Identification of REST Regulated Genes in Prostate Cancer Via High-Throughput Technologies

Identification of REST regulated genes in prostate cancer via high-throughput technologies

Yue Sun

Thesis to obtain the Master of Science Degree in Biotechnology

Examination Committee

Chairperson: Professor Arsénio do Carmo Sales Mendes Fialho Departamento de Bioengenharia (DBE) Supervisor: Professor Isabel Maria de Sá Correia Leite de Almeida (DBE) Co-supervisor: Professor Colin Collins (University of British Columbia/Vancouver Prostate Centre) Member of the Committee: Professor Miguel Nobre Parreira Cacho Teixeira (DBE)

December 2012

Acknowledgements

The experimental work reported in this thesis was carried out at the Laboratory for Advanced Genome Analysis (LAGA), Vancouver Prostate Centre (VPC), through he Erasmus Mundus Master's Program in Systems Biology (euSYSBIO).

I would like to express my gratitude to my supervisors Prof. Isabel Sá Correia (Instituto Superior Técnico IST) and Prof. Erik Aurell (Royal Institute of Technology KTH), for giving me the opportunity to make this thesis in an external institute.

I am very grateful to my supervisor Prof. Colin Collins (University of British Columbia/VPC), who supervised me and gave me good suggestions on this thesis work. I would also like to thank the staff VPC for their kind help through my work.

Within the two years study, the Erasmus Mundus euSYSBIO program navigated me through a cruise of the adventures of the Great Voyage, from the heavy snow in Sweden to the beach and sunshine of Portugal. I believe, being a student in this program will be a treasured experience for the rest of my life.

Dedication

For my parents and 天海祐希.

Abstract

Prostate cancer (PCa) is one of the most commonly diagnosed cancers in the world: males have 17% risk of developing PCa during lifetime. Androgen deprivation therapy (ADT) has been the leading form of PCa therapy [1]. Unfortunately, the effectiveness of this treatment is usually temporary, with a subpopulation of cells surviving. Growing evidence indicates that neuroendocrine differentiation (NED) occurs in 30-100% of disease recurrence after ADT [2, 3]. However, the mechanism is still poorly understood.

Repressor element-1 silencing transcription factor (REST) is responsible for restricting neuronal gene expression to the nervous system. REST was recently shown to be absent in various types of cancer. Blockade of REST function in human LNCaP cells is found to cause NED [4], indicating REST may be one of important determinants of neuroendocrine prostate cancer (NEPC). The aim of this study is to investigate the potential role of REST in PCa progression. We combined state-of-the-art sequencing technologies and bioinformatics with patient-derived next-generation models of PCa to deal with this problem. An integrated data analysis of RNA-seq on mouse xenograft model and expression microarray with RNA interference on human LNCaP cells was performed to identify potential REST targets. Experimentally, we validated most likely targets by QRT-PCR of LNCaP mRNA. In addition, we also estimated the clinical significance of the targets using public datasets. Our results help to elucidate the mechanism that drives development of androgen independent and aggressive NEPC after castration. Furthermore, it could provide a new prognosis factor for PCa and also provide a new target for the PCa treatment.

Key words: REST, prostate cancer, neuroendocrine differentiation, gene expression

Resumo O cancro da próstata (CaP) é um dos cancros mais diagnosticados em todo o mundo: Cerca de 17% dos homens correm o risco de desenvolver CaP durante a vida. A terapia de privação de andrógenios (ADT) tem sido a principal forma de tratamento do CaP [1]. Infelizmente, a eficácia deste tratamento é normalmente temporária, com uma sub-população de células sobreviventes. Várias evidências indicam que a diferenciação neuroendócrina (NED) ocorre em 30-100% das recorrências da doença após a ADT [2, 3]. No entanto, o mecanismo ainda é pouco compreendido.

O “Repressor element-1 silencing transcription factor ” (REST) é responsável por restringir a expressão do gene neuronal no sistema nervoso. Evidências recentes mostraram que o REST estava ausente em vários tipos de cancro. O bloqueio da função do REST em células humanas LNCaP causa NED [4], o que indica que o REST pode ser um importante determinantedo cancro da próstata neuroendócrino (NEPC). O objetivo deste estudo é investigar o potencial papel do REST na progressão do CaP, através da combinação de tecnologia de sequenciação e bioinformática com modelos baseados em pacientes de CaP. Para tal, foi realizada uma análise integrada de dados de RNA-seq em ratos modelo xenoenxerto e de expressão com RNA de interferência em células humanas LNCaP com o objectivo de identificar potenciais alvos de REST. Foram validados experimentalmente osalvos mais prováveis de QRT-PCR de mRNA de LNCaP. Além disso, também foi avaliada a importância clínica dos alvos utilizando conjuntos de dados públicos. Os nossos resultados ajudam a elucidar o mecanismo que conduz ao desenvolvimento das formas de NEPC androgénio-independente e agressiva após a castração. Além disso, pode fornecer um novo fator prognóstico para o CaP, bem comor um novo alvo para o tratamento do CaP. Palavras-chave: REST, câncer de próstata, diferenciação neuroendócrina, expressão gênica

Table of contents

Table of contents ...... I

List of tables ...... III

List of figures ...... IV

List of abbreviations ...... VI

1. Prostate cancer ...... 1

1.1 The prostate and the development of prostate cancer ...... 1

1.1.1 Gleason score ...... 2

1.1.2 Prostate specific antigen (PSA) ...... 4

1.2 Androgen receptor (AR) ...... 6

1.3 Castrate-resistance prostate cancer (CRPC) ...... 6

1.4 Neuroendocrine prostate cancer (NEPC) ...... 7

2 The transcription factor REST ...... 10

2.1 REST and its alternative isoforms ...... 10

2.2 RE1 motif ...... 12

2.3 REST Co-repressors ...... 14

2.4 REST and non-coding RNAs ...... 18

2.5 REST function and cancer ...... 19

3 Aims of research ...... 22

4 Materials and methods ...... 25

4.1 Cell lines ...... 25

4.2 Prostate tissue samples ...... 25

4.3 Mouse xenograft model ...... 25

4.4 Public dataset ...... 26

4.5 siRNA and shRNA experiments ...... 26

4.6 Microarray ...... 27

4.7 Transcriptome sequencing ...... 27

4.8 Statistics Analysis ...... 28

4.9 Data analysis ...... 28

4.10 Experimental RT-PCR validation ...... 28

5 Results ...... 30

5.1 Specific and efficient knock-down of REST mRNA in LNCaP cells ...... 30

5.2 High fidelity of mouse xenograft model for neuroendocrine

transdifferentiation ...... 32

5.3 Microarray and RNA-seq analysis for identification of differential

expressed genes ...... 33

5.4 Functional categorization analysis ...... 37

5.5 Network analysis ...... 38

5.6 Experimental RT-PCR validation ...... 43

5.7 Kaplan Meier survival analysis ...... 44

6 Discussion ...... 47

References ...... 50

List of tables

Table 2.1 Molecular weight of REST alternative isoforms (Uniprot.org) ...... 12

Table 4.1 RT-PCR primer sequence ...... 29

Table 5.1 Profiling of overlap genes differentially expressed ...... 35

Table 5.2 IPA-generated molecular networks ...... 39

III

List of figures

Figure 1.1 Human prostate gland...... 1

Figure 1.2 Gleason scale...... 3

Figure 1.3 Functions of PSA in the normal prostate and in carcinogenesis...... 5

Figure 1.4 Neuroendocrine cells in the development of prostate cancer...... 8

Figure 2.1 REST Gene in genomic location: bands according to Ensembl...... 10

Figure 2.2 Model of REST domains...... 11

Figure 2.3 REST mRNA and alternative splicing variants...... 11

Figure 2.4 Sequence logo of canonical RE1 motif...... 13

Figure 2.5 Histone modification on H3 N-terminus...... 15

Figure 2.6 Mechanism of repression/activation via REST-RE1 interaction...... 16

Figure 2.7 REST-ncRNA interactions in neural networks...... 18 Figure 2.8 Experimental models and possible routes for REST dysfunction in

cancers...... 20

Figure 5.1 siRNA knockdown of REST in LNCaP cells...... 30

Figure 5.2 shRNA knockdown of REST in LNCaP cells...... 31 Figure 5.3 The LTL-331 xenograft model of neuroendocrine transdifferentiation.

...... 32

Figure 5.4 Summary of microarray data on LNCaP model...... 33

Figure 5.5 Summary of RNA-seq data on xenograft model...... 34 Figure 5.6 The overlap between genes differentially expressed from microarray

data and from RNA-seq data...... 35 Figure 5.7 Functional categorization analysis of most significant pathways and

diseases...... 37

Figure 5.8 Ingenuity pathway analysis Network 1...... 41

Figure 5.9 Combination of Network 1, Network 2 and Network 3...... 42

Figure 5.10 BEX1 expression level in REST shRNA knockdown LNCaP cells. 43 Figure 5.11 SRRM3 expression level in REST shRNA knockdown LNCaP cells.

...... 44 Figure 5.12 Kaplan–Meier survival plots on 3 selective genes in a public dataset

prostate tumors (p<0.01)...... 46

List of abbreviations

ADT: androgen deprivation therapy AR: androgen receptor CNS: central nervous system CRPC: castrate-resistance prostate cancer DRE: digital rectal examination FC: fold-change HAT: histone acetyltransferase HDAC: histone deacetylase H3K4: histone H3 lysine 4 LNCaP: lymph node carcinoma of prostate LSD1: lysine-specific demethylase 1 miRNA: microRNA ncRNA: non-coding RNA NEPC: neuroendocrine prostate cancer NOC/SCID: nonobese diabetic/severe combined immunodeficiency NRSF: neuron-restricted silencing factor PCa: prostate cancer PSA: prostate-specific antigen RD: repression domains RE1: repressor element 1 REST: repressor element 1- silencing transcription factor RIN: RNA Integrity Number RISC: RNA induced silencing complex shRNA: short hairpin RNA siRNA: small interfering RNA

1. Prostate cancer

Prostate cancer (PCa) is the most common non-skin cancer in North American men and a leading cause of cancer death (http://www.pcf.org/faq). Approximately, 1 in 7 North American men will be diagnosed with prostate cancer in his lifetime. On average, everyday 73 Canadian men are diagnosed with PCa and on average 11 Canadian men die from the disease (http://www.prostatecancer.ca). Approximately 26,500 Canadians will be diagnosed with PCa in 2012.

1.1 The prostate and the development of prostate cancer

The prostate is a male reproductive gland. The physiological role of prostate in reproduction is to serve as a secretory gland, primarily to maintain semen gelation, coagulation and liquefaction for successful fertilization [5]. Prostate gland surrounds the urethra and is located inferior to the bladder and anterior to the rectum (Fig. 1.1). At birth, the size of the prostate in humans is approximately 1-2 grams and after puberty androgens act to mature the organ to its adult size of approximately 20 grams [6].

Figure 1.1 Human prostate gland. 1

Development of the prostate originates in utero from the ambisexual endodermal urogenital sinus (UGS) when androgens such as testosterone and dihydrotestosterone (DHT) stimulate the androgen receptor (AR) in the surrounding embryonic connective tissue, urogenital sinus mesenchyme (UGM) [7]. Androgen stimulation of the AR in the UGM supports the differentiation of stem cells in the UGS.

What differentiates a cancer cell from a normal cell is its ability to grow uncontrollably and invade surrounding tissues, as well as its potential to spread to other distant organs [8]. Within the stromal microenvironment an epithelial stem cell destined for differentiation into a normal basal or luminal secretory cell can instead be forced to undergo differentiation into a cancerous cell [9]. Many factors contribute to PCa development, including the disruption in the interactions between mesenchymal and epithelial cells, which is mediated through changes in growth factors expression and overexpression of proteases acting on the mesenchyme to increase metastasis [5, 10]. The presence of PCa may be indicated by symptoms, physical examination, and elevation of serum levels of prostate-specific antigen (PSA), which can result in biopsy. The severity of PCa, and therefore the treatment regime, can then be determined by Gleason score.

1.1.1 Gleason score

Gleason score is a histomorphometric method that was developed by Donald Gleason to classify adenocarcinoma of the prostate by evaluating the pattern of glandular differentiation [11]. The tumour grade is the sum of the dominant and secondary patterns, each numbered on a scale of 1 to 5 Gleason patterns are associated with the following features [12] (Fig. 1.2):

Figure 1.2 Gleason scale.

Pattern 1 - The cancerous prostate glands are small, well-formed, and closely packed. Pattern 2 - The gland tissues are still well-formed, but they are larger and have more space between them. Pattern 3 - The tissue still has recognizable glands, but the cells are darker. At high magnification, some cells have left the glands and are beginning to invade the surrounding tissue. Pattern 4 - The tissue has few recognizable glands. There are irregular masses of cells with few glands. Pattern 5 - The tissue does not have recognizable glands. There are often just sheets of cells throughout the surrounding tissue.

The pathologist assigns a grade to the most common tumor pattern, and a second grade to the next most common tumor pattern. The two grades are added together to get a Gleason score. The Gleason grade is also known as the Gleason pattern, and the Gleason score is also known as the Gleason sum. For example, if the most common tumor pattern was grade 3, and the next most common tumor pattern was grade 4, the

Gleason score would be 3+4 = 7. As Gleason pattern ranges from 1 to 5, with 5 having the worst prognosis, the Gleason score ranges from 2 to 10, with 10 having the worst prognosis. For Gleason score 7, Gleason 4+3 is a more aggressive cancer than Gleason 3+4.

1.1.2 Prostate specific antigen (PSA)

In North America, routine screening for PCa involves a digital rectal examination (DRE), a Prostate Specific Antigen (PSA) blood test and a biopsy. The DRE is relatively simple procedure: the doctor gently puts a lubricated, gloved finger of one hand into man’s rectum to check for problems with organs or other structures in the pelvis and lower belly. The PSA test, which was originally approved by the FDA in 1986, is a simple blood test screen for elevated levels of PSA. PSA, prostate specific member of kallikrein-related protease family, is produced specifically in the luminal secretory epithelial cells of the gland. There is no specific normal or abnormal level of PSA in the blood. In the past, most doctors considered PSA levels of 4.0 ng/mL and lower as normal. Therefore, if a man had a PSA level over 4.0 ng/mL, doctors would often recommend a prostate biopsy to determine whether PCa was present. But more recent studies have shown that some men with PSA levels below 4.0 ng/mL have PCa and that many men with higher levels do not have PCa [13]. Although experts’ opinions vary, there is no clear consensus regarding the optimal PSA threshold for recommending a prostate biopsy for men of any racial or ethnic group. In general, however, the higher a man’s PSA level, the more likely it is that he has PCa.

From the perspective of molecular pathology, PSA is highly expressed by normal prostatic epithelial cells, however overexpression of PSA accelerates carcinogenesis in the prostate. PSA plays many crucial roles in the development of prostate cancer (Fig. 2.3) [14]. Firstly, it is thought to cleave insulin-like growth-factor-binding protein 3 (IGFBP3), thereby liberating insulin-like growth-factor 1 (IGF1), which is a

4 mitogen to prostatic stromal and epithelial cells. Also, it can activate latent transforming growth factor (TGF) to stimulate cell detachment and facilitate tumour cell spread. Moreover, PSA can directly proteolyse components of the basement membrane, which leads to tumour-cell invasion and metastasis [14].

Figure 1.3 Functions of PSA in the normal prostate and in carcinogenesis.

(a). In the normal prostate, PSA acts to liquefy spermatozoa. (b). In carcinogenesis of the prostate, PSA acts on various molecules to potentially enhance proliferation, cell detachment, invasion and metastasis through increasing the availability of insulin-like growth-factor 1 (IGF1), through proteolysis of insulin-like growth-factor-binding protein (IGFBP) and activation transforming growth factor-β (TGF-β).

1.2 Androgen receptor (AR)

Androgen receptor (AR) is a member of the steroid and nuclear receptor superfamily, which is activated by the binding of androgens, mainly testosterone and 5α-dihydrotestosterone (DHT) [15].

In the normal prostate, the predominant role of AR is to promote differentiation of luminal epithelial cells and to regulate the transcription of genes encoding proteins necessary for prostate function, such as PSA. In PCa, the proliferation of cancer cells with a low concentration of androgen may involve the overexpression of the AR [16]. Furthermore, AR activity could modulate the expression of genes associated with cancer cell cycle regulation, survival and growth [17, 18]. In addition, AR is known to stimulate the expression of TMPRSS2-ERG [19]. TMPRSS2-ERG fusions are the predominant molecular subtype of PCa. Approximately 50% of PCa from PSA screened surgical cohorts are TMPRSS2-ERG fusion-positive, and greater than 90% of PCa overexpress TMPRSS2-ERG fusions [20]. High levels of ERG and other members of the ETS oncogene family contribute to PCa development.

1.3 Castrate-resistance prostate cancer (CRPC)

There are several different treatment options for PCa patients, depending on the grade and stage of the disease at diagnosis, and the aggressiveness of the tumor after first-line treatment. If the cancer is locally confined, treatments such as radical prostatectomy, radiation therapy or active surveillance may be employed. Radical prostatectomy involves surgically removing the prostate and closely surrounding tissues [21]. For patients with low to intermediate-risk disease, radiation therapy is often successful (70-90% at 10-year progression-free survival) [22]. If the cancer at diagnosis has spread well beyond the gland or is not responsive to these localized therapies, then androgen deprivation therapy

(ADT) is the most common treatment for patients.

Unfortunately, the effectiveness of this type of treatment is usually temporary due to the progression of surviving tumor cells which become castration-resistant prostate cancer (CRPC). This extremely painful form of PCa has no curative treatment options and a median life expectancy of around 18 months [23]. Globally, CRPC claims the lives of 32,000 men annually [24]. In actual fact, only a few chemotherapeutic agents that act to both prolong survival and improve quality of life are available for patients with metastatic CRPC. The cancer chemotherapeutic docetaxel has been used as treatment for CRPC with a median survival benefit of 2 to 3 months [25]. The oral drug ketoconazole is sometimes used as a way to further manipulate hormones with a therapeutic effect in CRPC. However, there are many side effects associated with ketoconazole usage. Abiraterone, a specific inhibitor of the cytochrome p450 enzyme 17A1, is likely to supplant ketoconazole as it shown to be more effective, with less toxic side effects [26]. On April 28, 2011, the U.S. FDA approved abiraterone acetate in combination with prednisone to treat patients with late-stage (metastatic) castration-resistant prostate cancer.

Only through understanding the molecular biology underlying disease progression have we been able to develop drugs, such as the ones listed above. Furthering our understanding by exploring known and yet to be understood mechanisms by which prostate cancer cells become resistant to castration will allow us to develop and optimize new, more effective drugs of CRPC.

1.4 Neuroendocrine prostate cancer (NEPC)

Neuroendocrine tumors have been recognized as a unique subcategory of cancers of various epithelial organs. Neuroendocrine prostate cancer (NEPC) is an aggressive subtype of PCa, evolving from preexisting prostate adenocarcinoma. NE cells express potent neuropeptides that mediate diverse biological processes such as cell growth, 7 differentiation, transformation and invasion. Although NE cells do not stain for proliferative antibodies, they may be a source of paracrine factors that support prostate cancer growth and progression (Fig. 1.4).

Figure 1.4 Neuroendocrine cells in the development of prostate cancer.

Prostate epithelium is interspersed with neuroendocrine cells. These cells produce specialized neuropeptides, including chromogranin-A (CgA), which stimulate and regulate the normal prostatic-cell secretory functions [27]. On the contrary, inappropriate neuroendocrine regulation might facilitate carcinogenesis, aiding proliferation and other tissue changes such as loss of basal cells, angiogenesis, piling up of prostatic luminal epithelium and invasion, which are characteristic of prostatic carcinoma.

Prostate cancer can be broadly divided into two groups at diagnosis, based on the AR expression level [28]. AR and PSA positive adenocarcinoma accounts for over 95% of initial prostate cancer diagnoses, while NEPC which are typically AR and PSA negative (also known as small cell carcinoma) accounts for 2-3% of primary prostate cancer. However, at the late stage of PCa, 30-100% of advanced type show enrichment of neuroendocrine-like cells [3]. One hypothesis shows that ADT may induce micro-environmental changes that increase the activity of these cells. Concomitantly, NE cells may allow continued PCa growth through paracrine

8 stimulation of neoplastic epithelial cells. Indeed, mitogenic and oncogenic activity has been demonstrated for many of the factors NE cells are known to produce.

Interestingly, blockade of REST function in the human LNCaP cells is found to cause NED [4]. As reported by Tawadros’ team, during NE phenotype acquisition of LNCaP, REST binding activity studies demonstrated a decreased binding activity even after 8h of differentiation. Furthering, they evaluated the regulation of the human synaptophysin (SYP) gene, one REST-target gene, which is also known to be a marker of NE differentiation in human prostate. The result shows that SYP is expressed at low level in LNCaP cells and is largely increased during NE phenotype acquisition. Therefore, these data indicate REST may be one of the important determinants of the NE phenotype in PCa [4].

2 The transcription factor REST

Repressor element 1- silencing transcription factor (REST), also known as neuron-restricted silencing factor (NRSF), is a prominent transcriptional repressor, regulating a myriad of genes throughout the human body. In 1995, two independent groups identified that REST/NRSF was a protein that binds a DNA sequence element, called the neuron-restrictive silencer element (NRSE) which represses neuronal gene transcription in non-neuronal cells [29, 30]. The REST gene was assigned to the chromosomic region 4q12 by fluorescence in situ hybridization (Fig. 2.1) [31]. REST has been shown to play critical roles during embryonic development and neurogenesis. However, during tumorigenesis REST shows paradoxical roles in different type of cells. REST acts as a tumor suppressor in human epithelial cancers while it plays an oncogenic role in the development of brain tumors and other non-epithelial cancers.

Figure 2.1 REST Gene in genomic location: bands according to Ensembl.

2.1 REST and its alternative isoforms

The structure of REST includes a DNA-binding domain and two repression domains (RD): one at the N-terminal, RD1 and one at the C-terminal, RD2 [32]. REST recognizes repressor element 1 (RE1) motifs within the control regions of target genes through its eight finger DNA-binding domain and recruits multiple co-factors, including mSin3 and CoREST via these two repression domains (Fig. 2.2).

Figure 2.2 Model of REST domains.

The REST gene spans 24 kb of genomic DNA. According to Entrezgene the REST is composed of four exons. However, the literature indicates that at least six different splice variants of REST exist (Fig.2.3) [33] and are associated with neural gene expression and certain disease states [34].

Figure 2.3 REST mRNA and alternative splicing variants.

The human REST gene is shown in Fig. 2.3 (top) illustrating the three main exons (dark blue), an alternative exon (light blue) and one of three alternative 5' non-coding exons (white). The approximate position of sequence encoding eight zinc fingers are illustrated by boxes, these are associated with DNA binding (white) and nuclear localization (red). REST reference sequence (NM_005612.3) codes the major isoform 1. REST 1 and REST-5FΔ code isoform 2 or isoform 4 respectively. The alternative exon in REST-N62, REST-N4 and sNRSF (N50) introduces a premature stop codon (white) and all three transcripts encode isoform 3 (REST4).

Table 2.1 Molecular weight of REST alternative isoforms (Uniprot.org) Isoform Amino acids Protein Molecular Weight (kDa) Isoform 1 1097 122 Isoform 2 (Δ314-1097), 313 35 Isoform 3 (Δ330-1097), 329 37 Isoform 4 (Δ304-326), 1074 119

Isoform 1 is the canonical sequence, which comprises two repression domains, lysine-rich (400-603) and proline-rich (595-815) domains, a DNA binding domain (159-412) of eight zinc fingers. As previously mentioned, REST is able to recruit a variety of transcriptional co-repressors through binding at RD1 or RD2; examples are illustrated including Sin3A and CoREST. Isoform 2, which lacks RD2, is truncated and retaining only zinc fingers 1-4 and localizes to the cytoplasm. Isoform 3, also called REST4 or sNRSF, is expressed in neuroblastoma and SCLC; it also lacks RD2 and has a truncated DNA binding domain, but it retains zinc fingers 1-5 and nuclear localization. Isoform 4 has selective deletion of zinc finger 5 [34, 35].

REST4 contains only five zinc fingers of the eight in full length REST and it lacks the C-terminal zinc finger repression domain. The biological function of REST4 is controversial. In fact, it does not act as a transcriptional repressor like REST but rather regulates the repressor activity of REST. In fact, REST4 inhibits the binding of REST to the RE1 sequence, instead of binding to the RE1 directly [36]. Furthermore, it has been proposed that different isoforms of REST can regulate distinct patterns of gene expression and, especially in certain cancer types, may function antagonistically.

2.2 RE1 motif

RE1 is a silencing element which was originally found in the 5’ flanking region of 12

NaV1.2 (voltage-gated sodium type II channel) and SCG10 (superior cervical ganglion 10) genes [37]. RE1 is a general DNA element for the control of neuron-specific gene expression in vertebrates.

REST interacts with 21-bp RE1 sites via its DNA-binding domain [30]. However, individual RE1 sites interact differentially with REST under a particular cellular content, which leads to various RE1-binding profiles. Several independent groups have performed genome-wide analysis of RE1s and potential REST target genes across different genomes [38-40].

Figure 2.4 Sequence logo of canonical RE1 motif.

The canonical RE1 sequence is actually a subset of a much larger, more complex motif. RE1 shows an uneven distribution to the binding of REST of the 21 positions (Fig. 2.4). The position 2 to 9 and 12 to 17 nucleotides are referred as core positions of RE1 while position 10 and 11 are two of the least-highly conserved nucleotides across mammalian species. Furthermore, to address transcriptional regulation by REST in a global and unbiased manner, serial analysis of chromatin occupancy (SACO) was utilized to identify the transcriptome regulated by REST [41]. There were novel REST-binding motifs discovered at the position of 10 and 11: one is the expanded site where additional nucleotides were inserted; the other one is the compressed site where nucleotides were removed. These results strongly suggest a complex mechanism for REST/RE1 interaction for normal neuronal gene repression.

2.3 REST Co-repressors

The nucleosome is the basic unit of DNA packaging in eukaryotes. The nucleosome core particle consists of approximately 147 base pairs of DNA wrapped in 1.67 left-handed superhelical turns around a histone octamer consisting of 2 copies each of the core histones H2A, H2B, H3, and H4. Fundamentally, nucleosomes form the basic repeating units of eukaryotic chromatin, which is used to pack the large eukaryotic genomes into the nucleus while still ensuring appropriate access to it. Since they were discovered in the mid-1960s, histone modifications have been predicted to affect cellular transcription in different ways [42]. Two of the most important types of histone modifications are acetylation and methylation of lysine residues [43]. Acetylation of histones on lysine residues neutralizes their positive charge, is a mark for active transcription. Typically, this kind of reaction is catalyzed by enzymes with histone deacetylase (HDAC) or histone acetyltransferase (HAT). On the other hand, methylation of lysine is more complex. Several lysine residues on the tails of histones H3 and H4 can be mono-, di- or even tri-methylated. The position of the lysine and the exact state of methylation (in conjunction with other epigenetic marks) determines whether the region is transcriptionally active, repressed, or silenced. The methylation of lysine prevents its acetylation. That is to say, acetylation and methylation may act antagonistically. Histone H3 lysine 4 (H3K4) can be mono-, di- or tri-methylated, and the two states of di- and tri-methyated H3K4 are hallmarks of transcriptional activity. H3K9 can be either methylated or acetylated. Acetylation of H3K9 inhibits the ability of LSD1 (lysine-specific demethylase 1) to demethylate H3K4 (Fig. 2.5) [44].

Figure 2.5 Histone modification on H3 N-terminus.

The structure of REST includes a DNA-binding domain and two repression domains (RD): one at the N-terminal, RD1 and one at the C-terminal, RD2 [32]. REST-mediated gene repression is achieved by the recruitment of two separate co-repressor complexes, mSin3 and CoREST, in addition to other co-repressors such as the histone H3 lysine 9 (H3K9) methylase (G9a) which enables methyl groups to add on lysine 27 as well as lysine 9 in H3 [45]. And the NADH-sensitive co-repressor, carboxyl-terminal binding protein (CtBP) is also included [46]. G9a and CtBP can be recruited directly by REST but also have been found associated with the CoREST complex. The mSin3 complex contains two class I histone deacetylases (HDACs), HDAC1 and HDAC2 while the CoREST complex contains HDAC1 and HDAC2, a histone H3K4 demethylase (LSD1), the chromatin-remodeling enzyme, (BRG1), also known as SMARCA4, as well as CtBP, G9a, and a methyl-CpG-binding protein (MeCP2) [44]. In summary, REST interacts with several complementary chromatin-modifying activities, with the recruitment of some of these enzymes being regulated by intracellular signals such as CAMK-mediated phosphorylation and NADH levels.

Figure 2.6 Mechanism of repression/activation via REST-RE1 interaction.

A. REST binds to RE1 sequence through eight zinc finger domains. The N-terminal domain of REST interacts with Sim3A, which in turn recruits HDAC and associated proteins to the promoter region. The C-terminal domain of REST interacts with CoREST which recruits HDAC1 and HDAC2, resulting in inhibition of RNA polymerase activity (RNA PII). B. Mechanism for activation of gene expression by derepression. REST4 inhibits the binding of REST to the RE1 sequence through association with Sin3 co-repressor complex. Then the CoREST-HDAC complex is sequestered away from interacting with REST, which fails to remove acetyl groups from nucleasomal core histones. Thus the RNA PII is able to form a transcriptionally active complex with the promoter.

2.4 REST and non-coding RNAs

A non-coding RNA (ncRNA) is a functional RNA that is not translated into a protein such as transfer RNA (tRNA) and ribosomal RNA (rRNA), as well as RNAs such as small nucleolar (snoRNA), microRNA (miRNA), short interfering RNA (siRNA) and long non-coding RNA (long ncRNA). It is increasingly clear that ncRNAs are far from being just transcriptional noise but are involved in chromatin accessibility, transcription and post-transcriptional processing, trafficking, or RNA editing. REST and its cofactor CoREST are both highly regulated by various ncRNAs.

Figure 2.7 REST-ncRNA interactions in neural networks.

A yet to be identified mechanism initiates the binding of BRG1 to REST, resulting in the exposure of the RE1 motif and the subsequent occupation of it by REST. With the presence of Sin3 and CoREST, the RE1-bound REST complexes recruit HDAC1/2 and MeCP2, to promote histone deacetylations, G9a and LSD1. REST can recruit additional histone methylases and demethylases to target lysine residues of histones; arrows link cofactors to their chromatin remodeling and modifying activities. Inhibition of these histone modifying components results in an attenuation of REST recruitment, a decrease in H3K9me2, an increase in H3K9/K14 and H4 acetylation, an increase in H3K4me3 and a decrease in binding of MeCP2 to methylated DNA. In neural progenitors, REST represses miR-124a expression, allowing the persistence of non-neuronal transcripts. Upon differentiation of the neuronal progenitors into neurons, REST dissociation from the miR-124a gene loci induces the activity of gene derepression, which leads to selective degradation of non-neuronal gene transcripts. 18

Initially, REST was found to modulate genes for neuronal differentiation, homeostasis and plasticity. A similar role has been demonstrated for miRNA and ncRNA networks in neuronal differentiation and plasticity [47]. The miRNA miR-124 was originally identified as a miRNA specifically expressed in the brain, and later was shown to negatively regulate the expression of many non-neuronal genes. Interestingly, expression of miR-124 is repressed by the transcriptional repressor REST [47]. Fig. 2.7 illustrates the major components of REST-ncRNA interactions in neural networks [48]. REST-mediated regulation of neuronal gene expression relies on the dissociation of the REST repressor complex from the RE1 site. In neural progenitors, REST represses miR-124a expression, allowing the persistence of non-neuronal transcripts. Upon differentiation of the neuronal progenitors into neurons, REST dissociation from the miR-124a gene loci induces the activity of gene derepression, which leads to degrade non-neuronal gene transcripts selectively. Therefore, REST links transcriptional and post-transcriptional events to fine-tune the balance of phenotype between neuronal and non-neuronal cells, partly by controlling miR-124 and other ncRNAs that act in a network associated with miR-124 [48].

2.5 REST function and cancer

Transcription factors often coordinately control complex programs of gene expression during development as well as the aberrant activation of developmental programs in cancer. As REST plays crucial roles in the regulation of gene expression, abnormal expression or dysfunction of REST has been associated with diverse diseases, including Down’s syndrome [49], Huntington’s disease [50] and various types of cancer [51]. Interestingly, REST appears to have opposite roles in non-nervous cancers and tumors in central nervous system (CNS): REST acts as a tumor suppressor to inhibit tumorigenesis in epithelial cells, while it plays an oncogenic role during the development of brain tumors.

Figure 2.8 Experimental models and possible routes for REST dysfunction in cancers. (A,B) In cell line models experimental strategies that reduce REST levels or inactivate REST are associated with transformation or progression of the cancer phenotype. (C) On the contrary, active REST is overexpressed in medulloblastoma. (D–F) In other cancers, genetic deletion or depletion of REST may be accompanied by mutation (D) or alternative splicing (E, F) generating different isoforms such as REST-FS, sNRSF or hREST-N62, lacking the C-terminal repression domain.

REST was identified as a human tumor suppressor in epithelial cells during an unbiased RNA interference (RNAi) library screening [52]. Functional inhibition of REST in human mammary epithelial cells increased cellular capability for oncogenic transformation. Furthermore, a splice variant of REST transcript was highly expressed in established cell lines and primary small cell lung cancer cultures as well as in some primitive neuroectodermal tumor biopsies [35]. Both breast and colon cancer can display some neuroendocrine features, where many neuroendocrine genes that would normally be restricted through RE1 motifs in non-neuronal cells are aberrantly expressed [51, 52]. All these findings are corroborated by the loss of REST function accompanying the acquisition of the neuroendocrine phenotype in LNCaP prostate cancer cells via transdifferentiation [4]. In human cell lines derived from 20 neuroblastoma, additional exons introduce a frameshift and premature termination, resulting in a truncated DNA-binding domain lacking the C-terminal repression domain. Therefore these tumors would employ independent strategies to generate the C-terminal deleted variants that eschew key transcriptional silencing mechanisms as an alternative route to modulate REST function (Fig. 2.8).

On the contrary, it has been demonstrated that REST acts as an oncogenic promoter in several types of brain tumors, including medulloblastoma [53] and glioblastoma. REST is upregulated in the development of medulloblastoma relative to differentiated neurons or neural progenitors. Ectopic expression of REST-VP16 mutant in medulloblastoma cells significantly reduced tumorigenic potential of these cancer cells in mice [54]. And inhibition of REST in medulloblastoma cells triggers apoptosis through the activation of caspase cascades [53]. Therefore, the effect of REST inhibition on cell survival in medulloblastoma cells and human epithelial cancer cells appears totally opposite. Inhibition of REST activity promotes survival of human epithelial cell but induces apoptosis of medulloblastoma cells. It is not well understood why REST plays differential roles in different cancer types. The precise molecular mechanisms associated with this cell type-specific role of REST need to be further elucidated. Probably, REST forms different dynamic repressor complex through recruiting other co-activators in different cell types to exert its diverse biological roles. For the importance of REST in regulating tumor development, more characterizations on cell type-specific transcription of REST as well as REST-mediated signaling pathways are necessary.

3 Aims of research

Prostate cancer (PCa) is one of the most commonly diagnosed cancers in the world: males have 17% risk of developing PCa during their lifetime. Pharmacological androgen deprivation therapy (ADT) has been the leading form of PCa therapy since Huggin’s discovery that castration induced the regression of PCa tumors in 1941[1]. Unfortunately, the effectiveness of this type of treatment is usually temporary, with a subpopulation of cells surviving and becoming castrate resistant prostate cancer (CRPC). This form of PCa is most often terminal, with a median life expectancy of

approximately 18 months [23]. Globally, CRPC claims the lives of 32,000 men annually [24]. The mortality rate of PCa could be lowered by more effective early detection due to identification of novel molecular markers and better understanding the molecular mechanisms underlying tumor progression and metastasis. Growing evidence indicates that neuroendocrine differentiation (NED) occurs in 30-100% of disease recurrence after ADT [2, 3]. However, the mechanism is still poorly understood.

Repressor element-1 silencing transcription factor (REST) is generally expressed throughout the body, which is in charge of restricting neuronal gene expression to the nervous system by silencing expression of neuronal genes in non-neuronal cells. REST expression was recently shown to be absent in various types of cancer, such as small cell lung cancer (SCLC) [55] and colorectal cancer [52]. Blockade of REST function in the human LNCaP cells is found to cause NED [4].

Here, we aim to investigate the possible role of REST in PCa progression from the following aspects: a. Use expression microarray with RNA interference (RNAi) technology to indentify potential REST targets under human LNCaP cell model

preliminarily. b. Sequence the transcriptome of the patient-derived mouse xenograft models at two different time points from castration to neuroendocrine differentiation. c. Detect overlapped genes differentially expressed between two platforms (microarray on LNCaP model and RNA-seq on xenograft model) and investigate these genes for pathway and function enrichment analysis. d. Experimentally validate the highest priority targets, using quantitative RT-PCR. e. Estimate the clinical significance of the REST targets using public datasets.

It is expected that sequencing-based approaches to measuring gene expression levels have more advantages than microarray. This ultra-high-throughput sequencing techniques enable thousands of megabases of DNA/RNA to be sequenced in a matter of day. Furthermore, RNA-Seq has much higher resolution, better dynamic range of detection and lower technical variation than microarray. Nevertheless, microarray represents a well-established and widely used technology through last decades. Thus, compared between these two platforms will provide us with more accurate information.

The unique patient-derived xenograft model of PCa we used here is based on sub-renal capsule xenografting of patient tumor tissue in NOC/SCID mice. The xenograft retains key biological properties of the original malignancies, including histopathological, genomic and transcriptomic characteristics, tumor heterogeneity and metastatic ability. Therefore, this high fidelity model is more clinical research relevant.

We believe all the results could contribute to elucidate the mechanism that drives

adenocarcinoma to escape AR blockade through NED to androgen independent and aggressive NEPC. Furthermore, it could provide a new prognosis factor for PCa and also provide a new target for the PCa treatment.

4 Materials and methods

4.1 Cell lines

LNCaP cells were kindly provided by Dr. Leland WK Chung (1992, University of Texas MD Anderson Cancer Center, Houston, TX, USA). LNCaP cells used for sequencing were provided by Pfizer (La Jolla, CA, USA). LNCaP cell line was authenticated using aCGH, expression microarrays, and whole-genome and whole-transcriptome sequencing on the Illumina Genome Analyzer IIx platform (Illumina, San Diego, CA, USA) at the time of data collection for the current study.

4.2 Prostate tissue samples

All patients signed a consent form approved by the Ethics Board (University of British Columbia UBC Ethics Board No: H09-01 628; VCHRI Nos: V09-0320 and V07-0058). Surgical samples were collected and snap-frozen (FF) at the Vancouver General Hospital (VGH). The remaining tissue was fixed in formalin and used for histopathological evaluation by three independent pathologists from the VGH pathology department and Vancouver Prostate Centre (VPC). Snap-frozen and formalin-fixed blocks are stored at the VPC Tissue Bank. [56].

4.3 Mouse xenograft model

LTL331 xenograft of human prostate cancer was developed by grafting tumor biopsy specimen (tumor 927) under the renal capsules of nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice [57]. The tumor line development protocol and use of human specimen use was approved by the Clinical Research Ethics Board of the UBC, in accordance with the Laboratory Animal Guidelines of the Institute of Experimental Animal Sciences. To collect PCa tissue after castration, fresh LTL-331 tumor tissues from the 5th generation of grafting were cut into 3mm×3mm×1mm

25 pieces and re-grafted into the subrenal capsules of male NOD/SCID mice.

4.4 Public dataset

Study data is deposited in NCBI GEO under accession GSE21032. The analyzed data can also be accessed and explored through the Memorial Sloan-Kettering Cancer Center (MSKCC) Prostate Cancer Genomics Data Portal: http://cbio.mskcc.org/prostate-portal/ [58]. This database contains a total of 218 tumor samples and 149 matched normal samples that were obtained from patients treated by radical prostatectomy at MSKCC. All patients provided informed consent and samples were procured and the study was conducted under MSKCC Institutional Review Board approval. Clinical and pathologic data were entered and maintained in our prospective prostate cancer database. Following radical prostatectomy, patients were followed with history, physical exam, and serum PSA testing every 3 months for the first year, 6 months for the second year, and annually thereafter. For all analyses described here, biochemical recurrence (BCR) was defined as increase in serum PSA concentration by ≥ 0.2 ng/ml on two occasions. At the time of data analysis, patient follow-up was completed through December 2008 [58].

4.5 siRNA and shRNA experiments siRNAs and shRNAs corresponding to the antisense sequence of human REST gene were purchased from Santa Cruz Biotechnology, Inc. REST siRNA (ID: sc-38129) is a pool of 3 target-specific 19-25 nucleotides siRNAs designed to knock down gene expression. REST shRNA Plasmid (ID: sc-38129-SH) is a pool of 3 target-specific lentiviral vector plasmids each encoding 19-25 nt (plus hairpin) shRNAs designed to knock down gene expression. Each plasmid contains a puromycin resistance gene for the selection of cells stably expressing shRNA. In addition, negative controls were purchased from the same company: control siRNA-A (sc-37007) and control shRNA

plasmid-A (sc-108060). The REST siRNA was transfected into LNCaP cells and allowed to grow for 48 h before harvest with Trizol. The mRNA from this experiment was prepared for gene expression microarray. The REST shRNA knock-down experiment was also transfected into LNCaP cells for analysis by RT-PCR.

4.6 Microarray

Total RNA was isolated from the LNCaP cells treated with 2 mg REST siRNA or 2 mg control siRNA for 48 h in duplicate using the Trizol reagent according to the manufacturer’s instructions. All conditions were done in duplicate. Then RNA quality was assessed with the Agilent 2100 bioanalyzer prior to microarray analysis. Samples with a RNA Integrity Number (RIN) value of greater than or equal to 8.0 were deemed to be acceptable for microarray analysis. Total RNA samples were prepared following Agilent One-Color Microarray-Based Exon Analysis Low Input Quick Amp WT Labeling Protocol Version 1.0. An input of 100ng of total RNA was used to generate Cyanine-3 labeled cRNA and samples were hybridized on Agilent SurePrint G3 Human 4x180K Microarrays (Design ID 028679). Arrays were scanned with the Agilent DNA Microarray Scanner at a 3um scan resolution, and data was processed with Agilent Feature Extraction 10.10 and analyzed by Agilent GeneSpring 11 software. Statistical analysis was performed using unpaired t-test considering Benjamini–Hochberg corrected p-value of 0.05. Differentially expressed genes with

over log2 fold change (FC) of 1.5 greater than or equal to 1.5 (up or down) were considered for the present study (supplementary 4.1).

4.7 Transcriptome sequencing

Transcriptome sequencing of the samples was performed at British Columbia Cancer Agency Michael Smith Genome Sciences Centre (Vancouver, BC, Canada) using the

the Illumina Genome Analyser II according to established protocols described in [59]. Junction aware alignment tool (Tophat [60]) was employed to map RNA-seq reads onto transcriptome and genome. Read counts for each gene, transcript, exon or junction were calculated according to gene model annotation (ENSEBML database, v. 62). Raw read counts were normalized by R-package DESeq [61], which is able to

remove experimental variability for RNA-seq data across different samples. Since there are only two time points, we simply calculated fold change for each gene. And

differentially expressed genes with log2 fold change (FC) greater than or equal to 1.5 (up or down) were considered for the present study.

4.8 Statistics Analysis

Results were reported as means ± standard deviations. To detect genes differentially expressed between subsets of tumor samples, a two-tailed Student's t-test was applied with equal variances with a P value level cut-off of 0.05. Survival curves were calculated using the Kaplan–Meier method, where the significance was determined by log-rank test.

4.9 Data analysis

Pathway and functional enrichment analysis was performed using Ingenuity (IPA) Knowledge Base 9 (Ingenuity®Systems, www.ingenuity.com).

4.10 Experimental RT-PCR validation

Total RNA was extracted from cultured cells after 48 hours of treatment using Trizol reagent (Invitrogen). Two micrograms of total RNA was reverse transcribed using the Transcriptor First Strand cDNA Synthesis Kit (Roche Applied Science). Real-time 28

monitoring of PCR amplification of cDNA was performed using DNA primers (Table 4.1 ) on ABI PRISM 7900 HT Sequence Detection System (Applied Biosystems) with SYBR PCR Master Mix (Applied Biosystems). Target gene expression was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels in

respective samples as an internal standard, and the comparative cycle threshold (Ct) method was used to calculate relative levels of target mRNAs. Each assay was performed in triplicate. Table 4.1 RT-PCR primer sequence Gene Oligonucleotide primer sequence (5’—3’) REST-F TGCGTACTCATTCAGGTGAGA REST-R TCTTGCATGGCGGGTTACTT Bex1-F GATAGGCCCAGGAGTAATGGAG Bex1-R CTTGGTTGGCATTTTCCATG Srrm3-F CATCCTTTTGAGCTGGCTGC Srrm3-R CCAAGAGTGAGGTGCCTAGC

5 Results

5.1 Specific and efficient knock-down of REST mRNA in LNCaP cells

The emergence and development of RNA interference (RNAi) as a biological mechanism to silence gene expression has enabled mammalian cells to lose function in a potentially genome-wide manner [38]. The applications of RNAi can be mediated through two types of molecules; the chemically synthesized double-stranded small interfering RNA (siRNA) or vector based short hairpin RNA (shRNA). We have utilized both siRNA and shRNA to identify genes regulated by REST in the human LNCaP cell model. The efficiency of the RNAi strategy for REST was evaluated by semi-quantitative real-time PCR in LNCaP cells treated with a REST siRNA compared with that treated with controls treated with a scrambled siRNA. As shown in Fig. 5.1, RT-PCR gel shows down-regulation of REST and concomitant up-regulation of NE differentiation marker SYP, whereas beta-actin was not changed.

Figure 5.1 siRNA knockdown of REST in LNCaP cells. The RT-PCR gel showed down-regulation of REST and concomitant up-regulation of NE differentiation marker SYP, whereas beta-actin (reference gene) was not changed.

Compared to siRNA, small hairpin RNA (shRNA) is an advantageous mediator of RNAi in that it has a relatively low rate of degradation and turnover. They are 30 synthesized in the nucleus of cells, further processed and transported to the cytoplasm, and then incorporated into the RNA induced silencing complex (RISC) for activity [62]. The experimental protocols between siRNA and shRNA share similarity except that shRNA needs puromycin dihydrochloride selection. Thus we selected a batch of puromycin-resistant clones (a, b, e, f, g, h, j) which may have varying levels of shRNA expression due to the random integration of the plasmid into the genome of the cell. As shown in Fig. 5.2, the RT-PCR result indicated that all expression levels of the shRNA knockdown samples were lower than the three controls (right side), whereas sample e showed the best silencing efficiency. Thus we continued to use e for the further studies.

Figure 5.2 shRNA knockdown of REST in LNCaP cells.

Sh(-a to -j) are the experimental groups which are LNCaP cells treated by REST shRNA while control-a, control-c and control-d are LNCaP cells treated by scrambled control RNA. Expression level is calculated from relative Ct value from experimental group and beta-actin (reference gene).

5.2 High fidelity of mouse xenograft model for neuroendocrine transdifferentiation

LTL331 xenograft of human prostate cancer was developed by grafting tumor biopsy specimen (tumor 927) under the renal capsules of SCID mice [57]. The patient-derived LTL331 xenograft retains the histological characteristics of the patient tumor and is strongly PSA positive and AR positive (Fig. 5.3 A, B). Androgen deprivation by castration of LTL 331 mice resulted decreased tumor volume and serum PSA. In less than 4 months, tumors recurs, accompany with PSA and AR negative (Fig. 5.3 C, D). This recurrent tumor model is re-named “LTL 331R” and expresses a range of neuroendocrine markers including most common one synaptophysin (SYP), as shown E. Judging from this pathological evidence, we hypothesized our treatment of the LTL 331 model had evolved into the neuroendocrine differentiated LTL 331R model in response to castration.

Figure 5.3 The LTL-331 xenograft model of neuroendocrine transdifferentiation.

A-E: Immunohistochemical stains before and after NE transdifferentiation.

5.3 Microarray and RNA-seq analysis for identification of differential expressed genes

Total RNA of LNCaP cells treated with REST siRNA or scramble control for 48 h was isolated for comparative expression analysis using human cDNA microarrays. Overall, 63 genes were found to be up-regulated and 59 genes were down-regulated over log2- fold change of 1.5 (P< 0.05) after knocking down REST in LNCaP cells (Figure 5.4 and Supplementary 4.1). The up-regulated genes including SCG3, BEX1, CHRNB2, SYP, UNC13A, RAB39A, VGF, GDAP1, KCNC1, SYT4, are mainly involved in cell-to-cell signaling and interaction, molecular transport, small molecular biochemistry and drug metabolism.

Figure 5.4 Summary of microarray data on LNCaP model.

63 genes were found to be up-regulated and 59 genes were down-regulated over log2- fold change of 1.5 (P< 0.05) after knocking down REST in LNCaP cells

As discussed previously, sequencing-based approaches to measuring gene expression levels have more advantages than hybridization-based approaches such as microarray. In principle, RNA-seq has much higher resolution, better dynamic range of detection

and lower technical variation. Moreover, massively parallel sequencing allows investigation of far more genome and transcriptome features (i.e. alternative splicing, fusion transcript etc.) than expression microarray which may be used better for the preliminary characterization. Therefore, we performed trancriptome sequencing on patient-derived mouse xenograft models at two different time points: before castration and after neuroendocrine differentiation. We identified 59.7% (18195/30485) genes as

differentially expressed (10381 is up-regulated and 7814 is down-regulated) over log2 fold change of 1.5, providing strong evidence that the majority of genes identified by sequence data are genuinely differentially expressed (Fig. 5.5).

Figure 5.5 Summary of RNA-seq data on xenograft model. 10381 genes were found to be up-regulated and 7814 genes were down-regulated over log2 fold change of 1.5 after castration on mouse xenograft model.

Interestingly, there was a significant overlap between the genes differentially expressed in the siRNA knockdown of REST in LNCaP cells and those in LTL-331R. Overall, we identified 34 genes up-regulated and 12 genes down-regulated overlapping between these two platforms (Fig. 5.6 Tab 5.1). This enrichment (p=0.035<0.05; Fisher’s exact test) is statistical significant to the different model systems. 34

Figure 5.6 The overlap between genes differentially expressed from microarray data and from RNA-seq data. The Venn diagram shows the number of up/down-regulated genes shared (red) and platform-specific genes (blue).

Table 5.1 Profiling of overlap genes differentially expressed Microarray RNA-seq Gene Entrez Gene Name Fold Fold change change BEX1 brain expressed, X-linked 1 44.72 10.41 SYT4 synaptotagmin IV 3.37 8.91 potassium voltage-gated channel, subfamily KCNH6 1.86 6.12 H, member 6 ganglioside induced differentiation GDAP1 6.34 5.94 associated protein 1 CHRNB2 cholinergic receptor, nicotinic, beta 2 17.10 5.88 potassium voltage-gated channel, KCNC1 4.19 5.61 Shaw-related subfamily, member 1 UNC13A unc-13 homolog A 13.01 5.61 FGFBP3 fibroblast growth factor binding protein 3 1.57 5.52 cadherin, EGF LAG seven-pass G-type CELSR3 2.40 5.28 receptor 3 SCG3 secretogranin III 143.70 5.17 SYP synaptophysin 16.17 5.10 OLFM1 olfactomedin 1 2.43 5.04 SRRM3 serine/arginine repetitive matrix 3 2.40 5.01 BSN bassoon (presynaptic cytomatrix protein) 2.05 4.48 STMN3 stathmin-like 3 2.80 4.12 RAB39 RAB39A, member RAS oncogene family 8.24 3.99

resistance to inhibitors of cholinesterase 3 RIC3 1.74 3.50 homolog GEMIN5 gem (nuclear organelle) associated protein 5 1.72 3.03 adaptor-related protein complex 3, beta 2 AP3B2 2.35 2.99 subunit TMEM198 transmembrane protein 198 3.13 2.87 CCDC64 coiled-coil domain containing 64 1.86 2.81 SYN1 synapsin I 1.72 2.72 sodium channel, voltage gated, type VIII, SCN8A 2.21 2.61 alpha subunit ECSIT ECSIT homolog 1.59 2.56 SECISBP2 SECIS binding protein 2 1.68 2.41 VGF VGF nerve growth factor inducible 6.71 2.06 KLHL2 kelch-like 2, Mayven 1.51 1.93 MMP24 matrix metallopeptidase 24 2.66 1.83 solute carrier family 2 (facilitated glucose SLC2A14 1.57 1.81 transporter), member 14 TMEM180 transmembrane protein 180 1.92 1.77 protein tyrosine phosphatase, receptor type, PTPRN2 1.86 1.76 N polypeptide 2 NOLC1 nucleolar and coiled-body phosphoprotein 1 1.69 1.69 MANEAL mannosidase, endo-alpha-like 1.87 1.67 DPYSL4 dihydropyrimidinase-like 4 2.09 1.55 PLAGL2 pleiomorphic adenoma gene-like 2 -1.70 -1.52 SP9 Sp9 transcription factor homolog -1.51 -1.87 cytidine monophosphate (UMP-CMP) CMPK2 -2.02 -2.07 kinase 2, mitochondrial C2orf50 chromosome 2 open reading frame 50 -2.04 -2.36 DAPK1 death-associated protein kinase 1 -1.66 -3.00 transmembrane protein with EGF-like and TMEFF2 -1.56 -3.20 two follistatin-like domains 2 SLITRK3 SLIT and NTRK-like family, member 3 -2.39 -3.33 FZD2 frizzled family receptor 2 -1.50 -3.59 IL28A interleukin 28A -2.14 -3.83 GBP4 guanylate binding protein 4 -1.76 -3.89 ARG1 arginase, liver -1.96 -4.43 CAV2 caveolin 2 -2.20 -4.51 REST RE1-silencing transcription factor -4.95 -5.38

5.4 Functional categorization analysis

Ingenuity pathway analysis (IPA) functional analysis for pathways and diseases revealed many pathways represented in our overlapping dataset, including cell-to-cell signaling and interaction, molecular transport, small molecule biochemistry, and cellular assembly and organization (Fig. 5.7). We overlapped our results with functional pathways or disease categories defined by IPA, the data in Fig.5.7 represent the most significant sub-level function in each category. Cell-to-cell signaling and interaction, molecular transport, small molecule biochemistry and neurological disease were the most significantly represented in our dataset. People found that cell-to-cell signaling influences the fate of prostate cancer stem cells and accelerates them into more aggressive tumors [63].

Figure 5.7 Functional categorization analysis of most significant pathways and diseases.

A functional categorization analysis of the most significant pathways and diseases represented in the microarray-and-RNA-seq-generated list of significantly different genes was generated using Ingenuity software. The enrichment significance for each function is provided as a Benjamini–Hochberg adjusted p-value. The Benjamini-Hochberg procedure aims to calculate the False Discovery Rate (FDR) for each of the p-values. Each bar represents the highest level of function for each category, each of which includes many sub-level functions. The bar was determined by the lowest p-value among sub-level functions for each category. Threshold indicates p-value adjusted with Benjamini-Hochberg cutoff of 0.05.

5.5 Network analysis

To investigate the interaction of these differentially regulated genes with other gene pathways and biological processes, IPA was used to form molecular networks based on functional relationship between gene products referenced in peer-reviewed literature. Other groups of genes, which were biologically relevant to the pathway but not identified in our experiment, were also included. The top 4 networks among the significant genes up/down regulated in our overlap dataset represent many pathways suspected to be involved in cellular assembly and organization, nervous system development and function, neurological disease, cellular compromise, cell cycle, hereditary disorder, and drug metabolism, lipid metabolism, molecular transport and so on. Each network was identified based on a numerical rank score according to the degree of relevance of the network to the molecules in the significant genes list and based on the hypergeometric distribution calculated as –log (p-value). In these network illustrations, genes or gene products are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). All edges are supported by at least one published reference.

Table 5.2 IPA-generated molecular networks

The score is a numerical rank of the degree of relevance of the network to the molecules in the significant genes list and is based on the hypergeometric distribution calculated as –log (Fisher’s exact test result). The bold genes were in our overlapping dataset with up-regulated genes in red arrow and down-regulated genes in green arrow.

Network 1 (Fig. 5. 8) received the highest score (44) and was assembled 18 focus genes that were differentially regulated in REST knockdown condition. In particular, Network 1 contains gene and molecules known to be involved in nervous system development and function (SYN1, UNC13A), neurological disease (BSN, CHRNB2, KCNC1, SYN1), small molecule biochemistry and cell-to-cell signaling and interaction (CHRNB2, SYN1, SYT4, UNC13A). REST acts as one of the hubs in this network, directly interacting with 6 genes (SYT4, SCG3, CHRNB2, KCNH6, SYP, SYN) which significantly up-regulated in our datasets. This network also reveals that REST acts on other 9 genes (PCLO, PHKB, TRPM2, SCN1A, B3GAT1, FGF12, NCAM2,

PHF21A, RCOR2) which were not shown in our over-expressed dataset.

Of the genes involved in the top ranked network which were not altered in our experiments, death-associated protein kinase 1 (DAPK1), which is directly regulated by ERK1/2 in the network, plays a crucial role in metastasis of carcinoma cell lines and delay in growth of tumor. The top canonical pathways derived from this network were bladder cancer signaling (DAPK1, ERK1/2, FGF12, MMP24), Huntington’s disease signaling (Creb, ERK1/2, RCOR2, REST), protein kinase A signaling (Calmodulin, Creb, ERK1/2, PHKB), ERK/MAPK signaling (Creb, ERK1/2, estrogen receptor) which is involved in the development of many cancers, and androgen signaling (Calmodulin, Creb, ERK1/2) which plays a critical role in the development, function and homeostasis of the prostate.

Figure 5.8 Ingenuity pathway analysis Network 1.

Ingenuity pathway analysis was used to assemble a network based upon 18 differentially expressed focus genes. Genes are represented as nodes, with node shape representing the functional class of the gene product (seen in legend). Node color depicts degree of overrepresentation in our experimental data; uncolored nodes are depicted based upon evidence in the IPA Knowledge Base indicating a strong biological relevance to the network. Corresponding fold change values are listed beneath each gene label.

Furthermore, in order to view all molecular interactions through a more comprehensive pathway network, we merged top 3 IPA-generated molecular networks: Network 1, Network 2 and Network 3 described in Table 5.2. This merge network (Fig. 5.9) covered 39 focus genes (18 in Network 1, 12 in Network 2 and 9 in Network 3) , in total 103 genes and molecules which were associated with cell-to-cell signaling and interaction, nervous system development and function, neurological disease, cellular function and maintenance and molecular transport. Specifically, this comprehensive network contains several cancer-related biomarkers: EGFR, NFkB, TNF, TH1 cytokine, FGF2.

Figure 5.9 Combination of Network 1, Network 2 and Network 3.

Ingenuity pathway analysis was used to assemble a merge network based upon Network 1, Network 2 and Network 3 (Table 5.2), covering 39 differentially expressed focus genes. Corresponding fold change values are listed beneath each gene label. All 103 gene names are defined in Supplementary 5.1.

5.6 Experimental RT-PCR validation

Representative genes BEX1 and SRRM3 were selected from overlap datasets of differentially up-regulated genes for validation via semi-quantitative RT-PCR, using the same LNCaP cell line used in the microarray. In the further research, we utilized REST knockdown LNCaP cell line via shRNA instead of siRNA. And as shown in Fig 5.2, among the batch of puromycin-resistant clones (a, b, e, f, g, h, j), sample e group shows the best silencing efficiency.

BEX1 is signaling adapter involved in p75 neurotrophin receptor/ Nerve Growth Factor Receptor (p75NTR/NGFR) signaling, which plays a role in cell cycle progression and neuronal differentiation. BEX1 increased by 44.7 fold change in our microarray platform while increasing by 10.4 fold change in the RNA-Seq platform. Interestingly, the semi-quantitative RT-PCR result (Fig. 5.10) showed that BEX1 significantly over-expressed in e which has better REST silencing efficiency than other groups. Thus, the RT-PCR result of BEX1 is consistent with both microarray and RNA-seq data.

Figure 5.10 BEX1 expression level in REST shRNA knockdown LNCaP cells.

Serine/arginine repetitive matrix protein 3 (SRRM3), belongs to the complexed with Cef1 protein 21 (CWC21) family. The biological function of SRRM is still unknown. However, we found SRRM3 might activate an alternative splicing on PHD finger protein 21A (PHF21A), a REST associated genes from our preliminary data. SRRM3 increased by 2.4 fold change in microarray platform while increasing by 5.0 fold change in the RNA-Seq platform. In the RT-PCR result, e group showed the significant increase of SRRM3 expression level (Fig. 5.11).

Figure 5.11 SRRM3 expression level in REST shRNA knockdown LNCaP cells. ShRNA_REST are the experimental groups which are LNCaP cells treated by REST shRNA while controla and controlc are LNCaP cells without any treatment. Expression level is calculated from relative Ct value from experimental group and beta-actin (reference gene).

5.7 Kaplan Meier survival analysis

In order to estimate their clinical significance of the preliminary set of potential REST targets we identified, we utilized the Kaplan–Meier analysis of the survival which

measures the fraction of patients living for a certain amount of time after treatment. However, prostate tumors show tremendous biological heterogeneity, coupled with its high prevalence. Genomic-based classification is needed for more informed clinical decision-making. Therefore we extended our research to an external public dataset which contains transcriptomes and copy-number alterations (CNA) in 218 prostate samples (181 primaries, 37 metastases) and 12 prostate cancer cell lines and xenografts through Memorial Sloan-Kettering Cancer Center (MSKCC) Prostate Cancer Genomics Data Portal: http://cbio.mskcc.org/prostate-portal/ [58].

After obtaining a general view of all gene alterations in a cohort of 216 tumors, we selected a small high priority groups of genes identified by microarray and RNA-seq based on their percentage of alteration and alteration trend. Three genes BEX1, SCG3, and UNC13A were found to be up-regulated in both LNCaP and LTL331 experimental systems. Interestingly, all three of these candidate genes show shortened disease free survival (Fig. 5.12).

Figure 5.12 Kaplan–Meier survival plots on 3 selective genes in a public dataset prostate tumors (p<0.01).

6 Discussion

Protate cancer is a clinically and biologically heterogeneous disease. Neuroendocrine prostate cancer is an aggressive subtype of prostate cancer that most commonly evolves from preexisting prostate adenocarcinoma. Remarkably, growing evidence indicates that neuroendocrine differentiation occurs 30-100% of prostate cancer recurrence especially after androgen deprivation treatment.

It must be stressed that we still need a more comprehensive system to acquire a more detailed understanding of neuroendocrine differentiated prostate cancer. We are still far away from identifying all the regulated genes or molecules involved in this phenotype; it is possible to do so through microarray and RNA-seq, as microarrays give a broad overview whereas RNA-seq detects subtle changes in gene expression and detects alternative aplicing, which occurs quite often in cancer progression.

RE1 silencing transcription factor REST is expressed throughout the body and it was recently shown to be absent in various types of cancer including prostate cancer. REST transcriptional complex acts as a master repressor of the neuronal phenotype. Blockade of REST function in the human LNCaP cells is found to cause NED, indicating REST may be one of the important determinants of the neuroendocrine phenotype in PCa.

In our study, an integrated data analysis of RNA-Seq on mouse xenograft model and expression microarray with RNA interference (RNAi) on human LNCaP cells was performed to identify potential REST targets: 34 genes were up-regulated and 12 genes were down-regulated. Afterwards, a functional categorization analysis of the most significant pathways and diseases represented by these significantly different genes was generated using Ingenuity software. To investigate the interaction of these differentially regulated genes with other gene pathways and biological processes, molecular networks were formed based on functional relationships between gene products described in peer-reviewed literature. We found that REST acts as one of the hubs in the most significant network, directly interacting with 6 genes (SYT4, SCG3, CHRNB2, KCNH6, SYP, SYN) significantly up-regulated in our datasets, as well as other 9 genes (PCLO, PHKB, TRPM2, SCN1A B3GAT1, FGF12, NCAM2, PHF21A RCOR2) which were not shown to be over-expressed in our dataset. We then extended our research to an external public dataset which contains transcriptomes and copy-number alterations (CNA) in 218 prostate samples (181 primaries, 37 metastases) and 12 prostate cancer cell lines and xenografts. In addition, in order to estimate their clinical significance, we utilized the Kaplan–Meier analysis of survival which measures the fraction of patients living for a certain amount of time after treatment, correlated to the expression of particular genes.

This preliminary study investigated the potential role of REST in prostate cancer progression. It is expected that during the neuroendocrine differentiation, REST may help to repress the expression of several genes not yet required by the differentiation program. And, REST may function different ways at different stages of cell differentiation as changing the expression of associated genes. Thus, further studies are needed to understand these problems.

We identified high consistency between microarray and RNA-seq platforms, thus encouraging the continual use of this method as a tool for differential gene expression

48 analysis. In conclusion, our study contributes to the elucidation of the basic biological mechanism that drives adenocarcinoma to escape AR blockade through NED to androgen independent and aggressive NEPC. Furthermore, REST may serve as a potential target for the treatment of prostate cancer and a novel prognosis factor for prostate cancer.

References

1. Huggins, C. and P. J. Clark, Quantitative Studies of Prostatic Secretion : Ii. The Effect of Castration and of Estrogen Injection on the Normal and on the Hyperplastic Prostate Glands of Dogs. J Exp Med, 1940. 72(6): p. 747‐62. 2. Nelson, E.C., et al., Clinical implications of neuroendocrine differentiation in prostate cancer. Prostate Cancer Prostatic Dis, 2007. 10(1): p. 6‐14. 3. Cindolo, L., et al., Neuroendocrine differentiation in prostate cancer: from lab to bedside. Urol Int, 2007. 79(4): p. 287‐96. 4. Tawadros, T., et al., IB1/JIP‐1 controls JNK activation and increased during prostatic LNCaP cells neuroendocrine differentiation. Cell Signal, 2005. 17(8): p. 929‐39. 5. Hayward, S.W. and G.R. Cunha, The prostate: development and physiology. Radiol Clin North Am, 2000. 38(1): p. 1‐14. 6. Isaacs, J.T., Role of androgens in prostatic cancer. Vitam Horm, 1994. 49: p. 433‐502. 7. Cunha, G.R., et al., Hormonal, cellular, and molecular regulation of normal and neoplastic prostatic development. J Steroid Biochem Mol Biol, 2004. 92(4): p. 221‐36. 8. Hanahan, D. and R.A. Weinberg, Hallmarks of cancer: the next generation. Cell, 2011. 144(5): p. 646‐74. 9. Kasper, S., Exploring the origins of the normal prostate and prostate cancer stem cell. Stem Cell Rev, 2008. 4(3): p. 193‐201. 10. Mueller, M.M. and N.E. Fusenig, Friends or foes ‐ bipolar effects of the tumour stroma in cancer. Nat Rev Cancer, 2004. 4(11): p. 839‐49. 11. Narayan, P., et al., The role of transrectal ultrasound‐guided biopsy‐based staging, preoperative serum prostate‐specific antigen, and biopsy Gleason score in prediction of final pathologic diagnosis in prostate cancer. Urology, 1995. 46(2): p. 205‐12. 12. Gleason, D.F. and G.T. Mellinger, Prediction of prognosis for prostatic adenocarcinoma by combined histological grading and clinical staging. J Urol, 1974. 111(1): p. 58‐64. 13. Thompson, I.M., et al., Prevalence of prostate cancer among men with a prostate‐specific antigen level < or =4.0 ng per milliliter. N Engl J Med, 2004. 350(22): p. 2239‐46. 14. Bok, R.A. and E.J. Small, Bloodborne biomolecular markers in prostate cancer development and progression. Nat Rev Cancer, 2002. 2(12): p. 918‐26. 15. Roy, A.K., et al., Regulation of androgen action. Vitam Horm, 1999. 55: p. 309‐52. 16. Visakorpi, T., et al., In vivo amplification of the androgen receptor gene and progression of human prostate cancer. Nat Genet, 1995. 9(4): p. 401‐6. 17. Xu, Y., et al., Androgens induce prostate cancer cell proliferation through mammalian target of rapamycin activation and post‐transcriptional increases in cyclin D proteins. Cancer Res, 2006. 66(15): p. 7783‐92. 18. Wang, Q., et al., Androgen receptor regulates a distinct transcription program in androgen‐independent prostate cancer. Cell, 2009. 138(2): p. 245‐56. 19. Perner, S., et al., TMPRSS2:ERG fusion‐associated deletions provide insight into the heterogeneity of prostate cancer. Cancer Res, 2006. 66(17): p. 8337‐41. 20. Tomlins, S.A., et al., Role of the TMPRSS2‐ERG gene fusion in prostate cancer. Neoplasia, 2008.

10(2): p. 177‐88. 21. Hammad, F.T., Radical prostatectomy. Ann N Y Acad Sci, 2008. 1138: p. 267‐77. 22. Mendenhall, W.M., R.H. Henderson, and N.P. Mendenhall, Definitive radiotherapy for prostate cancer. Am J Clin Oncol, 2008. 31(5): p. 496‐503. 23. Antonarakis, E.S., Abiraterone acetate for prostate cancer: a new era of hormonal therapies. Asian J Androl, 2011. 13(5): p. 663‐4. 24. Jemal, A., et al., Global cancer statistics. CA Cancer J Clin, 2011. 61(2): p. 69‐90. 25. Altavilla, A., et al., Medical strategies for treatment of castration resistant prostate cancer (CRPC) docetaxel resistant. Cancer Biol Ther, 2012. 13(11): p. 1001‐8. 26. Reid, A.H., et al., CYP17 inhibition as a hormonal strategy for prostate cancer. Nat Clin Pract Urol, 2008. 5(11): p. 610‐20. 27. Khan, M.O. and M.H. Ather, Chromogranin A‐‐serum marker for prostate cancer. J Pak Med Assoc, 2011. 61(1): p. 108‐11. 28. Beltran, H., et al., Molecular characterization of neuroendocrine prostate cancer and identification of new drug targets. Cancer Discov, 2011. 1(6): p. 487‐95. 29. Schoenherr, C.J. and D.J. Anderson, The neuron‐restrictive silencer factor (NRSF): a coordinate repressor of multiple neuron‐specific genes. Science, 1995. 267(5202): p. 1360‐3. 30. Chong, J.A., et al., REST: a mammalian silencer protein that restricts sodium channel gene expression to neurons. Cell, 1995. 80(6): p. 949‐57. 31. Cowan, J.M., J. Powers, and A.S. Tischler, Assignment of the REST gene to 4q12 by fluorescence in situ hybridization. Genomics, 1996. 34(2): p. 260‐2. 32. Ballas, N. and G. Mandel, The many faces of REST oversee epigenetic programming of neuronal genes. Curr Opin Neurobiol, 2005. 15(5): p. 500‐6. 33. Palm, K., et al., Neuronal expression of zinc finger transcription factor REST/NRSF/XBR gene. J Neurosci, 1998. 18(4): p. 1280‐96. 34. Palm, K., M. Metsis, and T. Timmusk, Neuron‐specific splicing of zinc finger transcription factor REST/NRSF/XBR is frequent in neuroblastomas and conserved in human, mouse and rat. Brain Res Mol Brain Res, 1999. 72(1): p. 30‐9. 35. Coulson, J.M., et al., A splice variant of the neuron‐restrictive silencer factor repressor is expressed in small cell lung cancer: a potential role in derepression of neuroendocrine genes and a useful clinical marker. Cancer Res, 2000. 60(7): p. 1840‐4. 36. Shimojo, M., et al., Protein kinase A regulates cholinergic gene expression in PC12 cells: REST4 silences the silencing activity of neuron‐restrictive silencer factor/REST. Mol Cell Biol, 1999. 19(10): p. 6788‐95. 37. Mori, N., et al., A common silencer element in the SCG10 and type II Na+ channel genes binds a factor present in nonneuronal cells but not in neuronal cells. Neuron, 1992. 9(1): p. 45‐54. 38. Bruce, A.W., et al., Genome‐wide analysis of repressor element 1 silencing transcription factor/neuron‐restrictive silencing factor (REST/NRSF) target genes. Proc Natl Acad Sci U S A, 2004. 101(28): p. 10458‐63. 39. Wu, J. and X. Xie, Comparative sequence analysis reveals an intricate network among REST, CREB and miRNA in mediating neuronal gene expression. Genome Biol, 2006. 7(9): p. R85. 40. Mortazavi, A., et al., Comparative genomics modeling of the NRSF/REST repressor network: from single conserved sites to genome‐wide repertoire. Genome Res, 2006. 16(10): p. 1208‐21. 51

41. Otto, S.J., et al., A new binding motif for the transcriptional repressor REST uncovers large gene networks devoted to neuronal functions. J Neurosci, 2007. 27(25): p. 6729‐39. 42. Allfrey, V.G., R. Faulkner, and A.E. Mirsky, Acetylation and Methylation of Histones and Their Possible Role in the Regulation of Rna Synthesis. Proc Natl Acad Sci U S A, 1964. 51: p. 786‐94. 43. Lachner, M. and T. Jenuwein, The many faces of histone lysine methylation. Curr Opin Cell Biol, 2002. 14(3): p. 286‐98. 44. Ooi, L. and I.C. Wood, Chromatin crosstalk in development and disease: lessons from REST. Nat Rev Genet, 2007. 8(7): p. 544‐54. 45. Tachibana, M., et al., Set domain‐containing protein, G9a, is a novel lysine‐preferring mammalian histone methyltransferase with hyperactivity and specific selectivity to lysines 9 and 27 of histone H3. J Biol Chem, 2001. 276(27): p. 25309‐17. 46. Zhang, Q., D.W. Piston, and R.H. Goodman, Regulation of corepressor function by nuclear NADH. Science, 2002. 295(5561): p. 1895‐7. 47. Ooi, L. and I.C. Wood, Regulation of gene expression in the nervous system. Biochem J, 2008. 414(3): p. 327‐41. 48. Rossbach, M., Non‐Coding RNAs in Neural Networks, REST‐Assured. Front Genet, 2011. 2: p. 8. 49. Lepagnol‐Bestel, A.M., et al., DYRK1A interacts with the REST/NRSF‐SWI/SNF chromatin remodelling complex to deregulate gene clusters involved in the neuronal phenotypic traits of Down syndrome. Hum Mol Genet, 2009. 18(8): p. 1405‐14. 50. Zuccato, C., et al., Huntingtin interacts with REST/NRSF to modulate the transcription of NRSE‐controlled neuronal genes. Nat Genet, 2003. 35(1): p. 76‐83. 51. Huang, Z. and S. Bao, Ubiquitination and deubiquitination of REST and its roles in cancers. FEBS Lett, 2012. 586(11): p. 1602‐5. 52. Westbrook, T.F., et al., A genetic screen for candidate tumor suppressors identifies REST. Cell, 2005. 121(6): p. 837‐48. 53. Lawinger, P., et al., The neuronal repressor REST/NRSF is an essential regulator in medulloblastoma cells. Nat Med, 2000. 6(7): p. 826‐31. 54. Fuller, G.N., et al., Many human medulloblastoma tumors overexpress repressor element‐1 silencing transcription (REST)/neuron‐restrictive silencer factor, which can be functionally countered by REST‐VP16. Mol Cancer Ther, 2005. 4(3): p. 343‐9. 55. Kreisler, A., et al., Regulation of the NRSF/REST gene by methylation and CREB affects the cellular phenotype of small‐cell lung cancer. Oncogene, 2010. 29(43): p. 5828‐38. 56. Lapuk, A.V., et al., From sequence to molecular pathology, and a mechanism driving the neuroendocrine phenotype in prostate cancer. J Pathol, 2012. 227(3): p. 286‐97. 57. Wang, Y., et al., Cell differentiation lineage in the prostate. Differentiation, 2001. 68(4‐5): p. 270‐9. 58. Taylor, B.S., et al., Integrative genomic profiling of human prostate cancer. Cancer Cell, 2010. 18(1): p. 11‐22. 59. Shah, S.P., et al., Mutational evolution in a lobular breast tumour profiled at single nucleotide resolution. Nature, 2009. 461(7265): p. 809‐13. 60. Trapnell, C., L. Pachter, and S.L. Salzberg, TopHat: discovering splice junctions with RNA‐Seq. Bioinformatics, 2009. 25(9): p. 1105‐1111. 61. Anders, S. and W. Huber, Differential expression analysis for sequence count data. Genome 52

Biol, 2010. 11(10): p. R106. 62. Cullen, B.R., RNAi the natural way. Nat Genet, 2005. 37(11): p. 1163‐5. 63. Salvatori, L., et al., Cell‐to‐cell signaling influences the fate of prostate cancer stem cells and their potential to generate more aggressive tumors. PLoS One. 7(2): p. e31467.