Identification and Pharmacological Inactivation of the MYCN Gene Network as a Therapeutic Strategy for Neuroblastic Tumor Cells*

Background: Neuroblastic tumors are often addicted to the MYCN protooncogene. Results: Using a genome wide shRNA screen, we have identified key MYCN synthetic lethal genes. Conclusion: Chemical inhibition of the newly identified MYCN synthetic lethal genes selectively kills MYCN-amplified cell lines. Significance: Decoding the MYCN gene network will help to develop drugs for the treatment of neuroblastic tumors with activated MYCN.

There are three well characterized members of the MYC family in mammalian cells, c-MYC, MYCN, and L-MYC (defined hereafter as MYC). Other less studied members are S-MYC and B-MYC. They interact with DNA through a consensus sequence called the E-box (CANNTG) and in concert with the partner MAX facilitate gene transcription. Experimental evidence indicates that MYC and MYCN are functionally interchangeable, with MYCN having a more restricted spatial and temporal role during development (1). The importance of MYC in cell biology is vast. MYC controls key cellular processes including: regulation of cell death (2,3), angiogenesis (4,5), metabolism (6), chromatin remodeling (7,8), and generation of pluripotent stem cells (9). MYC members are proto-oncogenes: on a par with p53 mutation, MYC activation is the most frequent molecular alteration observed in human cancer (2). How MYC mechanistically brings about all these functions is still a matter of study, but its classical role as transcriptional activator is being revised in the light of evidence suggesting that MYC is able to modify the chromatin by direct and indirect cross-talk to chromatin modifiers, for example DNA methyl-transferases (7). Along this line, we and others have recently observed that MYCs inhibits the transcription of tumor suppressor genes or microRNAs by physically recruiting the polycomb-repressive complex 2 to promoters in the proximity of MYC-binding sites (10 -12). This demonstrates that MYC recruits co-repressor molecules in a sequence-specific manner to induce methylation of histone H3 on lysine 27 and transcriptional repression of specific genes. Given the importance of MYC in the biology of cancer, many clinical researchers are trying to develop drugs that inhibit its activity. Although small molecule inhibitors of MYC are difficult to develop, a dominant-negative variant of MYC, called omomyc, has shown anticancer effects in vivo, validating the hypothesis that targeting MYC is highly relevant (13). However, the protein omomyc is unlikely to be used in patients; thus further research is needed to develop clinically viable MYC-targeting drugs. An alternative approach to MYC targeting is to develop drugs that inactivate key MYC partners. For example, it has been shown that MYC-dependent tumors are sensitive to inhibitors of BET family chromatin adaptors and containing bromodomains such as BRD4, required for transcriptional elongation of MYC (14,15). Furthermore, we and others have shown that inhibiting the MYCN associate EZH2 or other enzymes that modify the chromatin landscape causes reactivation of CLU and other potential tumor suppressor genes in cancer cells with therapeutic effects (16 -18). This indicates that blocking chromatin modifiers could be of clinical value in MYCN-overexpressing tumors such as neuroblastoma.
Neuroblastoma is the most common extracranial solid tumor in childhood and one of the major causes of cancer death in infancy. Neuroblastoma originates from cells of the neural crest, the embryonal structure that forms the peripheral nervous system. Clinically, we distinguish two types of neuroblastoma: a localized form, usually benign, and a high risk, metastatic form. High risk neuroblastoma has very poor prognosis, and the survival rate after 5 years is only ϳ40% despite chemoand radiotherapy (19). Indeed, metastatic neuroblastoma shows initial response to therapeutic interventions but typically relapses into an incurable form of the disease. It is notable that in the past 20 years there has been no substantial improvement in the outcome of high risk neuroblastoma, suggesting that new avenues of therapy are urgently needed. When activated by amplification in a fraction (ϳ30%) of neuroblastomas, the protooncogene MYCN is a direct cause of the disease. Transgenic expression of MYCN in the neuroectoderm causes neuroblastomas in mice with features similar to those seen in the human disease (20). Conversely, inhibition of MYCN by antisense approaches results in neuroblastoma regression in vivo and in vitro (21). Collectively these results demonstrate that MYCN is a key driver of tumorigenesis in neuroblastoma, suggesting that therapeutic efforts aimed at inhibiting its expression/activity should have an important clinical relevance.
Activation of MYC imposes an oncogenic stress to tumor cells that respond by increasing the expression of genes that enhance cell fitness. The intricate network of genes sustaining the oncogenic activity of MYC, the MYC network, if identified, could lead to the development of drugs for cancer therapy. To identify the MYCN network in neuroblastoma we used a global, genome-wide approach in which we carried out an shRNA drop-out screen in isogenic cell lines expressing MYCN or not. The prediction was that the introduction of the shRNAs targeting the MYCN network should trigger synthetic lethality in a MYCN-dependent manner. A similar approach has been recently used to identify shRNAs synthetic lethal to c-MYCoverexpressing cells in breast cancer and fibroblasts (22,23). In another study, the laboratory of Martin Eliers has identified, after the analysis of 97 MYCN target genes, Aurora A as a kinase critically required to stabilize MYCN and whose inhibition by small molecules has a strong impact on MYCN-driven tumors (24,25).
We describe here the identification of 536 genes whose knockdown is synthetically lethal to MYCN-overexpressing cells. To prioritize candidates, we selected for further analyses genes whose products are inhibited by small molecule drugs, are direct targets of MYCN, and predict poor survival in neuroblastoma patients. Using these criteria, we verified that CKS1B, AHCY, PKMYT1, and BLM could potentially be used as targets for the treatment of MYCN-overexpressing tumors.

MATERIALS AND METHODS
shRNA Screen and Data Analysis-Seven pools of 9,600 shRNAs were prepared from GIPZ Human Whole Genome shRNA Library (Thermo Scientific). The shRNA screen was carried out following published procedures (26) using a multiplicity of infection of 0.3 and a representation of ϳ1,000 cellular integrations per shRNA. GIMEN-EMPTY and GIMEN-MYCN cells were infected in triplicate with each pool and harvested at time point 1 (T ϭ 1; 48 h after puromycin selection) and at time point 2 (T ϭ 2; 2 weeks after time point 1). Genomic DNA from harvested cells was isolated using a blood and cell culture DNA mini kit (Qiagen) following the manufacturer's protocol. Unique barcode sequences were amplified by PCR and purified from agarose gel using Wizard SV gel a PCR clean-up system (Promega, Southampton, UK) following the manufacturer's protocol. Purified PCR products from each cell line and time points were combined and labeled with Cy5 and Cy3 dyes, using the Agilent genomic DNA labeling kit plus, following the Open Biosystem protocol adapted from Agilent oligonucleotide array-based CGH for genomic DNA analysis. Labeled PCR products were then competitively hybridized to custom microarrays containing the barcode sequences. Data extraction was carried out using Aglient's Feature Extraction Software. Analysis was performed using Bioconductor (27) and Limma (28) software. Probes that produced a signal lower than 1.5 times the mean intensity of control probes in at least two of the three replicates were removed. To identify MYCN synthetic lethal shRNAs, the mean log10 Cy5/Cy3 ratios of GIMEN-MYCN replicates was compared with that of GIMEN-EMPTY to derive the log10 ratio difference. Genes with fold change more than 1.5 and p values less than 0.05 were considered as potential MYCN synthetic lethal partners.
Bioinformatic Analysis-Biofunctions of the candidate shRNAs were assessed with the IPA software (Qiagen). Fisher's Exact test p value was used as a scoring method, and the threshold was set at p ϭ 0.05. The relationships of MYCN synthetic lethal genes identified in our screen with the three hubs of MYC synthetic lethal genes identified in previous studies (29) was established using the GeneMania software. Indirect interactions were extracted from Pathway commons-Reactome and physical interactions extracted from iRefIndex collection of databases. In silico analysis of gene expression in neuroblastoma patients was carried out using the databases Oncomine and Oncogenomics following the instructions and using the tools available at the websites.
Lentivirus Production-Lentiviruses were generated by transfecting 5 g of the pGIPZ-shRNA and packaging plasmids pPAX and pMDG2 into HEK-293FT cells using Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. Supernatants were harvested 48 h after transfection and filtered through a 0.45-mm filter unit. The sequences and codes of the pGIPZ-shRNA constructs used for shRNA knockdown studies are illustrated in Table 1.
Cell Culture-Human embryonic kidney cells 293FT, human Neuroblastoma cell lines SK-N-AS, SH-SY5Y, IMR-32, SK-N-BE(2), LA-N-1, and NB19 were obtained from the American Type Culture Collection (Teddington, Middlesex, UK). GI-M-EN cells were a kind gift from Mirco Ponzoni (Gaslini Hospital, Genova, Italy). SMS-KCNR Neuroblastoma cells were a kind gift from Dr. Andrew Stoker (Institute of Child Health, University College London, London, UK). HEK-293FT, LA-N-1, SK-N-AS, and SH-SY5Y were maintained in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine, and sodium pyruvate. IMR-32, NB19, SMS-KCNR, and GI-M-EN were cultured in RPMI medium 1640 containing 10% fetal bovine serum, 2 mM glutamine supplemented with sodium pyruvate and nonessential amino acids. All cell lines were incubated at 37°C and 5% CO 2 . Stable cell lines expressing the indicated shRNAs were generated by lentiviral transduction in the presence of 8 g/ml polybrene followed by selection with puromycin (1-4 g/ml). For drug treatments, cells were plated at a density of 10,000 -25,000 cells in 24-well plates in triplicate. After 24 h, the cells were exposed to Fluoxetine (10 mM), ML216 (15 mM), 3-deazaadenosine (20 mM), or PD166285 (0.05 or 0.5 mM) for 24, 48, or 72 h and counted using an hemocytometer or Couness Automated Cell Counter (Invitrogen) after Trypan Blue staining. For growth assays, GIMEN-EMPTY and GIMEN-MYCN cells were seeded into 12-well plates at 10,000 cells/well density in triplicate.
Generation of Isogenic Cell Lines Expressing MYCN-CMV-MYCN (16) and the empty pcDNA3.1(ϩ) vector were transfected into GI-M-EN cells using Lipofectamine 2000 (Invitrogen), as described in the manufacturer's protocol. Several G418-resistant clones were analyzed for MYCN expression by Western blot analysis, and the clone expressing the highest levels of MYCN was selected for further experiments.
Annexin V Staining-Apoptosis was detected using an annexin V Alexa Fluor 647 conjugate antibody (640912, Bio-Legend, Cambridge, UK) following the manufacturer's protocol. Briefly, 2 ϫ 10 6 cells were harvested and resuspended in 500 l of annexin V binding buffer (BioLegend) either with or without annexin V antibody (antibody was diluted 1:100 in annexin V binding buffer). After 1 h of incubation on ice, 10 l of a 200 g/ml DAPI solution was added to cells before the analysis using a BD LSRII flow cytometer.
Propidium Iodide Staining and FACS Analysis-Neuroblastoma cells were fixed on ice with 70% ethanol, then washed twice in phosphate-citrate buffer (0.2 M disodium phosphate and 0.1 M citric acid), and resuspended in PBS containing 2 g/ml propidium iodide, 0.1% Nonidet P-40, and RNase followed by analysis on BD CyAn ADP flow cytometer.

RESULTS
Genome-wide shRNA Drop-out Screen-The shRNA screen was carried out using an shRNA lentiviral library following a protocol described previously (26,31). The library consists of 67,200 lentiviral shRNA vectors targeting 19,851 human transcripts or 17,700 gene IDs. The library was divided into seven pools, each containing ϳ9,600 lentiviral shRNA vectors. Pools were calculated to introduce a single copy of the vector in at least 1,000 cells (Fig. 1A). Lentiviral pools were used to infect the control or MYCN-expressing GIMEN cell lines (Fig. 1B). To generate biological replicates, each pool was delivered in triplicate infections. Genomic DNA was isolated from the infected cells at early and late (2 weeks after puromycin selection) time points. The assumption was that a synthetic lethal shRNA should be depleted by the 2-week time point. We verified that MYCN overexpression led to increased growth rates of GIMEN cells as testified by cell growth assays and cell cycle analysis using flow cytometry (Fig. 1, C and D). These results are in line with previous reports in which MYCN overexpression was shown to promote cancer cell proliferation (32,33).
At the end of the screen, we identified 789 shRNAs depleted in GIMEN-MYCN cells targeting 537 genes (supplemental Data Set S1). To identify additional synthetic lethal shRNA, we inspected 4,429 hairpins that were added in further releases of the Open Biosystem library, focusing on genes implicated in the cell cycle. 60 genes were identified as involved in cell cycle regulation according to the Gene Ontology analysis by DAVID (34,35), and among these genes, 22 were previously identified as potential MYC-regulated genes. The promoter sequences of the 38 genes not already know as MYC target genes were inspected for the presence of MYCN binding sites using Eukaryotic Promoter Database and TFSEARCH (36,37). According to this analysis, nine genes contained potential MYCN binding sites in their promoters. Among these, we selected for further analysis PKMYT1 because it is a cell cycle kinase similar to WEE1, whose inactivation was previously demonstrated to be synthetic lethal to c-MYC (23, 38) (Fig. 1E).
Pathway analysis showed that the genes identified in the screen are involved in regulation of cell survival, DNA replication and damage, cell cycle, and cellular movements ( Fig. 2A). According to the Myc Target Gene Database (39), several known MYC target genes score high in the list such as ABCE1, BCL2L12, RBBP4, NAP1L1, EIF2S1, POLD1, BMP4, POLR2H, SUCLG1, HNRPA1, BIRC6, and AHCYL1 (supplemental Data Set S1). Notably, a number of the hits identified in our screen are functionally related to the MYC synthetic lethal genes organized in three major hubs recently described by Grandori and co-workers (29) on the basis of published RNAi screens. For example, we detected genes involved in transcriptional initiation and elongation (DDB1, BTAF1, TAF9B, TBPL1, and POLR2H), transcription and the MYC/MAX network (HDAC9,  CKS1b, CUL2, FBXL7, FBX040, FBX015, and FBX011) (supplemental Data Set S1 and Fig. 2B). Notably, we also identified AURKA, a kinase critically required to stabilize the MYCN protein and whose inactivation triggers synthetic lethality in MYCN-amplified neuroblastomas (24).  demonstrated that ablation of B-MYB triggers synthetic lethality in cells with amplified MYCN (40). Similarly to MYBL2, MYB expression is significantly increased in MYCN-amplified neuroblastomas predicting poor survival (Fig. 3, A and B). To verify whether c-MYB is required for fitness of MYCN-expressing cells, we infected GIMEN and GIMEN-MYCN cells with an shRNA lentiviral vector targeting MYB and observed that this caused massive activation of apoptosis in MYCN, but not non-MYCN, expressing cells (Fig. 3C). From a clinical perspective, MYB proteins are not optimal targets because small molecule inhibitors specific for this class of transcription factors have not been yet developed. To prioritize druggable genes and reduce   the number of candidates to a manageable number, we used the following criteria: (a) the genes are, or have the potential to be, direct MYC targets; (b) the genes have prognostic value in neuroblastoma; and (c) inhibitors are readily available. We therefore selected AHCY, BLM, PKMYT1, and CKS1B for further analysis.
The adenosylhomocysteinase (AHCY) gene encodes Sadenosylhomocysteine hydrolase (SAHH) 2 that metabolizes S-adenosyl homocysteine, an inhibitor of methylation reactions. It was shown previously that AHCY is a direct c-MYC target gene and that SAHH is critical for c-MYC metabolic effects and tumorigenic activity (41). The SAHH inhibitor 3-deazadenosine (3-DAZA) inhibits methylation reactions and has been shown to possess antiretroviral activity in preclinical experiments (42,43).
CKS1B encodes a cyclin-dependent kinase binding protein that plays a crucial role in cell-cycle regulation. CKS1B is overexpressed in many malignancies including breast, prostate, cervical cancers, nasopharyngeal carcinoma, and multiple myeloma, where it promotes cell proliferation (44 -49). CKS1B is transcriptionally induced by c-MYC and negatively regulates the cyclin-dependent kinase (Cdk) inhibitor p27 Kip1 in lymphoma, supporting c-MYC tumorigenesis (50). CKS1B has been shown to overcome the DNA damage response barrier triggered by activated oncoproteins, suggesting that in the context of activated MYC its expression could enhance cancer cell fitness 2 The abbreviations used are: SAHH, S-adenosylhomocysteine hydrolase; 3-DAZA, 3-deazadenosine. (45). A small molecule inhibitor of CKS1B, fluoexetine (also known as Prozac), is an antidepressant widely used in clinics. PKMYT1 encodes a member of the serine/threonine protein kinase family involved in cell cycle regulation. PKMYT1 inactivates cell division cycle 2 protein (CDK1) by promoting Thr 14 and Tyr 15 phosphorylation in concert with the WEE1 kinase (51). This phosphorylation event is critically required for progression of cells into mitosis. Ablation of WEE1 and PKMYT1 expression can cause mitotic collapse and apoptosis of cancer cells (38,52). A small molecule tyrosine kinase inhibitor, PD166285, has been shown to inhibit PKMYT1 at low nanomolar concentrations (53). The PKMYT1 promoter contains a potential E-box sequence near the transcription start site. ChIP assays confirmed that binding of MYCN could be detected in the MYCN-amplified neuroblastoma cell lines NB19 and LA-N-1, whereas it was absent in the non-MYCN-amplified SHSY5Y cell line (Fig. 4A).
Bloom syndrome, RecQ helicase protein (BLM), is required to maintain the intra-S-phase checkpoint, and its deregulation may be important for cancer progression (54 -57). People carrying mutations in the BLM gene are predisposed to different types of cancer and have a shortened lifespan (58). Because accelerated cell division imposed by activated MYC is associated with increased risk of replication errors and DNA damage, it is reasonable to postulate that cancer cells with activated MYC are particularly sensitive to reduced RecQ helicases such as BLM and WRN (59 -62). Recently, small molecule inhibitors of BLM have been developed that have shown anticancer activity in preclinical studies (63). Examination of the promoter of the BLM gene revealed a canonical E-box (CACGTG) located ϳ500 bp upstream from the transcriptional start site. We performed ChIP analysis to demonstrate that MYCN binds to the BLM promoter in MYCN-amplified LA-N-1 and NB19 neuroblastoma cells, whereas the binding was not detected in non-MYCN-amplified SH-SY5Y cells (Fig. 4B).
The significance of BLM, AHCY, PKMYT1, and CKS1B in the pathogenesis of neuroblastoma is indicated by their elevated expression in MYCN-amplified tumor samples and in MYCNoverexpressing cells and their correlation with poor patient survival (Fig. 5, A-C). To further validate the selected candidates, we transduced GIMEN-EMPTY and GIMEN-MYCN, with single pGIPZ-shRNAmir lentiviral vectors targeting PKMYT1, AHCY, BLM, and CKS1B. After 11 days in culture, GIMEN-MYCN cells infected with the shRNA targeting the candidate genes were more prone to apoptosis or growth arrest than parental cells (Fig. 6, A and B). Interestingly, concurrent inhibition of AHCY and BLM resulted in additive inhibition of the proliferation of MYCN-expressing cells, suggesting that targeting more than one gene at the time could have clinical value (Fig. 6B). As expected, the shRNAs caused reduced gene expression that varied between 90 and 50% in both cell lines (Fig. 6C). To confirm that down-regulating the selected genes has the potential to inhibit the survival and/or proliferation of cells with natural amplification of MYCN, we first assessed the expression of PKMYT1, BLM, AHCY, and CKS1B at the protein and mRNA levels in a panel of MYCN-amplified and nonamplified cell lines. We observed higher expression of the genes in MYCN-amplified cells, in line with the hypothesis that they are direct targets of MYCN (Fig. 7, A and B). Knockdown of the genes caused a selective increase of apoptosis in the majority of MYCN-amplified cells as opposed to cells with no MYCN amplification, confirming that expression of PKMYT1, BLM, AHCY, and CKS1B is crucial for the fitness of cells with activated MYC (Fig. 8, A and B).

BLM Limits the Oncogenic Stress and PKMYT1 Regulates MYCN Protein Levels in MYCN-amplified Neuroblastoma Cell
Lines-It has been reported that inhibition of BLM leads to increased DNA double-stranded breaks and apoptosis of cancer cells (54,63). To verify the role of BLM in neuroblastoma, we examined the expression of phosphorylated histone ␥-H2AX, used as a marker for DNA double-stranded breaks, after infection of neuroblastoma cells with a BLM shRNA. We observed that ablation of BLM expression caused increased expression of the ␥-H2AX marker in MYCN positive cell lines (Fig. 9A). Depletion of PKMYT1 caused a drastic reduction of MYCN protein in most of the neuroblastoma cell lines analyzed (Fig. 9B). This is consistent with the role of cdk1 as a kinase required to prime MYCN for proteosomal degradation (24). We hypothesized that ablation of PKMYT1 promotes cdk1/ GSK3-mediated phosphorylation and degradation of MYCN. Indeed, the PKMYT1 inhibitor PD166285 caused increased phosphorylation of MYCN at threonine 58 followed by sharp degradation of the MYCN protein that did not required de novo protein synthesis (Fig. 9, C and D). As a control, we verified that MYCN mRNA levels were unchanged after the treatments, ruling out an effect on gene expression (Fig. 9E).
PD166285 and 3-DAZA Cause Selective Apoptosis of MYCNamplified Cells-To assess whether chemical inhibitors of PKMYT1 and SAHH have the potential to destabilize the fitness of MYCN-amplified cells inducing synthetic lethality, we used PD166285 and 3-DAZA in survival assays. We observed that the compounds inhibited the growth of both MYCN positive and negative neuroblastoma cells. However, the drugs selectively killed MYCN positive cells (Fig. 10, A-D). Analysis by propidium iodide DNA staining and flow cytometry revealed that the drugs caused apoptosis of MYCN-positive, but only growth arrest in the G 1 phase of the cell cycle of MYCN-negative cells (data not shown).   concentration of each agent is lower than that required as a single drug, enhancing specificity and reducing toxicity. We tested the killing effect of PD166285, 3-DAZA, and inhibitors of BLM and CKS1B (ML216 and fluoxetine, respectively). When used as single agents at low micromolar (3-DAZA, ML216, and fluoexetine) or nanomolar (PD166285) concentrations, the drugs were ineffective or only caused a marginal inhibition of cell proliferation. Combining the drugs caused profound inhibition of proliferation of all neuroblastoma cell lines, independently from MYCN expression. However, a significant killing effect of the drugs was only observed in MYCN-amplified cells (Fig 11, A and B).

DISCUSSION
Oncogenic transcription factors are notoriously difficult to target therapeutically, and the hypothesis that we tested in this investigation is whether subverting a MYCN-dependent genetic program could be used as a strategy in treating cancers driven by this protooncogene. To identify key genes and pathways required for survival of cancer cells with activated MYC, FIGURE 11. Combinations of small molecule inhibitors targeting the MYCN interactome suppress neuroblastoma cell proliferation and induce selective killing of MYCN-amplified cells. Nonamplified (A) and MYCN-amplified (B) neuroblastoma cell lines were exposed to Flouxetine, ML216, 3-DAZA, and PD166285 alone or in combination as indicated. Live/death cells were harvested at the indicated times. The results represent the means of three independent experiments, each performed in triplicate. Error bars indicate standard deviations. *, p Ͻ 0.01; **, p Ͻ 0.001 (Student's t test).
we carried out an shRNA genome wide drop-out screen in cancer cell lines with or without expression of MYCN. Using this method, we identified hundreds of shRNAs depleted in MYCNexpressing cells that mainly inactivate genes involved in transcription, cell cycle, and apoptosis. The result of our analysis is consistent with previous RNAi screens in which genes synthetically lethal to c-MYC have been identified in mammalian cells. The Grandori laboratory has recently integrated the results of two major RNAi screen identifying three main hubs. The first hub includes genes involved in transcription initiation and elongation complexes; the second highlights both positive and negative regulators connected to the MYC/MAX network; the third hub includes kinases, ubiquitin, and sumoylation functions related to cell cycle checkpoint and DNA repair (29). Several of the MYCN synthetic lethal genes identified in our screen appear to be related to each of these hubs, in line with the hypothesis that MYCN and c-MYC functions are overlapping, and their role is mainly defined by different spatial and temporal physiopathological contexts. To prioritize interesting candidates, we have manually selected four genes that responded to the following criteria: (a) are direct MYC targets, (b) have prognostic value, and (c) are druggable. The genes selected for further analysis were AHCY, PKMYT1, CKS1B, and BLM. PKMYT1 is a serine-threonine kinase that in concert with WEE1 phosphorylates and inactivates cdk1 complexed with cyclinB, regulating entry into mitosis. Interestingly, the MYCN protein is destabilized by phosphorylation of serine 62 by the cdk1/cyclinB kinase required to prime phosphorylation of threonine 58 by gsk3b, initiating proteasomal degradation (24). We showed here that PKMYT1 is required to stabilize the MYCN protein in neuroblastoma cells by reducing T58 phosphorylation by the cdk1/gsk3 kinases, explaining the synthetic lethal effect of PKMYT1 inactivation in MYCN-amplified cells. CKS1B is overexpressed in several cancers and binds to cdk2, overcoming the replicative barrier imposed by activated oncoproteins. DNA replication stress induced by oncogenes such as Cyclin E or c-MYC can activate an intra-S-phase checkpoint that induces accumulation of inactive cdk2. CKS1B can override the intra-S-phase checkpoint by restoring cdk2 activity (45). Interestingly, inactivation of cdk2 has been previously shown to promote synthetic lethality in cells with amplified MYCN, confirming the relevance of this pathway for MYC tumorigenesis (64). BLM is similar to the Werner DNA helicase, WRN, required to limit replication stress imposed by accelerated S-phase imposed by activated oncogenes, including c-MYC (60,61). Our results support the hypothesis that BLM is critical for the survival of MYCN-amplified cells because it limits DNA damage promoted by oncogenic activation, as suggested by the induction of phosphorylated histone ␥-H2AX in MYCN-amplified cells after depleting BLM expression.
The majority of metastatic, high risk neuroblastomas are characterized by the presence of MYCN amplification. These highly proliferative tumors are treated with chemotherapeutic drugs that are potentially damaging and highly toxic in children. Dissecting the pathways required by cancer cells to tolerate oncogenic stress imposed by activated MYC potentially allows identification of druggable targets that would lead to a more specific and less toxic therapeutic treatments for MYC-driven tumors. To establish, as a proof of principle, that drugging the MYCN network has clinical value, we have used readily available small molecule inhibitors of BLM, CKS1B, PKMYT1, and SAHH. The compounds, either alone or in combination, caused death in the majority of MYCN-amplified cells, suggesting that clinically viable derivatives of these compounds could be used to treat high risk neuroblastomas and other tumors driven by MYC. Of course, the effects of the small molecule inhibitors will need to be validated in appropriate neuroblastoma mouse models prior to any potential clinical application. Mathematical modeling demonstrates that the simultaneous treatment of cancer cells with drugs targeting multiple pathways can improve patients outcome, avoiding drug resistance (65,66). Our observation that combining pharmacological concentrations of drugs targeting the MYCN network can activate synthetic lethality in MYCN-amplified tumor cells suggests that this approach has the potential to be successfully translated in the clinic.