Molecular Portraits of Non- Coding RNAs in Neuroblastoma

Transcription

1 Ghent University, Faculty of Medicine and Health Sciences Molecular Portraits of Non- Coding RNAs in Neuroblastoma this thesis is submitted as fulfilment of the requirements for the degree of Doctor in Biomedical Sciences by Pieter Mestdagh, 2011 promoter prof. dr.ir. Jo Vandesompele co- promoter prof. dr. Frank Speleman Center for Medical Genetics Ghent University Hospital, Medical Research Building De Pintelaan 185, 9000 Gent, Belgium

2 II

3 Thesis submitted to fulfil the requirements for the degree of Doctor in Biomedical Sciences III Promoter prof. dr. ir. Jo Vandesompele (Ghent University, Belgium) Co- promoter prof. dr. Frank Speleman (Ghent University, Belgium) Members of the examination committee: prof. dr. Mark Bracke (Ghent University, Belgium) prof. dr. Yves Van De Peer (Ghent University, Belgium) prof. dr. Jan Cools (Catholic University of Leuven, Belgium) prof. dr. Michel Georges (University of Liège, Belgium) dr. Pieter Rondou (Ghent University, Belgium) prof. dr. Ray Stallings (Royal College of Surgeons, Ireland) prof. dr. Jason Shohet (Baylor College, USA) De auteur en de promotoren geven de toelating deze scriptie voor consultatie beschikbaar te stellen en delen ervan te kopiëren voor persoonlijk gebruik. Elk ander gebruik valt onder de beperkingen van het auteursrecht, in het bijzonder met betrekking tot de verplichting uitdrukkelijk de bron te vermelden bij het aanhalen van resultaten uit deze scriptie. The author and the promoters give the permission to use this thesis for consultation and to copy parts of it for personal use. Every other use is subject to the copyright law, more specifically the source must be extensively specified when using results from this thesis. The research described in this thesis was conducted at the Centre for Medical Genetics, Ghent University Hospital, Ghent, Belgium This work was supported by the Ghent University Research fund (BOF 01D31406), the Fund for Scientific Research (grant number: G and ), the Belgian Kid s Fund, the Stichting tegen Kanker and GOA (01G01910). This article represents research results of the Belgian program of Interuniversity Poles of Attraction, initiated by the Belgian State, Prime Minister s Office, Science Policy Programming.

4 IV

5 Table of Contents LIST OF ABBREVIATIONS INTRODUCTION NEUROBLASTOMA NEUROBLASTOMA GENETICS MYCN THE MYC FAMILY TRANSCRIPTIONAL CONTROL ONE ONCOGENE, TWO FACES TARGETS TO TARGET NON- CODING RNAS MESSENGERS WITHOUT A MESSAGE THE BIOGENESIS OF MIRNAS ONCOMIRS AND TUMOUR SUPPRESSOR MIRNAS MIRNAS REGULATED BY MYC MIR- 34A: THE MISSING PIECE IN THE P53 NETWORK PUZZLE MIRNAS AND METASTASIS MECHANISMS OF DEREGULATED MIRNA EXPRESSION MIRNA SIGNATURES FOR IMPROVED DIAGNOSTIC AND PROGNOSTIC CLASSIFICATION MIRNA THERAPEUTICS QUANTIFICATION OF MIRNA EXPRESSION EXPLORING MIRNA FUNCTION FUNCTIONAL EXPLORATION OF GENE EXPRESSION PATTERNS REFERENCES RESEARCH OBJECTIVES RESULTS V PAPER 1: HIGH- THROUGHPUT STEM- LOOP RT- QPCR MIRNA EXPRESSION PROFILING USING MINUTE AMOUNTS OF INPUT RNA 30 PAPER 2: A NOVEL AND UNIVERSAL METHOD FOR MICRORNA RT- QPCR DATA NORMALIZATION 42 PAPER 3: MYCN/C- MYC- INDUCED MICRORNAS REPRESS CODING GENE NETWORKS ASSOCIATED WITH POOR OUTCOME IN MYCN/C- MYC- ACTIVATED TUMOURS 65 PAPER 4: THE MIR MICRORNA CLUSTER REGULATES MULTIPLE COMPONENTS OF THE TGF- Β PATHWAY IN NEUROBLASTOMA 101 PAPER 5: AN INTEGRATIVE GENOMICS SCREEN UNCOVERS NCRNA T- UCR FUNCTIONS IN NEUROBLASTOMA TUMOURS 132 PAPER 6: THE MICRORNA BODY MAP: DISSECTING MICRORNA FUNCTION THROUGH INTEGRATIVE GENOMICS 149 PAPER 7: OUTCOME PREDICTION OF CHILDREN WITH NEUROBLASTOMA USING MIRNA AND MRNA GENE EXPRESSION SIGNATURES 168 DISCUSSION AND FUTURE PERSPECTIVES REFERENCES SUMMARY SAMENVATTING CURRICULUM VITAE

6

7 1 List of abbreviations INSS: International Neuroblastoma Staging System INRGSS: International Neuroblastoma Risk Group Staging System GWAS: genome- wide association study SNP: single nucleotide polymorphism shrna: short- hairpin RNA DFMO: alpha- difluoromethylornithine ncrna: non- coding RNA mirna: micro RNA pirna: PIWI- interacting RNA endo- sirna: endogenous short- interfering RNA T- UCR: transcribed ultraconserved region snorna: small nucleolar RNA lncrna: long non- coding RNA pri- mirna: primary mirna transcript pre- mirna: premature mirna transcript RISC: RNA- induced silencing complex CLL: chronic lymphocytic leukemia CAGR: cancer- associated genomic region AML: acute myeloid leukemia RT- qpcr: reverse transcription quantitative polymerase chain reaction RIP- Chip: Ribonucleoprotein ImmunoPrecipitation- gene Chip HITS- CLIP: high- throughput sequencing of RNA isolated by crosslinking immunoprecipitation GO: gene ontology GSEA: gene set enrichment analysis LNA: locked nucleic acid

8 Introduction 2 Neuroblastoma Neuroblastoma is a childhood malignancy that accounts for 15% of pediatric cancer mortality 1. It is also the most frequently diagnosed extracranial tumour in children and is characterized by a remarkable heterogeneity, both in terms of clinical behavior and genetic aberrations. Despite intensive multimodal therapy, the cure- rate for high- risk patients is lower than 40%. A major determinant of the clinical course is the age of the patient at diagnosis. Typically, children older than 1 year have metastatic disease and a poor overall survival while infants present with localized tumours that can mature into a benign ganglioneuroma or regress spontaneously 2. In the 1990 s, the first therapeutic stratification system for neuroblastoma patients was based on age at diagnosis and a number of post- surgical parameters (International Neuroblastoma Staging System, INSS) 3, (Figure 1). Because surgical approaches differ between institutions, a uniform staging system (International Neuroblastoma Risk Group Staging System, INRGSS), based on an international consensus, was recently proposed 4. The INRGSS combines a series of non- surgical parameters - such as age at diagnosis and radiographic characteristics of the tumour - with several biological factors and forms the current basis for risk- related therapies. Neuroblastoma tumours originate from precursor cells of the sympathoadrenal lineage. Hence, tumours can develop anywhere in the sympathetic nervous system. At least half of the primary tumours are found in the adrenal medulla while others originate in the paraspinal sympathetic ganglia or in pelvic ganglia. Neuroblastoma genetics Most neuroblastomas, as most cancers, are supposed to result from accumulating somatic mutations. However, a limited fraction of neurobastoma tumours (1-2%) are inherited in an autosomal dominant fashion 5-7. Germline mutations in PHOX2B, a key regulator in the process of nervous system development, account for a small subset of hereditary cases of neuroblastoma 8. Recently, genome- wide linkage analysis of neuroblastoma pedigrees identified ALK as the major familial neuroblastoma predisposition gene 9. ALK mutations target the tyrosine receptor kinase domain, resulting in a constitutive phosphorylation, sufficient to drive oncogenic transformation 10. Unlike PHOX2B, ALK is also mutated in a substantial portion (6.9%) of sporadic cases 10 implicating it as a putative target for molecular therapy. Earlier, genome- wide linkage analysis in neuroblastoma pedigrees identified a region on chromosome 16p and one on 4p suggesting that additional hereditary neuroblastoma predisposition genes exist 11, 12. In an attempt to pinpoint genetic events associated with susceptibility to aggressive neuroblastoma in sporadic cases, Maris and colleagues performed a genome- wide association study (GWAS) of single nucleotide polymorphisms (SNPs) and identified common SNPs at three different loci, i.e. 6p22, BARD1 and LMO LMO1 copy number analysis of primary tumour cells demonstrated copy number gain in 12.4% of the cases. Both the LMO1 germline SNP and the somatic copy number gain were associated with increased LMO1 expression enhancing cell proliferation. Neuroblastoma tumours are characterized by a high number of somatically acquired genomic alterations. Typically, these genetic aberrations target large genomic regions encompassing hundreds of genes. Regions that are frequently associated with copy number loss include chromosome 1p, chromosome 3p, chromosome 6q and chromosome 11q while regions that are associated with copy number gain include chromosome 1q and chromosome 17q 16 (Figure 2). Several of these genomic aberrations serve as prognostic markers for risk stratification and are used to classify neuroblastoma tumours in genetic subgroups However, the large number of genes residing within these regions has thwarted researchers in their search for tumour driving events underlying the individual genomic aberrations. One way to identify tumour- driving events relies on the reverse genetics approach where short- hairpin RNA (shrna) libraries are used to perturb the expression of a high number of genes followed by a relevant phenotypic read- out. Hölzel and colleagues applied such shrna

9 b a 3 4S Figure 1 International neuroblastoma staging system. Localization of primary and metastatic tumours for each disease stage as defined by the international neuroblastoma staging system (INSS). Stage 1: Localised tumour with complete gross excision, with or without microscopic residual disease; representative ipsilateral lymph nodes negative for tumour microscopically. Stage 2a: Localised tumour with incomplete gross excision; representative ipsilateral non-adherent lymph nodes negative for tumour microscopically. Stage 2b: Localised tumour with or without complete gross excision, with ipsilateral non-adherent lymph nodes positive for tumour. Enlarged contralateral lymph nodes should be negative microscopically. Stage 3: Unresectable unilateral tumour infiltrating across the midline, with or without regional lymph node involvement; or localised unilateral tumour with contralateral regional lymph node involvement; or midline tumour with bilateral extension by infiltration (unresectable) or by lymph node involvement. Stage 4: Any primary tumour with dissemination to distant lymph nodes, bone, bone marrow, liver, skin, or other organs (except as defined by stage 4S). Stage 4S: Localised primary tumour in infants younger than 1 year (as defined for stage 1, 2A, or 2B), with dissemination limited to skin, liver, or bone marrow. Metastatic sites: (1) liver, (2) bone, (3) bone marrow, (4) distant lymph node, (5) skin. (source: and Maris, ) library to identify genes modulating neuroblastoma response to retinoic acid, a differentiation- inducing agent 21. They identified a crosstalk between the tumour suppressor NF1 and retinoic acid- induced differentiation and found NF1 microdeletions and mutations in neuroblastoma tumours. Often, tumour- driving genetic events are identified through focal genomic aberrations such as amplification of the ERBB2 oncogene in breast cancer 22 or homozygous deletion of the PTEN tumour suppressor in prostate cancer 23. Homozygous deletions have been described in neuroblastoma, amongst others for CDKN2A 24 and NF1 25, but are rare events. In contrast, several amplicons have been identified in neuroblastoma, encompassing known oncogenes such as MDM2 26, 27, ALK 28 and most importantly MYCN 29. Amplification of the MYCN oncogene occurs in 20-25% of neuroblastoma tumours and delineates a subgroup of patients with highly aggressive, metastatic disease and poor outcome. Tumours with MYCN amplification are cytogenetically characterized by double minute chromatin bodies or homogenous staining regions 30, 31, two typical manifestations of gene amplification. Although the amplicon can contain multiple genes, MYCN is the only gene that shows consistent amplification 32. In general, tumours with MYCN amplification express MYCN at much higher levels than tumours without MYCN amplification. However, it remains controversial whether MYCN overexpression is correlated to survival in tumours lacking MYCN

10 4 amplification 33, 34. Most tumours with MYCN amplification also have a deletion of chromosome 1p but not all tumours with a 1p- deletion have MYCN amplification suggesting that deletion of chromosome 1p occurs earlier in tumour development 2. Possibly, genes negatively regulating MYCN expression or inducing apoptosis in the presence of high MYCN levels need to be deleted in order for MYCN amplification to occur. The assumption that MYCN amplification is a tumour driving event in neuroblastoma was put to the test by Weiss and colleagues who developed transgenic mice that overexpress MYCN in sympathetic nervous system cells 35. The mice developed neuroblastoma- like tumours with chromosomal gains and losses that are syntenic with comparable abnormalities detected in human neuroblastoma tumours. These results suggest that MYCN amplification indeed contributes to the genesis of neuroblastoma tumours. MYCN The MYC family The MYC- family of proto- oncogenes or MYC- box genes encodes three different transcription factors: MYC, MYCN and MYCL. MYC (c- MYC) was originally isolated from chicken DNA due to its homology with the v- myc oncogene from the avian myelocytomatosis virus 36. MYC activation occurs either through chromosomal translocation, placing the MYC open reading frame under the control of a constitutively active promoter 37, 38, or through gene amplification 39, 40. MYCN and MYCL were detected as amplified DNA fragments with partial homology to MYC in neuroblastoma and small cell lung cancer respectively 29, 41. While MYC is activated in a wide variety of tumours, MYCN activation has a strong preference for tumours of neuroectodermal origin such as neuroblastoma, retinoblastoma and peripheral neuroectodermal tumours. Human tumour cells almost invariably express MYC but, depending on their derivation of embryonal cell lineage, hardly ever express MYCN 42. Analysis of tissues from different stages of the fetal and the developing mouse embryo further support these observations by showing that, during embryonal development, the overall expression pattern of MYC is relatively constant whereas MYCN displays a temporal and spatial expression pattern. Strikingly, Malynn and colleagues have shown that MYCN can functionally replace MYC in murine growth, development and differentiation 43. Mice in which the MYC coding sequences was replaced with the MYCN coding sequence were shown to survive into adulthood and reproduce with MYCN expression being similarly regulated as MYC expression. This functional redundancy suggests that MYC and MYCN can regulate identical cellular processes and that the MYC- family was evolved to facilitate differential expression patterns of the MYC- family genes. Of note, a reverse experiment where MYC replaces MYCN has not been reported so far. Transcriptional control MYCN encodes a nuclear protein of approximately 65 kda that is comprised of a basic DNA binding region (b), an α- helical protein protein interaction domain known as helix loop helix (HLH), a leucine zipper motif (Z) and an N- terminal transactivation domain containing two MYC- boxes termed MBI and MBII, all of which are conserved between the different MYC- family members. Through their bhlhz domain, MYC- family proteins form heterodimers with another bhlhz- protein called MAX and MYC(N)/MAX heterodimers have DNA binding and transcriptional activity 45 (Figure 3A). While MAX proteins can also homodimerize, these complexes are believed to be transcriptionally inert. MYC/MAX heterodimers specifically recognize and bind the hexameric DNA element CACGTG belonging to the larger class of E- box elements (CANNTG). Importantly, MYC proteins fail to activate transcription at E- box elements in the absence of MAX 46. MAX can also form heteroduplexes with bhlhz proteins from the MAD- family (MAD1, MXI1, MAD3 and MAD4).

11 5 mir-34a MYCN AMP ALKAMP, MUT PHOX2B MUT BARD1 GWAS SNP rs GWAS LMO1 GWAS MDM2 AMP NF1 HDEL AMP Amplification HDEL Homozygous deletion MUT Mutation GWAS Identified through GWAS chromosomal loss chromosomal gain Figure 2 Genetic defects in neuroblastoma tumours. Schematic representation of common genetic aberrations and genes associated with neuroblastoma. This non-exhaustive overview contains genes and chromosomal regios that are discussed in this thesis. Chromosomal regions of copy number gain are indicated in red, regions of loss are indicated in blue. Genes that are amplified, homozygously deleted, mutated or identified through GWAS are indicated by AMP, HDEL, MUT and GWAS respectively. (Adapted from Van Roy et al., )

12 Upon heteroduplex formation, MAD/MAX complexes bind the E- box consensus sequence where they act as transcriptional repressors hereby antagonizing transactivation by MYC proteins. Apart from a few examples, transcriptional repression by MYC proteins is believed to be independent of E- box binding and, compared to transcriptional activation, the underlying mechanisms are less well understood. Different models have been proposed that might explain MYC- induced transcriptional repression 47. One model suggests that MYC activates protein- coding transcriptional repressors and thereby acts indirectly on promoter elements. However, MYC mediated transcriptional repression has been observed in the presence of inhibitors of protein synthesis suggesting that repression does not require the synthesis of an intermediate protein 48. Second, repression might also be mediated through a direct binding of MYC to sequences in the target gene promoter as was shown for CDKN1B 49 (Figure 3B). As only few examples of direct MYC binding were reported, MYC- mediated repression is likely to occur through alternative mechanisms. A third possibility is that MYC is recruited to core promoter elements through protein- protein interactions without directly binding to DNA. Several candidate proteins, such as YY1, TFII- I, SP1 and MIZ- 1, have been proposed to target MYC to core promoter elements and repression of different protein- coding genes has indeed been shown to depend on the interaction between MYC and MIZ or SP1 48 (Figure 3C). In the results section of this work, a second mechanism explaining MYC mediated transcriptional repression of coding genes is presented. This mechanism relies on the induction of mirna expression by MYC- family proteins. MYC proteins have been thought to function only as classic transcription factors that bind to well defined regions in the promoters of specific target genes. Recent studies now suggest that MYC functions reach far beyond that of a classical bhlhz protein. In contrast to a classic bhlhz protein, MYC proteins potentially regulate up to 15% of all human protein- coding genes 53. Although widespread, the magnitude of transcriptional regulation of most target genes by MYC is nonetheless relatively modest compared with other bhlhz family members and transcription factors in general 54. In addition, MYC proteins have been shown to influence global chromatin state by regulating both acetylation and methylation of several histones 55, 56. These findings suggest a new model in which MYC genes function both locally and globally. 6 A MAX MAX B MAX MYC MAX MAX MAX MAD MYC MNT!"#$%& C '()"*+*,*-.& MAX MIZ-1 MYC /01*&210,0.*1& Figure 3 The MYC-MAX-MAD network. (A) Schematic representation of MAX-interaction partners and their effect on gene transcription upon E-box binding. (B-C) Models for MYC-mediated repression of gene transcription, either through binding of MYC-MAX complexes to INR-elements or through interaction with MIZ-1 at core promoters.

13 7 One oncogene, two faces One of the most important functions of MYC proteins is the regulation of cell growth and proliferation (Figure 4). Several MYC target genes are thought to play a role in the ability of MYC to promote cellular growth, including those associated with cellular metabolism, ribosomal and mitochondrial biogenesis, and protein and nucleic acid synthesis 57. MYC is also known to stimulate glycolysis through increased glucose transport and regulation of glycolytic genes resulting in the production of lactate under aerobic conditions 58, 59. This phenomenon is called the Warburg effect and is active in many cancer types. Cells with activated MYC often display a shortened G1 cell cycle phase and accelerated cell cycle progression. This is explained by MYC s ability to repress cell cycle checkpoint genes and inhibit cyclin dependent kinase inhibitors such as CDKN1A. MYC can also promote cell cycle progression by activating cyclines (CCND1, CCNE1), cyclin dependent kinases (CDK4) and members of the E2F transcription family. Through these mechanisms, deregulated MYC activity results in increased cell proliferation making cells vulnerable to additional oncogenic events that further accelerate tumourigenesis. MYC is also involved in triggering an angiogenic switch, amongst others through repression of THBS1 and CTGF 60, 61 and has been shown to promote chromosomal instability 62. Next to its role as a driver of oncogenesis, MYC can also trigger intrinsic tumour suppressor mechanisms including apoptosis and cellular senescence (Figure 4). Therefore, secondary mutations targeting components of the MYC tumour suppressor axis are under strong selective pressure during MYC- driven tumourigenesis 58. MYC sensitizes cells for apoptosis through two different mechanisms. One mechanism involves the induction of P14ARF by MYC, which in turn results in the activation of TP53 and transcription of pro- apoptotic genes such as BAX and PUMA 57. Alternatively, MYC directly represses expression of members from the anti- apoptotic BCL2- family such as BCL2 and BCLX that regulate BAX. The role of MYC in regulating apoptosis is further supported by studies in MYC- null cells showing that, in the absence of MYC, cells are resistant to apoptotic stimuli 57, 63. MYC- induced senescence is depending on an intact Arf- p53- p21 and p16ink4a- prb axis and the absence of protective factors such as Wrn and Cdk2 58. This was demonstrated in Eµ- MYC mice where lymphoma development was delayed in mice lacking functional Cdk2 due to the induction of senescence 64. However, in the presence of Cdk2, MYC is capable of repressing senescence. When overexpressed, many oncogenes such as BRAF and RAS, induce senescence in primary cells. This process is called oncogene- induced senescence and can be repressed by activated MYC, amongst others in melanomas with BRAF activation 65 and rat fibroblasts with RAS overexpression 66. Targets to target The high frequency of tumours with activation of a MYC family member and the plethora of functions that are controlled by MYC proteins suggests that MYC proteins are key drivers of tumourigenesis. In principle, this makes inhibition of MYC proteins an attractive pharmacological approach for cancer treatment. However, practical difficulties in designing MYC inhibitory drugs and concerns about possible side effects in proliferating normal tissues has tempered the enthusiasm 67. Therefore, researchers are trying to identify MYC target genes that are crucial in the process of MYC- driven tumourigenesis and that can be targeted pharmacologically. In neuroblastoma, several MYCN target genes have been identified that meet these requirements. For example, Slack and colleagues demonstrated that MDM2, a critical negative regulator of TP53, is directly activated by MYCN 68. MDM2 is an E3 ubiquitin ligase that negatively regulates p53 activity and stability by binding to its transactivation domain, hereby promoting its ubiquitination and degradation 69. Tumour cells with MYCN amplification overexpress MDM2 to escape TP53- mediated cell death, proliferate, and progress to invasive malignancy 68. As most neuroblastoma tumours have an intact TP53 signaling pathway, targeted disruption of the MDM2- TP53 interaction using the small- molecule nutlin- 3 activates the TP53 pathway resulting in an antiproliferative and cytotoxic effect 69, 70.

14 8 BCL2 CDKN2A APOPTOSIS MYC/MYCN E2F1/2/3 CDK4 CDKN1A metabolism nucleic acid synthesis THBS1 CTGF PROLIFERATION CELL GROWTH ANGIOGENESIS Figure 4 Opposing functions of the MYC family. MYC genes have the ability to induce apoptosis (upper part) or to stimulate cell growth, proliferation and tumour vascularization (lower part). Another attractive MYCN target gene is ODC1, a rate- limiting enzyme in polyamine biosynthesis that was shown to be upregulated in MYCN amplified neuroblastoma tumours 71. As polyamines are essential for cell survival and cancer progression, the authors evaluated the use of the ODC1 inhibitor alpha- difluoromethylornithine (DFMO) on neuroblastoma cell proliferation in vitro and in vivo. DFMO treatment was shown to inhibit proliferation of neuroblastoma cells in culture and prevented MYCN- induced oncogenesis in vivo. Of note, high- risk neuroblastoma tumours without MYCN amplification also overexpress ODC1 when compared to low- risk tumours suggesting that therapeutic treatment with DFMO is of potential relevance for the entire population of high- risk neuroblastoma patients. Westermann and colleagues defined a core set of MYC/MYCN target genes that show similar expression patterns to that of ODC1. They found that overexpression of these genes is predominantly driven by MYC in high- risk tumours without MYCN amplification and by MYCN in MYCN amplified tumours 72. These findings suggest that high- risk neuroblastoma tumours are characterized by a MYC/MYCN driven transcriptional program that is independent of the MYCN amplification status. Non- coding RNAs Messengers without a message The study of coding genes has been heavily pursued, amongst others through the use of high- throughput platforms enabling the profiling of alterations in entire (epi)genomes and transcriptomes, yielding insights into the complexity of cancer and disease biology in general. Recently, a complete new and unexpected piece of the transcriptome puzzle has emerged. In contrast to what has long been thought, protein- coding

15 genes only account for a fraction of the genomic DNA that is transcribed in a human cell. Genome- wide studies have shown that the vast majority of the mammalian genome is transcribed and produces many thousands of regulatory non- protein- coding RNAs (ncrnas). According to their size and function, ncrnas are divided into different classes including the small micrornas (mirnas), PIWI interacting RNAs (pirnas), endogenous small- interfering RNAs (endo- sirnas), small nucleolar RNAs (snornas), promoter- associated RNAs (PARs) and the longer transcribed ultraconserved regions (T- UCRs) and long non- coding RNAs 73, 74 (lncrnas) (Table 1). Of the classes identified until now, mirnas have been most thoroughly investigated. The discovery that mirnas regulate gene expression and protein translation is one of the most exciting new findings in biological and medical sciences of the past decade, rightfully assigning the 2006 Nobel Prize in Physiology or Medicine to Andrew Fire and Craig Mello. Initially described in 1993, mirnas were thought to be unique to the tiny roundworm Caenorhabditis elegans in which they were shown to be implicated in the process of developmental regulation 75. Soon, homologues were identified in other organisms including humans. Currently, over 1000 human mirnas have been reported and more are awaiting experimental validation, making mirnas one of the largest classes of gene regulators Table 1. Overview of established non- coding RNA classes ncrna Long (regulatory) noncoding RNAs (lncrnas) Transcribed ultraconserved regions (T- UCRs) Small interfering RNAs (sirnas) micrornas (mirnas) PIWI-interacting RNAs (pirnas) Promoter-associated RNAs (PARs) Small nucleolar RNAs (snornas) Adapted from Taft et al., characteristics The broadest class, lncrnas, encompass all non-protein-coding RNA species > 200 nt, including mrna-like ncrnas. Their functions include epigenetic regulation, acting as sequence-specific tethers for protein complexes and specifying subcellular compartments or localization Specific class of long non-coding RNAs, transcribed from genomic regions that are at least 200 nt in length and 100% conserved between human, mouse and rat species. Their function and exact length is unknown. Small RNAs nt long, produced by Dicer cleavage of complementary dsrna duplexes. sirnas form complexes with Argonaute proteins and are involved in gene regulation, transposon control and viral defence Small RNAs 22 nt long, produced by Dicer cleavage of imperfect RNA hairpins encoded in long primary transcripts or short introns. They associate with Argonaute proteins and are primarily involved in post-transcriptional gene regulation Dicer-independent small RNAs nt long, principally restricted to the germline and somatic cells bordering the germline. They associate with PIWI-clade Argonaute proteins and regulate transposon activity and chromatin state A general term encompassing a suite of long and short RNAs, including promoterassociated RNAs (PASRs) and transcription initiation RNAs (tirnas) that overlap promoters and TSSs. These transcripts may regulate gene expression Traditionally viewed as guides of rrna methylation and pseudouridylation. However, there is emerging evidence that they also have gene-regulatory roles The biogenesis of mirnas Upon transcription of the mirna gene in the nucleus, primary mirna (pri- mirna) transcripts ( nucleotides) are formed and further processed by different endonucleases 77 (Figure 5). First, the pri- mirna transcripts are cleaved in the nucleus by Drosha into ~70 nucleotide precursors called premature mirnas (pre- mirnas). These hairpin precursors are exported into the cytoplasm by XPO5 where they are further processed by DICER1 into small imperfect double stranded RNA duplexes (mirna- mirna*) that contain both the mature mirna strand and its complementary strand (mirna*). The duplex is then loaded into the mirna- associated multiprotein RNA- induced silencing complex (mirisc) with retention of the mature mirna strand. DICER1, TARBP2 and Argonaute proteins mediate RISC assembly 78. The mature mirna guides the complex towards complementary sites in the target mrna to regulate gene expression.

16 10 Imperfect binding between the mirna and the mirna recognition element in the 3 UTR of the target gene ultimately results in degradation of the target mrna or inhibition of protein translation. Because of the ability to bind with incomplete complementarity, only part of the mirna sequence, the seed, is used to identify its target mrnas. This seed sequence encompasses bases 2-8 of the mature mirna and is established as important for biological function and stability. Recent bioinformatic studies argue that as much as 60% of all human protein- coding genes could be under the control of at least one mirna 79. This raises the possibility that mirnas control a large number of genetic pathways and that deregulated mirna expression contributes to disease, including cancer. nucleus cytoplasm AGUAGGUUGUAUAGUUGGA UCCAACUAUACAACCUACU GU Pri-miRNA Pre-miRNA DICER mirna-mirna* duplex AUAGAGGGUCACCACCCACACUGGGAUUGA U GAG UGGGA CUC UCCCA Unwind mature mirna mirna gene mirisc assembly Imperfect complementarity ORF Figure 5 mirna biogenesis pathway. Overview of the different steps in the processing of primary mirna transcripts to functional, mature mirnas by endonucleases. (adapted from Mestdagh et al., ) Oncomirs and tumour suppressor mirnas Although we have yet to learn most of the specific functions associated with a given mirna, at present we know mirnas are implicated in developmental timing, differentiation, cell proliferation, growth control, apoptosis and stem cell maintenance, all aspects of normal cell function known to be deregulated in cancer. One of the first direct links between particular mirnas and cancer came from observations by the group of Carlo Croce upon the investigation of chronic lymphocytic leukaemia (CLL), the most common form of adult leukaemia in the Western world. More than half of CLL cases present with deletions at the chromosome band 13q14, an abnormality that also occurs in 50% of mantle cell lymphomas, 16 40% of multiple myelomas, and in 60% of prostate cancers. Although strongly suggestive for tumour suppressor

17 Table 2. Overview of mirnas frequently inactivated in cancer mirna deregulation in cancer functions let-7 family Downregulated in lung, breast, gastric, ovary, prostate and colon cancers, chronic lymphoid leukemia, leiomyomas Represses cell proliferation and growth. Promotes angiogenesis. Negative regulation of oncoproteins RAS, MYC and HMGA2. mir-15a, mir-16-1 mir-29 family mir-34 family mir-143, mir-145 cluster mir-200 family mir-26a mir-125a, mir- 125b mir-101 Adapted from Spizzo et al., Downregulated in chronic lymphoid leukemia, diffuse large B-cell lymphoma, multiple myeloma, pituitary adenoma, prostate and pancreatic cancer. Germline mutations in B-cell chronic lymphoid leukemia patients. Downregulated in chronic lymphoid leukemia, colon, breast, and lung cancer, and cholangiocarcinomas Downregulated in pancreatic cancer and Burkitt s lymphoma without MYC translocation. Hypermethylation in colon cancer. Downregulated in neuroblastoma with 1p deletion Downregulated in colon adenoma/carcinoma, in breast, lung, and cervical cancer, in B cell malignancies Downregulated in clear-cell carcinoma, metastatic breast cancer Downregulated in hepatorcellular carcinoma, breast cancer, Burkitt lymphoma and anaplastic thyroid carcinoma. Downregulated by MYC Downregulated in glioblastoma, breast, prostate and ovarian cancer Downregulated in prostate cancer, hepatocellular carcinoma, bladder cancer and gastric cancer Induces apoptosis in leukemia cells by repression of BCL2. Regulates cell cycle by downregulating G0/G1 proteins. Induces aberrant methylation in lung cancer. Induces apoptosis. induces upregulation of p53, downregulation of E2F in colon cancer. Negative regulation of MYC and MYCN. Negative regulation of MYC and KRAS oncoproteins. Promotes invasion. Involved in TGFβ- mediated EMT. Negative regulation of ZEB1 and ZEB2 Induces apoptosis through negative regulation of MTDH and EZH2. Induces cell-cycle arrest associated with direct targeting of CCND2 and CCNE2. Negative regulation of ERBB2, ERBB3 and LIN28 oncoproteins Induces alterations in global chromatin structure via repression of EZH2. Inhibits proliferation, migration and invasion. Sensitizes tumour cells to radiation 11 activity, involvement of any of the protein- coding genes in the deleted region could not be demonstrated. Interestingly, two mirnas, mir- 15a and mir- 16-1, resided within a critically deleted 30- kb region at 13q14 and were reduced in expression in two thirds of CLL cases 81. Both mirnas were shown to negatively regulate BCL2, an anti- apoptotic gene frequently overexpressed in leukaemias, lymphomas and carcinomas. Down regulation of mir- 15a and mir is therefore believed to result in an increase of BCL2 expression and hence the inactivation of the intrinsic apoptosis pathway 82. The role of mir- 15a and mir as tumour suppressors in CLL is further supported by the presence of a pathogenic mutation in the mir- 15a mir genes in two patients, leading to a decreased mir expression 83. The sequence abnormalities that were identified in the mirna genes were not present in 160 normal control individuals and in several instances they were also found in DNA from normal cells of the patient. As CLL (as well as other cancers) is a disease with known familial occurrence (5% to 10% of patients have an inherited susceptibility to CLL), mirna mutations may be a predisposing factor to cancer, especially for those familial cases where the culprit genes are still unknown. Soon after this discovery, additional mirnas with a putative tumour suppressor function were described (Table 2). The mirnas that are encoded by the let- 7 family were shown to negatively regulate the expression of RAS through a direct interaction with its 3 - UTR 85. The RAS oncogene regulates proliferation and differentiation and is commonly mutated in human cancers, including lung cancer. When comparing lung tumours to normal adjacent cells, taken from patients

18 with squamous- cell carcinoma of the lung, let- 7 mirnas were found to be down regulated in the tumours whereas RAS was highly expressed. Besides RAS, let- 7 has also been shown to coordinate the repression of another growth promoting gene implicated in cancers, HMGA2 86. HMGA2 functions at the transcriptional level by altering chromatin structure and is primarily expressed in proliferating cells during embryogenesis and in a wide variety of tumours. In many of these tumours the HMGA2 open reading frame is truncated through chromosomal translocations resulting in a loss of the C- terminal domain. Together with loss of the C- terminal domain, such translocations also replace the 3 - UTR hereby disrupting the let- 7 coordinated repression of HMGA2 and promoting anchorage independent growth, a characteristic of oncogenic transformation. Moreover, transgenic mice that overexpressed wild- type Hmga2 had similar phenotypes to those expressing the truncated protein indicating that the disruption of a single mirna- target interaction can be sufficient to produce a clinical phenotype in vivo. 12 Table 3. Overview of mirnas frequently activated in cancer mirna deregulation in cancer functions mir cluster Overexpression in lung and colon cancer, lymphoma, multiple myeloma, medulloblastoma, neuroblastoma. Upregulated by MYC and MYCN Increases tumour growth and proliferation by negative regulation of cell cycle inhibitors. Promotes tumour angiogenesis through repression of THBS1 and CTGF. Anti-apoptotic activity through repression of BIM. Induces lymphoproliferative disease and autoimmunity. mir-106b-25 cluster mir-10b mir-21 mir-125a, mir- 125b mir-155 Overexpression in gastric, colon, and prostate cancer, neuroblastoma, multiple myeloma Overexpressed in metastatic breast cancer, nasopharyngeal carcinoma and malignant peripheral nerve sheath tumours Overexpression in glioblastoma, breast, lung, prostate, colon, stomach, esophageal, and cervical cancer, uterine leiomyosarcoma, diffuse large B-cell lymphoma, head and neck cancer Upregulation in myelodysplastic syndrome and acute myeloid leukemia with t(2;11)(p21;q23), urothelial carcinoma Overexpressed in pediatric Burkitt s lymphoma, Hodgkin s lymphoma, primary mediastinal lymphoma, diffuse large B-cell lymphoma, breast, lung, colon, pancreatic cancer Reduces apoptotic response after TGFβ stimulation via BIM. Increases tumour growth and proliferation by negative regulation of cell cycle inhibitors. Activates cell migration and extracellular matrix remodelling through negative regulation of HOXD10. Promotes tumourigenesis through negative regulation of NF1 Induces invasion and metastasis. Inhibits apoptosis through negative regulation of PDCD4. Inhibits negative regulators of the RAS/MEK/ERK pathway. Negative regulation of the TP53 tumour suppressor protein. Induces pre-b-cell proliferation, lymphoblastic leukemia and high-grade lymphoma. Promotes chemosensitivity through repression of FOXO3A. mir-26a Upregulated in glioblastoma Induces tumourigenesis by repression of PTEN, RB1, and MAP3K2/MEKK2 mir-181 family mir-221, mir-222 cluster mir-372, mir-373 cluster Adapted from Spizzo et al., Overexpressed in breast, pancreas, prostate cancer and hepatocellular carcinoma. Upregulated by MYCN in neuroblastoma Overexpressed in chronic lymphoid leukemia, thyroid papillary carcinoma, glioblastoma. Overexpression in testicular germ cell tumours, thyroid andenomas, esophageal cancer and metastatic breast cancer Enhances proliferation, migration and invasion. Promotes cancer cell proliferation. Impairs TRAIL-dependent response. Antagonizes p53-mediated CDK inhibition through repression of LATS2. Stimulates migration and invasion through negative regulation of CD44

19 13 Apart from tumour suppressor mirnas, several mirnas displaying oncogenic properties have also been identified (Table 3). The only mirna found to be overexpressed in almost any type of solid tumour (breast, colon, lung, prostate, stomach, pancreas, glioblastoma and uterine leiomyoma) is mir Functional data supporting the oncogenic role of mir- 21 came from a study by Chan and colleagues in glioblastoma. They observed that knock down of mir- 21 in glioblastoma cell lines induced a caspase- mediated apoptotic response 88. Further, studies in breast cancer cells showed that, upon transfection with anti- mir- 21, cell growth in vitro and tumour growth in vivo were suppressed due to increased apoptosis and decreased cell proliferation. The mir- 21 targets include the tumour suppressors TPM1 and PTEN, a gene frequently mutated in a variety of advanced tumours and implicated in the AKT survival pathway central in cancer development 89, 90. Another mirna, mir- 155, has been shown to accumulate in various types of B- cell malignancy (Hodgkin lymphomas and Burkitt lymphomas) where it functions as an oncogene in cooperation with MYC 91. Deregulated mir- 155 expression is an early event in oncogenesis. By creating the first transgenic mouse overexpressing a mirna gene, Costinean and colleagues demonstrated that mir- 155 overexpression in B- cells leads to the development of pre- leukemic B- cell proliferation followed by B- cell malignancy 92. mirnas regulated by MYC Several oncogenic mirnas are linked to the increased expression of MYC proteins. The best studied MYC regulated mirnas are the ones located within the mir cluster, residing in a non- protein- coding RNA on chromosome 13 (C13orf25) encoding six different mirnas (mir- 17, mir- 18a, mir- 19a, mir- 20a, mir- 19b and mir- 92a). Since this genomic region is frequently amplified in a subset of B- cell lymphomas, He and colleagues postulated that increased expression of this cluster contributes to cancer formation. This was tested with a mouse model for human B- cell lymphoma, driven by the presence of the MYC oncogene. Enforced expression of the mir b- 1 cluster (the vertebrate portion of the mir cluster) significantly accelerated the onset of tumour formation in the transgenic animals (~51 days vs. 3-6 months) 93. This was not the case when individual mirnas from the mir b- 1 cluster were introduced nor were there any effects with other unrelated mirnas. Lymphomas resulting from both MYC and mir b overexpression presented increased cell proliferation and decreased cell death. These results clearly point towards an oncogenic cooperation of several mirnas within the truncated mir cluster. The oncogenic properties of the mir cluster were confirmed using antisense technology in lung cancer 94. A microarray screen for MYC regulated mirnas identified elevated mir expression in a human B- cell line overexpressing MYC 95. Chromatin immunoprecipitation experiments confirmed MYC binding to elements in the mir promoter entailing direct regulation by MYC. Two mirnas in this cluster were shown to negatively regulate the expression of E2F1, an essential regulator of the initiation of DNA- replication during the cell cycle. Given the fact that E2F1 activity is also stimulated directly by MYC, this reveals a mechanism through which MYC simultaneously activates E2F1 transcription and limits its translation, allowing a tightly controlled proliferative signal. Dews and colleagues demonstrated that mir mirnas promote an angiogenic switch in MYC- activated tumours. Transduction of cells with a mir encoding retrovirus reduced TSP1 and CTGF levels and cells formed larger, better- perfused tumours 60. In neuroblastoma, mirna expression profiling identified increased mir expression in MYCN amplified tumours and MYCN was shown to bind the mir promoter Overexpression of mir in neuroblastoma cells without MYCN amplification strongly increased their in vitro proliferation and in vivo tumourigenesis through mir- 17- mediated repression of CDKN1A and BIM. In addition, mir- 18a and mir- 19a were shown to target and repress the expression of ESR1, a transcription factor implicated in neuronal differentiation 99. Inhibition of mir- 18a in neuroblastoma cells led to severe growth retardation, outgrowth of neurites, and induction of neuronal sympathetic differentiation markers. MYC and MYCN have also been shown to activate the expression of mir In breast cancer cells, mir- 9 directly targets

20 14 CDH1 leading to increased cell motility and invasiveness. In addition, mir- 9- mediated CDH1 down regulation results in the activation of beta- catenin signaling, up regulation of VEGF and ultimately increased tumour angiogenesis. As beta- catenin activation also results in transcriptional upregulation of MYC, MYC- induced mir- 9 expression might induce a feed- forward loop resulting in increased MYC expression. MYC has also been shown to repress mirna expression. Through the analysis of B- cell lymphoma models with inducible MYC expression, Chang and colleagues unexpectedly found a widespread repression of mirnas upon MYC activation 101. Much of the repressed mirnas were known to have tumour suppressor functions such as let- 7 family members, mir- 15 and mir- 34a. Chromatin immunoprecipitation revealed that much of the repression was likely to be a direct result of MYC binding to mirna promoters. These mirnas were shown to be relevant in the process of tumourigenesis as enforced expression of repressed mirnas diminished the tumourigenic potential of lymphoma cells. These studies clearly demonstrate that activation of MYC genes has a profound effect on the mirna transcriptome and that deregulated mirnas contribute to tumourigenesis. MiR- 34a: the missing piece in the p53 network puzzle The TP53 gene plays a central role as a stress sensor for the cell. It can arrest cell growth in order to allow DNA repair or can eliminate cells with severe damage by activating the apoptotic pathway. Given this important protective role, it is not surprising that TP53 is functionally inactivated in most if not all tumours, either through mutations or through inhibition of its function by alterations in up or down stream mediators. Of further interest, TP53 mutations are often found in aggressive tumours with poor prognosis. Neuroblastoma tumours harbor relatively few TP53 mutations at diagnosis whereas tumours from relapse patients often present with abnormalities in the TP53 pathway 102. Despite intensive efforts, certain issues concerning TP53 regulation remained unresolved. One such enigma was the fact that evidence was found for a role of TP53 as transcriptional repressor, although all available insights clearly pointed at the gene being a transcriptional activator. The answer to this contradiction came from mirna studies. First, investigators profiling mirnas in neuroblastoma found decreased expression of mir- 34a. MiR- 34a is located on the short arm of chromosome 1 which is often deleted in these tumours 103. Soon after, several other studies demonstrated that mir- 34a was directly regulated by TP53 104, 105. Therefore, TP53 could indeed lead to indirect transcriptional repression of target genes, most likely due to down regulation by mir- 34a. Up regulation of mir- 34a induced cell cycle arrest, senescence or apoptosis, all known TP53 pathway mediated effects. These effects are mediated by a wide range of mir- 34a targets such as BCL2, CCND1, CCNE2, CDK4, CDK6 and many others 106. Of note, mir- 34a was also shown to target MYCN in neuroblastoma 107. This might explain the close association between 1p deletion and MYCN amplification as loss of mir- 34a would be necessary to maintain high levels of MYCN protein. Even before the mir- 34a discovery, two other mirnas were found to interfere with the TP53 pathway. Testicular germ cell tumours, known to have functional TP53, displayed increased expression of mir- 372 and mir- 373, overriding TP53 mediated cell cycle arrest 108. Such findings are critically important, as they unveil mechanisms that suppress normal TP53 function in cancer cells with otherwise functionally intact TP53 protein and thus offer insights to alternative therapeutic strategies for treatment of this substantially large group of TP53 intact tumours. mirnas and metastasis The most life- threatening characteristic of cancer cells is their possibility to acquire the capacity to invade and metastasize to other organs. In contrast to tumour initiation, our understanding of the alterations of genes controlling invasion and metastasis is still poor. Therefore, the recent findings as to how mirnas might coordinate some of the gene expression programs controlling these phenomena attracted much attention. Based on previously available mirna gene expression data in breast cancer, mir- 10b was found to be correlated with vascular invasion 109. TWIST, a gene known to control tumour cell motility, up

21 15 regulated mir- 10b which in its turn caused down regulation of HOXD10 expression, thereby affecting other genes involved in metastasis. Much like the TP53 work, this study showed how mirnas play crucial roles in signaling networks implicated in cancer. Further work on breast cancer also revealed a role for mir- 335 and mir- 126 as metastasis suppressors and association of low expression for these genes with poor distal metastasis- free survival 110. Mechanisms of deregulated mirna expression So far, four different mechanisms of aberrant mirna expression have been identified. These mechanisms can function both independently and in concert to disturb mirna expression patterns in human tissues and are described below. 1. The location of mirnas at cancer associated genomic regions About half of all mirnas reside within cancer associated genomic regions (CAGRs) 111. These regions are frequently altered in cancer cells through copy- number alterations and are thought to harbour tumour suppressor genes or oncogenes, depending on whether the particular region is deleted or amplified. This association emphasizes the importance of mirnas in tumour biology and is supported by functional evidence, directly linking mirnas at CAGRs with tumour progression. 2. Epigenetically regulated mirnas Like many coding genes, mirnas can be located near or within a CpG island. This led to the hypothesis that mirna expression could be under epigenetic control. Indeed, recent studies have shown that aberrant DNA methylation as well as chromatin modifications may serve as a mechanism for deregulated mirna gene expression in cancer. First, Saito and colleagues demonstrated that about 5% of investigated mirnas are up regulated by treatment of T24 bladder cancer cells with a DNA demethylating agent in combination with a histone deacetylase inhibitor 112. In particular, reactivation of mir- 127 led to the down regulation of its predicted target BCL6, a proto- oncogene implicated in the pathogenesis of B cell lymphoma. Meanwhile, additional studies have shown that mirna expression can be silenced by epigenetic mechanisms, e.g. mir- 124a 113. On the other hand, Brueckner and colleagues observed that mirna let- 7a- 3, embedded in a CpG island located on 22q12.31, is hypomethylated in some lung adenocarcinomas, while it is heavily methylated in normal human tissues 114. Epigenetic reactivation of this non- coding gene resulted in enhanced tumour types and oncogenic transcription profiles, which suggests that let- 7a- 3 acts as an oncogene. 3. Defects in the mirna- processing machinery Various proteins that are implicated in mirna- processing and mirna directed regulation of protein- coding mrnas have been linked to tumourigenesis. When examining the expression of DROSHA and DICER1 in 67 non- small lung cancer samples, Karube and colleagues found a reduced expression of DICER1 that correlated with shortened post- operative survival 115. As DICER1 is a crucial protein in the production of mature mirnas, reduction of its expression could potentially result in a decrease of tumour suppressor mirnas in the lung 116. Surprisingly, complete loss of DICER1 expression is selected against during tumourigenesis suggesting that DICER1 functions as a haploinsufficient tumour suppressor gene 117, 118. This was shown in different mouse models of cancer including retinoblastoma and lung cancer. Monoallelic loss of Dicer1 dramatically increased tumourigenesis whereas complete loss of Dicer1 did not. Defects in other components of the mirna processing machinery have also been reported. Melo and colleagues identified mutations in TARBP2, encoding an integral component of a DICER1- containing complex, in sporadic and hereditary carcinomas 119. These cancer types were also shown to harbour mutations in other mirna processing genes including AGO2, TNRC6A and XPO5 120, Mutations in mirna genes and mirna binding sites Although still scarce, evidence for pathogenic mutations in mirna genes suggests that these mutations might influence mirna expression and function, ultimately contributing to tumour onset or progression 83, 122. Thus far, the number of reported mirna mutations is relatively low compared to the total number of human mirnas. Apart from mutations