nico B, Di Paolo D, Cioni M, et al. Overexpression of DR-nm23, a protein encoded by a member of the Effect of bortezomib on human neuroblastoma cell growth, apoptosis, and nm23 gene family, inhibits granulocyte differentiation and induces angiogenesis. J Natl Cancer Inst 2006;98:1142–57. apoptosis in 32Dc13 myeloid cells. Proc Natl Acad Sci U S A 1995; 63. Matthay KK, Villablanca JG, Seeger RC, Stram DO, Harris RE, 92:7435–9. RamsayNK,etal.Treatmentofhigh-riskneuroblastomawith 44. Vogel C, Marcotte EM. Insights into the regulation of protein abundance intensive chemotherapy, radiotherapy, autologous bone marrow from proteomic and transcriptomic analyses. Nat Rev Genet 2012;13: transplantation, and 13-cis-retinoic acid. Children's Cancer Group. 227–32. N Engl J Med 1999;341:1165–73. 45. Westermann F, Muth D, Benner A, Bauer T, Henrich KO, Oberthuer A, et al. 64. Osada T, Chen M, Yang XY, Spasojevic I, Vandeusen JB, Hsu D, et al. Distinct transcriptional MYCN/c-MYC activities are associated with spon- Antihelminth compound niclosamide downregulates Wnt signaling and taneous regression or malignant progression in neuroblastomas. elicits antitumor responses in tumors with activating APC mutations. Genome Biol 2008;9:R150. Cancer Res 2011;71:4172–82. 46. Sun L, Song L, Wan Q, Wu G, Li X, Wang Y, et al. cMyc-mediated 65. Liu C, Lou W, Zhu Y, Nadiminty N, Schwartz CT, Evans CP, et al. activation of serine biosynthesis pathway is critical for cancer pro- Niclosamide inhibits androgen receptor variants expression and over- gression under nutrient deprivation conditions. Cell Res 2015;25: comes enzalutamide resistance in castration-resistant prostate cancer. 429–44. Clin Cancer Res 2014;20:3198–210. 47. Valentijn LJ, Koster J, Haneveld F, Aissa RA, van Sluis P, Broekmans ME, 66. Tao H, Zhang Y, Zeng X, Shulman GI, Jin S. Niclosamide ethanolamine- et al. Functional MYCN signature predicts outcome of neuroblastoma induced mild mitochondrial uncoupling improves diabetic symptoms in irrespective of MYCN amplification. Proc Natl Acad Sci U S A 2012;109: mice. Nat Med 2014;20:1263–9. 19190–5. 67. Xu M, Lee EM, Wen Z, Cheng Y, Huang WK, Qian X, et al. Identification 48. Walz S, Lorenzin F, Morton J, Wiese KE, von Eyss B, Herold S, et al. of small-molecule inhibitors of Zika virus infection and induced Activation and repression by oncogenic MYC shape tumour-specific gene neuralcelldeathviaadrugrepurposingscreen.NatMed2016;22: expression profiles. Nature 2014;511:483–7. 1101–7. 49. Locasale JW. Serine, glycine and one-carbon units: cancer metabolism in 68. Ren X, Duan L, He Q, Zhang Z, Zhou Y, Wu D, et al. Identification of full circle. Nat Rev Cancer 2013;13:572–83. niclosamide as a new small-molecule inhibitor of the STAT3 signaling 50. Yang M, Vousden KH. Serine and one-carbon metabolism in cancer. pathway. ACS Med Chem Lett 2010;1:454–9. Nat Rev Cancer 2016;16:650–62. 69. Chen M, Wang J, Lu J, Bond MC, Ren XR, Lyerly HK, et al. The anti- 51. Liu M, Xia Y, Ding J, Ye B, Zhao E, Choi JH, et al. Transcriptional profiling helminthic niclosamide inhibits Wnt/Frizzled1 signaling. Biochemistry reveals a common metabolic program in high-risk human neuroblastoma 2009;48:10267–74. and mouse neuroblastoma sphere-forming cells. Cell Rep 2016;17: 70. Martinez R, Venturelli D, Perrotti D, Veronese ML, Kastury K, Druck T, et al. 609–23. Gene structure, promoter activity, and chromosomal location of the DR- 52. Nilsson R, Jain M, Madhusudhan N, Sheppard NG, Strittmatter L, nm23 gene, a related member of the nm23 gene family. Cancer Res 1997; Kampf C, et al. Metabolic enzyme expression highlights a key role for 57:1180–7. MTHFD2 and the mitochondrial folate pathway in cancer. 71. Flentie K, Gonzalez C, Kocher B, Wang Y, Zhu H, Marasa J, et al. Nucleoside Nat Commun 2014;5:3128. diphosphate kinase-3 (NME3) enhances TLR5-induced NFkappaB activa- 53. Hsu TY, Simon LM, Neill NJ, Marcotte R, Sayad A, Bland CS, et al. The tion. Mol Cancer Res 2018;16:986–99. spliceosome is a therapeutic vulnerability in MYC-driven cancer. Nature 72. Tsao N, Yang YC, Deng YJ, Chang ZF. The direct interaction of NME3 with 2015;525:384–8. Tip60 in DNA repair. Biochem J 2016;473:1237–45.

OF16 Clin Cancer Res; 2019 Clinical Cancer Research

Therapeutic Targeting of Non-oncogene Dependencies in High-risk Neuroblastoma

Chen-Tsung Huang, Chiao-Hui Hsieh, Wen-Chi Lee, et al.

Clin Cancer Res Published OnlineFirst April 5, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-18-4117

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2019/04/05/1078-0432.CCR-18-4117.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2019/05/03/1078-0432.CCR-18-4117. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.