Transcription Factor Activating Protein 4 Is Synthetic Lethal and a Master Regulator of MYCN Amplified Neuroblastoma

Transcription Factor Activating Protein 4 is synthetic lethal and a master regulator of MYCN amplified neuroblastoma

Shuobo Zhang

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Under the Executive Committee of the Graduate School of Arts and Sciences

COLUMBIA UNIVERSITY 2015

Shuobo Zhang

ABSTRACT

Transcription Factor Activating Protein 4 is synthetic lethal and a master regulator of

MYCN amplified neuroblastoma

Shuobo Zhang

Despite the identification of MYCN amplification as an adverse prognostic marker in neuroblastoma, no drugs that target MYCN have yet been developed. Here, by combining a whole genome shRNA library screen and Master Regulator Inference Algorithm (MARINa) analysis, we identified Transcription Factor Activating Protein 4 (TFAP4) as a novel synthetic lethal interactor with MYCN amplification in neuroblastoma. Silencing TFAP4 selectively inhibits MYCN amplified neuroblastoma growth both in vitro and in xenograft mice models.

TFAP4 expression is inversely correlated with patient survival in MYCN-high neuroblastoma.

Mechanistically, silencing TFAP4 induces neuroblastoma differentiation, as seen by increased neurite outgrowth, and up-regulation of neuronal markers. TFAP4 regulates a downstream signature similar to the signature of the oncogene anaplastic lymphoma kinase (ALK). Taken together, our results validate TFAP4 as an important master regulator in MYCN amplified neuroblastoma and a novel synthetic interactor with MYCN amplification. Thus, TFAP4 may be a novel drug target for neuroblastoma treatment. TABLE OF CONTENTS

List of tables and figures ...... ii

Acknowledgements ...... v

Chapter 1. Introduction ...... 1

Chapter 2. Materials and Methods ...... 12

Chapter 3. TFAP4 is a master regulator and potential synthetic lethal candidate for

MYCN amplified neuroblastoma ...... 21

Chapter 4. TFAP4 is synthetic lethal with MYCN amplification in neuroblastoma ...... 39

Chapter 5. TFAP4 inhibits differentiation of MYCN amplified neuroblastoma ...... 61

Chapter 6. Discussion ...... 80

References ...... 87

i LIST OF TABLES AND FIGURES

Chapter 1

Figure 1.1 Kaplan-Meier survival curve of infants <1 year old with metastatic

disease (Stage IV)...... 3

Figure 1.2 X-ray structure of MYC-MAX heterodimer binding on to

DNA E-box ...... 4

Figure 1.3 Illustration of synthetic lethal agents selectively killing cells with

gene A mutation...... 7

Figure 1.4 Schematic illustration of MARINa...... 10

Chapter 3

Figure 3.1 Neuroblastoma cell line and shRNA structure used in the synthetic lethal

screen...... 27

Figure 3.2 Whole genome shRNA library screen identifies 218 synthetic lethal

candidates with MYCN amplification...... 28

Figure 3.3 MARINa analysis identifies top 25 master regulators in

MYCN amplified neuroblastoma in NRC dataset...... 29

Figure 3.4 MARINa analysis identifies top 25 master regulators in MYCN amplified

neuroblastoma in NCI-TARGET dataset...... 30

Figure 3.5 Venn diagram of overlapping activated master regulators ...... 31

Figure 3.6 Venn diagram of overlapping transcriptional regulators in the

synthetic lethal screen and master regulators in MYCN amplified

neuroblastoma ...... 32

ii Figure 3.7 Rank of the overlapping master regulators in shRNA screen and in

MARINa analysis...... 33

Figure 3.8 TFAP4 expression is inversely correlated with survival in

MYCN-high neuroblastoma ...... 34

Chapter 4

Figure 4.1 MYCN directly upregulates TFAP4...... 46

Figure 4.2 TFAP4 is upregulated in MYCN amplified neuroblastoma patients .....47

Figure 4.3 SHEP21N competition assay...... 48

Figure 4.4 c-MYC expression is significantly higher in stage 4 MYCN non-

amplified neuroblastoma patients ...... 49

Figure 4.5 MYC and MYCN expression in neuroblastoma cell lines...... 50

Figure 4.6 Silencing TFAP4 by siRNA...... 51

Figure 4.7 Silencing TFAP4 by siRNA only inhibits MYCN amplified

neuroblastoma cell growth ...... 52

Figure 4.8 Knocking down of TFAP4 by two dox-inducible shRNAs in MYCN

amplified cells ...... 53

Figure 4.9 TFAP4 knocked down by dox-inducible shRNAs inhibits MYCN

amplified neuroblastoma cell growth ...... 54

Figure 4.10 Silencing TFAP4 inhibits colony formation of MYCN amplified

neuroblastoma ...... 55

Figure 4.11 Silencing of TFAP4 inhibits growth of MYCN amplified tumors ...... 56

Figure 4.12 Tumor weight of xenograft mice and TFAP4 protein level ...... 57

Figure 4.13 Silencing TFAP4 inhibits growth of established tumors ...... 58

iii Chapter 5

Figure 5.1 Family tree of helix-loop-helix transcription factors...... 62

Figure 5.2 Silencing TFAP4 induces neurite outgrowth in MYCN amplified

neuroblastoma ...... 68

Figure 5.3 Knocking down TFAP4 upregulates neuronal marker GAP43

expression...... 69

Figure 5.4 Silencing TFAP4 slows G1/S progression...... 70

Figure 5.5 Knocking down TFAP4 does not change CDKN1A expression...... 71

Figure 5.6 Differentially expressed genes in MYCN amplified neuroblastoma

after silencing TFAP4...... 72

Figure 5.7 Silencing TFAP4 represses CCNE2 gene expression...... 73

Figure 5.8 Silencing TFAP4 represses ROCK2 and PAK4 gene expression...... 74

Figure 5.9 TFAP4 induces similar signature as anaplastic lymphoma kinase...... 75

Chapter 6

Figure 6.1 Schematic diagram of TFAP4 signaling in MYCN amplified

neuroblastoma...... 85

iv ACKNOWLEDGEMENTS

First and foremost, I would like to thank my mentor, Dr. Darrell Yamashiro, for his guidance, mentorship and incredible support through the course of my graduate study. I am indebted to

Darrell for giving me such a wonderful experience to learn, explore and grow as a scientist.

Thank you also to my thesis committee members, Drs. Andrea Califano, Jan Kitajewski, Ken

Olive, and Jose Silva, for all your advice on my project. I am grateful for the Pathology and

Molecular Medicine graduate program at Columbia University for their support as well.

I would then like to thank all the members of the Yamashiro lab, past and present, for supporting me through this wonderful journey. I could not have imagined working with such a diverse group of colleagues and friends. It has been a great joy to learn from every one of you. Thank you to

Dr. Debarshi Banerjee, who always provided me great insights and suggestions. Thanks to Drs.

Angela Kadenhe-Chiweshe and Alejandro Garcia, who taught me and performed the intrarenal xenograft surgeries in such a professional way. Thank you also to Ashish Jani, Jason Mitchell,

John Andrews, Emily Sbiroli, Roderick Alfonso and Na Wei, for the assistance whenever I needed and for making lab so much fun. I wish you all every success in future. I am also extremely grateful for the support from Drs. Jessica Kandel and Angela Kadenhe-Chiweshe.

Additionally, Briana Fitch and Stanley Cho, worked with me as summer students. They both made significant contributions in validating the synthetic lethal candidates.

I extend a heartfelt thank you to the Califano lab, especially to my collaborator Dr. Gonzalo

Lopez, who often intrigued me with a different perspective. It has been my pleasure to work with you and this work could not have been so great without your contribution. Thanks also to Jiyang

v Yu, who introduced me to the bioinformatics world and donated his weekends to analyze the screen data for science and friendship. I would also want to thank the Silva lab, especially to Dr.

Jose Silva for your great mentorship when I first started at graduate school and for guiding me through this project, and to Ruth Rodríguez-Barrueco and David Llobet-Navas for teaching me all the molecular biology techniques. Finally I would like to thank the Califano, Kitajewski, and

Diacovo labs for lending me reagents, helping me with experiments, and making the ICRC 9th floor such a joyful place to work.

Starting graduate school in a foreign country is not easy and I am incredibly fortunate to have so many friends to help me along the way, to support me through every obstacle and to share all happiness with me. A special thank to my friends Yang Ou and Charles Wei for always being around, as my emergency contact, my dinner buddy, my travel buddy, etc. I wouldn’t have made it through without the fun you brought into my life. Thank you.

Finally I want to give my utmost thanks to my wonderful parents, June Zhang and Licun Bao.

Thank you for always believing in me, supporting me and inspiring me to fearlessly pursuit my dreams. I love you both!

Chapter 1

Introduction

Neuroblastoma

Neuroblastoma is the most common extracranial solid tumor in childhood and counts for 13% of all pediatric cancer mortality. In the United States, there were 10.7 cases per 1,000,000 persons aged 0-14 years diagnosed in 2010 (Louis and Shohet, 2015). It is a malignancy derived from the ventrolateral neural crest cells (Betters et al., 2010). The neural crest precursor cells are similar to other pluripotent cells with the ability to self-renew and generate multiple tissue types. During the early stages of embryogenesis, these neural crest cells migrate away from the neural tube to different parts of the body. Subsequent cascading of signaling gradients and ligands drive the pluripotent cells to differentiate into numerous derivatives including neurons and glia of the peripheral nervous system, melanocytes, portions of the cardiac outflow tract, craniofacial bone and cartilage, and smooth muscle of major blood vessels (Bronner and

LaBonne, 2012; Sauka-Spengler and Bronner, 2010). Inhibition at any stage of this neural crest differentiation process may predispose the precursor cells to malignancy development.

As a result, neuroblastoma presents a wide range of biologic heterogeneity and clinical outcome at diagnosis (Betters et al., 2010). The primary tumor is frequently located in tissues originating from the sympathetic nervous system, adrenal medulla, or paraspinal ganglia (Maris et al.,

2007). Metastasis occurs in ~50% of neuroblastoma patients at diagnosis, with the most frequent locations being the bone marrow (~70%), bone (~55%), lymph nodes (~30%), liver

(~30%) and brain (~18%) (DuBois et al., 1999).

Depending on the clinical and biological factors at diagnosis, neuroblastoma patients are classified into very low-, low-, intermediate- and high-risk groups (Cohn et al., 2009). Very low, low- and intermediate-risk patients generally have a favorable outcome with ~80%-95%

event-free survival rate. However, high-risk patients have <50% overall survival, despite intensive therapy with high dose chemotherapy, surgery, stem cell transplantation, and immunotherapy (Maris et al., 2007; Matthay et al., 2012). The high-risk patients are generally with MYCN amplification and/or >18 months and with metastatic disease.

MYCN in neuroblastoma

MYCN amplification is found in ~25% of neuroblastoma cases. Amplification of MYCN correlates significantly with poor prognosis (Figure 1.1) (Brodeur, 2003). The extra copies of

MYCN typically reside as double minutes or homogenously staining region amplicons (Cohn et al., 2009).

Figure 1.1: Kaplan-Meier survival curve of infants <1 year old with metastatic disease (Stage IV). Percentage of three-year event free survival of MYCN nonamplified neuroblastoma patients was 93% whereas for MYCN amplified patients it was only 10%. Figure from (Brodeur, 2003)

MYCN was first identified in neuroblastoma in 1983 as a c-myc related oncogene (Kohl et al.,

1983; Schwab et al., 1983). MYC and MYCN belong to the same basic-helix-loop-helix transcription factor family. Structurally, the coding regions of MYC and MYCN are highly

homologous, with gene products of similar sizes (~50-55 kDa) (Kohl et al., 1986; Stanton et al.,

1986). Functionally, MYC and MYCN can compensate for each other to certain degree. MYCN knocked-in at the MYC locus rescues embryonic lethality and restores the immune function of the MYC knockout mice, however, the MYCN knocked-in mice were smaller with differences in response to growth signals in some cell types (Malynn et al., 2000).

MYC protein forms a heterodimer with another bHLHZ protein MAX (Blackwood and

Eisenman, 1991; Blackwood et al., 1992). This heterodimerization is essential for MYC proteins to associate with its E-box DNA sequence CACGTG and stimulate transcription at promoter proximal E-boxes (Figure 1.2) (Amati et al., 1993; Kretzner et al., 1992; Nair and

Burley, 2003). The structure of MYC-MAX heterodimer presents a large solvent accessible surface area at the dimeric HLHZ region, which allows other cofactors to bind (Cheng et al.,

1999; Nair and Burley, 2003; von der Lehr et al., 2003).

Figure 1.2: X-ray structure of MYC-MAX heterodimer binding on to DNA E- box sequence CACGTG at resolution 19nm. MYC protein is in blue, and MAX protein is in red. Figure is taken from (Nair and Burley, 2003).

Under physiological condition, MYC functions as a sensor: integrating multiple cellular signals and mediating transcriptional activities in response to the cellular signals. Approximately 15% of genomic loci have been detected to bind MYC directly (Fernandez et al., 2003). In line with this observation, MYC has been shown to regulate many biological processes, including cell proliferation, cell death, DNA replication, metabolism, self-renewal and differentiation (Dang,

1999; Dang, 2012; Felsher, 2003; van Riggelen et al., 2010). Coordination of these processes is achieved by interacting with a wide variety of cofactors with diverse activities (other transcription factors, chromatin remodeling deacetylases, methyltransferases, antipausing factors, etc.) (Hann, 2014). The precise downstream transcription signal change may depend on the cofactors recruited, other transcription factors proximal to the binding site, and the chromatin organization of the target gene (Eilers and Eisenman, 2008; Guccione et al., 2006).

Overexpressing MYC may break this carefully balanced orchestra, predisposing cells to malignancy. Indeed, MYC overexpression and/or activation is found in more than half of human cancers, making it one of the most commonly activated oncogenes in cancer (Boxer and Dang,

2001; Escot et al., 1986; Kawate et al., 1999; Ladanyi et al., 1993).

MYCN also heterodimerizes with MAX and recognizes the same E-box binding sequence

CACGTG as MYC (Meyer and Penn, 2008). Since MYCN can compensate for MYC function, it is not surprising that MYCN recruits cofactors and regulates many biological processes in the same way as MYC. MYCN expression is up to 300-fold higher in MYCN amplified neuroblastoma than in normal tissue (Brodeur et al., 1984). Such high MYCN expression drives all facets of neuroblastoma tumorigenesis: promoting cell proliferation, inhibiting differentiation, influence immune surveillance, and promoting metastasis (reviewed in (Huang

and Weiss, 2013). MYCN transgenic mouse models confirm that MYCN overexpression in neural crest progenitor cells is sufficient to drive neuroblastoma tumorigenesis with high penetrance (Weiss et al., 1997). Therefore, MYCN has been seen as an attractive drug target for neuroblastoma treatment.

Synthetic lethality

Despite the discovery of MYCN amplification as an adverse prognostic marker for two decades, no clinically approved drug has been developed to target MYCN or MYC directly. This is likely due to MYCN and MYC being transcription factors composed of two extended alpha helices with no apparent surfaces for small molecule binding (Nair and Burley, 2003). An alternative approach is to target genes that are only essential for MYCN amplified tumors. Silencing of these genes would thus be “synthetically lethal” with MYCN amplified tumors.

Synthetic lethality is defined as two genetic mutations, either of which is innocuous to cells, but causes cell death when both are present in the same cell. This concept was first used in

Saccharomyces cerevisiae to map genetic interactions that affect cell viability (Dixon et al.,

2009). Later on, synthetic lethality was adapted to discover novel therapeutic targets in cancer.

The synthetic lethal agents, by definition, can selectively kill cells with certain genetic mutations (Figure 1.3).

Figure 1.3: Illustration of synthetic lethal agents selectively killing cells with gene A mutation. Synthetic lethal agents introduce a second mutation, which only leads to cell death when present with gene A mutation in the same cell. Figure is taken from (Fang, 2014).

The proof of concept example is the discovery of synthetic lethality of the repair enzymes poly

(ADP-ribose) polymerase (PARP) with the breast-cancer susceptibility genes 1 and 2 (BRCA1 and BRCA2). Cells with BRCA1 and BRCA2 mutation are dramatically sensitized to PARP inhibitors and PARP inhibitors selectively kills tumor without affecting normal tissues in mice models (Bryant et al., 2005; Farmer et al., 2005). PARP inhibitors are currently in clinical trials with encouraging results for the treatment of ovarian and breast cancers (Fong et al., 2009).

PARP1 and BRCA1/2 have well-established functions in homologous DNA recombination and

DNA repair, which logically lead to the examining synthetic lethality between the genes. Due to the expansion of our knowledge of the human genome and fast development of RNAi technology in mammalian cells, we can now employ a more systematic approach to unravel novel synthetic lethality interactions. Synthetic lethal screens were performed with either synthetic siRNA reagents transfecting into cancer cells individually in wells or with viral delivery of pooled shRNA libraries (Bernards, 2014). Changes in cell viability or shRNA 7

abundance were compared between cells with and without specific genetic mutations at the end point of the experiment. The siRNA reagents generate transient silencing lasting about 3-5 days, which limits the time length of the screen. shRNA infection on the other hand is able to deliver long-term stable gene silencing, and enrichment of shRNA can be analyzed by customized microarrays. Commercial shRNA libraries provide shRNAs set against the entire human genome or a specific subset such as genes expressed in one tissue, druggable genes, or kinome.

A few synthetic lethal screens on MYC/MYCN driven cancers have been reported so far

(Kessler et al., 2012; Liu et al., 2012; Otto et al., 2009; Toyoshima et al., 2012). Some synthetic lethal candidates were further validated in vivo, suggesting a synthetic lethality screen could potentially identify novel therapeutic target for MYCN amplified neuroblastoma.

Master Regulator Inference Algorithm (MARINa) analysis

Computational analysis of genome wide expression profiling has also been widely applied to identify novel drug targets associated with specific cancer phenotypes. With the development of high throughput profiling technologies and collaborative effort of cancer genome projects, resources from large-scale cancer patient studies have increased dramatically in recent years.

Early studies have been focused on exploiting individual genetic markers that associate with a phenotype of cancer (Hindorff; Kruglyak, 1999). However, few studies identified biomarkers successfully (reviewed in (Cooper and Shendure, 2011). The majority of the analyses found that enriched genetic regions lie in intergenic or intronic regions (Hindorff). Such regions are likely to influence gene regulation, but they are difficult to use as prognostic indicators or therapeutic targets.

Many bioinformatics studies have also sought gene signatures, with the results often including hundreds of genes in a gene signature and frequently with no statistically significant overlap between two signatures for one cancer phenotype (Ein-Dor et al., 2005; Michiels et al., 2005).

Such common variations cannot simply be explained as technical variation, caused by microarray data collection or computation analysis methods. One hypothesis is that the gene signatures identified are only the “passengers” of the cancer (Lefebvre et al., 2010; Lim et al.,

2009). Genetic mutations within cancer regulate such passenger genes, and as a result their expression is enriched in a subgroup of patients. Since these genes may not play any role in cancer development, their expression is unlikely to correlate with cancer phenotype on a large scale. To discovery more robust prognosis markers and/or therapeutic targets, one should aim to identify the “drivers” of the cancer whose activity leads to the development of a certain cancer phenotype.

Recently, a Master Regulator Inference Algorithm (MARINa) analysis was developed by

Califano and colleagues to identify key master regulators of a cancer phenotype (Lefebvre et al.,

2010; Lim et al., 2009) (Figure 1.4). First, a patient gene expression dataset is processed by an extensively validated algorithm ARACNe (Algorithm for the Reconstruction of Accurate

Cellular Networks) to infer the genome wide transcriptional interaction network (Margolin et al., 2006). The enrichment of each transcription factor’s regulon genes is then accessed by the

Gene Set Enrichment Analysis (GSEA) method (Subramanian et al., 2005).

Figure 1.4: Schematic illustration of MARINa. Transcriptional network is first reverse engineered utilizing ARACNe network and neuroblastoma patient database. Enrichment of each transcriptional regulator is analyzed by GSEA. Top ranked transcriptional regulators (depicted as red dots) are the master regulators of a cancer phenotype.

The differential expression of the transcription factor’s regulon between two prognostic groups, representing the activity of the transcription factor, is then ranked. Top ranked transcription regulators are master regulators that drive the progression of the cancer phenotype. It was shown that a small number of top master regulators gives a higher accuracy in predicting breast cancer metastasis than a 70 gene signature set. Moreover, the top master regulators inferred from different datasets are far more consistent than the gene signatures (Lim et al., 2009). Using

MARINa analysis, master regulators have been identified for B-cells proliferation in the germinal center, mesenchymal transformation of brain cancers and the transformation of malignant prostate cancer. In all cases, two synergistic master regulators were efficient to drive the transformation, providing a more robust prognostic indicator (Aytes et al., 2014; Carro et al., 2010; Lefebvre et al., 2010).

Summary of Research

Combining a whole genome shRNA screening and MARINa analysis, we identified TFAP4 as the top synthetic lethal candidate and master regulator for MYCN amplification in neuroblastoma. We showed that TFAP4 is an important direct downstream effector of MYCN, and that silencing TFAP4 inhibits MYCN-amplified cell growth both in vitro and in vivo. We also showed that TFAP4 regulates a similar downstream gene signature as ALK. TFAP4 functions to promote proliferation, migration and inhibit differentiation in MYCN amplified neuroblastoma.

Chapter 2

Materials and Methods

Cell culture and reagents

Neuroblastoma cell lines NGP and SHEP21N (gift of Dr. Garrett Brodeur), SH-SY5Y (ATCC) were cultured in RPMI medium (Gibco) supplemented with 10% fetal bovine serum (FBS) and

1% penicillin/streptomycin (Invitrogen). Neuroblastoma cell lines SK-N-DZ and CHP212

(ATCC) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco) with 10% FBS and 1% penicillin/streptomycin. All cells were incubated humidified at 37˚C and 5% CO2.

shRNAs were constructed in pGIPZ (non-inducible) plasmid or pTRIPZ (dox-inducible) plasmid (Open Biosystems). Both plasmids contain puromycin resistant gene for selection. pGIPZ plasmid contains a GFP marker that expresses constitutively. pTRIPZ plasmid contains shRNA and a RFP marker both of whose expression are induced by doxycycline (1 µg/ml).

Transfections:

Phoenix packaging cells were cultured in DMEM medium (Sigma) supplemented with 10%

FBS, 1% penicillin and streptomycin (Invitrogen). 6 µg of DNA was transfected with either 3

µg of lentiviral helpers, CMV and PMD (Salmon et al., 2000), using 40 µl PEI (Polysciences) transfection reagent in a 10cm tissue culture plate following manufacturer’s instructions.

Lentivirus in the medium were collected, filtered through a 220 µm filter and stored at -80˚C.

Infection:

Cells were seeded in 6 well plates overnight. Appropriate amount of lentivirus was added into medium. The plate was centrifuged at 1000 rpm for 60 min in the presence of 5 µg/ml polybrene

(Millipore) to increase infection efficiency. Stable cells with lentiviral infection were selected by puromycin (2 µg/ml).

shRNA screen

The shRNA library contains shRNAs targeting 18,661 genes in the human genome (gift from

Dr. Jose Silva). The pooled shRNA screen was carried out at a representation of 1000 cellular integrations per shRNA. From previous experience, an infection rate of 30% yields one shRNA infection per cell. SHEP21N cells were infected with shRNA library packed by lentivirus in independent duplicates. After puromycin selection, each replicate was divided into two populations and cultured in the absence or presence of 1µg/ml doxycycline for 10 doubling times. Cells are harvested at the end of incubation and genomic DNA was extracted. For each pair of DNA samples, shRNA region were PCR amplified from genomic samples and hybridized to a microarray containing the corresponding probes. The customized microarrays with complementary sequence to shRNA were from Agilent. Statistical analysis method of the microarray data is described in (Yu et al., 2013).

Master Regulator Inference analysis

Neuroblastoma patient gene expression profile is from National Cancer Institute (NCI)

TARGET database (http://target.nci.nih.gov) and European Neuroblastoma Research Network database (https://www.siopen-r-net.org/). Interactomes in MYCN-amplified neuroblastoma and stage 1 neuroblastoma were assembled using the ARACNe algorithm (Margolin et al., 2006).

Master regulator analysis was then performed using MARINa algorithm (Carro et al., 2010;

Lefebvre et al., 2010).

Cell proliferation

Competition assay

SHEP21N cells were infected with pGIPZ-shTFAP4 (Open Biosystems) or pGIPZ-control plasmid. GFP+ infected cells were sorted out by flow cytometry. The GFP+ pGIPZ-shTFAP4 or the pGIPZ control SHEP21N cells were mixed in equal amounts with the parental SHEP21N cells (GFP-) and plated in 10cm tissue culture plates. Cells were collected at day 0, 4, 7, 11 and

14, centrifuged at 1000 rpm for 5 min, resuspended in PBS, and percentage of GFP in the population was analyzed by FACS analyzer (LSRII Flow cytometer, BD).

siRNA Dimscan

ON-TARGETplus siRNA constructs (Dharmacon) were used to transfect neuroblastoma cells, four different siRNA constructs per targeted gene. Neuroblastoma cells were seeded into 96 well plates. After 24 hr incubation, cells were transfected in quadruplicate with siRNA (50nM) packed in DharmaFECT reagents (0.15 µl per well). Fresh medium was changed 24 hr after transfection. Total mRNA was collected 36 hr after transfection to measure efficiency of siRNA silencing. Cells were stained with 10µg/ml fluorescein diacetate and 0.1% Eosin Y 96 hours after transfection. Stained live cells were imaged and fluorescence was measured by Dimscan, as described in (Keshelava et al., 2005).

shRNA

Neuroblastoma cells NGP, SKNDZ, SY5Y and CHP212 were infected with doxycycline (Dox) inducible pTRIPZ-shRNA (Open Biosystems) and puromycin (2 µg/ml) selected post-infection.

shRNA infection efficiency was maximized by sorting out the top 20% RFP+ cells after 16 hours doxycycline induction.

For neuroblastoma cell proliferation assay, infected cells were seeded into 3.5 cm plate with or without 1 µg/ml doxycycline. At day 0, 2, 4 and 6, cells were harvested and centrifuged at 1000 rpm for 5 minutes. Cell nuclei were stained with trypan blue and live cells were counted using automated cell counter (Invitrogen).

For clonogenic assays, cells were plated in semi-solid media as follows in 6 well plates: 35mm plates were layered with 0.6% agar (Spectrum Chemical, AG110) and media, neuroblastoma cells infected with pTRIPZ-shRNA were seeded at 20,000 cells per well in triplicate in the second layer of 0.3% agar and media, and cultured in appropriate growth media with or without doxycycline (1 µg/ml). Cells were re-fed with 2ml medium every two days until colonies were macroscopic. The colonies were then stained with 1mg/ml Thiazolyl Blue Tetrazolium Blue

(MTT) (Sigma). Photos of the stained colonies were taken and pixels of colonies were quantified using Adobe Photoshop software.

ChIP assay

NGP cells were cultured as described above. For cross-linking formaldehyde (electron microscopy sciences) was added to a final concentration of 1%. The reaction was stopped by addition of glycine at a final concentration of 0.125M. Processing of cells was done following manufacturer’s instruction using truChIP chromatin shearing kit (Covaris). Chromatin was sheared by sonication (Covaris, S-220 series) to generate DNA fragments with an average size

of 500bp. Pre-clearing, incubating with antibodies and reversal of cross-linking was performed as described in EZ-ChIP immunoprecipitation kit (Millipore). MYCN antibody (Santa-Cruz

Biotechnology) or rabbit anti-mouse IgG (Sigma).

Primer sequences used are as following:

TFAP4 binding site Fwd: CGCGACGTTTGTAAATTGC

TFAP4 binding site Rev: CTCAGATCCCGAGGAAGGA

TFAP4 negative control Fwd: TCTCAGTGGTTCGTCCCTGT

TFAP4 negative control Rev: GGAGGCGGTGTCAGAGGT

PMA positive control binding site Fwd: ATCTTGTGTGGCACAGGT

PMA positive control binding site Rev: TCGTCTCTGGAGCCAGTTGG

mRNA expression by real time PCR

Total RNA was isolated from cells according to manufacturer’s instruction using the RNeasy

Mini kit (Qiagen), and the quantity and quality of the RNA analyzed by Nanodrop. One microgram of total RNA was reverse transcribed using Verso cDNA Kit (Thermo Scientific) in a 20ul volume. Resultant cDNA was used for template in quantitative real time PCR with ABI

7300 real time PCR System and pre-designed Taqman probes (TFAP4: Hs01558245_m1;

ACTB: 4326315E) (Applied Biosystems) or designed primers with Fast Start SYBR Green

Master (Roche).

Designed forward and reverse sequences for Fast Start SYBR Green system were as follows:

ACTB Fwd: ATTGGCAATGAGCGGTTC

ACTB Rev: CGTGGATGCCACAGGACT

PAK4 Fwd: CCACCGGGACATCAAGAG

PAK4 Rev: CAGAACCCAAAGTCTGACAGC

ROCK2 Fwd: CGCTGATCCGAGACCCT

ROCK2 Rev: TTGTTTTTCCTCAAAGCAGGA

CCNE2 Fwd: GGGAAACATTTTATCTTGCACA

CCNE2 Rev: CTGCAAGCACCATCAGTGAC

Each gene expression sample was done in triplicate measurement, and normalized with triplicate measurement of house keeping genes cyclophilin A and β-actin from the same samples.

Western blot

Cell

Cells were lysed in 1X protein lysis buffer (10x RIPA buffer (Sigma), 1mM PMSF and proteases inhibitor) and centrifuged at 13,200 rpm for 20 min at 4°C. Total protein in cell lysate was measured and 40µg of protein was boiled at 95°C for 10 min. Lysed protein was run on

SDS-PAGE gel and then transfer to nitrocellulose membrane. Non-specific binding was blocked by incubation with 5% non-fat milk and TBST (20 mM Tris-HCl pH 7.4, 150mM NaCl, 0.1%

Tween-20). Membranes were incubated with the primary antibodies overnight at 4°C followed by appropriate secondary antibody incubation for 1 hr at room temperature. The following antibodies were used for Western blotting: anti-MYCN (1:200; Santa-Cruz Biotechnology),

anti-TFAP4 (1:200; Santa-Cruz Biotechnology), anti-GAP43 (1:20000; Novus Biologicals), and anti-β-actin (1:5000; Cell Signaling).

Tumor

Tumor tissue was snap frozen in liquid nitrogen then homogenized in protein lysis buffer (10x

RIPA buffer (Sigma), 1mM PMSF and protease inhibitor cocktail). Total lysate centrifuged at

13,200 rpm for 20 min at 4°C and protein was collected in the aqueous layer. 40 µg of protein was taken to do Western blot as described above.

Xenograft tumors

Procedures were approved by Columbia University IACUC. NGP and CHP212 infected with pTRIPZ-shRNA and fuw-luc were counted and resuspended in phosphate buffered saline. For intrarenal tumors, 4-6 week old female nude mice (Taconic) were anesthetized with ketamine and xylazine, an incision made at the left flank, and 106 cells injected into the renal parenchyma

(Huang et al., 2003). Three days or 24 days after implantation, mice were randomized and half of the mice were giving drinking water with 2 mg/ml Dox. Luminescence of tumor cells was measured twice a week using a bioluminescence imaging system (Xenogen) until luminescence reaches threshold of 6x109 photons/sec. Mice were sacrificed and tumors were fixed or frozen for further analysis.

Immunohistochemical staining

Tumor tissues were fixed in normal buffered formalin and subsequently processed to paraffin.

Tissue serial section of 5 µm were deparaffinized, rehydrated, and heat mediated antigen

retrieval was performed using trypsin solution. Endogenous biotin was blocked with

Avidin/Biotin blocking kit (Vector), followed by CAS-block (Invitrogen) for 30 min at room temperature. Tissue was stained with anti-TFAP4 primary antibody (1:100; Thermo Scientific) at 4°C overnight and appropriate secondary antibodies (Vector). Slides were developed with

AEC single solution (Invitrogen) and counterstained with hematoxylin.

RNA-seq

Cells were treated with or without doxycycline for 40 hr. Total RNA was extracted from cells using RNAse extraction kit (Qiagen). Quality of mRNA was analyzed by bioanalyzer (Agilent).

RNA sequencing was conducted at Columbia genome center

(http://genomecenter.columbia.edu) using Illumina platform. It was performed at 30 million reads per sample. Expression data was normalized and analyzed using the limma package as previously described (Anders and Huber, 2010; Ritchie et al., 2015). Gene Set Enrichment

Analysis (GSEA) was performed using database and methods developed by Broad Institute

(Mootha et al., 2003; Subramanian et al., 2005)

Chapter 3

TFAP4 is a master regulator and potential synthetic lethal candidate for

MYCN amplified neuroblastoma

Introduction

To identify synthetic lethal interactions with MYCN amplification, we utilized a whole genome shRNA library developed by Silva and colleagues (Silva et al., 2005). The library contains

58,493 shRNA targeting 18,661 genes, which encompasses almost the entire human genome.

This allows for unbiased screening when identifying novel synthetic lethal genes. In the shRNA library, there are multiple different shRNA constructs targeting the same gene in order to minimize shRNA off-target effects. The shRNAs are constructed with a microRNA mir-30 backbone (Lee et al., 2003; Lee et al., 2004). The plasmid was packed by lentivirus and integrated into the genome. Once transcribed, the microRNA backbone of shRNA is recognized by endogenous microRNA processing machinery to generate a 20-30 nucleotide siRNA product silencing the target gene. This provides a more efficient RNAi production (Paddison et al.,

2004; Silva et al., 2005).

The use of an unbiased genome wide synthetic lethal screen, however, will generate a lengthy list of candidates. We hypothesized that the “best” synthetic lethal gene would be one that would regulate multiple genes and be the “driver” for that cancer malignancy. Thus another way to discover synthetic lethal candidates is to identify master regulators for MYCN amplified neuroblastoma. There are several neuroblastoma patient expression profiling databases available from large scale studies. One of the resources is the National Cancer Institute sponsored

TARGET neuroblastoma project (http://target.nci.nih.gov). It contains expression data for 250 patients, 66 stage 4 MYCN amplified patients, 145 stage 4 MYCN non-amplified patients and 33 low-risk stage 1 patients. Data were obtained from patients enrolled in various Children’s

Oncology Group biology and clinical trials. Another resource is the dataset from European

Neuroblastoma Research Network (https://www.siopen-r-net.org/). It includes 44 stage 4 MYCN amplified patients, 59 stage MYCN nonamplified patients and 50 stage 1 low-risk patients.

Using MARINa analysis, we interrogated both neuroblastoma datasets, reconstructed the transcriptional network, and identified the top master regulators. Overlapping results from two different datasets, we can minimize master regulator selected due to data sample bias.

Combining both the computational MARINa analysis approach and functional shRNA whole genome screening, we identified the top synthetic lethal candidate TFAP4.

Results

As a first step to identify novel synthetic lethal candidate with MYCN amplification, we performed a whole genome shRNA screen. A pooled shRNA lentiviral library consisting 58,493 shRNA-mirs targeting 18,661 known human genes was used to infect a neuroblastoma cell line

SHEP21N (Silva et al., 2005). The shRNA backbone vector co-expresses the targeting shRNA- mir, a puromycin resistance gene, and GFP (Figure 3.1a). SHEP21N constitutively expresses exogenous MYCN a level similar to the MYCN amplified neuroblastoma cell line NGP. MYCN expression can be switched off by adding doxycycline, which decreases MYCN to an expression level similar to the MYCN nonamplified cell line SH-SY5Y (Figure 3.1b) (Lutz and

Schwab, 1997). Cells infected with the shRNA-mir library were puromycin selected and split evenly into two populations: one without doxycycline treatment thus MYCN ON and the other one with doxycycline treatment thus MYCN OFF. Total DNA of these two populations was collected after ten cell doubling times. We identified shRNA candidates using a customized microarray and statistical analysis methods as previously described (Yu et al., 2013). Synthetic lethal candidates were the shRNAs that were significantly depleted in MYCN ON population only (Figure 3.2a). We identified 396 shRNAs that were differentially expressed between the two populations (P<0.01), of which 218 were significantly depleted in the MYCN ON population - thus our synthetic lethal candidates (Figure 3.2b).

We hypothesized that the best synthetic lethal candidates would be genes that regulate multiple genes, e.g. master regulator (MR) genes. A novel algorithm, MARINa, was used to interrogate both the NCI TARGET neuroblastoma dataset and European neuroblastoma research network dataset. We reverse engineered the transcriptional network in MYCN amplified neuroblastoma 24

specifically and created a MR rank list comparing MR regulon expression in MYCN amplified patient group with the stage 1 low-risk patient group. The top 25 master regulators in both NRC dataset (Figure 3.3) and NCI-TARGET database (Figure 3.4) were listed. We identified 90 MRs in MYCN amplified patients from NRC dataset and 86 MRs from NCI-TARGET dataset

(P<0.01). Among them, 56 MRs were common between both datasets (Fisher’s exact test p- value = 1.31e-56, Figure 3.5).

Correlating the activated MRs (49 genes) with the transcriptional regulators in MYCN-SLG list

(25 genes), we identified four candidates that are synthetic lethal and master regulators in

MYCN amplified neuroblastoma. These are transcriptional factor activating enhancer binding protein 4 (TFAP4), mitochondrial recycling factor (MRRF), protein kinase DNA activated catalytic polypeptide (PRKDC), and zinc finger protein (ZNF77) (Figure 3.6). shRNAs against these four genes were significantly depleted in the MYCN ON population in the whole genome shRNA screen (Figure 3.7a). They are robust MRs in MYCN amplified neuroblastoma patients

(Figure 3.7b), but not in MYCN nonamplified stage 4 patients compared to low-risk stage 1 patients (Figure 3.7c). Among the four candidates, TFAP4 ranked highest as a key master regulator in both neuroblastoma databases (1st in NRC dataset and 6th in NCI-TARGET dataset). shRNA against TFAP4 were also significantly depleted in MYCN ON population (ranked 150th in MYCN-SLC).

If inhibition of TFAP4 is synthetic lethal with MYCN, we would predict that MYCN-high expressing human neuroblastoma tumors with high TFAP4 expression would have poorer survival. To test this hypothesis we analyzed the TARGET neuroblastoma data set. Tumors

were stratified on the basis of MYCN expression levels (MYCN-high/MYCN-low), and then

TFAP4 expression levels (TFAP4-high/TFAP4-low). In patients with MYCN-high tumors, those with a higher-level expression of TFAP had a significantly worse survival then those with lower-level expression with P=0.022 (log-rank test). In contrast, the expression of TFAP4 did not correlate with outcome in patients with MYCN-low tumors (Figure 3.8). These results indicate that TFAP4 is important for tumorigenesis only in neuroblastoma with MYCN amplification.

a MYCN

β-actin

NGP SY5Y SHEP21N -Dox +Dox

Figure 3.1: Neuroblastoma cell line and shRNA structure used in the synthetic lethal screen. A) Doxycycline (1 µg/µl) switched off MYCN expression in SHEP21N, to level comparable MYCN nonamplified SH-SY5Y cells. Proteins were collected three days after doxycycline induction. B) structure of the shRNA used in the shRNA library screen;

Figure 3.2: Whole genome shRNA library screen identifies 218 synthetic lethal candidates with MYCN amplification. a) Scheme of whole genome shRNA screening in SHEP-21N neuroblastoma cells. After cells infected by whole genome shRNA library, SHEP21N were split evenly to two populations, one treated with doxycycline (MYCN OFF), one without doxycycline (MYCN ON). The two populations then grew separately for ten passages. Abundance of an individual shRNA was determined by hybridization to a customized microarray. Synthetic lethal candidates are the shRNAs depleted in MYCN ON population only; b) Heatmap of differentially expressed genes in MYCN ON vs MYCN OFF conditions. Abundance of shRNAs against 396 genes significantly changed (p-value < 0.01), with 218 genes depleted in the MYCN ON population.

NRC

Figure 3.3: MARINa analysis identifies top 25 master regulators in MYCN amplified neuroblastoma in NRC dataset. Master regulators’ positive targets (activated) are shown in red bar and negative targets (repressed) are shown in blue bars. The ranks of differential activity (DA) and differential expression (DE) are shown by the shaded boxes.

NCI-TARGET

Figure 3.4: MARINa analysis identifies top 25 master regulators in MYCN amplified neuroblastoma in NCI-TARGET dataset. Master regulators’ positive targets (activated) are shown in red bar and negative targets (repressed) are shown in blue bars. The ranks of differential activity (DA) and differential expression (DE) are shown by the shaded boxes.

Figure 3.5: Venn diagram of overlapping activated master regulators. Master regulators activated in MYCN amplified tumors (p-value < 0.01) overlap between TARGET (red circle) and NRC (yellow circle) cohorts. There are 1313 genes tested in both datasets.

Figure 3.6: Venn diagram of overlapping transcriptional regulators in the synthetic lethal screen and master regulators in MYCN amplified neuroblastoma. 25 transcriptional regulators that were significantly depleted in the shRNA screen (pink circle) and 49 activated master regulators identified by MARINA algorithm (red circle). There are 1077 genes tested in both datasets.

Figure 3.7: Rank of the overlapping master regulators in shRNA screen and in MARINA analysis. A) Four master regulators in MYCN amplified neuroblastoma were depleted in the shRNA screen. shRNA differential change was ranked by z-score with the most significantly depleted shRNA from the left tail end of the curve; b) Four master regulators in MYCN amplified neuroblastoma. TFAP4 ranked highest. 1344 transcriptional regulators were ranked by the significance of differential expression of its regulon. The curve represents the distribution with the most repressed master regulators at left tail end and the most activated master regulators at right tail end. c) None of the identified synthetic lethal master regulators in MYCN amplified neuroblastoma are master regulator in MYCN nonamplified neuroblastoma patients.

Figure 3.8: TFAP4 expression is inversely correlated with survival in MYCN-high neuroblastoma (P=0.022, log-rank test). NCI-TARGET neuroblastoma patient dataset was stratified based on the MYCN expression evenly to “MYCN-high” and “MYCN-low” group. In each group, patients were then stratified by TFAP4 expression evenly to “TFAP4-high” (red) and “TFAP4-low” (blue).

Discussion

Several synthetic lethal screens with MYC overexpression have been reported (Kessler et al.,

2012; Liu et al., 2012; Toyoshima et al., 2012; Zhou et al., 2014). However, BRD4 was the only common hit among the screens. BRD4 belongs to the family of bromodomain proteins, which associate with acetylated chromatin and facilitate transcriptional activation (Rahman et al.,

2011). Preclinical data showed that the small molecule inhibitor of bromodomain proteins, JQ1 downregulates the MYC transcriptional network and inhibits tumor growth of MYC-driven multiple myeloma (Delmore et al., 2011), lymphoma (Bhadury et al., 2014) and neuroblastoma

(Wyce et al., 2013). This suggests that synthetic lethal screens can identify novel targets for

MYC driven cancers. Several bromodomain inhibitors are currently in phase I clinical trials.

However, the anticancer activity of bromodomain inhibitors, have shown much broader effects than down-regulating the MYC network and toxicity in patients remains to be tested (Shi and

Vakoc, 2014).

On a closer look, it is not surprising that there is only one common hit among these reported screens. The screens in Liu et al. (Liu et al., 2012) and Toyoshima et al (Toyoshima et al., 2012) were done with siRNA technologies for a short length of selection times, 48 hr and 7 days after transfection, respectively. Screens with such short selection time do not identify long-term synthetic lethal genes. The screens also used a selective set of siRNA library: human kinome siRNAs or druggable siRNAs (Liu et al., 2012; Toyoshima et al., 2012). The screen by Zhou et al (Zhou et al., 2014) used a selective kinase pooled shRNA library. Only the screen by Kessler et al (Kessler et al., 2012) was performed with a whole genome shRNA screen for 12 doubling time of cells. Similarly, we employed a whole genome shRNA library to screen for synthetic

lethal genes after 10 cell doubling time (~30 days). The longer selection time would allow identification of the long-term synthetic lethal genes with MYC/MYCN overexpression.

The whole genome library screen is a validated unbiased functional screening to discover novel synthetic lethality interactions. However, the technology has several limitations: 1) variation in efficiency to silence genes; 2) short-term in vitro selection process; 3) a large number of candidates generated by microarray analysis which requires secondary selection. Therefore, shRNA screens should be used as one filter to select synthetic lethal genes and prioritization of genes should not be only be based on statistical analysis of shRNA screens alone.

Another way to discover novel synthetic lethal interaction is by computational analysis of patient expression databases. Employing an algorithm to analyze the genome wide post- translational modulators of transcription activity (MINDy) in MYC-driven cancers (Wang et al.,

2009), the novel candidate CSNK1e was identified. In one of the MYC-driven synthetic lethal screens, CSNK1e was also identified as one of the 49 synthetic lethal candidates from the ~3300 druggable siRNAs screen. CSNK1e was the only candidate that was further validated in vivo, because its expression correlated with MYCN amplification in neuroblastoma (Toyoshima et al.,

2012). This demonstrated that combining functional biological screens with computational analysis provides much greater power to identify key synthetic lethal interactors in cancer development.

Therefore, we also employed a computational approach - the master regulator inference algorithm (MARINa) to interrogate neuroblastoma patient expression profile data and identify

the top master regulators in MYCN amplified neuroblastoma. MARINa and MINDy are two different approaches to reverse engineer the transcriptional network. MINDy analyzes transcriptional regulation of all genes and identifies modulators at upstream of MYC in MYC- driven cancers. MARINa has the power to compare two cancer phenotypes, e.g. MYCN amplification verses stage 1 neuroblastoma and ranks the master regulators according to differential expression of their regulons in two phenotypes. The master regulators may act upstream, downstream or parallel with MYCN amplification to drive tumor progression. We interrogated two neuroblastoma patient datasets from different resources, so that results due to sample bias can be minimized. Out of the 86 MRs identified in NRC dataset and 90 MRs in

TARGET dataset, 56 were common (P = 1.31e-56). This suggests MARINa analysis is a robust algorithm to predict MRs. Overlapping the top MRs in MYCN amplified neuroblastoma with the

25 transcriptional regulators in MYCN-SLG list we identified four candidates, MRRF, PRKDC,

ZNF77 and TFAP4. PRKDC was previously described as the top synthetic lethal gene for

MYC-dependent cancer in a pooled kinase shRNA screen ((Zhou et al., 2014)). This result confirms the validity of the approach of combining shRNA library screen and master regulator computational analysis to identify novel targets.

We selected TFAP4 as our top synthetic lethal and master regulator candidate, because it is the highest ranking master regulator among the four candidate genes (1st in NRC dataset, 6th in

TARGET dataset). shRNA against TFAP4 was also significantly depleted in MYCN ON population. TFAP4 was the top ranking MR in MYCN amplified neuroblastoma patients, but not in stage 4 MYCN nonamplified neuroblastoma patients. Moreover, homozygous TFAP4-null mice were grossly normal and fertile, suggesting that knocking down TFAP4 does not affect

normal cells (Egawa and Littman, 2011). When we stratified the patient survival in NCI-

TARGET database, the Kaplan-Meier survival curve showed TFAP4 expression is significantly correlated with poor survival only in MYCN high group (P=0.022) (Figure 3.8). Taken together, our results suggest that TFAP4 is a synthetic lethal master regulator for MYCN amplified neuroblastoma.

Chapter 4

TFAP4 is synthetic lethal with MYCN amplification in neuroblastoma

Introduction

TFAP4 is a ubiquitously expressed transcription factor belonging to the basic helix-loop-helix leucine zipper group of proteins, which recognizes the symmetrical DNA core sequence

CAGCTG (Hu et al., 1990). Elevated expression of TFAP4 mRNA has been reported in many malignant cancers, including invasive adenocarcinoma (Cao et al., 2009), pancreatic cancer

(Friess et al., 2003), colorectal adenomas (Jackstadt et al., 2013b), hepatocellular cancer (Hu et al., 2013) and gastric cancer (Xinghua et al., 2012). Xenograft experiments showed that overexpressing TFAP4 alone cannot drive tumorigenesis (Jackstadt et al., 2013b).

Overexpression of TFAP4 along with the RAS oncogene, however, increased the frequency of tumor formation of RAS expressing MEFs to a similar extent as overexpressing c-MYC and RAS in MEFs. This suggests that TFAP4 can perform additional tumorigenic functions (Jackstadt et al., 2013a).

Previous literature has shown that TFAP4 is a direct and conserved target of c-Myc (Jung et al.,

2008). TFAP4 is upregulated when MYC expression is induced and TFAP4 protein is expressed at regions of breast tumor tissue where MYC is overexpressed. Since MYC and MYCN recognize the same E-box sequence and regulate similar downstream genes, we show here that

TFAP4 is direct target of MYCN in neuroblastoma.

Several papers have also reported that MYCN and MYC regulate each other at the level of transcript expression (Breit and Schwab, 1989; Helland et al., 2011; Rosenbaum et al., 1989;

Westermann et al., 2008). MYC is frequently upregulated in MYCN-non-amplified neuroblastoma. We examined TFAP4 synthetic lethality with MYCN in five neuroblastoma cell

lines. We showed that TFAP4 is synthetic lethal with MYCN amplification in vitro and in vivo, despite upregulation of MYC in MYCN non-amplified neuroblastoma.

Results

The neuroblastoma cell line SHEP21N constitutively expresses exogenous MYCN, which can be switched off by adding doxycycline (Lutz and Schwab, 1997). When MYCN expression is switched off by doxycycline in SHEP21N, TFAP4 is downregulated at both transcriptional and translational level (Figure 4.1a, b). One MYC/MYCN E-box binding sequence was identified within the first intron of the TFAP4 gene. ChIP-PCR assay demonstrated that MYCN directly binds to the predicted binding site (Figure 4.1c), indicating TFAP4 is directly regulated by

MYCN. In neuroblastoma patients, TFAP4 expression level is significantly higher in the MYCN amplified patient group compared to both the MYCN non-amplified patient group and the stage

1 low-risk patient group (P = 2.54e-12 in TARGET dataset, P = 6.01e-12 in NRC dataset)

(Figure 4.2a). Immunohistochemistry staining of TFAP4 on neuroblastoma patient samples also showed that TFAP4 is highly expressed in the four MYCN amplified primary tumors but not in the four stage 1 tumors (Figure 4.2b). Taken together, we showed that TFAP4 is upregulated in

MYCN amplified neuroblastoma.

We then validated TFAP4 synthetic lethality with MYCN amplification in the shRNA library screening cell line SHEP21N. A multicolor competition assay was performed to access the growth rate of SHEP21N with TFAP4 knocking down compared to the growth rate of

SHEP21N with empty vector control. Both shRNA constructs against TFAP4 and empty vector pGIPZ expresses GFP. We mixed equal number of infected GFP+ SHEP21N cells and uninfected GFP- SHEP21N cells. The percentage of GFP+ cells in each population was analyzed by FACS at different time points (Figure 4.3a). TFAP4 knocked down by the pGIPZ- shTFAP4 depleted in the shRNA screen was confirmed (Figure 4.3b). After two weeks, GFP+

percentage was significantly decreased by 35% when TFAP4 was knocked down and MYCN overexpression was on (-dox), whereas GFP+ percentage was only reduced by 5% at MYCN off

(+dox) condition (Figure 4.3c). This suggested that silencing TFAP4 only inhibits SHEP21N cell proliferation when MYCN is high.

MYCN and MYC have been shown to regulate each other (Breit and Schwab, 1989; Helland et al., 2011; Rosenbaum et al., 1989; Westermann et al., 2008). In line with these previous reports,

MYC expression is significantly higher in Stage 4 MYCN nonamplified neuroblastoma patients compared to stage 4 MYCN nonamplified patients in both the NCI-TARGET and NRC neuroblastoma datasets (Figure 4.4). We chose two MYCN amplified human neuroblastoma cell lines (NGP and SK-N-DZ) and two MYCN non-amplified human neuroblastoma cell lines

(CHP212 and SH-SY5Y) to further validate synthetic lethality of silencing TFAP4 with MYCN amplification. MYCN expression in the MYCN amplified cells were ~200 fold higher than

MYCN non-amplified cells (Figure 4.5a, d). In contrast, C-MYC mRNA expression and protein levels in the two MYCN nonamplified cell lines were significantly higher, especially in CHP212 whose C-MYC mRNA was ~200 folds higher than the level in MYCN amplified cells (Figure

4.5b, d). Since MYC and MYCN both regulate TFAP4 expression at transcriptional level, there was no significant difference in TFAP4 mRNA expression and protein level between MYCN amplified cell lines and MYCN non-amplified cell lines (Figure 4.5c, d).

Four different constructs of siRNAs were used to silence TFAP4 in order to exclude possible off-target effect. All siRNAs and shRNAs constructs knocked down TFAP4 (Figure 4.6). 96 hours after siRNAs transfection, silencing TFAP4 significantly reduced survival of the MYCN

amplified cell line NGP cell by 50% and SK-N-DZ by 40%, while cell survival of MYCN nonamplified lines were not affected (Figure 4.7).

We generated stable cell lines with doxycycline inducible shRNAs against TFAP4. Effective knocking down of TFAP4 was confirmed by both real-time PCR and Western blot (Figure 4.8).

Silencing TFAP4 significantly reduced cell growth of MYCN amplified line NGP by 25% and

SK-N-DZ by 35% 6 days after shRNA induction, while MYCN nonamplified cell growth was not affected (Figure 4.9). We further validated the results by the soft agar colony formation assay. Knocking down TFAP4 significantly reduced colony formation of NGP to 60% and SK-

N-DZ to 50% after three weeks, while there was no difference of CHP212 colony formation

(Figure 4.10). (SH-SY5Y did not form colonies in soft agar.)

We validated TFAP4 synthetic lethality with MYCN amplification in a xenograft model. 106 luciferase expressing NGP shTFAP4 cells were implanted intrarenally in nude mice. Mice were randomized after 7days, and half the mice were given drinking water with doxycycline to induce silencing of TFAP4. Tumor growth was monitored by bioluminescence imaging. From week 3 onwards, we observed significantly reduced tumor growth in the TFAP4 silenced group

(+dox). By week 5, tumors have established in all mice in the –dox control group, while tumor growth is significantly inhibited in +dox TFAP4 silenced group (Figure 4.11a,b). Mice were sacrificed when the primary tumor luciferase flux reached 6x109 photons/sec. Knocking down

TFAP4 in MYCN amplified NGP significantly inhibited tumor growth with the median survival prolonged by 12.5 days (P = 0.0058)(Figure 4.11c). Decreased expression TFAP4 in +dox

tumors was verified by Western blot (Figure 4.12b), demonstrating that tumor progression was not due to escape from silencing of TFAP4.

In contrast, silencing TFAP4 in MYCN nonamplified CHP212 did not affect tumor growth in the intrarenal xenograft model. By week 4, tumor growth was established in all mice with similar luciferase readings and there was no difference in survival rate between two groups

(Figure 4.11d-f). Tumor weights did not differ between –dox and +dox group in both NGP and

CHP212 xenograft mice, indicating luciferase reading correlated well with actual tumor growth

(Figure 4.12a).

We performed a regression study by inducing TFAP4 silencing after NGP tumors were established, 24 days after intrarenal implantation. Tumor growth was monitored by bioluminescence imaging and mice sacrificed when the primary tumor flux reached 6x109 photons/sec. There was no difference in tumor flux with TFAP4 silencing for the 24 days following addition of doxycycline (Figure 4.13a). Silencing of TFAP4, however, did prolong overall survival with a median survival day significantly prolonged by 13 days (P = 0.03, Figure

4.13b). This result indicates that TFAP4 is important for the growth of established tumors.

Figure 4.1: MYCN directly upregulates TFAP4. a) quantitation of MYCN and TFAP4 mRNA in SHEP21N cells with or without doxycycline. Expression was analyzed by quantitative PCR 72 hrs after doxycycline (1 µg/µl) addition. b) SHEP21N cells were treated with or without doxycycline (1 µg/µl) for 72 hrs. Expression of MYCN, TFAP4, and a loading control β-actin was determined by immunoblotting. c) ChIP assay showing that MYCN binds to the predicted binding site in the first intron of TFAP4, but not at a non-binding site in TFAP4.

Figure 4.2: TFAP4 is upregulated in MYCN amplified neuroblastoma patients. A) Violin plot of TFAP4 expression level in stage 4 MYCN amplified patients (red), stage 4 MYCN nonamplified patients (yellow) and stage 1 low-risk patients (green). Neuroblastoma database are from NCI-TARGET (left) and NRC (right). B) immunohistochemistry staining on neuroblastoma patient tissues. TFAP4 (red) protein is highly expressed in stage 4 MYCN amplified patients. Picture represents staining of 4 patient samples in each group.

Figure 4.3: SHEP21N competition assay. A) Schematic illustration of cell growthcompetition assay. SHEP21N cells infected with either shTFAP4-GFP or an empty vector control pGIPZ- GFP were mixed with equal number of uninfected SHEP21N cells. The infected cells to WT cells ratio is measured by FACS; b) Western blot of TFAP4 protein level in SHEP21N-pGIPZ control cells and SHEP21N expressing shTFAP4. c) %GFP ratio measured at day 4, 7, 10, and 14. Experiments were performed in triplicate. Mean + stddev.

Figure 4.4: c-MYC expression is significantly higher in stage 4 MYCN nonamplified neuroblastoma patients. Violin plot of TFAP4 expression level in stage 4 MYCN amplified patients (red), stage 4 MYCN nonamplified patients (yellow) and stage 1 low-risk patients (green). Neuroblastoma database are from NCI-TARGET (left) and NRC (right).

Figure 4.5: MYC and MYCN expression in neuroblastoma cell lines. A) Gene expression of MYCN; b) Gene expression of C-MYC; c) Gene expression of TFAP4. All gene expression were measured by quantitative PCR. Error bar depicts standard deviation. D) Western blot of MYCN, CMYC, TFAP4 protein levels in different neuroblastoma cell lines. β-actin is the loading control.

Figure 4.6: Silencing TFAP4 by siRNA. Neuroblastoma cell lines were transfected with four different constructs of siRNA against TFAP4, and TFAP4 gene expression is measured by RT- PCR 40 hours after transfection. Error bar depicts standard deviation.

Figure 4.7: Silencing TFAP4 by siRNA only inhibits MYCN amplified neuroblastoma cell growth. a) % of neuroblastoma cell survival 96 hours after siRNA mediated down regulation of TFAP4. Neuroblastoma cell lines were transfected with four different constructs of siRNA against TFAP4. Cell viability was quantified by Dimscan. Effect of TFAP4 silencing on cell survival was compared to cells transfected with scrambled siRNA. The experiments were performed in quadruplicate. Mean + stddev.

Figure 4.8: Knocking down of TFAP4 by two dox-inducible shRNAs. (a) MYCN amplified cell lines and (b) MYCN nonamplified cell lines. TFAP4 gene expression was measured by quantitative PCR 3 days after doxycycline induction. Error bar depicts standard deviation. Total proteins were collected 3 days after doxycycline induction and measured by Western blot.

Figure 4.9: TFAP4 knocked down by dox-inducible shRNAs inhibits MYCN amplified neuroblastoma cell growth. Neuroblastoma cell lines were infected with two different dox- inducible shRNAs against TFAP4 as well as the empty vector control pTRIPZ. shRNA was induced by 1 µg/µl doxycycline at day 0. Cells were counted on day 0, 2, 4, and 6. Experiments were performed in triplicate. Mean + stddev.

a a

Figure 4.10: Silencing TFAP4 inhibits colony formation of MYCN amplified neuroblastoma. (a) MYCN amplified, (b) MYCN nonamplified cells. Neuroblastoma cells were plated in semisolid agar media for 21 days and stained with MTT. Colonies were quantified by using Adobe Photoshop software. Experiments were done in triplicate. Mean + stddev.

Figure 4.11: Silencing of TFAP4 inhibits growth of MYCN amplified tumors. (a-c) Tumor growth measurement of NGP xenografts; (d-f) Tumor growth measurement of CHP212 xenografts. (a, d) Representative images of tumor load quantification via in vivo by bioluminescence imaging. 106 luciferase labeled neuroblastoma cells with dox-inducible shTFAP4 were implanted into the nude mice (n=10 mice per group). Mice were randomized and half were given drinking water with doxycycline. Bioluminescence imaging was taken once a week. (b, e) Luciferase activity of the tumor over time. *P<0.05; **P< 0.01; (c, f) Kaplan-Meier curve of mice with TFAP4 silenced (+Dox) or control (-Dox). Mice were sacrificed when luciferase activity reached 6x109 photons/sec.

Figure 4.12: Tumor weight of xenograft mice and TFAP4 protein level at sacrifice point. (a) Tumor weights do not different between –dox and +dox group in both NGP and CHP212 xenograft mice (n=10). Mean + stddev. (b) Western blot of TFAP4 in xenograft tumors. TFAP4 was knocked down in +dox group at the end point of the experiment.

Figure 4.13: Silencing TFAP4 prolongs survival of mice after tumors established. (a)106 luciferase labeled NGP cells with dox-inducible shTFAP4 were implanted into the nude mice. Bioluminescence imaging was taken once a week. Mice were randomized 24 days after tumor implantation when tumor established. Half mice were given drinking water with doxycycline. (b) Kaplan-Meier curve of mice with TFAP4 knocked down tumor (+Dox) or control (-Dox). Mice were sacked when luciferase activity reached 6x109 photons/sec.

Discussion

MYC and MYCN presents distinct spatiotemporal expression pattern. MYCN is primarily expressed during the early developmental stages with highest expression in pre-B cells, kidney, forebrain, and hindbrain, and its expression is virtually absent in all adult tissues. MYC expression is more generalized, with highest expression found in thymus, liver and spleen during development, and it is continuously expressed in many adult tissues (Zimmerman et al.,

1986). Due to this difference in expression pattern, knockout of either MYC or MYCN results in embryonic lethality at approximately E10.5-E11.5 (Charron et al., 1992; Davis et al., 1993;

Sawai et al., 1993), suggesting MYC and MYCN play different roles during development.

Moreover, there is evidence that specifically during neuronal development, MYCN function cannot be replaced by MYC. In neural stem and progenitor cells, conditional knockout of

MYCN reduced cerebellar granule neural precursors proliferation, despite the upregulation of

MYC. In contrast, knocking out of MYC did not affect neural progenitor cell proliferation

(Hatton et al., 2006). Moreover, the expansion of cerebellar granule neural progenitors driven by sonic hedgehog signaling is associated with MYCN, but not MYC (Kenney et al., 2003).

Mice with conditional deletion of MYCN in neural progenitor cells also showed decreased brain size and a substantial increase in neuronal differentiation (Knoepfler et al., 2002).

Neuroblastoma is a malignancy derived from neural crest progenitor cells (Betters et al., 2010).

High MYCN in neuroblastoma may play a distinct role from MYC in driving tumorigenesis.

Here we show that TFAP4 is a direct target of MYCN and its expression is upregulated in neuroblastoma cell lines and patient tumors with MYCN amplification. MARINa analysis

suggested TFAP4 is a master regulator only in MYCN amplified neuroblastoma, which could be explained by the elevated level of TFAP4 activating oncogenic pathways.

We also validated TFAP4 synthetic lethality in the screen cell line SHEP21N and in two MYCN amplified cell lines compared with two MYCN nonamplified cell lines. Consistent with previously reported literature (Breit and Schwab, 1989; Helland et al., 2011; Rosenbaum et al.,

1989; Westermann et al., 2008), MYC expression is upregulated in MYCN non-amplified neuroblastoma. It is particularly dramatic in CHP212, where C-MYC expression level is ~200 fold higher than that in MYCN amplified neuroblastoma cell lines. However, we demonstrated that silencing TFAP4 only inhibits growth of MYCN amplified neuroblastoma in vitro and in vivo, suggesting TFAP4 is synthetic lethal with MYCN overexpression but not MYC. In line with our results, TFAP4 is not a master regulator in MYCN non-amplified neuroblastoma patients

(Figure 3.7c) even though TFAP4 is also significantly upregulated in Stage 4 MYCN nonamplified neuroblastoma patients comparing to low-risk stage 1 patients (Figure 3.7d).

Previous reports showed that overexpression of TFAP4 cannot drive tumorigenesis (Jackstadt et al., 2013a). However, overexpressing TFAP4 with the RAS oncogene increased the frequency of tumor formation of RAS expressing MEFs to a similar extent as overexpressing C-MYC and

RAS in MEFs (Jackstadt et al., 2013a). Thus it is possible that in MYCN amplified neuroblastoma, TFAP4 has additional oncogenic functions in to drive tumorigenesis.

Chapter 5

TFAP4 inhibits differentiation of MYCN amplified neuroblastoma

Introduction:

TFAP4 belongs to the leucine zipper subgroup of the basic-helix-loop-helix family very closely related to MYC proteins (Figure 6.1). TFAP4 recognized the core DNA sequence CAGCTG

(Hu et al., 1990).

Figure 5.1: Family tree of helix-loop-helix transcription factors. Figure from (Ledent et al., 2002)

Unlike MYC proteins, TFAP4 predominantly form homodimers via several interaction surfaces

(Hu et al., 1990). Thus, the homodimer potentially can be targeted by small molecules. TFAP4 contains several protein-protein-interacting domains, in addition to the bHLH domains. It also has two distinct LR domains, LR1 and LR2, a Q/P rich domain and an acidic domain (Hu et al.,

1990). These domains allow TFAP4 to interact with a diverse range of cofactors to achieve transcriptional regulation.

One class of the cofactors are transcription factors, including the stimulatory protein 1 (Sp1) and the activating protein 1 (AP1). Sp1 and AP1 binding motifs are frequently located proximal to occupied TFAP4 binding motifs (Jackstadt et al., 2013b). Both AP1 and Sp1 function in promoting cell proliferation and cancer progression (Li and Davie, 2010; Lopez-Bergami et al.,

2010). It is possible that TFAP4 interacts with these transcription factors to achieve regulation of common downstream targets in tumorigenesis. Another class of cofactors are the chromatin modifiers, such as histone deacetylase 1 (HDAC1) and histone deacetylase 3 (HDAC3). It has been shown that TFAP4 recruits Geminin and HDAC3 in nonneuronal cells to repress expression of neuronal genes (Kim et al., 2006). The ability to recruit chromatin modifiers indicates TFAP4 can alter global transcriptional regulation via chromatin modification.

Reports have shown that TFAP4 regulates a number of biological processes including cell cycle

(D'Annibale et al., 2014; Jung et al., 2008), senescence (Jackstadt et al., 2013a), immune response (Egawa and Littman, 2011), epithelial to mesenchymal transition (Shi et al., 2014), and maintaining stemness (Jackstadt et al., 2013b; Kim et al., 2006). Many of these processes are related to oncogenesis, thus overexpression of TFAP4 may play an important role in tumorigenesis.

Here we profiled silencing TFAP4 induced genome wide transcriptional change by RNA sequencing of transcripts. We showed that TFAP4 shared significant downstream targets with

the oncogene anaplastic lymphoma kinase (ALK). TFAP4 upregulates genes involved in cell cycle progression and epithelial to mesenchymal transition. We also observed neurite outgrowth and upregulation of neuronal marker after silencing TFAP4, suggesting TFAP4 functions to inhibit neuroblastoma differentiation.

Results:

We sought to determine of TFAP4 was involved in differentiation of tumor cells. After eight days doxycycline treatment, we observed increase neurite outgrowth from both NGP and SK-N-

DZ cells, suggesting that the cells were undergoing neuronal differentiation (Figure 6.2).

Western blots showed that the neuronal marker GAP43 was upregulated when TFAP4 was knocked down, both in MYCN amplified neuroblastoma cells and in NGP xenografts (Figure

6.3a, b). In neuroblastoma patients, expression of GAP43 was also inversely correlated with

TFAP4 expression (Figure 6.3c).

TFAP4 has been reported to repress cell cycle checkpoint gene CDKN1A expression in MCF7 breast cancer cells (Jung et al., 2008). Our cell cycle profiling of neuroblastoma cells lines showed that silencing TFAP4 increased the percentage of cell population in G0/G1 phase by

10% (Figure 6.4). We did not, however, observe a change in CDKN1A mRNA level (Figure

6.5).

To identify genes that are regulated by TFAP4 in MYCN amplified neuroblastoma, we profiled genome wide transcriptional changes by RNAseq 40 hours after shRNA against TFAP4 was induced. We found the expression of 415 genes was significantly upregulated and 457 genes downregulated over this course of time (P<0.01). Most significantly differentially expressed genes are listed in (Figure 6.6, P< 0.001). Several other genes involved in the G1/S checkpoint were differentially expressed. We validated one of the top gene involved G1/S cell cycle progression - CCNE2 (ranked 34th of 14893 genes analyzed in RNA sequencing profiling, P =

3.76e-04). Silencing TFAP4 reduces CCNE2 expression by 40% in both NGP and SK-N-DZ

(Figure 6.7). CCNE2 encodes cyclin E2, which is a CDK2 partner in the late G1 and S phase of mammalian cell cycle and overexpression of cyclin E2 accelerates G1 progression (Lauper et al., 1998).

We validated top genes involved in focal adhesion and axon guidance - PAK4 (ranked 25th, P =

1.36e-04) and ROCK2 (ranked 16th, P = 3.45e-05). Silencing TFAP4 repressed ROCK2 expression by 30% and PAK4 expression by 25%, 48 hours after doxycycline induction of shRNA (Figure

6.8). Both ROCK2 and PAK4 are signaling molecules in cytoskeleton rearrangement and cell migration (Dart and Wells, 2013; Schofield and Bernard, 2013). Their expression is frequently upregulated in metastatic cancers (Kamai et al., 2003; Kesanakurti et al., 2012; Zhou et al.,

2003)

We performed GSEA to compare the TFAP4 downstream transcriptional signature with 85 oncogene signatures in MSigDB. We found that the TFAP4 signature correlates the best with the downstream signature of anaplastic lymphoma tyrosine kinase (ALK) (Wiederschain et al.,

2007) (ranked 1st, P = 4.5e-04) (Figure 6.9a, b). High ALK mRNA expression is associated with a poor prognosis in neuroblastoma (De Brouwer et al., 2010; Passoni et al., 2009; Schulte et al.,

2011) and amplification of ALK and point mutations within the kinase domain are also found in neuroblastoma (Chen et al., 2008; George et al., 2008; Janoueix-Lerosey et al., 2008; Mosse et al., 2008).

Taken together, our results show that TFAP4 regulates a broad range of genes in MYCN amplified neuroblastoma that contribute to cell cycle progression, cell migration and block

neuronal differentiation. The downstream gene signature showed significant similarity to the oncogenic gene ALK signature. TFAP4, thus appears to have an important role in tumorigenesis for MYCN amplified neuroblastoma.

Figure 5.2: Silencing TFAP4 induces neurite outgrowth in MYCN amplified neuroblastoma cell lines. a) Silencing TFAP4 induces neurite outgrowth in MYCN amplified neuroblastoma cell. B) No change in neurite outgrowth in MYCN nonamplified cells. Cells were grown in culture for 8 days, with 1 µg/µl doxycycline (shRNA induced) or without doxycycline.

Figure 5.3: Silencing TFAP4 upregulates the neuronal marker GAP43. a) MYCN amplified NGP cells were cultured with or without doxycycline for 6 days. Protein expression of TFAP4, GAP43 and a loading control β-actin were measured by immunoblotting. b) Protein was extracted from NGP xenograft tumors. Expression of TFAP4, GAP43 and β-actin were analyzed by immunoblotting. C) Box plots of GAP43 gene expression levels in neuroblastoma patients of TFAP4 high to low groups using NCI-TARGET dataset. Groups were stratified into quartiles based on TFAP4 mRNA level. Data were analyzed using two-tailed student’s t-test with P-value indicated.

Figure 5.4: Silencing TFAP4 slows G1/S progression. Bar graph of percentage of cells in G0/G1 phase (purple), G2 phase (red) and S phase (blue) in a population. Cells were collected three days after doxycycline induction. Experiments were performed in triplicate. Graph showed mean percent.

Figure 5.5: Knocking down TFAP4 does not change CDKN1A expression. Gene expression of TFAP4 and CDKN1A were measured by quantitative PCR at different time points after induction of shRNA against TFAP4. Mean + stddev.

Figure 5.6: Differentially expressed genes in MYCN amplified neuroblastoma after silencing TFAP4. Forty hours shRNA induction by doxycycline, total mRNA was collected from NGP and SK-N-DZ and analyzed by RNAseq. mRNA most significantly alters by shTFAP4 induction were shown (P<0.001). Genes activated by TFAP4 were shown on the left panel, repressed by TFAP4 on the right.

Figure 5.7: Silencing TFAP4 represses CCNE2 gene expression. Gene expression was measured by quantitative PCR 48 hours after induction of shRNA against TFAP4. Error bar depicts standard deviation.

Figure 5.8: Silencing TFAP4 represses ROCK2 and PAK4 gene expression. Gene expression was measured by quantitative PCR 48 hours after induction of shRNA against TFAP4. Error bar depicts standard deviation.

Figure 5.9: TFAP4 induces similar signature as ALK. A) TFAP4 showed most significant similarity with ALK signature. Gene set enrichment analysis was used to compare the TFAP4 signature with the 85 oncogene signature in the Broad Institute oncogene database. Gene signature similarity was ranked by false discovery rate. B) Genes in ALK signature were repressed (red) or activated (blue) by silencing ALK are shifting in the same direction as shTFAP4 mediated silencing. Each bar represents a gene in ALK signature. Curve was plotted based on their differential expression in shTFAP4 RNA sequencing data.

Discussion

We demonstrated that in MYCN-amplified neuroblastoma cells, reducing TFAP4 expression induces a differentiated morphology. TFAP4 expression is inversely correlated with the neuronal marker GAP43, both in MYCN amplified cells and in patient tumors. Consistent with this observation, TFAP4 has been reported to inhibit differentiation, as TFAP4 upregulated the expression of two colorectal cancer stem cell markers, CD44 and LGR5 (Jackstadt et al.,

2013b). In the colon, immunohistochemical staining showed that TFAP4 protein expression is restricted to the base of human colonic crypts, an area that is populated by nondifferentiated, stem and progenitor cells. TFAP4 level was found to gradually decline during development, from embryonic to adult brain (Kim et al., 2006), suggesting that TFAP4 represses neuronal differentiation genes. In nonneuronal cells, TFAP4 forms a repressor complex with Geminin and HDAC3 to downregulate the neuronal gene PAHX-AP1 (Kim et al., 2006). TFAP4 also represses the neuronal DBH genes by interaction with GATA-3 and Sp1 (Hong et al., 2008).

Cells undergoing differentiation frequently have a slower cell cycle progression. TFAP4 has shown to repress the cell cycle checkpoint gene CDKN1A in a breast cancer cell line MCF7

(Jung et al., 2008). Our cell cycle profiling results show that silencing TFAP4 inhibited G1 cell cycle progression. CDKN1A mRNA expression, however, did not change upon TFAP4 silencing. One possible reason is that CDKN1A is already directly repressed by MYCN. We performed a RNA sequencing profiling and identified a few potential TFAP4 downstream target involved in G1/S checkpoint progression. We validated that one of the top TFAP4 targets

CCNE2, encoding cyclin E2 protein, is activated by TFAP4. Cyclin E2 partners with CDK2 at the G1/S checkpoint and overexpression accelerates G1 cell cycle progression (Dulic et al.,

1992). CCNE2 has not been reported to be direct MYC/MYCN target and does not contain

MYC/MYCN binding sites in its coding region. Thus TFAP4 may contribute to accelerating cell cycle progression via upregulating CCNE2.

Genes involved in blocking differentiation frequently function in epithelial to mesenchymal transition (EMT). Pathways involved in EMT promotes the loss of cell adhesion and gain migratory abilities (Hanahan and Weinberg, 2011). A genome wide expression study showed that numerous genes implicated in EMT were differentially regulated by TFAP4 (Jackstadt et al., 2013b). Among them, the regulation of the EMT-associated genes CDH1 and SNAIL were further validated in colorectal cancer cell lines DLD-1 and HT29. We showed TFAP4 upregulates the top two genes involved in epithelial to mesenchymal transition, PAK4 and

ROCK2. Both ROCK2 and PAK4 are signaling molecules in cytoskeleton rearrangement and cell migration (Dart and Wells, 2013; Schofield and Bernard, 2013). Their expression is frequently upregulated in metastatic cancers (Kamai et al., 2003; Kesanakurti et al., 2012; Zhou et al., 2003). Moreover, knocking down PAK4 has shown to reduce neuronal progenitor cell proliferation and self-renewal ability (Tian et al., 2011). Small molecule inhibitors for both kinases are available (Murray et al., 2010; Zhang et al., 2012). Although SNAIL and CDH1 were not identified in our RNAseq analysis, it is possible that TFAP4 regulates a different set of

EMT effectors, thus contributing to block MYCN-amplified neuroblastoma differentiation and promote metastasis.

We performed GSEA and found that the downstream signature of TFAP4 most closely resembles the downstream signature of ALK among 85 oncogene signatures in the database.

ALK belongs to the insulin receptor family of trans-membrane receptor tyrosine kinases (Morris et al., 1994). In normal physiological condition, ALK activates various downstream pathways:

JAK/STAT, RAS/MAPK, and PI3K, which in turn promotes cell growth and development of neuronal and central nervous system (Palmer et al., 2009). ALK oncogenic function has been described in lung cancer (Wang et al., 2011), thyroid cancer (Murugan and Xing, 2011), glioblastoma (Grzelinski et al., 2009), rhabdomyosarcoma (van Gaal et al., 2012) and neuroblastoma (Hasan et al., 2013). In neuroblastoma, amplification of wild type ALK has been seen 2-6% of cases and activating mutations within the tyrosine kinase domain is found in 8% of cases (Carpenter and Mosse, 2012; Chen et al., 2008; George et al., 2008; Janoueix-Lerosey et al., 2008; Mosse et al., 2008). Moreover, amplified wild type ALK and the F1174L mutation occur predominantly in MYCN amplified neuroblastoma (De Brouwer et al., 2010). Transgenic mouse models have shown that overexpressing the ALK(F1174L) mutation alone can drive neuroblastoma tumor growth, and ALK(F1174L) and MYCN overexpression work cooperatively to promote neuroblastoma tumorigenesis (Berry et al., 2012; Heukamp et al.,

2012). This cooperative effect is supported by neuroblastoma patient data indicating that the combined occurrence of MYCN amplification and the F1174L mutation is associated with poorer patient survival (De Brouwer et al., 2010).

The precise signaling pathways of ALK in neuroblastoma are unclear. ALK has been reported to be a MYCN target, and overexpression of ALK promotes cell cycle progression and cell migration (Hasan et al., 2013). The top gene repressed in both TFAP4 and ALK signature is

CHAC1, which encodes a protein that promotes neurogenesis by blocking Notch (Chi et al.,

2012). CHAC1 is the 5th most significantly repressed gene by TFAP4 in our RNAseq analysis

(p-value = 2.92e-06). TFAP4/ALK may, therefore, function to inhibit neuroblastoma differentiation.

In summary, our results demonstrate that TFAP4 functions to inhibit neuroblastoma differentiation, promote cell cycle progression and cell migration, in part by regulating genes that are not directly regulated by MYCN. TFAP4 induces a downstream gene signature similar to the downstream signature of ALK. TFAP4 may thus play additional oncogenic roles in

MYCN amplified neuroblastoma.

Chapter 6

Discussion

Although it has now been two decades since the MYCN amplification has been identified as a marker of poor prognosis for neuroblastoma, no therapeutic drug has been developed to target

MYCN directly. Here we combined whole genome shRNA library screening and computational master regulator inference analysis to identify novel drug targets for MYCN amplified neuroblastoma.

Compared to previous MYC/MYCN driven synthetic lethal screens, our approach has several advantages: 1) we used a whole genome shRNA library rather than a subset of shRNAs. This unbiased selection of library allowed us to discover novel synthetic lethal interactors whose function was previous not well studied; 2) we passaged the infected neuroblastoma cells for 10 cell doubling time (~30 days) before analyzing changes of shRNA abundance, enabling us to select for long-term synthetic lethal genes; 3) MARINa analysis is an algorithm that interrogates patient expression database and predicts a short list of master regulators that drives development of a certain disease phenotype. Correlating the functional shRNA screen results and master regulators of MYCN amplification neuroblastoma results allowed us to overcome the limitations of in vitro shRNA screening and gave us the power to produce a short list of synthetic lethal master regulators specific for MYCN amplified neuroblastoma.

By overlapping shRNA library candidates with activated MRs in MYCN amplified neuroblastoma, we were able to identify four “common” MRs: activating enhancer binding protein 4 (TFAP4), mitochondrial recycling factor (MRRF), protein kinase DNA activated catalytic polypeptide (PRKDC), and zinc finger protein (ZNF77). Among them, PRKDC was previously validated as a synthetic lethal gene discovered in a screen with kinase shRNA pool

in MYC-overexpressed cells (Zhou et al., 2014), a finding that supports the robustness of our approach.

We selected TFAP4 as our top candidate to validate because: 1) TFAP4 was the highest ranking

MR in the MARINa analysis (1st in NRC dataset, 6th in TARGET dataset) in MYCN amplified neuroblastoma, but not a MR in MYCN nonamplified neuroblastoma; 2) shRNA against TFAP4 was significantly depleted in MYCN ON population in the screen; 3) homozygous TFAP4-null mice are normal and fertile, suggesting that TFAP4 is not a lethal gene (Egawa and Littman,

2011); 4) stratifying TARGET patient survival data showed that TFAP4 expression is inversely correlated with survival in MYCN high patient group only. Taken together, the data support

TFAP4 as a synthetic lethal gene with MYCN amplification.

TFAP4 has been previously shown to be a direct target of MYC (Jung et al., 2008). Here we demonstrated that TFAP4 is also a direct target of MYCN and its expression is significantly upregulated in stage 4 MYCN amplified neuroblastoma. We further validated synthetic lethality of TFAP4 with MYCN amplification in a range of neuroblastoma cell lines. We showed both in vitro and in xenograft mouse model that silencing TFAP4 only inhibits MYCN amplified neuroblastoma tumor growth, despite the upregulation of MYC.

Even though MYCN and MYC can functionally compensate each other (Malynn et al., 2000), there is evidence that during neuronal development, MYCN function cannot be replaced by

MYC. In neural stem and progenitor cells, conditional knock out of MYCN reduced cerebellar granule neural precursors proliferation, despite the upregulation of MYC, while knocking out of

MYC did not affect neural progenitor cell proliferation (Hatton et al., 2006). The expansion of cerebellar granule neural progenitors driven by sonic hedgehog signaling is associated with

MYCN, but not MYC (Kenney et al., 2003). Neuroblastoma is a malignancy derived from the neural crest progenitor cells, which migrate away from neuronal tube and differentiate into numerous derivatives including neurons and glia of the peripheral nervous system (Betters et al., 2010; Bronner and LaBonne, 2012). Patients with MYCN amplification display an undifferentiated histology (Louis and Shohet, 2015). Since TFAP4 is synthetic lethal with

MYCN despite upregulation of MYC, TFAP4 may play additional oncogenic function specifically associated with MYCN function in neuronal development.

In line with the hypothesis, we observed that in MYCN amplified neuroblastoma cells, knocking down TFAP4 induces increased neurite outgrowth and upregulation of neuronal marker GAP43.

Moreover, TFAP4 has previously been shown to suppress neuronal genes expression (REF).

TFAP4 levels were found to gradually decline during development, from embryonic to adult brain (Kim et al., 2006). In non-neuronal cells, TFAP4 forms a repressor complex with Geminin and HDAC3 to downregulate neuronal gene PAHX-AP1 (Kim et al., 2006). It also represses the neuronal DBH genes by interaction with GATA-3 and Sp1 (Hong et al., 2008).

To understand TFAP4 function in MYCN amplified tumor growth, we performed RNA sequencing and observed the expression of 415 genes was significantly upregulated and 457 genes down-regulated (p-value <0.01). Since cells undergoing differentiation frequently slow cell cycle progression and gain cell adhesion, we validated that TFAP4 activated genes involved in cell cycle progression (CCNE2) and cytoskeleton rearrangement (ROCK2 and PAK4). We

compared the TFAP4 downstream signature with the oncogene signatures in the database from the Broad Institute. The result showed that TFAP4 induces transcriptional changes that are most similar to the oncogene ALK signature. High ALK mRNA expression is associated with a poor prognosis in neuroblastoma (De Brouwer et al., 2010; Passoni et al., 2009; Schulte et al., 2011) and amplification of ALK and point mutations within kinase domain are also found in neuroblastoma (Chen et al., 2008; George et al., 2008; Janoueix-Lerosey et al., 2008; Mosse et al., 2008). The mechanistic relationship between ALK and TFAP4 remains to be determined.

The top gene repressed in both the TFAP4 and ALK signatures was CHAC1, which encodes a protein that promotes neurogenesis by blocking Notch (Chi et al., 2012). This suggests the possibility that one of the functions of both TFAP4 and ALK to inhibit differentiation of neuroblastoma.

We propose a model which posits that in MYCN amplified neuroblastoma, high levels of

MYCN upregulates the master regulator TFAP4, which in turn activates downstream oncogenic signaling pathways to promote cell cycle progression, metastasis and inhibit neuroblastoma differentiation (Figure 6.1). TFAP4 and MYCN may regulate different sets of genes involved in these processes. As a result, TFAP4 may play additional oncogenic roles in MYCN amplified neuroblastoma tumorigenesis. Since TFAP4-null mice are normal and fertile, TFAP4 is not essential for development or normal physiological function. Thus, TFAP4 is an attractive therapeutic target, as it is both synthetically lethal and a master regulator for MYCN amplified neuroblastoma.

Figure 6.1: Schematic diagram of TFAP4 signaling in MYCN amplified neuroblastoma. High level of MYCN in MYCN amplified neuroblastoma upregulates the master regulator TFAP4, which in turn activate downstream oncogenic signaling pathways to promote cell cycle progression (by activating CCNE1 and CCNE2), cell migration (by activating ROCK2 and PAK4) and inhibit neuroblastoma differentiation (by repressing CHAC1).

Although there is no drug currently available to target TFAP4, there are several ways one could potentially inhibit TFAP4 activity. First, TFAP4 predominantly forms homodimers (Hu et al.,

1990). In contrast to MYC/MYCN which consists almost entirely of alpha helices with no obvious surface for small molecule binding, the TFAP4 homodimer interacts with each other at multiple surface areas. Small molecules could be designed that would bind to the interaction surface, thus blocking TFAP4 homodimer formation. Secondly, even though TFAP4 is upregulated in MYCN amplified neuroblastoma, its expression is not as high as MYCN. We could screen for drugs that reduce TFAP4 expression/activity in neuroblastoma.

Therapeutic drugs targeting the ALK oncogene, with different activating mutations, are currently in clinical trials for the treatment of neuroblastoma (Barone et al., 2013). We have shown that TFAP4 induces a similar downstream signature as the ALK signature. Further study is needed to understand the connection between TFAP4 and ALK. Given their common down stream pathways, targeting both TFAP4 and ALK may be synergistic in MYCN amplified neuroblastoma.

REFERENCES

Amati, B., Brooks, M. W., Levy, N., Littlewood, T. D., Evan, G. I., and Land, H. (1993). Oncogenic activity of the c-Myc protein requires dimerization with Max. Cell 72, 233-245.

Anders, S., and Huber, W. (2010). Differential expression analysis for sequence count data. Genome biology 11, R106.

Aytes, A., Mitrofanova, A., Lefebvre, C., Alvarez, M. J., Castillo-Martin, M., Zheng, T., Eastham, J. A., Gopalan, A., Pienta, K. J., Shen, M. M., et al. (2014). Cross-species regulatory network analysis identifies a synergistic interaction between FOXM1 and CENPF that drives prostate cancer malignancy. Cancer cell 25, 638-651.

Barone, G., Anderson, J., Pearson, A. D., Petrie, K., and Chesler, L. (2013). New strategies in neuroblastoma: Therapeutic targeting of MYCN and ALK. Clinical cancer research : an official journal of the American Association for Cancer Research 19, 5814-5821.

Bernards, R. (2014). Finding effective cancer therapies through loss of function genetic screens. Current opinion in genetics & development 24, 23-29.

Berry, T., Luther, W., Bhatnagar, N., Jamin, Y., Poon, E., Sanda, T., Pei, D., Sharma, B., Vetharoy, W. R., Hallsworth, A., et al. (2012). The ALK(F1174L) mutation potentiates the oncogenic activity of MYCN in neuroblastoma. Cancer cell 22, 117-130.

Betters, E., Liu, Y., Kjaeldgaard, A., Sundstrom, E., and Garcia-Castro, M. I. (2010). Analysis of early human neural crest development. Developmental biology 344, 578-592.

Bhadury, J., Nilsson, L. M., Muralidharan, S. V., Green, L. C., Li, Z., Gesner, E. M., Hansen, H. C., Keller, U. B., McLure, K. G., and Nilsson, J. A. (2014). BET and HDAC inhibitors induce similar genes and biological effects and synergize to kill in Myc-induced murine lymphoma. Proceedings of the National Academy of Sciences of the United States of America 111, E2721-2730.

Blackwood, E. M., and Eisenman, R. N. (1991). Max: a helix-loop-helix zipper protein that forms a sequence-specific DNA-binding complex with Myc. Science (New York, NY) 251, 1211-1217.

Blackwood, E. M., Luscher, B., and Eisenman, R. N. (1992). Myc and Max associate in vivo. Genes & development 6, 71-80.

Boxer, L. M., and Dang, C. V. (2001). Translocations involving c-myc and c-myc function. Oncogene 20, 5595-5610.

Breit, S., and Schwab, M. (1989). Suppression of MYC by high expression of NMYC in human neuroblastoma cells. Journal of neuroscience research 24, 21-28.

Brodeur, G. M. (2003). Neuroblastoma: biological insights into a clinical enigma. Nature reviews Cancer 3, 203-216.

Brodeur, G. M., Seeger, R. C., Schwab, M., Varmus, H. E., and Bishop, J. M. (1984). Amplification of N-myc in untreated human neuroblastomas correlates with advanced disease stage. Science (New York, NY) 224, 1121-1124.

Bronner, M. E., and LaBonne, C. (2012). Preface: the neural crest--from stem cell formation to migration and differentiation. Developmental biology 366, 1.

Bryant, H. E., Schultz, N., Thomas, H. D., Parker, K. M., Flower, D., Lopez, E., Kyle, S., Meuth, M., Curtin, N. J., and Helleday, T. (2005). Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP-ribose) polymerase. Nature 434, 913-917.

Cao, J., Tang, M., Li, W. L., Xie, J., Du, H., Tang, W. B., Wang, H., Chen, X. W., Xiao, H., and Li, Y. (2009). Upregulation of activator protein-4 in human colorectal cancer with metastasis. International journal of surgical pathology 17, 16-21.

Carpenter, E. L., and Mosse, Y. P. (2012). Targeting ALK in neuroblastoma--preclinical and clinical advancements. Nature reviews Clinical oncology 9, 391-399.

Carro, M. S., Lim, W. K., Alvarez, M. J., Bollo, R. J., Zhao, X., Snyder, E. Y., Sulman, E. P., Anne, S. L., Doetsch, F., Colman, H., et al. (2010). The transcriptional network for mesenchymal transformation of brain tumours. Nature 463, 318-325.

Charron, J., Malynn, B. A., Fisher, P., Stewart, V., Jeannotte, L., Goff, S. P., Robertson, E. J., and Alt, F. W. (1992). Embryonic lethality in mice homozygous for a targeted disruption of the N-myc gene. Genes & development 6, 2248-2257.

Chen, Y., Takita, J., Choi, Y. L., Kato, M., Ohira, M., Sanada, M., Wang, L., Soda, M., Kikuchi, A., Igarashi, T., et al. (2008). Oncogenic mutations of ALK kinase in neuroblastoma. Nature 455, 971-974.

Cheng, S. W., Davies, K. P., Yung, E., Beltran, R. J., Yu, J., and Kalpana, G. V. (1999). c-MYC interacts with INI1/hSNF5 and requires the SWI/SNF complex for transactivation function. Nature genetics 22, 102-105.

Chi, Z., Zhang, J., Tokunaga, A., Harraz, M. M., Byrne, S. T., Dolinko, A., Xu, J., Blackshaw, S., Gaiano, N., Dawson, T. M., and Dawson, V. L. (2012). Botch promotes neurogenesis by antagonizing Notch. Developmental cell 22, 707-720.

Cohn, S. L., Pearson, A. D., London, W. B., Monclair, T., Ambros, P. F., Brodeur, G. M., Faldum, A., Hero, B., Iehara, T., Machin, D., et al. (2009). The International Neuroblastoma Risk Group (INRG) classification system: an INRG Task Force report. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 27, 289-297.

Cooper, G. M., and Shendure, J. (2011). Needles in stacks of needles: finding disease-causal variants in a wealth of genomic data. Nature reviews Genetics 12, 628-640.

D'Annibale, S., Kim, J., Magliozzi, R., Low, T. Y., Mohammed, S., Heck, A. J., and Guardavaccaro, D. (2014). Proteasome-dependent degradation of transcription factor activating

enhancer-binding protein 4 (TFAP4) controls mitotic division. The Journal of biological chemistry 289, 7730-7737.

Dang, C. V. (1999). c-Myc target genes involved in cell growth, apoptosis, and metabolism. Molecular and cellular biology 19, 1-11.

Dang, C. V. (2012). MYC on the path to cancer. Cell 149, 22-35.

Dart, A. E., and Wells, C. M. (2013). P21-activated kinase 4--not just one of the PAK. European journal of cell biology 92, 129-138.

Davis, A. C., Wims, M., Spotts, G. D., Hann, S. R., and Bradley, A. (1993). A null c-myc mutation causes lethality before 10.5 days of gestation in homozygotes and reduced fertility in heterozygous female mice. Genes & development 7, 671-682.

De Brouwer, S., De Preter, K., Kumps, C., Zabrocki, P., Porcu, M., Westerhout, E. M., Lakeman, A., Vandesompele, J., Hoebeeck, J., Van Maerken, T., et al. (2010). Meta-analysis of neuroblastomas reveals a skewed ALK mutation spectrum in tumors with MYCN amplification. Clinical cancer research : an official journal of the American Association for Cancer Research 16, 4353-4362.

Delmore, J. E., Issa, G. C., Lemieux, M. E., Rahl, P. B., Shi, J., Jacobs, H. M., Kastritis, E., Gilpatrick, T., Paranal, R. M., Qi, J., et al. (2011). BET bromodomain inhibition as a therapeutic strategy to target c-Myc. Cell 146, 904-917.

Dixon, S. J., Costanzo, M., Baryshnikova, A., Andrews, B., and Boone, C. (2009). Systematic mapping of genetic interaction networks. Annual review of genetics 43, 601-625.

DuBois, S. G., Kalika, Y., Lukens, J. N., Brodeur, G. M., Seeger, R. C., Atkinson, J. B., Haase, G. M., Black, C. T., Perez, C., Shimada, H., et al. (1999). Metastatic sites in stage IV and IVS neuroblastoma correlate with age, tumor biology, and survival. Journal of pediatric hematology/oncology 21, 181-189.

Dulic, V., Lees, E., and Reed, S. I. (1992). Association of human cyclin E with a periodic G1-S phase protein kinase. Science (New York, NY) 257, 1958-1961.

Egawa, T., and Littman, D. R. (2011). Transcription factor AP4 modulates reversible and epigenetic silencing of the Cd4 gene. Proceedings of the National Academy of Sciences of the United States of America 108, 14873-14878.

Eilers, M., and Eisenman, R. N. (2008). Myc's broad reach. Genes & development 22, 2755- 2766.

Ein-Dor, L., Kela, I., Getz, G., Givol, D., and Domany, E. (2005). Outcome signature genes in breast cancer: is there a unique set? Bioinformatics (Oxford, England) 21, 171-178.

Escot, C., Theillet, C., Lidereau, R., Spyratos, F., Champeme, M. H., Gest, J., and Callahan, R. (1986). Genetic alteration of the c-myc protooncogene (MYC) in human primary breast

carcinomas. Proceedings of the National Academy of Sciences of the United States of America 83, 4834-4838.

Fang, B. (2014). Development of synthetic lethality anticancer therapeutics. Journal of medicinal chemistry 57, 7859-7873.

Farmer, H., McCabe, N., Lord, C. J., Tutt, A. N., Johnson, D. A., Richardson, T. B., Santarosa, M., Dillon, K. J., Hickson, I., Knights, C., et al. (2005). Targeting the DNA repair defect in BRCA mutant cells as a therapeutic strategy. Nature 434, 917-921.

Felsher, D. W. (2003). Cancer revoked: oncogenes as therapeutic targets. Nature reviews Cancer 3, 375-380.

Fernandez, P. C., Frank, S. R., Wang, L., Schroeder, M., Liu, S., Greene, J., Cocito, A., and Amati, B. (2003). Genomic targets of the human c-Myc protein. Genes & development 17, 1115-1129.

Fong, P. C., Boss, D. S., Yap, T. A., Tutt, A., Wu, P., Mergui-Roelvink, M., Mortimer, P., Swaisland, H., Lau, A., O'Connor, M. J., et al. (2009). Inhibition of poly(ADP-ribose) polymerase in tumors from BRCA mutation carriers. The New England journal of medicine 361, 123-134.

Friess, H., Ding, J., Kleeff, J., Fenkell, L., Rosinski, J. A., Guweidhi, A., Reidhaar-Olson, J. F., Korc, M., Hammer, J., and Buchler, M. W. (2003). Microarray-based identification of differentially expressed growth- and metastasis-associated genes in pancreatic cancer. Cellular and molecular life sciences : CMLS 60, 1180-1199.

George, R. E., Sanda, T., Hanna, M., Frohling, S., Luther, W., 2nd, Zhang, J., Ahn, Y., Zhou, W., London, W. B., McGrady, P., et al. (2008). Activating mutations in ALK provide a therapeutic target in neuroblastoma. Nature 455, 975-978.

Grzelinski, M., Steinberg, F., Martens, T., Czubayko, F., Lamszus, K., and Aigner, A. (2009). Enhanced antitumorigenic effects in glioblastoma on double targeting of pleiotrophin and its receptor ALK. Neoplasia (New York, NY) 11, 145-156.

Guccione, E., Martinato, F., Finocchiaro, G., Luzi, L., Tizzoni, L., Dall' Olio, V., Zardo, G., Nervi, C., Bernard, L., and Amati, B. (2006). Myc-binding-site recognition in the human genome is determined by chromatin context. Nature cell biology 8, 764-770.

Hanahan, D., and Weinberg, R. A. (2011). Hallmarks of cancer: the next generation. Cell 144, 646-674.

Hann, s. R. (2014). MYC Cofactors: molecular switches controlling diverse biological outcomes. Cold Spring Harb Perspect Med 4.

Hasan, M. K., Nafady, A., Takatori, A., Kishida, S., Ohira, M., Suenaga, Y., Hossain, S., Akter, J., Ogura, A., Nakamura, Y., et al. (2013). ALK is a MYCN target gene and regulates cell migration and invasion in neuroblastoma. Scientific reports 3, 3450.

Hatton, B. A., Knoepfler, P. S., Kenney, A. M., Rowitch, D. H., de Alboran, I. M., Olson, J. M., and Eisenman, R. N. (2006). N-myc is an essential downstream effector of Shh signaling during both normal and neoplastic cerebellar growth. Cancer research 66, 8655-8661.

Helland, A., Anglesio, M. S., George, J., Cowin, P. A., Johnstone, C. N., House, C. M., Sheppard, K. E., Etemadmoghadam, D., Melnyk, N., Rustgi, A. K., et al. (2011). Deregulation of MYCN, LIN28B and LET7 in a molecular subtype of aggressive high-grade serous ovarian cancers. PloS one 6, e18064.

Heukamp, L. C., Thor, T., Schramm, A., De Preter, K., Kumps, C., De Wilde, B., Odersky, A., Peifer, M., Lindner, S., Spruessel, A., et al. (2012). Targeted expression of mutated ALK induces neuroblastoma in transgenic mice. Science translational medicine 4, 141ra191.

Hindorff, L. A., MacArthur, J., Morales, J., Junkins, H.A., Hall, P.N., Klemm, A.K., and Manolio, T.A. A Catalog of Published Genome-Wide Association Studies. http://wwwgenomegov/gwastudies.

Hong, S. J., Choi, H. J., Hong, S., Huh, Y., Chae, H., and Kim, K. S. (2008). Transcription factor GATA-3 regulates the transcriptional activity of dopamine beta-hydroxylase by interacting with Sp1 and AP4. Neurochemical research 33, 1821-1831.

Hu, B. S., Zhao, G., Yu, H. F., Chen, K., Dong, J. H., and Tan, J. W. (2013). High expression of AP-4 predicts poor prognosis for hepatocellular carcinoma after curative hepatectomy. Tumour biology : the journal of the International Society for Oncodevelopmental Biology and Medicine 34, 271-276.

Hu, Y. F., Luscher, B., Admon, A., Mermod, N., and Tjian, R. (1990). Transcription factor AP- 4 contains multiple dimerization domains that regulate dimer specificity. Genes & development 4, 1741-1752.

Huang, J., Frischer, J. S., Serur, A., Kadenhe, A., Yokoi, A., McCrudden, K. W., New, T., O'Toole, K., Zabski, S., Rudge, J. S., et al. (2003). Regression of established tumors and metastases by potent vascular endothelial growth factor blockade. Proceedings of the National Academy of Sciences of the United States of America 100, 7785-7790.

Huang, M., and Weiss, W. A. (2013). Neuroblastoma and MYCN. Cold Spring Harb Perspect Med 3, a014415.

Jackstadt, R., Jung, P., and Hermeking, H. (2013a). AP4 directly downregulates p16 and p21 to suppress senescence and mediate transformation. Cell death & disease 4, e775.

Jackstadt, R., Roh, S., Neumann, J., Jung, P., Hoffmann, R., Horst, D., Berens, C., Bornkamm, G. W., Kirchner, T., Menssen, A., and Hermeking, H. (2013b). AP4 is a mediator of epithelial- mesenchymal transition and metastasis in colorectal cancer. The Journal of experimental medicine 210, 1331-1350.

Janoueix-Lerosey, I., Lequin, D., Brugieres, L., Ribeiro, A., de Pontual, L., Combaret, V., Raynal, V., Puisieux, A., Schleiermacher, G., Pierron, G., et al. (2008). Somatic and germline activating mutations of the ALK kinase receptor in neuroblastoma. Nature 455, 967-970.

Jung, P., Menssen, A., Mayr, D., and Hermeking, H. (2008). AP4 encodes a c-MYC-inducible repressor of p21. Proceedings of the National Academy of Sciences of the United States of America 105, 15046-15051.

Kamai, T., Tsujii, T., Arai, K., Takagi, K., Asami, H., Ito, Y., and Oshima, H. (2003). Significant association of Rho/ROCK pathway with invasion and metastasis of bladder cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 9, 2632-2641.

Kawate, S., Fukusato, T., Ohwada, S., Watanuki, A., and Morishita, Y. (1999). Amplification of c-myc in hepatocellular carcinoma: correlation with clinicopathologic features, proliferative activity and p53 overexpression. Oncology 57, 157-163.

Kenney, A. M., Cole, M. D., and Rowitch, D. H. (2003). Nmyc upregulation by sonic hedgehog signaling promotes proliferation in developing cerebellar granule neuron precursors. Development 130, 15-28.

Kesanakurti, D., Chetty, C., Rajasekhar Maddirela, D., Gujrati, M., and Rao, J. S. (2012). Functional cooperativity by direct interaction between PAK4 and MMP-2 in the regulation of anoikis resistance, migration and invasion in glioma. Cell death & disease 3, e445.

Keshelava, N., Frgala, T., Krejsa, J., Kalous, O., and Reynolds, C. P. (2005). DIMSCAN: a microcomputer fluorescence-based cytotoxicity assay for preclinical testing of combination chemotherapy. Methods in molecular medicine 110, 139-153.

Kessler, J. D., Kahle, K. T., Sun, T., Meerbrey, K. L., Schlabach, M. R., Schmitt, E. M., Skinner, S. O., Xu, Q., Li, M. Z., Hartman, Z. C., et al. (2012). A SUMOylation-dependent transcriptional subprogram is required for Myc-driven tumorigenesis. Science (New York, NY) 335, 348-353.

Kim, M. Y., Jeong, B. C., Lee, J. H., Kee, H. J., Kook, H., Kim, N. S., Kim, Y. H., Kim, J. K., Ahn, K. Y., and Kim, K. K. (2006). A repressor complex, AP4 transcription factor and geminin, negatively regulates expression of target genes in nonneuronal cells. Proceedings of the National Academy of Sciences of the United States of America 103, 13074-13079.

Knoepfler, P. S., Cheng, P. F., and Eisenman, R. N. (2002). N-myc is essential during neurogenesis for the rapid expansion of progenitor cell populations and the inhibition of neuronal differentiation. Genes & development 16, 2699-2712.

Kohl, N. E., Kanda, N., Schreck, R. R., Bruns, G., Latt, S. A., Gilbert, F., and Alt, F. W. (1983). Transposition and amplification of oncogene-related sequences in human neuroblastomas. Cell 35, 359-367.

Kohl, N. E., Legouy, E., DePinho, R. A., Nisen, P. D., Smith, R. K., Gee, C. E., and Alt, F. W. (1986). Human N-myc is closely related in organization and nucleotide sequence to c-myc. Nature 319, 73-77.

Kretzner, L., Blackwood, E. M., and Eisenman, R. N. (1992). Myc and Max proteins possess distinct transcriptional activities. Nature 359, 426-429.

Kruglyak, L. (1999). Prospects for whole-genome linkage disequilibrium mapping of common disease genes. Nature genetics 22, 139-144.

Ladanyi, M., Park, C. K., Lewis, R., Jhanwar, S. C., Healey, J. H., and Huvos, A. G. (1993). Sporadic amplification of the MYC gene in human osteosarcomas. Diagnostic molecular pathology : the American journal of surgical pathology, part B 2, 163-167.

Lauper, N., Beck, A. R., Cariou, S., Richman, L., Hofmann, K., Reith, W., Slingerland, J. M., and Amati, B. (1998). Cyclin E2: a novel CDK2 partner in the late G1 and S phases of the mammalian cell cycle. Oncogene 17, 2637-2643.

Ledent, V., Paquet, O., and Vervoort, M. (2002). Phylogenetic analysis of the human basic helix-loop-helix proteins. Genome biology 3, Research0030.

Lee, Y., Ahn, C., Han, J., Choi, H., Kim, J., Yim, J., Lee, J., Provost, P., Radmark, O., Kim, S., and Kim, V. N. (2003). The nuclear RNase III Drosha initiates microRNA processing. Nature 425, 415-419.

Lee, Y., Kim, M., Han, J., Yeom, K. H., Lee, S., Baek, S. H., and Kim, V. N. (2004). MicroRNA genes are transcribed by RNA polymerase II. The EMBO journal 23, 4051-4060.

Lefebvre, C., Rajbhandari, P., Alvarez, M. J., Bandaru, P., Lim, W. K., Sato, M., Wang, K., Sumazin, P., Kustagi, M., Bisikirska, B. C., et al. (2010). A human B-cell interactome identifies MYB and FOXM1 as master regulators of proliferation in germinal centers. Molecular systems biology 6, 377.

Li, L., and Davie, J. R. (2010). The role of Sp1 and Sp3 in normal and cancer cell biology. Annals of anatomy = Anatomischer Anzeiger : official organ of the Anatomische Gesellschaft 192, 275-283.

Lim, W. K., Lyashenko, E., and Califano, A. (2009). Master regulators used as breast cancer metastasis classifier. Pacific Symposium on Biocomputing Pacific Symposium on Biocomputing, 504-515.

Liu, L., Ulbrich, J., Muller, J., Wustefeld, T., Aeberhard, L., Kress, T. R., Muthalagu, N., Rycak, L., Rudalska, R., Moll, R., et al. (2012). Deregulated MYC expression induces dependence upon AMPK-related kinase 5. Nature 483, 608-612.

Lopez-Bergami, P., Lau, E., and Ronai, Z. (2010). Emerging roles of ATF2 and the dynamic AP1 network in cancer. Nature reviews Cancer 10, 65-76.

Louis, C. U., and Shohet, J. M. (2015). Neuroblastoma: molecular pathogenesis and therapy. Annual review of medicine 66, 49-63.

Lutz, W., and Schwab, M. (1997). In vivo regulation of single copy and amplified N-myc in human neuroblastoma cells. Oncogene 15, 303-315.

Malynn, B. A., de Alboran, I. M., O'Hagan, R. C., Bronson, R., Davidson, L., DePinho, R. A., and Alt, F. W. (2000). N-myc can functionally replace c-myc in murine development, cellular growth, and differentiation. Genes & development 14, 1390-1399.

Margolin, A. A., Nemenman, I., Basso, K., Wiggins, C., Stolovitzky, G., Dalla Favera, R., and Califano, A. (2006). ARACNE: an algorithm for the reconstruction of gene regulatory networks in a mammalian cellular context. BMC bioinformatics 7 Suppl 1, S7.

Maris, J. M., Hogarty, M. D., Bagatell, R., and Cohn, S. L. (2007). Neuroblastoma. Lancet 369, 2106-2120.

Matthay, K. K., George, R. E., and Yu, A. L. (2012). Promising therapeutic targets in neuroblastoma. Clinical cancer research : an official journal of the American Association for Cancer Research 18, 2740-2753.

Meyer, N., and Penn, L. Z. (2008). Reflecting on 25 years with MYC. Nature reviews Cancer 8, 976-990.

Michiels, S., Koscielny, S., and Hill, C. (2005). Prediction of cancer outcome with microarrays: a multiple random validation strategy. Lancet 365, 488-492.

Mootha, V. K., Lindgren, C. M., Eriksson, K. F., Subramanian, A., Sihag, S., Lehar, J., Puigserver, P., Carlsson, E., Ridderstrale, M., Laurila, E., et al. (2003). PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nature genetics 34, 267-273.

Morris, S. W., Kirstein, M. N., Valentine, M. B., Dittmer, K. G., Shapiro, D. N., Saltman, D. L., and Look, A. T. (1994). Fusion of a kinase gene, ALK, to a nucleolar protein gene, NPM, in non-Hodgkin's lymphoma. Science (New York, NY) 263, 1281-1284.

Mosse, Y. P., Laudenslager, M., Longo, L., Cole, K. A., Wood, A., Attiyeh, E. F., Laquaglia, M. J., Sennett, R., Lynch, J. E., Perri, P., et al. (2008). Identification of ALK as a major familial neuroblastoma predisposition gene. Nature 455, 930-935.

Murray, B. W., Guo, C., Piraino, J., Westwick, J. K., Zhang, C., Lamerdin, J., Dagostino, E., Knighton, D., Loi, C. M., Zager, M., et al. (2010). Small-molecule p21-activated kinase inhibitor PF-3758309 is a potent inhibitor of oncogenic signaling and tumor growth. Proceedings of the National Academy of Sciences of the United States of America 107, 9446- 9451.

Murugan, A. K., and Xing, M. (2011). Anaplastic thyroid cancers harbor novel oncogenic mutations of the ALK gene. Cancer research 71, 4403-4411.

Nair, S. K., and Burley, S. K. (2003). X-ray structures of Myc-Max and Mad-Max recognizing DNA. Molecular bases of regulation by proto-oncogenic transcription factors. Cell 112, 193- 205.

Otto, T., Horn, S., Brockmann, M., Eilers, U., Schuttrumpf, L., Popov, N., Kenney, A. M., Schulte, J. H., Beijersbergen, R., Christiansen, H., et al. (2009). Stabilization of N-Myc is a critical function of Aurora A in human neuroblastoma. Cancer cell 15, 67-78.

Paddison, P. J., Silva, J. M., Conklin, D. S., Schlabach, M., Li, M., Aruleba, S., Balija, V., O'Shaughnessy, A., Gnoj, L., Scobie, K., et al. (2004). A resource for large-scale RNA- interference-based screens in mammals. Nature 428, 427-431.

Palmer, R. H., Vernersson, E., Grabbe, C., and Hallberg, B. (2009). Anaplastic lymphoma kinase: signalling in development and disease. The Biochemical journal 420, 345-361.

Passoni, L., Longo, L., Collini, P., Coluccia, A. M., Bozzi, F., Podda, M., Gregorio, A., Gambini, C., Garaventa, A., Pistoia, V., et al. (2009). Mutation-independent anaplastic lymphoma kinase overexpression in poor prognosis neuroblastoma patients. Cancer research 69, 7338-7346.

Rahman, S., Sowa, M. E., Ottinger, M., Smith, J. A., Shi, Y., Harper, J. W., and Howley, P. M. (2011). The Brd4 extraterminal domain confers transcription activation independent of pTEFb by recruiting multiple proteins, including NSD3. Molecular and cellular biology 31, 2641-2652.

Ritchie, M. E., Phipson, B., Wu, D., Hu, Y., Law, C. W., Shi, W., and Smyth, G. K. (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic acids research.

Rosenbaum, H., Webb, E., Adams, J. M., Cory, S., and Harris, A. W. (1989). N-myc transgene promotes B lymphoid proliferation, elicits lymphomas and reveals cross-regulation with c-myc. The EMBO journal 8, 749-755.

Salmon, P., Oberholzer, J., Occhiodoro, T., Morel, P., Lou, J., and Trono, D. (2000). Reversible immortalization of human primary cells by lentivector-mediated transfer of specific genes. Molecular therapy : the journal of the American Society of Gene Therapy 2, 404-414.

Sauka-Spengler, T., and Bronner, M. (2010). Snapshot: neural crest. Cell 143, 486-486.e481.

Sawai, S., Shimono, A., Wakamatsu, Y., Palmes, C., Hanaoka, K., and Kondoh, H. (1993). Defects of embryonic organogenesis resulting from targeted disruption of the N-myc gene in the mouse. Development 117, 1445-1455.

Schofield, A. V., and Bernard, O. (2013). Rho-associated coiled-coil kinase (ROCK) signaling and disease. Critical reviews in biochemistry and molecular biology 48, 301-316.

Schulte, J. H., Bachmann, H. S., Brockmeyer, B., Depreter, K., Oberthur, A., Ackermann, S., Kahlert, Y., Pajtler, K., Theissen, J., Westermann, F., et al. (2011). High ALK receptor tyrosine kinase expression supersedes ALK mutation as a determining factor of an unfavorable

phenotype in primary neuroblastoma. Clinical cancer research : an official journal of the American Association for Cancer Research 17, 5082-5092.

Schwab, M., Alitalo, K., Klempnauer, K. H., Varmus, H. E., Bishop, J. M., Gilbert, F., Brodeur, G., Goldstein, M., and Trent, J. (1983). Amplified DNA with limited homology to myc cellular oncogene is shared by human neuroblastoma cell lines and a neuroblastoma tumour. Nature 305, 245-248.

Shi, J., and Vakoc, C. R. (2014). The mechanisms behind the therapeutic activity of BET bromodomain inhibition. Molecular cell 54, 728-736.

Shi, L., Jackstadt, R., Siemens, H., Li, H., Kirchner, T., and Hermeking, H. (2014). p53-induced miR-15a/16-1 and AP4 form a double-negative feedback loop to regulate epithelial- mesenchymal transition and metastasis in colorectal cancer. Cancer research 74, 532-542.

Silva, J. M., Li, M. Z., Chang, K., Ge, W., Golding, M. C., Rickles, R. J., Siolas, D., Hu, G., Paddison, P. J., Schlabach, M. R., et al. (2005). Second-generation shRNA libraries covering the mouse and human genomes. Nature genetics 37, 1281-1288.

Stanton, L. W., Schwab, M., and Bishop, J. M. (1986). Nucleotide sequence of the human N- myc gene. Proceedings of the National Academy of Sciences of the United States of America 83, 1772-1776.

Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette, M. A., Paulovich, A., Pomeroy, S. L., Golub, T. R., Lander, E. S., and Mesirov, J. P. (2005). Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America 102, 15545-15550.

Tian, Y., Lei, L., and Minden, A. (2011). A key role for Pak4 in proliferation and differentiation of neural progenitor cells. Developmental biology 353, 206-216.

Toyoshima, M., Howie, H. L., Imakura, M., Walsh, R. M., Annis, J. E., Chang, A. N., Frazier, J., Chau, B. N., Loboda, A., Linsley, P. S., et al. (2012). Functional genomics identifies therapeutic targets for MYC-driven cancer. Proceedings of the National Academy of Sciences of the United States of America 109, 9545-9550. van Gaal, J. C., Flucke, U. E., Roeffen, M. H., de Bont, E. S., Sleijfer, S., Mavinkurve- Groothuis, A. M., Suurmeijer, A. J., van der Graaf, W. T., and Versleijen-Jonkers, Y. M. (2012). Anaplastic lymphoma kinase aberrations in rhabdomyosarcoma: clinical and prognostic implications. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 30, 308-315. van Riggelen, J., Yetil, A., and Felsher, D. W. (2010). MYC as a regulator of ribosome biogenesis and protein synthesis. Nature reviews Cancer 10, 301-309. von der Lehr, N., Johansson, S., Wu, S., Bahram, F., Castell, A., Cetinkaya, C., Hydbring, P., Weidung, I., Nakayama, K., Nakayama, K. I., et al. (2003). The F-box protein Skp2 participates 97

in c-Myc proteosomal degradation and acts as a cofactor for c-Myc-regulated transcription. Molecular cell 11, 1189-1200.

Wang, K., Saito, M., Bisikirska, B. C., Alvarez, M. J., Lim, W. K., Rajbhandari, P., Shen, Q., Nemenman, I., Basso, K., Margolin, A. A., et al. (2009). Genome-wide identification of post- translational modulators of transcription factor activity in human B cells. Nature biotechnology 27, 829-839.

Wang, Y. W., Tu, P. H., Lin, K. T., Lin, S. C., Ko, J. Y., and Jou, Y. S. (2011). Identification of oncogenic point mutations and hyperphosphorylation of anaplastic lymphoma kinase in lung cancer. Neoplasia (New York, NY) 13, 704-715.

Weiss, W. A., Aldape, K., Mohapatra, G., Feuerstein, B. G., and Bishop, J. M. (1997). Targeted expression of MYCN causes neuroblastoma in transgenic mice. The EMBO journal 16, 2985- 2995.

Westermann, F., Muth, D., Benner, A., Bauer, T., Henrich, K. O., Oberthuer, A., Brors, B., Beissbarth, T., Vandesompele, J., Pattyn, F., et al. (2008). Distinct transcriptional MYCN/c- MYC activities are associated with spontaneous regression or malignant progression in neuroblastomas. Genome biology 9, R150.

Wiederschain, D., Chen, L., Johnson, B., Bettano, K., Jackson, D., Taraszka, J., Wang, Y. K., Jones, M. D., Morrissey, M., Deeds, J., et al. (2007). Contribution of polycomb homologues Bmi-1 and Mel-18 to medulloblastoma pathogenesis. Molecular and cellular biology 27, 4968- 4979.

Wyce, A., Ganji, G., Smitheman, K. N., Chung, C. W., Korenchuk, S., Bai, Y., Barbash, O., Le, B., Craggs, P. D., McCabe, M. T., et al. (2013). BET inhibition silences expression of MYCN and BCL2 and induces cytotoxicity in neuroblastoma tumor models. PloS one 8, e72967.

Xinghua, L., Bo, Z., Yan, G., Lei, W., Changyao, W., Qi, L., Lin, Y., Kaixiong, T., Guobin, W., and Jianying, C. (2012). The overexpression of AP-4 as a prognostic indicator for gastric carcinoma. Medical oncology (Northwood, London, England) 29, 871-877.

Yu, J., Putcha, P., Califano, A., and Silva, J. M. (2013). Pooled shRNA screenings: computational analysis. Methods in molecular biology (Clifton, NJ) 980, 371-384.

Zhang, J., Wang, J., Guo, Q., Wang, Y., Zhou, Y., Peng, H., Cheng, M., Zhao, D., and Li, F. (2012). LCH-7749944, a novel and potent p21-activated kinase 4 inhibitor, suppresses proliferation and invasion in human gastric cancer cells. Cancer letters 317, 24-32.

Zhou, J., Zhao, L. Q., Xiong, M. M., Wang, X. Q., Yang, G. R., Qiu, Z. L., Wu, M., and Liu, Z. H. (2003). Gene expression profiles at different stages of human esophageal squamous cell carcinoma. World journal of gastroenterology : WJG 9, 9-15.

Zhou, Z., Patel, M., Ng, N., Hsieh, M. H., Orth, A. P., Walker, J. R., Batalov, S., Harris, J. L., and Liu, J. (2014). Identification of synthetic lethality of PRKDC in MYC-dependent human cancers by pooled shRNA screening. BMC cancer 14, 944. 98

Zimmerman, K. A., Yancopoulos, G. D., Collum, R. G., Smith, R. K., Kohl, N. E., Denis, K. A., Nau, M. M., Witte, O. N., Toran-Allerand, D., Gee, C. E., and et al. (1986). Differential expression of myc family genes during murine development. Nature 319, 780-783.