Investigation of Novel Progression-Related Methylation Events and HOXD Genes in Prostate Cancer

Home , HOXD3, HOXD8, IRX1

Investigation of novel progression-related methylation events

and HOXD genes in prostate cancer

Ken Kron

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Department of Laboratory Medicine and Pathobiology University of Toronto

Investigation of novel progression-related methylation events and HOXD genes in prostate

cancer

Ken Kron

Doctor of Philosophy

Department of Laboratory Medicine and Pathobiology

University of Toronto

2012

Abstract

Aberrant DNA methylation in gene promoters causes gene silencing and is a common event in prostate cancer development and progression. While commonly identified methylated genes have been analyzed for their potential clinical utility in a variety of cancers, few studies have attempted a genome-wide methylation approach to discover new and possibly improved biomarkers for prostate cancer.

In order to identify DNA methylation changes associated with aggressive prostate cancer, we performed a genome-wide analysis of 40 prostate cancers using Agilent human CpG island

microarrays. Methylation profiles of candidate genes were validated using quantitative

MethyLight technology in an independent series of 219 radical prostatectomies and compared to

clinicopathological parameters. The effects of methylation on expression of HOXD3 and

HOXD8 and the possible role of HOXD8 in progression of PCa were also investigated.

We discovered previously unidentified methylation in the HOXD cluster of genes, namely

HOXD3 and HOXD8, as well as TGFβ2 and GENE X as potential prognostic biomarkers.

Furthermore, unsupervised clustering of samples by methylation signature indicated ERG oncogene expression as significantly different between clusters. Within the independent cohort, we observed strong correlations between Gleason score (GS) and HOXD3 as well as GENE X, while HOXD3 and HOXD8 methylation were associated with ERG expresson. TGFβ2 was an independent predictor of disease recurrence using Cox multivariate regression analysis. In gene expression studies, both HOXD3 and HOXD8 were elevated in cancers with poor prognosis, while DNA methylation did not correlate with expression levels. Both genes were found to contain alternative transcription start sites, explaining the poor correlation between methylation and expression. Finally, knockdown of HOXD8 expression did not have any effect on viable cells or cell motility in an in vitro model.

These results indicate that a panel of novel DNA methylation markers distinguish indolent prostate cancers from aggressive ones, and that expression of HOXD3 and HOXD8 is regulated by mechanisms including, but not dependent on, DNA methylation.

iii

Acknowledgements

I would like to begin by thanking my supervisor, Dr. Bharati Bapat, for the training, guidance, and support not only in lab and research related topics but essentially in all aspects of life. It is my belief that you took a chance on hiring a young, inexperienced scientist-in-training, which has been of immeasurable benefit to me. I am truly grateful for everything you have done for me and the experience you have provided.

I would also like to thank my thesis advisory members for their crucial input into my project over the years. Dr. Neil Fleshner and Dr. Rob Bristow, you have both provided immense support to my training and have constantly made me think, be critical, and improve upon the work that I did. I would also like to extend a special thank you to Dr. Theo van der Kwast, who aside from the support, teaching, and training has always made the time to talk about my project, new and exciting papers, and essentially anything I had in mind when I would wander over to his office. Your constant cheer and good humour is contagious.

To everyone I have had the pleasure of meeting and spending time with in LMP and CLAMPS over the years, it has truly been wonderful. Dr. Elsholtz and all administrative staff, you have provided great support and have always made the time to answer any question I may have. To the old coffee time crew including Menat, Golnessa, Igal and Shawna, let’s meet up again soon and remember to always keep in touch! You too Justin...

To all members of the Bapat lab, both former and current, thank you. Joyce, my grant time friend, thank you for everything. Sheron, you provided me with a lot of feedback, tips, and tricks along the way. George, you taught me many things throughout the years both research-related and not. Let’s get some early morning hoops going again. Miralem, my old bench neighbour, we had a lot of fun together. I agree to the continuing corruption of young, impressionable minds and forever will respond to a number of nicknames. Liyang, my project partner, thank you for all of the work you did and for constantly being fun. Ekaterina,you have been a great support and someone I can always talk to about anything. To Jamie and Andrea, you have been exceptional labmates and even better friends. Vaiju and Darko, I value the time in the lab we had together and will always carry fond memories. I look forward to keeping in touch with all of you in the future.

Finally, I would like to thank my friends and family for everything throughout the past five years. I love you all. Nick, Ryan, Steve and Mike, you are fantastic people with an unbelievable sense of humour, irregardless of what people say about you. Ergo, I value our friendship immensely. To my sister Melissa, brother-in-law Austin, and my special nieces Haven and Brynn, you provide motivation to work hard each and every day. To Grandma Grace and Grandma Verna, I truly hope I have made you proud. Dad, all I can say is thank you for your unwavering help and support. Mom, I would not be anywhere close to where I am today without you. There are too many things to count or name that I need to thank you for. And lastly, to my wife, Jill: you have been with me the entire way throughout this journey. There are not enough words or expressions of gratitude to thank you. I hope I provide even a fraction of the love and support that you show me. All that I have accomplished I owe to you and is shared between us.

Table of Contents

Abstract ...... ii Acknowledgements ...... iv Table of Contents ...... v List of Tables ...... ix List of Figures ...... xi List of Appendices ...... xiii List of Abbreviations ...... xiv Chapter 1: General Introduction ...... 1 1.1 Anatomy and function of the prostate gland ...... 1 1.2 Prostate pathology ...... 2 1.3 Prostate cancer ...... 3 1.3.1 Epidemiology of prostate cancer ...... 3 1.3.2 Diagnostic PCa screening ...... 7 1.3.3 PCa pathology ...... 8 1.3.4 Treatment and prevention ...... 11 1.4. Signaling pathways in prostate development and PCa ...... 19 1.4.1 Androgen signaling ...... 19 1.4.2 PI3K/Akt signaling ...... 22 1.4.3 TGFβ signaling ...... 25 1.4.4 Wnt signaling...... 25 1.5. Genetics of PCa ...... 26 1.5.1 Prevalent somatic mutations ...... 26 1.5.2 Familial and hereditary PCa ...... 31 1.6 Epigenetics of PCa ...... 34 1.6.1 Non-coding RNAs ...... 36 1.6.2 Histone modifications ...... 38 1.6.3 DNA methylation ...... 42

1.7 Homeobox genes ...... 51 1.7.1 Role in development ...... 54 1.7.2 Role in cancer ...... 55 1.7.3 Epigenetics of HOX clusters ...... 57 1.8 Rationale, hypothesis and objectives ...... 58 Chapter 2: Discovery of novel methylated loci associated with PCa progression ...... 62 2.1 Summary ...... 64 2.2 Introduction ...... 64 2.3 Materials and methods ...... 66 2.3.1 Patient Samples ...... 66 2.3.2 DNA isolation ...... 66 2.3.3 Differential methylation hybridization and CpG island microarrays ...... 67 2.3.4 Statistical Analyses ...... 68 2.3.5 MassARRAY EpiTYPER analyses ...... 69 2.3.6 ERG immunohistochemnistry ...... 70 2.4 Results ...... 70 2.4.1 GS6 versus GS8 analysis ...... 71 2.4.2 GS7 specimens and incorporation of recurrence data ...... 76 2.4.3 EpiTYPER analysis ...... 81 2.4.4 Total GS analysis ...... 84 2.4.5 ERG stratification ...... 87 2.4.6 Hierarchical clustering ...... 89 2.4.7 Functional annotation ...... 91 2.5 Discussion ...... 94 Chapter 3: Validation of novel methylated loci associated with PCa progression ...... 100 3.1 Summary ...... 102 3.2 Introduction ...... 103 3.3 Materials and methods ...... 105 3.3.1 Patients and pathology ...... 105 3.3.2 DNA extraction ...... 106

3.3.3 Sodium bisulfate modification and MethyLight ...... 106 3.3.4 Tissue microarray construction ...... 107 3.3.5 ERG immunohistochemistry ...... 108 3.3.6 Statistical analysis ...... 109 3.4 Results ...... 110 3.4.1 Clinical and pathological variables ...... 110 3.4.2 DNA methylation in tumor adjacent benign tissue versus PCa ...... 112 3.4.3 DNA methylation and GS/pathological stage ...... 114 3.4.4 DNA methylation and GP ...... 116 3.4.5 DNA methylation association with ERG ...... 118 3.4.6 Multivariate regression model for GS and pathological stage ...... 121 3.4.7 DNA methylation and biochemical progression-free survival ...... 124 3.4.8 ERG, DNA methylation and biochemical progression-free survival ...... 133 3.5 Discussion ...... 137 Chapter 4: Expression of novel methylated HOXD loci and putative role of HOXD8 in PCa progression...... 142 4.1 Summary ...... 143 4.2 Introduction ...... 144 4.3 Materials and methods ...... 146 4.3.1 5-aza-2-deoxycytidine treatment and RNA extraction of cells ...... 146 4.3.2 Reverse transcription, RT-PCR, and qRT-PCR ...... 146 4.3.3 5’ and 3’ RACE ...... 147 4.3.4 MSKCC expression data ...... 148 4.3.5 Generation of constructs ...... 148 4.3.6 Western blots ...... 149 4.3.7 Cell viability assay ...... 149 4.3.8 Cell motility assay ...... 150 4.4 Results ...... 150 4.4.1 Methylation affects expression of HOXD3 and HOXD8 ...... 150 4.4.2 HOXD3 and HOXD8 expression in PCa ...... 154

vii

4.4.3 5’/3 RACE and RT-PCR of HOXD3 and HOXD8 ...... 158 4.4.4 Expression of HOXD3 and HOXD8 variants ...... 162 4.4.5 Cell viability ...... 164 4.4.6 Cell motility ...... 164 4.5 Discussion ...... 166 Chapter 5: Discussion and future directions ...... 173 5.1 Novel DNA methylation events associated with PCa progression ...... 173 5.2 Validation of candidate genes ...... 176 5.3 Role of the HOXD cluster in PCa progression ...... 177 Appendices ...... 182 References ...... 189

viii

List of Tables

Chapter 1

Table 1.1 Mammalian histone lysine acetyl- and methyltransferases and the residues they modify ...... 41 Table 1.2. Commonly methylated genes in PCa ...... 45

Chapter 2

Table 2.1 Top 25 differentially methylated probes comparing GS6 versus GS8...... 73 Table 2.2 Top 25 differentially methylated homeobox probes comparing GS6 versus GS8 .....75 Table 2.3 Top 25 differentially methylated probes comparing GS6 versus GS7...... 78 Table 2.4 Top 25 differentially methylated probes comparing recurrent versus non-recurrent cases ...... 80 Table 2.5 Top 25 differentially methylated probes across all GS ...... 85 Table 2.6 Top 25 differentially methylated peaks across all GS ...... 86 Table 2.7 Top 25 differentially methylated peaks comparing ERG positive versus ERG negative ...... 88 Table 2.8 Significantly encriched functional terms for differentially methylated genes according to GS ...... 92 Table 2.9 Significantly encriched functional terms for differentially methylated genes according to ERG status ...... 93

Chapter 3

Table 3.1 Clinicopathological characteristics of MethyLight cohort ...... 111 Table 3.2. Average PMR values and p-values from cancer versus benign analysis of four genes115 Table 3.3. Average PMR values for four genes according to GS and pathological stage ...... 115 Table 3.4. Wilcoxon paired analysis of pattern 3 and pattern 4 specimens ...... 117 Table 3.5 Multivariate regression models for pathological stage and GS...... 123 Table 3.6 Percentage of patients with biochemical recurrence stratified by PMR quartile .....129 Table 3.7 Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and HOXD3 methylation ...... 131

Table 3.8 Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and TGFβ2 methylation ...... 131 Table 3.9 Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and HOXD8 methylation ...... 131 Table 3.10 Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and GENE X methylation ...... 132 Table 3.11 Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and three gene methylation signature ...... 132 Table 3.12 Multivariate Cox regression model for biochemical recurrence with clinicopathological variables, HOXD8 methylation, and ERG expression status ...... 136

List of Figures

Chapter 1

Figure 1.1 Time trends of incidence and mortality for PCa in Canada ...... 5 Figure 1.2 Gleason grading system ...... 10 Figure 1.3 Androgen signaling in the prostate ...... 21 Figure 1.4 PI3K/Akt signaling pathway ...... 24 Figure 1.5 Epigenetics in transcriptional activation and silencing ...... 35 Figure 1.6 Organization of the human HOX clusters ...... 53

Chapter 2

Figure 2.1 Genomic location of differential methylation comparing GS6 versus GS8 ...... 74 Figure 2.2 Genomic location of differential methylation comparing GS6 versus GS7 ...... 79 Figure 2.3 EpiTYPER quantitative methylation profiling ...... 83 Figure 2.4 Unsupervised hierarchical clustering of all cases and 193 probes showing the greatest variation ...... 90

Chapter 3

Figure 3.1 Receiver-operator curve analysis ...... 113 Figure 3.2 Average PMR values for each GP across four genes ...... 117 Figure 3.3. Representative cores from tissue microarray ERG immunohistochemistry ...... 120 Figure 3.4 Proportion of cases with positive ERG expression stratified by pathological variable ...... 120 Figure 3.5 Boxplots of PMR values stratified by ERG expression status...... 122 Figure 3.6 Kaplan-Meier curves and log-rank p-values with biochemical recurrence as outcome for clinicopathological variables ...... 126 Figure 3.7 Kaplan-Meier curves and log-rank p-values for methylated genes stratified by quartiles ...... 127 Figure 3.8 Kaplan-Meier curves and log-rank p-values for methylated genes stratified by high methylation and low methylation ...... 128 Figure 3.9 Kaplan-Meier curves and log-rank p-value for ERG expression status ...... 134

Figure 3.10 Kaplan-Meier curves and log-rank p-values for methylated genes stratified by ERG expression status ...... 135

Chapter 4

Figure 4.1 Methylation of HOXD3 and HOXD8 in PCa cell lines ...... 152 Figure 4.2 HOXD3 and HOXD8 expression following demethylating treatment ...... 153 Figure 4.3 Methylation and mRNA expression of HOXD3 and HOXD8 in frozen tissues ....155

Figure 4.4 MSKCC cohort normalized log2 expression values ...... 156 Figure 4.5 Kaplan-Meier curves of HOXD3 and HOXD8 expression association with biochemical recurrence ...... 157 Figure 4.6 5’ RACE, 3’RACE and RT-PCR composite map of HOXD3 and HOXD8 ...... 159 Figure 4.7 HOXD8 variant protein expression ...... 161 Figure 4.8 Expression of HOXD8 variants in frozen tissue ...... 163 Figure 4.9 HOXD8 does not affect cell viability and motility ...... 165 Figure 4.10 Methylation profile of HOXD8 and HOXD3/HOXD4 genomic region ...... 169

xii

List of Appendices

Table A.1 MethyLight results from watchful waiting cohort ...... 182 Table A.2 HOXD3 and TGFβ2 MethyLight results from biopsy specimens ...... 183 Figure A.1 HOXD cluster heatmap and dendogram of cases ...... 184 Figure A.2 Polycomb/trithorax modifications in HOXD3 and HOXD8 regions in PCa cells .185 Figure A.3 Activating histone marks in HOXD8 region ...... 186 Figure A.4 Androgen affects HOXD gene expression ...... 187 Figure A.5 MSKCC cohort HOXD gene expression ...... 188

xiii

List of Abbreviations

25OHD 25 hydroxyvitamin D

5-ARI 5-alpha reductase inhibitor

5caC 5-carboxylcytosine

5fC 5-formylcytosine

5hmC 5-hydroxymethylcytosine

5mC 5-methylcytosine

Akt v-akt murine thymoma viral oncogene homolog (also known as protein kinase B)

APC adenomatous polyposis coli

AR androgen receptor

ARE androgen response element

ATBC alpha-tocpherol, beta-carotene

AUC area under the curve

BAX BCL2-associated X protein

BCL2 B-cell CLL/lymphoma 2

BER base excision repair

Bp basepair

BPH benign prostatic hyperplasia

BRCA1 breast cancer 1, early onset

BRCA2 breast cancer 2, early onset c-FLIP CASP8 and FADD-like apoptosis regulator

CAB complete androgen blockade

CBP CREB binding protein

xiv cDNA complementary DNA

CHK1 checkpoint kinase 1

CIMP CpG island methylator phenotype

CMYC v-myc myelocytomatosis viral oncogene homolog

CpG cytosine-guanine dinucleotide

CRPC castration resistant prostate cancer

CTCF CCCTC-binding factor

CTLA-4 cytotoxic T-lymphocyte-associated protein 4

CYP17A cytochrome P450, family 17, subfamily A

DAC 5-aza-2-deoxycytidine, decitabine

DAVID database for annotation, visualization and integrated discovery

DNA deoxyribonucleic acid

DD3 differential display 3

DHT dihydrotestosterone

DNMT DNA methyltransferase

DMH differential methylation hybridization

DMR differentially methylated region dNTP deoxynucleotide triphosphate

DRE digital rectal exam

EDTA ethylenediaminetetraacetic acid

EMT epithelial-mesenchymal transition

ER estrogen receptor

ERG v-ets erythroblastosis virus E26 homolog

ETS v-ets erythroblastosis virus E26 oncogene

ETV ETS variant

EZH2 enhancer of zeste homolog 2

FBS fetal bovine serum

FISH fluorescence in-situ hybridization

FOXO forkhead box O

GM-CSF granulocyte macrophage colony stimulating factor

GP Gleason pattern

GS Gleason score

GSK3β glycogen synthase kinase 3 beta

GSTP1 glutathione S-transferase, pi 1

H&E hematoxylin and eosin

H3Ac histone 3, acetylated

H3K4me1/2/3 histone 3, lysine 4 mono/di/trimethylation

H3K9me2/3 histone 3, lysine 9 di/trimethylation

H3K27Ac histone 3, lysine 27 acetylation

H3K27me3 histone 3, lysine 27 trimethylation

H4Ac histone 4, acetylated

HDAC histone deacetylase

HM high methylation

HOTAIR HOX transcript antisense RNA

HOTTIP HOXA distal transcript antisense RNA

HOX HOX family of homeobox genes

xvi

HPC1 hereditary prostate cancer 1

HRP horseradish peroxidase

IGF insulin-like growth factor

IHC immunohistochemistry kDa kilodalton

KLK kallikrein-related peptidase family

LHRH luteinizing hormone releasing hormone

LEF lymphoid enhancer-binding factor lincRNA long intergenic non-coding RNA

LINE1 long interspersed nuclear element 1 lncRNA long non-coding RNA

LM low methylation

LRP low density lipoprotein receptor-related protein

MAb monoclonal antibody

MBD methyl-CpG binding domain

Mdm2 murine double minute 2 miRNA micro RNA

MgCl2 magnesium Chloride

MLL myeloid/lymphoid or mixed lineage leukemia

MSKCC Memorial Sloan Kettering Cancer Center mTOR mammalian target of rapamycin ncRNA non-coding RNA

NGS next-generation sequencing

xvii

NKX3.1 NK3 homeobox 1

ORF open reading frame

PAP prostatic acid phosphatase

PBMC peripheral blood mononuclear cells

PCa prostate cancer

PCA3 prostate cancer antigen 3

PCPT prostate cancer prevention trial

PCR polymerase chain reaction

PDK 3-phosphoinositide-dependent kinase

PI3K phosphoinositide-3-kinase

PIN prostatic intraepithelial neoplasia

PIP2 phosphatidylinositol 4,5 bisphosphate

PIP3 phosphatidylinositol 3,4,5 triphosphate

PITX2 paired-like homeodomain 2

PIWIL1 piwi-like 1

PLCO prostate, lung, colorectal, ovarian

PMR percent of methylated reference

PRC2 polycomb repressive complex 2

PRNCR1 prostate cancer non-coding RNA 1

PSA prostate specific antigen

PTEN phosphatase and tensin homolog qRT-PCR quantitative reverse transcription polymerase chain reaction

RACE rapid amplification of cDNA ends

xviii

RANKL receptor activator of nuclear factor kappa B ligand

RASSF1 ras association domain family member 1

RARβ2 retinoic acid receptor beta 2

REDUCE reduction by dutasteride of prostate cancer events

RNA ribonucleic acid

RNASEL ribonuclease L

ROC receiver-operator curve

RP radical prostatectomy

RRBS reduced representation bisulfite sequencing

RT radiation therapy

RT-PCR reverse transcription polymerase chain reaction

SELECT selenium and vitamine E cancer prevention trial

SFRP1 secreted frizzled-related protein 1

SHH sonic hedgehog

SMAD sma and mothers against decapentlegic homolog

SNP single nucleotide polymorphism

SRE skeletal related event

TALE three amino acid loop extension

TCF T-cell factor

TET ten-eleven translocation

TGFβ transforming growth factor beta

TMA tissue microarray

TMPRSS2 transmembrane protease, serine 2

xix

TSS transcription start site

TRAMP transgenic adenocarcinoma of the mouse prostate

UTR untranslated region

WIF1 wnt inhibitory factor 1

Wnt wingless-type MMTV integration site family

XMRV xenotropic MuLV-related virus

Chapter 1

General Introduction

1.1.Anatomy and function of the prostate gland

The prostate gland is a male reproductive organ, found only in mammals, that lies inferior to the bladder and surrounds the segment of the urethra that extends from the bladder neck. The prostate is contained within a fibromuscular structure comprised primarily of collagen, elastin, and smooth muscle. Seminal vesicles, which function to secrete a portion of the seminal fluid, connect to the base of the prostate gland. Functionally, the prostate produces the remaining portion of the ejaculate that protects sperm cells and contains proteolytic enzymes necessary for fertilization.

Comprised of approximately 70% glandular epithelium and 30% fibromuscular stroma, the prostate can be divided into three distinct zones: a central zone which surrounds the urethra

(~25% of prostate volume), a transition zone (~5-10% of prostate volume) which is lateral to the urethra, and a peripheral zone (~70% of prostate volume) (1).

Proper development and functioning of the male prostate gland is dependent on androgens (2). In humans, embryonic testosterone created by the fetal testes begins prostate morphogenesis, with epithelial buds from the urogenital sinus epithelium branching into the urogenital sinus mesenchyme (3). Development of the prostate gland continues postnatally with ductal branching morphogenesis and differentiation of epithelial and stromal cells. Stromal cells then begin to differentiate, surrounding forming ductal structures consisting of epithelial cells that themselves differentiate into two distinct cell types – basal cells, marked by expression of cytokeratins 5, 14

2 and p63, and luminal cells marked by expression of cytokeratins 8 and 18 (4). The prostate then enters a quiescent state until puberty when serum testosterone levels begin to increase. At this point the prostate grows rapidly in size and prostate specific secretory proteins increase accordingly (5).

As mentioned above, proper development of the prostate is dependent on androgen signaling. In the absence of androgens the urogenital sinus will form the lower part of the vagina and urethra, while if female mice are exposed to androgens during embryogenesis a prostate will begin to form (6). 5-α reductase enzymes are a necessary component to this process as they convert testosterone to dihydrotestoserone (DHT), which is a much more potent androgen. Binding of

DHT to the androgen receptor (AR) subsequently causes translocation into the nucleus and activation of a wide range of genes necessary for prostate development and function including

SHH and NKX3.1 (7, 8). Aside from androgen signaling, other critical pathways for prostate development include WNT, NOTCH, and TGFβ signaling, which are discussed in more detail below.

1.2 Prostate pathology

There are three well-described prostate pathologies that commonly occur in men: prostatitis

(both chronic and acute), benign prostatic hyperplasia (BPH), and prostate cancer (PCa).

The term prostatitis refers to inflammation of the prostate gland. Acute prostatitis is generally considered the result of bacterial infection from E. coli, Klebsiella, Proteus, as well as others, and is treated with appropriate antibiotics. Chronic prostatitis, however, can be considered either bacterial or non-bacterial, with causes of the latter syndrome not very well understood (9).

BPH occurs in about 50% of men over the age of 60, with the likelihood increasing further as a function of age (10). It is characterized by an increased number of both epithelial and stromal cells in the transition zone of the prostate, causing enlargement of the gland that in turn creates urinary symptoms (increased urinary frequency and urgency). As androgens are a necessary factor for the development of BPH, administration of the 5-α reductase inhibitors dutasteride and finasteride (both of which act to inhibit production of dihydrotestosterone) are common treatment options (11).

Prostate cancer (PCa) is the most common male malignancy in Canada, and in the vast majority of cases is an adenocarcinoma, or glandular cancer. A small proportion (0.5-2%) of PCa cases have a small cell morphology with neuroendocrine characteristics (12, 13). Approximately 60-

80% of adenocarcinomas arise in the peripheral zone while the remaining proportion occur within the transition and central zone or exist in multiple zones (14-16).

1.3 Prostate cancer

1.3.1 Epidemiology of prostate cancer

In 2011, it is estimated that 25,500 Canadian men were diagnosed with prostate cancer (PCa) and

4,100 men died from the disease (Canadian Cancer Society 2011). This translates into a rate of

122.5 per 100,000 men, which is an increase of 57.3% compared to the rate in 1982. The single most important factor contributing to this increase is the introduction and acceptance of prostate- specific antigen (PSA) screening in the clinic. Clinical availability of the PSA test created a peak in the number of diagnosed cancers (140.7 per 100,000; Figure 1.1A), likely due to the

4 identification of men either already harbouring significant PCa that was not previously detected by digital rectal exam (DRE) or who had not previously been screened. Additionally, increases in life expectancy and the proportion of the population ≥ 55 years of age may also be contributing factors to the overall increase in PCa incidence.

Figure 1.1. Time trends of incidence and mortality for PCa in Canada. (A) Age-standardized incidence rate for PCa as depicted by the bolded dashed-dotted line. (B) Age-standardized mortality rate for PCa as depicted by the bolded dashed-dotted line. Adapted from Canadian Cancer Society’s Steering Committee on Cancer Statistics. Canadian Cancer Statistics 2011. Toronto, ON: Canadian Cancer Society; 2011.

The mortality rate for PCa has steadily declined from a peak of 31.2 per 100,000 men in 1991 to an estimated 21.2 per 100,000 men in 2011 (Figure 1.1B). This may be due to improved treatment regimens for men diagnosed with localized PCa that includes increasingly accurate radiation treatment and nerve-sparing radical prostatectomy. In addition, PSA testing may be a contributing factor, as the European Randomized Trial of Screening for Prostate cancer has shown modest but significant reductions in mortality for screening versus control groups (17).

The majority of prostate cancers are diagnosed between the ages of 60 and 69 years of age

(estimated to be 10,100 in Canada in 2011) while death from disease most commonly occurs in men over 80 years old (an estimated 2,200). This discrepancy is likely due to a long latency period between the initial development/diagnosis of PCa and the progression to metastatic disease, which is often > 15 years (18).

Significant racial differences also exist in the rates of incidence and mortality of PCa. In general,

Asian populations have the lowest frequency of incidence and mortality, followed by

Caucasians, with the highest frequency observed in African populations in North America (19).

There is currently debate as to whether this is caused by genetics, environmental/dietary influences, or societal differences in screening and treatment. Data from the US National Cancer

Institute Surveillance Epidemiology and End Results (SEER) program revealed significant differences in rates of PCa and mortality from PCa among African Americans, European

Americans, and American Indian or Alaskan Natives (American Cancer Society 2007 Cancer facts and Figures 2007), suggesting that genetic differences between ethnicities does indeed play a role (20-22). It should be noted, however, that screening rates and environmental differences in the form of diet, for example, do still exist within the US between ethnic groups (23, 24).

1.3.2 Diagnostic PCa Screening

Prostate specific antigen (PSA) screening is currently the most widely accepted non-invasive clinical test that suggests the presence of PCa. PSA is a member of the kallikrein-related peptidase family (KLK3), is secreted from prostate epithelial cells, and is responsible for the liquefaction of semen. Neither transcription nor translation of PSA is upregulated in PCa cells, but instead is secreted into the blood at higher levels than usual only as the result of a greater than usual number of prostate epithelial cells. Thus, the specificity of PSA testing is quite low

(approximately 60%; (25)) as proliferation of prostate epithelial cells commonly occurs in men over the age of 50 as a result of BPH. The sensitivity of PSA as a diagnostic test is also questionable, as studies indicate there is not a sufficiently low level at which PCa can be ruled out. Indeed, recent large scale screening studies have questioned the efficacy of PSA testing due to either a complete lack of deaths prevented or the lack of a significant number of deaths prevented as compared to the number screened (17, 26). A number of variations to the standard

PSA test also exist, including free/total PSA, PSA density, and PSA kinetics. The free/total PSA ratio, which measures the amount of PSA that exists in an unbound form (“free”) compared to

PSA bound to other proteins, is lower in men with PCa as compared to those with BPH, and thus improves the diagnostic specificity (27). PSA density is a measure of the PSA concentration compared to total prostate volume. Initial studies indicated a higher density in men with PCa versus BPH (28) and suggest it can be used to better predict poor prognostic variables prior to biopsy (29). The need for a transrectal ultrasound for determination of PSA density is a practical limitation, however, as this is typically not performed prior to biopsy. Finally, studies in the dynamic changes of PSA concentration over time have been suggested as better indicators then

PSA concentration at any one time point. The PSA velocity (ng/ml/year) was initially shown to differentiate between PCa and BPH with a specificity of 90% (30) and has also been proposed to predict shorter time to death from PCa (31).

PCA3 (or DD3) is a non-coding RNA transcribed from the 9q21-22 region with unknown function. Overexpression of PCA3 was initially discovered and described by Bussemakers et al. in 1999 (32). Using differential display analysis and RT-PCR, PCA3 was shown to be highly overexpressed (10-100 fold) in 53/56 tumour specimens analyzed and was also prostate specific.

Further work assessing the diagnostic potential of PCA3 showed an AUC of 0.98 and a median

34 fold increase of mRNA in PCa tissues compared to benign prostate (33). Detection in urine was then proven possible with a sensitivity of 67% and negative predictive value of 90% (34).

Currently, the Progensa PCA3 assay, which normalizes PCA3 intensity to PSA (as a means to control for contaminating non-prostate RNA) is available in Canada and Europe and is seeking

FDA approval in the United States.

1.3.3 PCa pathology

The diagnosis of PCa is most commonly made following histological evaluation of needle biopsy tissue but can also be made after digital-rectal exam during which an unequivocal palpable mass is present. In fact, these two procedures are crucial in determining pre-treatment prognosis, the former providing histological and clinical staging information while the latter helps to further determine the clinical stage.

The histological grading system for PCa was established by DF Gleason in 1966 (35) and remains one of the best prognostic variables utilized today. A Gleason pattern (GP), ranging

9 from 1-5, is assigned by a pathologist to each tumour based on the amount of glandular

“dedifferentiation” (Figure 1.2). A GS is given based on the sum of the two most prevalent GPs.

For example, a tumour with varying proportions of GP3 and GP4, is a GS7 (either 3+4 or 4+3).

If the cancer is composed entirely of one GP, that GP is doubled to give the GS (for example,

3+3 is GS6). GS ≤ 6 prostate tumours are considered low grade and have a good prognosis, GS7 tumours are intermediate grade and have a variable outcome, while GS ≥ 8 tumours are high grade, have a poor outcome, and are not as common as GS7 or less tumours.

The GS and clinical stage assigned at biopsy, often along with patient age, family history, and

PSA levels, assist in determining appropriate treatment strategies for PCa.

Figure 1.2. Gleason grading system. (Above) illustration GPs ranging from 1 (well differentiated) to 5 (poorly differentiated). (Below) Histological samples of GPs 1-5. Adapted from http://www.prostatecancer.ca/PCCN/Prostate-Cancer/diagnosis/clinical-testing-and-the- Gleason-grade.

1.3.4 Treatment and prevention

Standard treatments

The most common therapeutic approaches for PCa include surgical resection of the gland, radiation therapy (external beam/brachytherapy), watchful waiting, hormone therapy, and chemotherapy. The majority of organ-confined tumours are treated surgically by removal of the entire prostate gland (radical prostatectomy, RP), by radiation therapy (RT), or by administering a watchful waiting approach. The latter of these options is generally only considered for those with D’Amico low risk characteristics; that is, having a GS <7, clinical stage

RP is associated with a significant amount of morbidity including bleeding, infection, urinary incontinence, and erectile dysfunction. Nerve-sparing approaches are an option to maintain sexual function, however, in men with no evidence of spread to the nerve bundles (37).

Following RP, additional prognostic variables can be assessed to determine the likelihood of disease recurrence. These include pathological stage (which provides a more accurate indication of disease spread than clinical stage) as well as surgical margin status (whether or not the entire cancer has been successfully resected). The presence of local invasion or a positive surgical margin may indicate the need for adjuvant therapies.

RT can be achieved through external beam therapy or through “seeded” radiation

(brachytherapy). Brachytherapy consists of small radioactive seeds planted into the prostate and

12 is typically only considered for patients with low risk characteristics (GS ≤ 6, PSA < 10). For

PCa with intermediate risk characteristics external beam RT is more suitable, which consists of a focussed beam of high energy electromagnetic radiation delivered by an external source into the prostate. This treatment type, however, is associated with increased morbidity in the form of diarrhea, urinary incontinence, impotence, and bleeding from the rectum or bladder.

Hormone treatment of PCa refers to ridding the body of circulating androgens that are required for proper prostate function and survival of prostate epithelial cells. There are two types of common chemical approaches to achieving reduced/abrogated androgen levels: luteinizing hormone-releasing hormone (LHRH) agonists and antiandrogens (androgen antagonists). The former of these act by interfering with the production of androgens via binding to the cognate

LHRH receptor (38). As LHRH is released in an intermittent fashion to avoid desensitization of the LHRH receptor, constantly elevated levels of LHRH analogues eventually results in receptor desensitization after an initial increase surge in LH production (39). In Canada, the most commonly prescribed LHRH agonists include gosrelin (Zoladex), leuprolide (Lupron or

Eliguard), and buserelin (Suprefact). Antiandrogens can be separated into two subtypes: steroidal and non-steroidal. Both types function by blocking the binding of androgen to AR, with steroidal antiandrogens acting similar to female sex hormones and having a structure similar to progesterone (40). Non-steroidal antiandrogens are often given in conjunction with LHRH agonists to achieve complete androgen blockade (CAB) and to avoid side effects associated with initial androgen surges in those receiving LHRH agonists.

As PCa is a generally a slow growing disease, there is a limited role for chemotherapeutics in treatment. Therefore, chemotherapy is often only considered as a late stage treatment for castration-resistant PCa (CRPC). In Canada, there are two drug combinations consisting of

13 chemotherapies: mitoxantrone plus prednisone and estramusine plus etoposide. Mitoxantrone is the chemotherapeutic drug in the first combination, and while the combination with prednisone

(a steroid) does not improve survival time it does improve overall quality of life (41). Both estramusine and etoposide are chemotherapies in the second combination, and the use of both drugs together reduces PSA levels in about half of the patients who have the drugs administered.

It should be noted, however, that there is no evidence for increased survival with these drugs

(42). Docetaxel is yet another chemotherapy that is often used to treat breast, ovarian, and lung cancers, and in the United States is approved for treatment of metastatic, hormone refractory PCa

(41, 43, 44) .

Recent treatment advances and clinical trials for metastatic CRPC

Sipuleucel-T

Sipuleucel-T is a dendritic cell vaccine that first gained approval in the US in 2010 for the treatment of metastatic CRPC. The study first supporting its use was a phase III IMPACT trial studying the potential use of this vaccine in men with metastatic CRPC who had a minimum expected survival time of 6 months. The median survival time of men treated with Sipuleucel-T versus placebo was 25.8 months versus 21.7 months (p-value = 0.03), with few morbidities that often cleared within 1 to 2 days of treatment (45).

The first step in creating the vaccine is drawing a blood sample from the patient and extracting peripheral blood mononuclear cells (PBMCs). The PMBCs are then incubated with synthetically produced antigen which is a recombinant protein that consists of prostatic acid phosphatase

(PAP) protein and granulocyte macrophage colony stimulating factor (GM-CSF). This creates loaded antigen presenting cells that are then infused into the patient in three separate doses (46).

Abiraterone

Abiraterone acetate is a derivative and structural analog of pregnenelone, a steroid which binds the CYP17A enzyme and is eventually converted to androgens/estrogens. Abiraterone binds and irreversibly inhibits the CYP17A enzyme, thus inhibiting androgen synthesis in the testes, adrenal glands and prostate (47, 48).

Results from phase I and II clinical studies revealed that abiraterone reduced PSA levels and decreased the number of circulating tumour cells (47, 49, 50). Subsequent phase III clinical trials in men with post-docetaxel metastatic CRPC displayed increased survival times in men treated with abiraterone plus prednisone versus placebo plus prednisone (14.8 months median survival versus 10.9; p-value < 0.001) (51).

Denosumab

Denosumab is a monoclonal antibody directed towards RANKL, a cytokine that binds the receptor activator of NF-κB and activates osteoclast differentiation and survival (52).

Denosumab treatment is often prescribed to patients with osteoporosis, rheumatoid arthritis and bone metastatic cancers (53-57). A phase III trial of denosumab in metastatic CRPC significantly delayed skeletal-related events (SRE, defined as pathological fracture, radiation therapy, surgery to bone, or spinal cord compression) with median times of 20.7 months to SRE in the denosumab

15 group versus 17.1 months for zoledronic acid (p-value = 0.0002), which is a bisphosphanate also used to treat bone-metastatic cancers (58).

Cabazitaxel

Cabazitaxel is a novel member of the taxane family of chemotherapeutic drugs that inhibit cell division by targeting tubulin and arresting cells in their mitotic phase (59). A Phase III clinical trial of men with metastatic CRPC who were previously treated with docetaxel was performed comparing cabizataxel plus prednisone versus mitoxantrone plus prednisone. Those who received cabazitaxel plus prednisone versus mitoxantrone plus prednisone had a median overall survival of 15.1 months versus 12.7 months for those who received mitoxantrone plus prednisone (p-value < 0.0001) (60).

Other treatments currently in phase III trials

MDV3100 is a selective AR antagonist chemically designed to target PCa cells overexpressing

AR (61). In phase I/II trials, MDV3100 induced tumour regression in mouse CRPC models and resulted in a sustained reduction of PSA values < 50% in 43% of metastatic CRPC patients (62).

Ipilimumab is a monoclonal antibody therapy designed to recognize the CTLA-4 antigen.

CTLA-4 inactivates T-cells by disrupting the binding between T-cells and stimulatory molecular present on the surface of APCs (63) . Phase I/II studies of this immunotherapy in 45 men with metastatic CRPC resulted in PSA declines < 50% in 22% of patients, with 38% reporting adverse immune-related effects (46).

There are a number of other treatments currently in phase III clinical trials for metastatic CRPC including radioisotope therapies (64), clusterin inhibitors (OGX-011; (65)), and kinase inhibitors

(Cabozantinib, Dasatinib) (66, 67), to name a few.

PCa Prevention

Chemopreventatives for PCa

The Prostate Cancer Prevention Trial (PCPT) enrolled over 18,000 men to study the effects of the 5-α reductase inhibitor (5-ARI) finasteride on the PCa prevention (68). The primary outcome was diagnosed PCa at any point during the seven year study. In the finasteride treated group, the prevalence of PCa was 18.4% versus 24.4% in the placebo treated group, or an overall reduction of 24.8% (p-value < 0.001). Importantly, however, there was a significant increase in the number of intermediate and high-grade (GS7-10) PCa in the finasteride treated group (37.0% versus

22.2%, p-value < 0.001).

Important questions raised from the results of this trial were whether finasteride was responsible for increasing the likelihood of developing high grade PCa or whether it increased the rate of detection of high grade PCa. Follow-up analysis of the data presented in the initial publication indicated that the latter scenario was the most likely. Lucia et al. showed that within the finasteride treated group there was a significant reduction of number of positive cores and percent positive cores upon biopsy within each GS (69). Furthermore, when comparing RP specimens of men who went on to receive surgical treatment following diagnosis, there were no differences in the rates of lymph node involvement or pathological stage.

The REDUCE trial also sought to determine the effect of a 5-ARI on preventing PCa. In this trial, however, the 5-ARI used was dutasteride, which inhibits both isoforms (type 1 and 2) as opposed to finasteride, which only targets isoform 1. In this trial, 8,231 men were enrolled and were administered dutasteride or placebo over a 4 year period and diagnosed PCa upon biopsy was the endpoint. Again, the dutasteride treated arm had a relative reduction in PCa of 23%, similar to what was observed with finasteride in the PCPT trial (70). There were no significant differences in the rates of intermediate or high grade PCa between the two arms in this trial.

Selenium and vitamin E

Selenium is a trace element necessary for the reduction of antioxidant enzymes such as glutathione peroxidises. A secondary analysis of the Nutritional Prevention of Cancer trial illustrated a relative risk for developing PCa of 0.51 with 200 µg supplement selenium (71), providing a foundation for inclusion in PCa prevention trials. Vitamin E is a fat soluble antioxidant found in many plant oils that helps to prevent damage caused by free radicals produced from lipid oxidation. Secondary analysis of results from the Alpha-Tocopherol, Beta- carotene (ATBC) study to understand potential protective effects of vitamin E and beta-carotene on lung cancer development, found a 32% decrease in PCa incidence and 41% decrease in PCa mortality (72).

These results led to the formation of the Selenium and Vitamin E Cancer Prevention Trial

(SELECT), which enrolled over 35,000 men to receive selenium, vitamin E, selenium plus vitamin E, or placebo to study the effect on PCa development. The trial ended prematurely in

2008, as interim results revealed a statistical trend (p-value = 0.06) towards increased PCa

18 incidence in men receiving vitamin E alone and a non-significant increase in type II diabetes found for men receiving selenium alone (p-value = 0.16) (73). There were also no benefits found in men receiving both selenium and vitamin E.

Lycopene

Lycopene is a carotene with antioxidant properties that is found abundantly in tomatoes. In 2002, the Health Professionals Follow-up Study published a report suggesting a link between the consumption of tomato products and reduced PCa incidence in a retrospective study of over

47,000 men (74). The authors of this study found a relative risk of PCa of 0.84 comparing those with the highest intake (top quintile) versus those with the lowest intake (bottom quintile).

In 2006, a similar large-scale retrospective study was published with participants from the prostate, lung, colorectal, and ovarian cancer (PLCO) trial. This study failed to find a significant link between dietary lycopene and reduced PCa incidence, although the authors noted statistically non-significant inverse associations between consumption of tomato sauce and PCa incidence (75). Additionally, significant reductions in PCa incidence were found in men who had a family history of PCa and increased dietary lycopene. Similar studies on a smaller scale have revealed inconsistent results (76-78), suggesting that the benefits (in terms of PCa prevention) of lycopene supplementation or increased consumption of tomato products are minimal at best.

Vitamin D

The role of vitamin D in cancer prevention has garnered a great deal of attention in recent years based on numerous studies showing inverse associations between serum 25 hydroxyvitamin D

(25OHD), vitamin D intake and cancer incidence or mortality (79, 80) as well as inverse associations between ultraviolet B exposure and cancer incidence (81). With respect to PCa, vitamin D metabolites have been shown to have both a pro-apoptotic and anti-proliferative effect on cancer cell lines in vitro (82, 83) and also to have anti-metastatic effects in vivo (84).

Several epidemiological studies have failed to find a link between dietary vitamin D intake and

PCa incidence (85-87), although one study found that increased supplementation of vitamin D (>

600 IU) corresponded to a 40% reduction in PCa risk (88). With respect to serum levels of vitamin D metabolites, conflicting results have also been shown with some studies reporting increased PCa risk with low serum 25OHD (89) while others have shown no association (90).

The anti-carcinogenic properties of soy (isoflavones), green tea (epigallocatechin), curcumin, and resveratrol have all been suggested from in vitro and in vivo studies (91-97), but their abilities in PCa prevention require additional studies.

1.4 Signaling pathways in prostate development and PCa

Deregulation of embryonic and growth factor signaling networks has been well established in multiple cancers, including PCa. Abnormalities of numerous pathways have been implication in

PCa development and progression, including the androgen, PI3K/Akt, TGFβ, and WNT signaling networks.

1.4.1 Androgen signaling

While androgen signaling is vital to normal prostate function, it also plays a role in PCa development and progression. Conventional androgen signaling in the prostate begins with binding of DHT to AR followed by the release of inhibitory heat-shock proteins from AR (98).

DHT bound AR then translocates into the nucleus where it binds androgen response elements

(AREs) and recruits chromatin remodeling enzymes including the SWI/SNF complex, histone modifying enzymes such as CBP, and RNA polymerase II (Figure 1.3) (99, 100).

There are a wide range of AR target genes that mediate numerous biological processes including secretory pathways, steroid metabolism, lipogenesis, and polyamine synthesis (101-104). With respect to PCa development and progression, AR activates genes including c-FLIP and Mdm-2

(105, 106), resulting in inhibition of apoptosis and cell proliferation.

The kallikrein family of genes located on chromosome 19 contains at least three members which are upregulated by androgen signaling - KLK2, KLK3/PSA, and KLK4. KLK3/PSA is expressed in the prostate but is not known to play a role in the carcinogenic process (107). KLK4, however, has recently been shown as a marker overexpressed in PCa which plays a role in cell proliferation and ECM degradation (108-110). Other kallikreins with potential responsiveness to androgens include KLK5, 12, and 15, although the role they play in PCa (if any) is unknown

(111-113).

Figure 1.3. Androgen signaling in the prostate. Testosterone (T) produced by the testes is converted to dihydrotestosterone (DHT) by 5-alpha reductase enzymes (5αR) in prostate epithelial cells. Binding of DHT to the androgen receptor (AR) occurs, displacing heat shock proteins (HSP) and leading to translocation of AR into the nucleus. AR then binds androgen response elements (AREs) and recruits co-activator (Co) proteins to activation gene transcription. Adapted from Tindall, D et al., The Journal of Urology, 2008; 179(4): 1235-1242.

TMPRSS2, a transmembrane serine protease with unknown function, is also an androgen regulated target gene. As discussed below, TMPRSS2 is fused to oncogenic ETS family genes in approximately 50% of PCa cases (114, 115). Interestingly, it has recently been shown that this fusion may be the result of AR dependent co-localization of TMPRSS2 with ETS family members, at least with respect to ERG Fusions (116, 117). Bastus et al. showed that prolonged androgen exposure of non-malignant prostate epithelial cells (PNT1a and PNT2) resulted in

TMPRSS2:ERG fusions, and that short term exposure in malignant AR-responsive PCa cells also results in translocations (118). It was also determined that this fusion was correlated with low expression of PIWIL1 and perhaps also with the number of CAG repeats found in exon 12 of AR.

There are a number of pathways that have been shown to crosstalk with androgen signaling and affect PCa progression, including the IGF and Wnt pathways (119). The insulin-like growth factor 1 (IGF-1), for example, is found in high serum quantities in men with increased risk of developing PCa (120), and IGF-1 has been shown to result in translocation of the AR into the nucleus, while AR also increases transcription of the IGF-1 receptor (121).

1.4.2 PI3K/Akt signaling

Signaling through the serine/threonine protein kinase Akt family members results in a number of oncogenic processes including increased proliferative ability through activation of mTOR (122), decreased apoptosis through inactivation of FOXO (which activates BIM and FASL) (123), genomic instability through sequestration of CHK1 (124). Phosphatidylinositol 4,5 bisphosphate

(PIP2) is converted into phosphatidylinositol 3,4,5 triphosphate (PIP3) via PI3K, and activation

via phosphorylation of Akt occurs through PIP3 and 3-phosphoinositide-dependent kinases

(PDKs) (125) (Figure 1.4). Phosphorylated Akt (pAkt) is typically undetectable in normal

prostate tissue and is correlated with a poor prognosis when present in PCa (126). The best

characterized mechanism of deregulated Akt signaling in PCa is the loss of the phosphatase and

tensin homolog (PTEN) through a number of mechanisms which are discussed below.

Figure 1.4. PI3K/Akt signaling pathway. PI 3-kinase (PI3K) phosphorylates phosphoinositide 4, 5 bisphosphate (PIP2) to phosphoinositide 3, 4, 5 trisphosphate (PIP3), leading to phosphorylation of Akt and and activation of signaling through mammalian target of rapaymycin (TOR), for example. PTEN dephosphorylates PIP3, abrogating this signalling cascade. GPCR, G protein coupled receptor; RTK, receptor tyrosine kinase; PDK1, Phosphinositide-dependent kinase 1; TSC1/TSC2, tuberous sclerosis complex 1 and 2; RHEB, RAS homolog enriched in brain. Adapted from Vogt, P et al.. Current Opinion in Genetics & Development, 2009; 19(1):12-17.

1.4.3 TGFβ signaling

TGFβ signaling is activated through binding of TGFβ cytokines to TGFβRII, followed by heterodimerization of TGFβRII to TGFβRI, and phosphorylation of serine and threonine residues on TGFβRI (127). TGFβRI then acts as a kinase, phosphorylating receptor SMAD proteins 2/3, which then binds SMAD4, with the entire complex then translocating to the nucleus and activating downstream target genes (128). TGFβ signaling acts to suppress proliferation on normal prostate epithelial cells by activating cell cycle regulators p15, p21, and p27 and repressing the MYC oncogene (129-131). TGFβ also induces apoptosis by upregulating BAX transcription and downregulating BCL-2 (132, 133). It is important to note, however, that in many instances PCa appears to become unresponsive to the inhibition of proliferation and apoptotic effects that TGFβ exerts on normal epithelium. In fact, TGFβ signaling switches from a growth suppressive role to a growth promoting role, mediated by an increase in TGFβ cytokine expression and a concomitant decrease in levels of TGFβRI and TGFβRII (134, 135).

1.4.4 Wnt Signaling

There is both a canonical and non-canonical Wnt signaling pathway. In the canonical form, Wnt ligand binds Frizzled receptors and LRP co-receptors followed by recruitment of Dishevelled to the plasma membrane (136). LRP co-receptors are then phosphorylated and Axin is recruited to the plasma membrane, which leads to inactivation of the beta-catenin destruction complex (Axin,

APC, and GSK3β) followed by translocation of beta-catenin into the nucleus, binding to

TCF/LEF, and activation of target genes (136). Non-canonical Wnt signaling can occur through

26 the actions of Wnt ligands Wnt4, Wnt5a, and Wnt 11, which can affect intracellular calcium levels and modulate cell movement and planar cell polarity, for example (136).

In the developing prostate, Wnt signalling via the Wnt5a ligand is required for proper branching morphogenesis, luminal cell differentiation, and proliferation of progenitor epithelial cells (137).

In PCa, activation of canonical and non-canonical Wnt pathways can occur in a number of different ways. Mutations of beta-catenin have been found in 5/104 PCa specimens, with 4/5 mutations occurring in a domain necessary for beta-catenin degradation (138). Loss of heterozygosity (LOH) and downregulation of the destruction complex member APC has also been described, with loss of expression also at least partially attributable to DNA methylation

(139-141). Furthermore, overexpression of Wnt5a and downregulation of Wnt inhibitory secreted factors SFRP1 and WIF1 have all been reported in PCa (142-144).

1.5 Genetics of PCa

1.5.1 Prevalent somatic mutations

TMPRSS:ETS gene fusions

TMPRSS2:ETS gene family fusions were the first commonly occurring gene fusions discovered in epithelial malignancies (145). The most common of these fusions, TMPRSS2:ERG, occurs in approximately 50% of PCa cases, while TMPRSS2:ETV1 and TMPRSS2:ETV4 are found at a much lower frequency (114, 115, 146). The TMPRSS2:ERG fusion event most often arises from deletion of an interstitial fragment of chromosome 21, resulting in the androgen responsive

27 promoter and 5’ end of TMPRSS2 driving the expression of the ERG oncogene (114). Recent work has shown that TMPRSS2:ERG is present in both low grade and high grade prostatic intraepithelial neoplasia (PIN)(147), indicating that this alteration is an early event in PCa development. Indeed, mouse models of TMPRSS2:ERG overexpression develop PIN lesions, but fail to develop PCa without concomitant activation of the PI3-kinase pathway (148, 149).

Conflicting results regarding the prognostic value of TMPRSS2 fusions have also arisen, with some reports claiming a positive association with clinicopathological features such as disease stage and metastases as well as disease recurrence (150), while others have found no association with these variables (151).

PTEN deletions

PTEN deletions that result in aberrant activity of the PI3-kinase/Akt pathway are one of the most well recognized mutations found in PCa. They were initially described in a series of publications that identified a candidate tumour suppressor gene within the 10q23 region, as deletions of this locus were commonly found in glioblastoma, breast and prostate tumours (152-154). Li et al. described PTEN mutations in a panel of 4 PCa cell lines (including LNCaP, DU-145, and PC-3) and in a variety of gliobastoma and breast cancer cell lines and xenografts (153). Depending on the study, genomic deletions of PTEN have been found in 10-70% of PCa and is often found to correlate with aggressive disease characteristics (155-159). In addition, prostate specific biallelic

PTEN knockout mice develop PCa in 100% of mice with characteristics of increased cell proliferation (160).

8q/CMYC copy gain

Coupled with loss of 8p21, amplifications of the q24 region of chromosome 8 is one of the most well described recurrent chromosomal abnormalities in PCa. The 8q24 locus contains the CMYC oncogene which is commonly amplified and/or overexpressed in hepatocellular and breast carcinomas (161, 162). Amplification of the 8q24 region in PCa was initially described by Bova et al. in a series of 32 primary and lymph node metastatic tumours (163). In 1994, Van Den Berg et al. reported amplification of 8q24 in 4/44 tumour specimens using fluorescence in-situ hybridization (FISH) (164). Four of these tumour specimens were positive for lymph node metastasis, and of those four, three had an 8q24 amplification. In 1997, Jenkins et al. used FISH to interrogate 8q24/CMYC amplification and described an association between the number of

CMYC copies and progression from PIN, to primary PCa and further to metastasis (165).

Associations between 8q24/CMYC copy number and GS have also been shown by numerous groups (166-168).

Prior to discovery of 8q24 amplifications, overexpression of the c-myc oncogene in PCa was first discovered by Fleming et al (169). They discovered significant overexpression of c-myc transcripts in PCa versus benign tissue, as well as a group of PCa patients that harboured RNA levels 2-fold higher than mean PCa levels. Attempts at protein expression verification have yielded somewhat confusing results, insofar as localization of c-myc protein (cytoplasmic versus expected nuclear)(165) and equal levels of protein in benign versus cancer tissues (170, 171).

Utilization of a different antibody for detecting myc protein has more recently yielded promising results, showing an increased expression in PIN and PCa tissue versus benign epithelium, although there was no association with 8q24 copy number (172).

8p (NKX3.1) deletions

The initial description of 8p loss existing in a subset of PCa was published by Kunimi et al. and

Bergerheim et al. in 1991, where they found deletions occurring in 50% and 65% of tumours analyzed, respectively (173, 174). Further work by Bova et al. described the 8p21-22 region as being the most common deletion location within the 8p arm (163), while others have shown frequent deletions spanning 8p11-21 (175, 176). With respect to clinicopathogical associations,

Matsuyama et al. reported an association of 8p22 and 8p23-pter loss with moderate and poor versus well differentiated tumours (177, 178). Also, El Gammal et al recently published an analysis of 8p loss (and 8q gain) in a TMA consisting of tumour material from a large cohort of

1,882 PCa patients (179). They described strong associations between 8p loss and pathological stage as well as GS (p-values < 0.0001).

The developmentally important NKX3.1 homeobox gene maps to chromosome 8p21 and was discovered by He et al. to be highly expressed in adult prostate tissue (180). Early research into

NKX3.1 expression changes between normal and PCa tissue yielded conflicting results, with some reporting no change or overexpression (181, 182) while others found decreases in expression that in fact correlated with tumour progression (183). Since then, however, the role of

NKX3.1 loss/underexpression in cancer initiation has been generally well established (184-187), while there appears to be contentious results regarding loss of expression correlating with tumour progression (188-190). It has also been shown that there are other mechanisms controlling

NKX3.1 expression, including DNA methylation (191), and there may be other important undefined tumour suppressor/modifier genes located on the short arm of chromosome 8. Thus, discrepant results regarding 8p loss, NKX3.1 expression, and aggressive PCa may due to

30 variable epigenetics and the concordant loss of other important factors existing in 8p that are deleted.

Other somatic mutations

Recent advances in methods to detect mutations that occur at lower frequencies (i.e. next generation sequencing (NGS)) have resulted in the identification of a number of previously uncharacterized fusions and point mutations. For example, gene fusions SLC45A3:ETV1,

C15ORF21:ETV1, and KLK2:ETV5 have all recently been described in PCa (192). Non-ETS gene fusion events including CDKN1A-CD9 and TNPO1-IKBKB have also recently been identified through RNA-seq (193). Exome sequencing of 112 PCa specimens also identified recurrent point mutations in MED12 , FOXA1, CDKN1B, p53 and SCN11A (194). SPOP was the most common mutation found, occurring in 6-15% of PCa cases across multiple cohorts and existed in a mutually exclusive fashion with TMPRSS2:ETS gene fusions. Such mutations were found to occur in the substrate binding cleft of SPOP (a ubiquitin ligase), and transfection of mutant SPOP increased the invasive capabilities of PCa cell lines in vitro (194).

XMRV

Ribonuclease L (RNASEL) encodes an endoribonuclease that plays a crucial role in interferon viral response, cleaving single stranded RNA and ultimately leading to cytochrome c release from the mitochondria and apoptosis. Further characterization of the RNASEL gene in PCa found that one such variant, R462Q, reduced the catalytic ribonuclease activity of RNASEL 3

31 fold and created deficiencies in apoptosis when cells were exposed to the RNASEL activating, apoptosis inducing agent 2-5A (195, 196).

The correlation between RNASEL variants and PCa risk, as well as the role for RNASEL in innate virus immunity, led to the discovery and initial characterization of Xenotropic MuLV- related virus (XMRV) (197). Urisman et al. reported the presence of XMRV RNA in 40% (8/20) of RNASEL R462Q homozygous patients, compared to just 1/66 heterozygous or homozygous wildtype patients (197). A follow-up study conducted by Fischer et al., however, demonstrated rare prevalence of XMRV in sporadic PCa cases (1%) (198). In addition, the PCa cell line 22Rv1 was found to contain multiple integrated copies of XMRV and produce high titer virus (199).

Since then, there has been conflicting evidence concerning XMRV in human PCa, with some groups reporting low or non-existent infection rates (200, 201) while others continued to support an association with PCa (202). Recent work, however, has pointed to a role for contamination of human samples with XMRV itself or its corresponding nucleic acids (203-205).

1.5.2 Familial and hereditary PCa

HPC1/RNASEL mutations

Approximately 10% of PCa cases are of a hereditary nature, with the remaining 90% being of a sporadic origin (206). Hereditary Prostate Cancer 1 (HPC1) was initially identified as a locus on chromosome 1q24-25 associated with hereditary PCa through a genome wide scan of repeat markers in a total of 91 affected families (207). Follow up studies in additional cohorts

32 confirmed the linkage of HPC1 with hereditary PCa (208), while further work identified variants in RNASEL as the causative factor (209, 210).

BRCA gene mutations

The BRCA1 and BRCA2 genes play important roles in genome stability, particularly through mediating the homologous recombination pathway for DNA repair, although BRCA1 is known to act in a number of important cellular processes including chromatin remodelling (211).

Research in the early 1990’s revealed a distinct increased risk of developing PCa for relatives of women with breast cancer and a clustering of breast and PCa within families (212, 213). This led researchers to candidate genes BRCA1 and BRCA2, two genes where mutations are known to be responsible for hereditary breast cancer (214). BRCA2 mutations have recently been shown to cause a 4.65-8.6 fold increase in risk for PCa for men ≤ 65 years of age, while BRCA1 mutations cause an increase risk of 1.82-3.5 fold (215). In addition to increasing the risk of PCa development, BRCA2 mutated PCa has also has a poor prognosis. An initial study analyzing an

Icelandic PCa cohort revealed earlier age of onset, advanced tumour stage, poorly differentiated tumours, shorter prostate cancer-specific survival times (216). Other studies have reported similar results with respect to stage, differentiation, and overall survival rates (217, 218).

HOXB13 mutations

In 2009, Norris et al. reported the physical interaction between HOXB13 and AR and also illustrated a role for HOXB13 and AR synergy in activating genes that contain a HOX DNA

33 element adjacent to androgen response elements (AREs) (219). Further work showed that

HOXB13 promotes androgen independent signaling in LNCaP cells by activating the E2F pathway via inhibition of p21 (220). Early linkage analysis in PCa susceptibility identified

17q21-22 region as a locus associated with increased risk (221, 222). In 2012, Ewing et al. sequenced this genomic region, consisting of more than 200 genes, in 94 unrelated PCa patients and their family members with family history of the disease and linkage to the 17q region. They discovered four families with a G84E HOXB13 mutation that was present in all 18 men from these families who had PCa (223). In a separate cohort of 5,083 PCa patients, they discovered a mutation rate of 1.4% which was significantly greater than the 0.1% found in healthy controls

(223). Additionally, the mutation was present in 3.4% of early onset familial cases versus 0.6% of late onset sporadic cases.

Single Nucleotide Polymorphisms

Over the past decade there has been a large focus on identifying single nucleotide polymorphisms (SNPs) that are associated with many cancers, include PCa. A recent review summarizes over 40 polymorphisms shown to be associated with increased risk of PCa development (224), and includes SNPs adjacent to NKX3.1, TET2, TERT, and

KLK2/KLK3(PSA), to name a few. Perhaps the most consistently identified SNPs are found on chromosome 8q24, approximately 200-700 kb upstream of the MYC oncogene (225-227), and act independently of each other to affect risk. In addition to PCa, these SNPs are associated with colon, breast, and bladder cancers (228-230). With respect to PCa, odds ratios for these loci range from 1.21 to 1.86, and depending on the specific SNP they have been replicated in up to 5

34 separate studies (224). A unique aspect of the 8q24 SNPs is the follow up functional work that has suggested they lie in MYC enhancer regions that are able to interact with the promoter and promote MYC expression (231-233). Importantly, though, it has not been shown that these SNPs actually affect enhancer activity (234).

1.6 Epigenetics of PCa

Epigenetics can be regarded as heritable changes in gene expression patterns without associated changes to DNA sequence. These include DNA methylation, post-translational histone modifications, and regulation via non-coding RNAs (for example, miRNA) (Figure 1.5).

Aberrant epigenetic changes are now known to be commonplace in neoplasia and can lead to the deregulation of tumour suppressors and/or oncogenes and can cause genomic instability (235)

Figure 1.5. Epigenetics in transcriptional activation and silencing. (Upper panel) Open chromatin formation is mediated by histone H3 lysine 4 trimethylation (H3K4me3, mediated by K4 histone methyltransferases, K4 HMTs), histone acetylation (mediated by histone acetyltransferases, HATs), a lack of DNA methylation, transcription factors (TFs) and co-activating proteins (CO- ACT). TATA-binding protein (TBP), TBP-associated factor (TAF), and RNA polymerase II (RNA-pII) bind to open chromatin and transcribe genes. (Lower panel) Silencing of gene transcription is achieved by removal of acetyl groups from histones (mediated by histone deacetylases, HDACs), H3K9 methylation (mediated by K9 methyltransferases, K9 HMTs), and DNA methylation (depicted as red circles with “M” on black DNA strand). Methyl-binding domain proteins (MBDs), co-repressors (CO-REP), heterochromatin protein 1 (HP1) and chromatin assembly factor 1 (CAF1) are also involved in the silencing process. Adapted from Laird, P. Human Molecular Genetics, 2005; 14(1): R65-R76.

1.6.1 Non-coding RNAs

miRNAs

Micro RNAs (miRNAs) are a recently identified class of non-coding RNAs that range in size from about 21-22 nucleotides and mediate gene expression in a post-transcriptional manner (236,

237). They typically act by base-pairing to the 3’UTRs of mRNA molecules, either resulting in mRNA degradation in cases of full complementarity (238) or translational blocking in cases of partial complementarity (239). To date, there have been numerous miRNAs identified as both up and downregulated in various cancers, with important implications as one miRNA can target dozens of genes belonging to the same or different pathways (240).

Compared to DNA methylation, the field of miRNA biomarker discovery is a new and to date under-studied field in PCa. Several groups have employed genome-wide miRNA profiling techniques in order to discover under and/or overexpression of miRNAs that are associated with aggressive and recurrent PCa, while the results from these groups have had little overlap.

Recently, a study by Schaefer et al. profiled miRNA expression using microarrays in 76 prostatic carcinomas spanning various GSs and pathological stages (241). After selecting the most promising candidates from array studies, they further validated expression in an independent cohort of 79 PCa specimens. Underexpression of miR-31 and miR-205 and overexpression of miR-96 were correlated with higher GSs while underexpression of miR-125b, miR-205 and miR-

222 were correlated with advanced stage (241). miR-221, a candidate miRNA that was significantly underexpressed in PCa tissue but not associated with clinicopathological variables in the study by Schaefer et al., however, was shown to be significantly underexpressed in higher

GSs by a separate group (242). Interestingly, circulating levels of miR-221 were shown to be elevated in the blood of 51 patients with PCa compared to a benign cohort and further elevated in patients with metastatic PCa (243).

Deregulation of other microRNAs that have recently been shown as associated with adverse PCa features include miR-34c (244), miR-20a (245), miR-143 and miR-145 (246) and serum levels of miR-195, let-7i , miR-141, and miR-375 (243, 247, 248).

Long non-coding RNAs

Long non-coding and long intergenic non-coding RNA molecules (lnc/lincRNAs) are non- coding RNAs ≥200 nucleotides that can influence gene expression through several different mechanisms that include classic double stranded RNA interference (similar to miRNAs), splicing regulation, and serving as a guide molecule for epigenetic modifying proteins (249). The number of functional lnc/lincRNAs and the extent of the biological role that they may play has only recently begun to be appreciated through genome wide RNA mapping studies, which began in earnest with high density tiling array analysis and has since continued with RNA-seq studies

(250, 251) and functional work. For example, the lincRNA HOTAIR was initially identified through high-resolution tiling analysis of the HOX loci and was mapped to the HOXC cluster of genes (250). Rinn et al. went on to show that HOTAIR binds the polycomb repressive complex 2

(PRC2) and guides histone H3 lysine 27 trimethylation (H3K27me3. Other groups have shown a similar mechanism of lincRNAs guiding histone modifications with HOTTIP guiding MLL complexes to activate HOXA gene transcription (252).

Several lncRNAs have been identified as aberrantly expressed in PCa, although there are a limited number of studies that assess the function of these molecules. Perhaps the best known is prostate cancer antigen 3 (PCA3), which is in clinical use as a marker for PCa diagnosis

(discussed above) (32). Aside from PCA3, Srikantan et al. discovered a prostate specific lncRNA, PCGEM1, that was overexpressed in 84% of PCa samples analyzed (253). Further work identified this ncRNA as significantly overexpressed in African American men with PCa and revealed a role in promoting cell proliferation and inhibiting apoptosis (254, 255). The

PRNCR1 transcript is a 13 kilobase intronless lncRNA that is transcribed from the 8q24 locus and upregulated in PCa (256). siRNA knockdown of PRNCR1 ncRNA resulted in reduced cell viability and reduced AR transactivation, implicating this transcript in oncogenesis through regulation of AR signaling. Further to these specific examples, recent RNA-seq of 102 prostate cell lines and specimens identified 121 unannotated ncRNAs associated with PCa (257). One of these transcripts, PCAT-1, is a regulator of cell proliferation that is targeted by the EZH2 histone methyltransferase.

1.6.2 Histone modifications

Coupled with DNA methylation, post-translational histone modifications act to regulate gene expression at the transcriptional level. These modifications are wide-ranging and include, for example, the addition of acetyl, methyl, and phosphate groups to protruding histone tails. The acetylation of lysine residues reduces the interaction of positively charged histone proteins with negatively charged DNA, creating an open chromatin conformation conducive to active gene transcription (258). Methylation of histone lysine residues is more complex than acetylation as it

39 can occur in mono-, di-, or trimethylated forms, and the effects are largely dependent on which lysine residue is modified. Promoter region H3K4 di- and trimethylation is associated with actively transcribed genes while H3K9 and H3K27 di- and trimethylation is associated with repressed genes (259). Furthermore, H3K4 mono- and dimethylation is associated with tissue- type specific gene enhancer regions.

In 2009, Ellinger et al. published a study analyzing global levels of histone modifications

H3K4me1/2/3, H3K9me1/2/3, as well as H3 and H4 acetylation (H3Ac and H4Ac) (260). They found reduced levels of H3K4me1, H3K9me2, H3K9me3, H3Ac, and H4ac in PCa compared to benign tissue. Additionally, reduced H3K4me1 was a significant predictor of biochemical recurrence, while reduced levels of each type of histone methylation were strongly correlated with GS. Gene specific mapping of histone modifications has also revealed intriguing patterns in both benign and cancerous prostate cells. For example, numerous loci are marked as “poised” in benign EP156T and PC3 cells, having both an active H3K4me3 signature and a repressed

H3K27me3 signature, and the most frequent change between these two cells lines was gain or loss of H3K27me3 which had little or no effect on gene expression (261). Furthermore, Yu et al. performed genome-wide mapping of H3K27me3, identifying a list of genes that were also polycomb group targets in embryonic stem cells (262). By integrating this data with publicly available gene expression data, they developed a “polycomb repression signature” associated with poor prognosis.

Histone modifying enzymes

There are a wide range of enzymes involved in dynamically altering histone modification status that include acetyltransferases, deacetylases, methyltransferases, demethylases, kinases, and ubiquitilases (Table 1.1). Histone deacetylases (HDACs) remove acetyl groups from histone proteins, thereby acting to repress gene transcription (263). In PCa, altered expression of

HDACs, and in particular HDAC1, has been well described and correlates with GS, pathological stage, biochemical recurrence, and CRPC (264-266). SIRT1, another deacetylase that has histone and non-histone targets, is also significantly elevated in PCa (267) and affects cell growth, motility and invasion of PCa cells (268, 269).

Enzymes that Modify Histones Residues Modified Acetyltransferase HAT1 H4 (K5, K12) H3 (K14, K18) H4 (K5, K8) H2A (K5) H2B (K12, CBP/P300 K15) PCAF/GCN5 H3 (K9, K14, K18) TIP60 H4 (K5, K8, K12, K16) H3 K14 HB01 H4 (K5, K8, K12) Lysine Methyltransferase SUV39H1 H3K9 SUV39H2 H3K9 G9a H3K9 ESET/SETDB1 H3K9 EuHMTase/GLP H3K9 CLL8 H3K9 MLL1 H3K4 MLL2 H3K4 MLL3 H3K4 MLL4 H3K4 MLL5 H3K4 SET1A H3K4 SET1B H3K4 ASH1 H3K4 SET2 H3K36 NSD1 H3K36 SYMD2 H3K36 DOT1 H3K79 Pr-SET 7/8 H4K20 SUV4 20H1 H4K20 SUV420H2 H4K20 EZH2 H3K27 RIZ1 H3K9

Table 1.1. Mammalian histone lysine acetyl- and methyltransferases and the residues they modify. Enzymes which specifically modify one/a few residues have been included. Modified from Kouzarides, T. Cell. 2007; 128(4): 693-705.

EZH2 is a subunit of the polycomb repressive complexes 2, 3 and 4, functioning as the enzyme responsible for trimethylation of lysine residue 27 of histone H3 (270). In 2002, Varambally et al. first described overexpression of EZH2 in metastatic CRPC and in a subset of primary PCa cases which had a poor prognosis (271). Furthermore, ectopic expression of EZH2 resulted in transcriptional repression of a specific subset of genes that was dependent on the SET domain (a necessary domain for methyltransferase activity). Knockdown of endogenous EZH2 also resulted in reduced cell proliferation. Additional work also elucidated a role for EZH2 in invasion (272) and found that ERG expression upregulates EZH2 which then represses transcription of the

NKX3.1 tumour suppressor (273, 274). Interestingly, EZH2 has also been linked to aberrant

DNA methylation patterns in PCa (275), and in particular represses DAB2IP, a gene which inhibits epithelial-mesenchymal transition (276, 277).

1.6.3 DNA methylation

The presence of 5-methylcytosine in mammalian DNA is essential for mammalian development

(278, 279) and was first proposed to serve as an epigenetic mark in mammals in the 1970s (280,

281). In mammals, 5-methylcytosine occurs primary in the context of CpG dinucleotides, although it was recently found to also exist in CpNpG (where N is A, T, or C) in human tissue as well (282). Overall, CpG dinucleotides are found to occur at a lower than expected frequency throughout the human genome (283). CpG islands (pockets of GC rich DNA that contain a higher than expected CpG frequency), however, are found scattered throughout the human genome, with approximately half of all genes containing these islands within their promoter regions (283). Methylation occurring at the 5’ carbon of cytosine within CpG dinucleotides is

43 known to contribute to gene silencing when it occurs excessively in the promoter region of genes

(284, 285).

The role of histone acetylation and methylation events have been well described in carcinogenesis (for review see (286)) and are intricately linked to DNA methylation status in maintaining either an active or repressed state of transcription (287). Aberrant CpG dinucleotide methylation is a hallmark of many cancers and is thought to contribute to cancer development.

DNA methylation is thought to contribute to carcinogenesis on two levels: an overall hypomethylated genomic state leading to chromosomal instability and a tumour suppressor/oncogene specific differential methylated state that often leads to alterations in gene expression (288). The latter of these two usually occurs in CpG islands within the promoter region of genes, although recent work has identified tumour specific methylation events in CpG islands outside of the promoter region as well as CpG “shores”, which lie adjacent to CpG islands and are less dense in terms of CpG content (289). Gene silencing through promoter hypermethylation is thought to be the result of methylated cytosine binding proteins (MBDs) binding in conjunction with the recruitment of histone deacetylases and other epigenetic factors

(158), although it has also been shown to directly inhibit transcription factor binding (290). In addition, CpG shore methylation has been shown as associated with transcription levels on a global and gene specific level (289). The consequence of DNA methylation in CpG islands/shores outside of canonical promoters is an active area of research and is likely context dependent. For example, gene body methylation correlates with active transcription (291, 292) and does not block transcription elongation in mammals as it does in other organisms (293). One possible role for intragenic DNA methylation may be to silence transcription beginning at alternative transcription start sites (TSSs) (294). In addition to this role, others have shown that

44 gene body methylation can affect RNA splicing. In the CD45 gene body, for example, CTCF binds intragenically and promotes pausing of RNA polymerase II as it transcribes, which in turn results in inclusion of exon 5 (295). If this region is methylated, however, CTCF cannot bind and exon 5 is spliced out of the mRNA product. There is considerably less known regarding the function of DNA methylation at intergenic regions. One possibility is that a hitherto uncharacterized lncRNA or miRNA resides adjacent to differentially methylated regions and thus their expression is regulated by methylation, similar to H19 regulation by DNA methylation

(296). Yet another possibility is that DNA methylation occurs at enhancer regions of DNA which exist distally to genes that they activate. Indeed, differential methylation has been shown at enhancer regions, in particular for genes involved in differentiation (297, 298). Intergenic DNA methylation can also occur at insulator regions (which can block enhancer/promoter interactions) of the genome and lead to inhibition of CTCF binding and subsequent activation of transcription

(299).

Methylated loci in PCa

To date, there have been over 50 genes identified as aberrantly methylated in PCa (see Table 1.2 for representative list). The most well studied methylated gene, GSTP1, was discovered by Lee et al. in 1994 (300) and has since been established as a very promising marker for early diagnosis

(301).

Gene Designation Frequency, % AR Androgen receptor 15–39 ESR1 Estrogen receptor 1 19–95 ESR2 Estrogen receptor 2 83–92 RARβ2 Retinoic acid receptor β2 68–95 RARRES1 Retinoic acid receptor responder 1 (TIG1) 55–96 CCND2 Cyclin D2 32–99 CDKN2A Cyclin-dependent kinase inhibitor 2A (p16) 3–77 EDNRB Endothelin receptor type B 15–100 RASSF1A Ras association domain family protein 1 isoform A 53–99 RUNX3 Runt-related transcription factor 3 27–44 APC Familial adenomatous polyposis 27–100 CDH1 E-cadherin 27–69 CDH13 Cadherin 13 45–54 CD44 Cluster differentiation antigen 44 19–72 TIMP3 TIMP metallopeptidase inhibitor 3 0–97 GSTP1 Glutathione S-transferase P1 79–95 MGMT O-6-methylguanine DNA methyltransferase 0–76 ASC Apoptosis-associated Speck-like protein containing a CARD 37–78 BCL2 B cell lymphoma 2 52–87 DAPK Death-associated kinase 0–36 MDR1 Multidrug resistance receptor 1 51–100 PTGS2 Prostaglandin endoperoxidase synthase 2 18–88 HIC Hypermethylated in cancer 99–100

Table 1.2. Commonly methylated genes in PCa. Modified from Jeronimo, C et al.. European Urology, 2011; 60(4): 753-766.

With respect to grade and stage associated DNA methylation events, several well studied genes include APC, RASSF1A, and RARβ2. Each of these genes was initially discovered to be methylated in malignancies other than PCa (302-304) and subsequently reported in PCa by

Maruyama et al. in 2002 (140). A number of studies have shown an association between APC methylation and aggressive PCa. In 2004, Kang et al. examined methylation of APC in a series of 37 PCa tissues and discovered an association with high PSA and high GS (305), while Cho et al. showed that APC methylation significantly correlated with pathological stage pT3/pT4 and intermediate/high grade PCa (306). More recently, increased levels of APC methylation have been found in organ confined pT2 versus pT3 cancers, intermediate GS 7 versus ≤6 cancers, and in high grade GPs 4 and 5 versus low grade pattern 3 (141).

An association between RASSF1A methylation and grade and stage has also been verified in numerous studies. Methylation of the RASSF1A promoter was associated with Gleason ≥7 cancers compared to GS 4-6 cancers in a cohort of 52 PCa patients (307). Using quantitative methylation profiling approaches, Jeronimo et al. demonstrated a correlation of RASSF1A methylation with increased tumour stage but not with tumour grade (Jeronimo 2004).

Furthermore, in a study of 219 prostate carcinomas using the quantitative methylation profiling technique MethyLight, a higher degree of methylation was found in GP 4 and 5 carcinoma compared to GP 3, although no association was observed between RASSF1A methylation and overall GS or pathological stage (141).

Using genome-wide methylation profiling techniques, many groups have published novel candidates for aggressive PCa markers. In 2007, Cottrell et al. hybridized DNA from 304 PCa specimens to an array representing 256 CpG sites throughout the genome and discovered methylation of GPR7, NOTCH, ABHD9, KBTBD6, and an expressed sequence on chromosome

3 that was significantly greater in high GS (8-10) versus low GS (2-6) cancers (308). Using an in silico approach, Perry et al. utilized publicly available transcriptome data along with CpG island analysis to identify IGFBP3 as potentially methylated in PCa (309). Real-time methylation analysis in a validation set of tumours revealed that IGFBP3 is methylated and that this correlates with GS ≥7 PCa. Perhaps one of the strongest cases for DNA methylation in terms of

PCa prognosis lies with the paired-like homeodomain 2 (PITX2) gene. Methylation of PITX2 was initially described in acute myeloid leukemia (310) and has since been discovered in breast cancer (311), lung cancer (312), and astrocytomas (313). In 2009, two groups suggested that

PITX2 was associated with recurrent PCa (314, 315). A multi-centre evaluation of the marker confirmed these findings (316), and a diagnostic Affymetrix microarray was developed to test specifically for PITX2 methylation (317).

DNA hypomethylation in events in PCa development and progression have been poorly characterized. Hypomethylation of the cancer/testis associated gene was the first example of a gene-specific hypomethylation event described in PCa, although the frequency was <40% (318).

Subsequent studies found gene-specific hypomethylation of heparanase (319), CYP1B1 (320),

XIST (321), and uPA (322) to name a few. Loss of imprinting and hypomethylation have also been found as a field-effect in benign tissue adjacent to PCa (323). In contrast to gene-specific hypomethylation, genome-wide hypomethylation, has been well characterized since its initial discovery in PCa in 1987 (324). In particular, hypomethylation of the retrotransposable element

LINE-1 was first discovered in a panel of 32 PCa specimens and increased in frequency with tumour stage (325). Further work by Schulz et al. showed a strong correlation between 8q loss,

LINE-1 hypomethylation, and the presence of metastases (326). Consistent with these results,

Yegnasubramanian et al. showed that genome wide hypomethylation arises later in PCa

48 progression than certain gene specific methylation events using several different methylation detection techniques (327).

Recent work in DNA methylation profiling is bringing the field beyond gene specific, promoter specific, and CpG island specific methylation identification through the use of next-generation sequencing. For instance, reduced representation bisulfite sequencing (RRBS) has identified aberrant methylation patterns in HOX loci and WNT signaling pathway genes in chronic lymphocytic leukemia (328), while Meissner et al. similarly employed RRBS to uncover methylation patterns of pluripotent and differentiated cells (329). With respect to PCa, Kim et al. used MethylPlex-next-generation sequencing to interrogate the methylome of 22 tissue specimens and 6 cell lines (330). They characterized 2,481 differentially methylated promoter regions as PCa-specific and found that DNA methylation and H3K4me3 might work to silence and activate, respectively, alternative promoters within genes. Furthermore, stratification of PCa cases into ERG fusion positive versus ERG fusion negative revealed greater LINE-1 methylation in ERG positive PCa.

DNA methyltransferases

The DNA methyltransferases (DNMTs – DNMT1, DNMT3A, DNMT3B) are responsible for enzymatic addition of methyl groups to cytosine in CpG dinucleotides. DNMT1 is considered the maintenance methyltransferase as it shows a significant preference for hemi-methylated double stranded DNA (331), while DNMT3A and DNMT3B are considered de novo methyltransferases, preferentially targeting completely unmethylated DNA (279, 332). These definitions should not be considered absolute, though, as recent work points to overlapping roles for DNMTs (333,

334). In addition to the DNMTs listed above, DNMT2 and DNMT3L were also identified as belonging to the DNA methyltransferase family by homology, but both proteins were found to lack enzymatic activity (335, 336). It was found that DNMT2 is an RNA methyltransferase that specifically methylates aspartic acid tRNA (335), while DNMT3L interacts with DNMT3A and

DNMT3B to regulate maternal methylation imprinting and interacts with HDAC1 to repress gene transcription (336, 337).

Aberrant expression of DNMTs has previously been shown in PCa, suggesting this may be a contributing factor to abnormal DNA methylation patterns. Overexpression of the DNMT1 enzyme in PCa specimens was initially shown in conjunction with HDAC1 overexpression in

2001 (264). In the transgenic adenocarcinoma of the mouse prostate (TRAMP) model, DNMT1 expression was significantly increased in PIN lesions and adenocarcinomas compared to benign prostate (338). Furthermore, Zhang et al. demonstrated increased DNMT1 expression correlating with adverse pathological features and biochemical recurrence, and that blocking of TGFβ signaling resulted in a reduced in DNMT1 expression (339). DNMT3A and DNMT3B expression have also been shown as upregulated in PCa (275, 340, 341). Interestingly, there are conflicting reports regarding the associations between DNMT overexpression and DNA methylation in PCa. In 2008, Morey Kinney et al. reported a weak correlation between DNMT1 expression and locus-specific methylation with no correlation between DNMT3A or DNMT3B expression and methylation (338). Hoffmann et al. demonstrated a lack of association with respect to DNMT1 and DNMT3B expression and methylation (275), while Kobayashi et al. have shown correlations between expression of the DNMT3A splice variant DNMT3A2 and

DNMT3B with hypermethylation of 5,912 genes in PCa (341). In the latter study EZH2 was also associated with DNA hypermethylation, verifying an earlier observation by Hoffmann et al..

DNA demethylation and hydroxymethylation

While much has been elucidated regarding the mechanism of DNA methylation in the past 40 years, comparatively little has been discovered in the understanding of DNA demethylation. In plants, the means through which active removal of methyl groups from 5-methylcytosine (5mC) is achieved has been clear for decades and involves removal of 5mC via 5mC specific DNA glycosylases (342). In mammals, both a passive and active mechanism likely exist to achieve demethylation. Evidence for the former method exists in the maternal genome following fertilization, where demethylation occurs following several cleavage divisions (343). The paternal genome, however undergoes active demethylation within 6-8 hours following fertilization (343) while primordial germ cells also exhibit rapid demethylation during embryonic development (344). Specific loci also undergo active demethylation in somatic cells, such as the

IL-2 promoter following T-lymphocyte activation (345). Early work attempting to understand how active DNA methylation is achieved in mammals identified several putative pathways including direct enzymatic removal of 5mC through MBD2 (346), the base excision repair

(BER) involving T DNA glycosylase (TDG) or MBD4 with or without prior deamination of

5mC to T (347-349), and the nucleotide excision repair pathway involving GADD45A (350).

The promise proved to be short-lived, however, as the above mechanisms have been shown to either have poor activity towards removal of 5mC versus T, for instance, or loss of function studies have resulted in no observable DNA methylation defects (349, 351).

The recent characterization of the ten-eleven translocation (TET) family of proteins in mediating conversion of 5mC to 5-hydroxymethylcytosine (5hmC) has emerged as the most likely mechanism for active DNA methylation. In 2009, Tahiliani et al. reported the presence of 5hmC

51 in embryonic stem (ES) cells and illustrated that 5hmC levels increase upon depletion of TET1

(352). It was subsequently shown that all three members of the TET Family (TET1, TET2, and

TET3) convert 5mC to 5hmC (353). In addition, Bocker et al. analyzed 5mC, 5hmC and expression levels of the HOXA cluster upon retinoic acid induced differentiation, demonstrating a TET2 dependent increase in 5hmC and reduction in 5mC that mirrored collinear activation of the HOXA cluster (354). The exact mechanism that leads to unmodified cytosine from 5hmC remains unknown, but may involve subsequent oxidation of 5hmC into 5-formylcytosine (5fC) and 5fC into 5-carboxylcytosine (5caC), along with involvement of the AID/APOBEC family of cytidine deaminases and/or DNA glycosylases/BER (355).

The field of 5hmC profiling and active DNA demethylation is in its infancy with respect to cancer research. TET1 was initially discovered as a fusion gene present in patients with mixed lineage leukemia (MLL), although the potential for interruption of DNA demethylation was unknown at the time (356). Mutations in TET2 have only recently been identified in myeloid cancers (357-359) and shown to associate with specific 5mC and 5hmC profiles in chronic myelomonocytic leukemia (CML) patients (360). In a 2011 publication by Haffner et al., IHC evaluation revealed a marked reduction in global 5-hmC in breast, colon, and prostate carcinoma as compared to benign tissue (361). One subsequent study has supported the loss of 5hmC in various cancers and also indicated substantial loss of expression of all three TET genes (362).

1.7 Homeobox genes

Homeobox genes are a family of transcription factors that contain a highly conserved 180 bp region knows as the homeodomain, which encodes for a 60 amino acid helix-turn-helix DNA

52 binding motif responsible for sequence specific recognition of DNA elements (363). These genes can be found in genomic clusters and often are expressed in a spatial and temporal manner, playing key roles during developmental stages of an organism (364). The class I homeobox

(HOX) genes are divided into 4 groups (HOXA-D) on different chromosomes, with each group consisting of 9-11 genes (39 in total) that are particularly important in determining anterior/posterior patterning. Numbering of HOX genes begins at the 3’ end of the cluster and extends to the 5’ end for a total of 13 paralagous groups with each of the four clusters missing various members of the 13 genes (see Figure 1.6). Typical class I HOX genes consist of two exons surrounding a sole intron with the homeodomain located within the second exon, although this gene architecture is not always found.

Figure 1.6. Organization of the human HOX clusters. Colour code represents spatial collinearity where yellow genes are expressed in a posterior setting and purple coloured genes in anterior regions. Adapted from Shah et al.. Nature Reviews Cancer, 2010; 10: 361:371.

HOX genes exert their transcriptional effects in combination with other classes of homeodomain containing proteins, namely the TALE class of homeodomain proteins (365, 366). The TALE class of genes consist of PBX and MEIS gene families which can interact with a hexapeptide domain adjacent and N-terminal to the homeodomain of HOX genes. PBX and MEIS proteins interacting with HOX proteins can result in either an activating or repressing effect on gene transcription (367). There are numerous other proteins that physically interact with HOX proteins, such as SMAD4 with HOXC9 (368), SMAD4 with HOXA9 (369), and CBP with

HOXB6 (370).

Aside from the well described and delineated nature of protein coding genes, recent work has identified a plurality of transcribed RNA originating from HOX loci without coding potential and with unknown function. HOTAIR and HOTTIP are two such ncRNAs that have a characterized function in targeting of epigenetic machinery (described above). In the initial study characterizing the HOTAIR ncRNA, though, 230 other transcribed regions were discovered that demarcated histone modification patterns (250). Whether these play a role in altering the chromatin state of surrounding HOX genes, altering the chromatin state in trans across the genome, or are simply byproducts of active HOX gene transcription remains unknown. There are also 5 miRNA genes encoded within HOX clusters (mir-10a, mir-10b, mir-196a-1, mir-196a-2, and mir-196b) at two distinct homologous positions shared across all four clusters which act to regulate expression of HOX genes (238, 371).

1.7.1 Role in development

Homeobox genes were initially identified in Drosophila in the early 1980s (363) and were initially characterized by their ability to cause homeotic transformations (alterations in positional identity of body segments) when mutated. In mammals, HOX cluster genes are essential for proper development along the anterior-posterior axis (372). During embryogenesis, there is a specific spatial and temporal pattern of HOX expression from the 3’ to 5’ end of each cluster.

Typically, 3’ end genes (e.g. HOXD1) are expressed in anterior regions and early in development while 5’ genes (e.g. HOXD13) are expressed in posterior regions and later during development. In addition, HOX genes of the 5’ end (posterior genes) display a dominant phenotypic effect as compared to 3’ end genes, a term known as posterior prevalence. In

Drosophila, for example, mutations causing derepression of the homeotic cluster results in a phenotype consistent with overexpression of the most posterior gene (Abd-B), and in the absence of Abd-B the phenotype is consistent with overexpression of the next most 5’ gene, abd-A (373).

HOX genes also play a role in limb development and organogenesis, as HOXA and HOXD genes are required for limb, digit, and genitalia development (374-376). For example, mutations of HOXD13 are associated with synpolydactyly (a fusion of fingers) and brachydactyly

(shortness of fingers and toes), while HOXA13 mutations cause hand-foot-genital syndrome

(resulting in deformities of the hands and feet). Group 13 Hox genes including Hoxa13 and

Hoxd13 are also essential for proper growth of the mouse prostate, as loss-of-function studies have revealed reduced branching of prostate ducts and improper lobe development (377-379).

1.7.2 Role in cancer

Importantly, expression of specific HOX genes is not limited to embryonic development but has also been detected in adults in a highly tissue specific fashion (380), indicating a role for these genes in maintaining a differentiated tissue phenotype. For example, expression of HOXC4 is associated with lymphoblastic differentiation while expression of HOXC11 and a novel HOXC6 isoform causes neuroblastoma cells to differentiate into Schwann cells (381, 382). Thus, controlled expression of specific HOX genes is essential and carried out under an intricate and multifaceted regulatory system consisting of signaling pathways and epigenetics (383-386), which is discussed in detail below.

Deregulated expression of homeobox genes has been observed in a variety of cancers including leukemia, lung and prostate carcinomas (387-389). In leukemia, upregulation of HOXA9,

HOXA10, and HOXC6 have been shown as the result of MLL fusions (390), with increased cell proliferation and evasion of apoptosis as a result of HOXA9 overexpression (391). NUP98-

HOXA9 fusions are also found in a subset of myeloid leukemias resulting in aberrant transcription profiles (392, 393). Overexpression of HOXB7 and HOXB13 have been shown in breast cancer, with the former causing EMT with increased invasive capabilities and the latter being associated with resistance to tamoxifen treatment in ER positive tumours (394, 395). In

PCa, NKX3.1, a non-HOX homeobox gene residing on 8p21, has been identified as frequently deleted and downregulated in PCa (described in detail above) (180). HOXC4, HOXC5, HOXC6 and HOXC8 have also been shown as overexpressed in PCa (389), with HOXC8 overexpression correlating with loss of differentiation, increasing invasion in vitro, and blocking the interaction between AR and steroid receptor coactivator 3 (396, 397). Conversely, HOXB13 is downregulated in a subset of PCa, resulting in abnormal androgen signalling and growth promoting effects (398, 399). This can be explained via a direct interaction between HOXB13

57 and AR, which mediates downstream androgen signalling in PCa (219). Other examples of abnormal HOX gene expression include downregulated HOXA4 corresponding with shorter overall survival in acute myeloid leukemia (400), HOXA5 inactivation via promoter methylation in breast cancer (401), and HOXB13 overexpression in ovarian cancer (402).

1.7.3 Epigenetics of HOX clusters

Epigenetic regulation of the HOX cluster genes is essential in maintaining proper cell-type specific expression patterns. Much of this is mediated by trithorax and polycomb group proteins, which activate expression in HOX clusters via H3K4 trimethylation and repress transcription via

H3K27 trimethylation (403, 404). DNA methylation and ncRNA regulation are also key factors in establishing a transcriptional state across HOX clusters (405, 406).

One of the most well known oncogenic gene fusion events in leukemia occurs when a trithorax group member, MLL, is fused to any one of several possible gene partners (407). These fusions are found in virtually all mixed lineage leukemias and a subset of acute lymphoblastic and acute myeloid leukemias (408). This results in aberrant epigenetic modifications in HOX gene clusters which can affect multiple tumourigenic processes including proliferation and differentiation

(discussed above) (409). EZH2, a PRC2 protein complex member, is overexpressed and is a marker of aggressive disease in PCa and breast cancer (271, 410), with overexpression also being described in a number of other malignancies (411-413). Co-operation between EZH2 and other epigenetic modifying enzymes has been shown to regulate HOX gene expression (414), suggesting that the phenotype observed in EZH2 overexpressing cancers may be in part due to aberrant HOX gene silencing.

Both miRNA and lincRNAs have also been shown to regulate HOX gene expression patterns. In

2007, Han et al. showed that mir-10a potentially inhibits HOXA3 and HOXD10 (415), while mir-10b has also been shown to target HOXD10 in breast cancer with concomitant increases in

RHOC, which is a well known gene involved in metastasis (416). Interestingly, mir-10a has also been shown to result in silencing of HOXD4, although this effect appears to be transcriptional silencing via DNA methylation and H3K27me3 (417). With respect to lincRNAs, the HOXC transcribed HOTAIR RNA causes gene silencing in posterior HOXD genes through targeting of the PRC2 complex and H3K27me3 (250). HOTTIP functions in a similar manner to HOTAIR, but acts in a converse manner to activate gene transcription of 5’ HOXA genes via targeting of

MLL complexes and H3K4me3 (252).

Atypical patterns of DNA methylation in HOX gene clusters have been described in a number of cancers including lung (418-420), leukemia (420), cholangiocarcinoma (421), and gastric cancers (422). While the mechanism behind targeting of DNA methylation in the HOX clusters is unclear, recent work has suggested interplay between histone modifications and DNA methylation. For example, H3K4me3 was shown to protect HOXA9 from aberrant DNA methylation (423), while having a bivalent chromatin state (that is, the colocalization H3K4me3 and H3K27me3) without H3K9me3 in embryonic stems cells protects from DNA methylation

(424). Other groups have shown that genes marked with H3K27me3 undergo de novo DNA methylation in cancer (425) which may be a result of direct interactions between EZH2 and

DNMTs (426).

1.8 Rationale, hypothesis and objectives

Genetic mutations, such as TMPRSS:ETS gene fusions, PTEN deletions, and NKX3.1 deletions have well described roles in the intiation of PCa, as do epigenetic lesions such as GSTP1 and

APC methylation. Given the current “overdiagnosis” problem in PCa, however, it is important to better delineate aberrant genetic and epigenetic events that may relate to disease progression and prognosis.

Prior studies have attempted to discern genetic differences that may exist in well differentiated

GP3 to poorly differentiated GP4 (427, 428). These studies had a limited degree of success, however, as the study by Skacel et al. found chromosomal changes in high grade patterns (427), but was limited by small sample size and the study by van Dekken et al. did not find genetic differences in patterns 3 and 4 of GS 7 tumours using an array based CGH technique (428). The role that epigenetics may play in poorly differentiated from well differentiated PCa has not been well established. Specifically, DNA methylation may have a role as both genomic hypomethylation and gene specific hypermethylation have been shown to increase in tumour progression. Importantly, DNA methylation events may serve as very useful biomarkers (429) for a number of reasons: (i) they result in a positively detectable signal, (ii) methylation is a stable modification to DNA, and (iii) DNA can be easily isolated and analyzed from archived paraffin embedded material with sensitive detection techniques. Accordingly, recent studies have analyzed methylated genes in PCa along with prognostic indicators, each identifying an independent set of markers (314, 430, 431). However, most studies have not investigated the PCa

“methylome” in primary prostate tissue, limiting their analysis to a set of previously identified genes.

Methylation profiling and validation of candidate genes in primary PCa tissue, as presented below in Chapters 2 and 3, revealed HOXD3 and HOXD8 as potential prognostic indicators in

PCa. Given the strong prognostic associations between DNA methylation of HOXD3 and

HOXD8, we sought to analyze the expression of both genes in PCa and the prognostic implications, as well as analyzing the putative role of HOXD8 in PCa progression. Currently,

HOXD3’s role in carcinogenesis has only been characterized in lung cancer and melanoma, where overexpression contributes to motility and invasion (432-434). Downregulation of

HOXD8 expression has been shown in liver metastatic colon cancer (435) and DNA methylation has been shown in mantle cell lymphomas (436), but the functional effect of abnormal HOXD8 expression has not been explored in any cancer. In lymphatic endothelial cells, however,

HOXD8 has been shown to contribute to lymphangiogenesis through altered expression of

PROX1 (437). The HOXD8 paralog, HOXC8, has been functionally characterized in PCa where it promotes invasion and may lead to androgen independent cell proliferation (397, 438).

Therefore, HOXD8 may play a similar role in proliferation and cell motility/invasion of PCa.

1.8.1 Hypothesis and Objectives

I hypothesize that a panel of methylation markers is capable of distinguishing GPs and/or score, and may be used in conjunction with GS in establishing patient prognosis. Moreover, I hypothesize that aberrant HOXD3 and HOXD8 contribute to the pathogenesis of the disease, with HOXD8 contributing to invasive disease characteristics.

In order to address this hypothesis I have the following aims:

1. To discover novel methylated genes correlating with the Gleason grading system

2. To validate these genes in an independent set of PCa cases and assess their suitability as potential biomarkers

3. To assess the expression patterns of HOXD3 and HOXD8 and uncover the functional role of

HOXD8 in PCa

Chapter 2

Discovery of novel methylated loci associated with PCa progression

Ken Kron1,2, Vaijayanti Pethe1, Dominique Trudel3, Nino Demetrashvili1,4, Michael Nesbitt5,

John Trachtenberg6, Neil Fleshner6, Laurent Briollais1, Theodorus van der Kwast2,3, and Bharati

Bapat1, 2

1. Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON, Canada

2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON,

Canada

3. Department of Pathology, University Health Network, University of Toronto, Toronto, ON,

Canada

4. Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada

5. University Health Network, University of Toronto, Toronto, ON, Canada

6. Department of Surgical Oncology, Division of Urology, University Health Network,

University of Toronto, Toronto, ON, Canada

The work in this chapter was primarily contributed by Ken Kron. VP, a research assistant, performed DNA extraction and DMH for the GS6 and GS8 microarray cases. DT, a pathology

resident, read the ERG immunohistochemistry. ND, an MSc student, and LB, a principal

investigator, performed probe level analysis of CpG island array data. Clinical information for

patients was obtained by MN (clinical research coordinator) and JT (a surgeon), while samples

were retrieved by NF (a surgeon) and TVDK (a pathologist).

The GS8 and GS8 probe level analysis of this chapter has been published in PLoS ONE: Kron K,

Pethe V, Briollais L, Sadikovic B, Ozcelik H, Sunderji A, et al. Discovery of novel hypermethylated genes in prostate cancer using genomic CpG island microarrays. PLoS One.

2009;4(3):e4830.

Chapter 2

Discovery of novel methylated loci associated with PCa progression

2.1 Summary

Promoter and 5’ end methylation regulation of tumour suppressor genes is a common feature of many cancers. Such occurrences often lead to the silencing of these key genes and thus they may contribute to the development of cancer, including prostate cancer. In order to identify methylation changes in prostate cancer, we performed a genome-wide analysis of DNA methylation using Agilent human CpG island arrays. Using computational approaches we have identified a large number of potential epigenetic biomarkers, enriched in the homeobox gene family, that are associated with aggressive PCa. Specific genes including HOXD3, TGFβ2,

HOXD8 and GENE X were selected based on methylation correlations with GS and biochemical recurrence. Further work identified a subset of genomic loci with differential DNA methylation

This study identifies novel epigenetic biomarkers of PCa progression, ERG oncogene expression, and provides a global assessment of DNA methylation changes in PCa. Thus, incorporating DNA methylation markers have the potential to be incorporated into clinical testing, and could serve to identify a subset of patients who require aggressive disease treatment.

2.2 Introduction

CpG island hypermethylation in PCa is a common event with over 50 hypermethylated loci currently identified. The best characterized of these events, GSTP1 promoter methylation, occurs in >90% of cancers and 70% of precursor high grade prostatic intraepithelial neoplasia (PIN) lesions and can also be detected in blood and urine samples. Thus, GSTP1 methylation may serve as a useful diagnostic marker for PCa. Recently, substantial progress has been made in the high-throughput epigenomic screening for the identification of novel targets of DNA methylation. Subsequently, other well characterized hypermethylated genes have been identified in PCa (see Table 1.2). However, no gene studied to date has been identified as a specific diagnostic/prognostic biomarker in PCa similar to GSTP1.

As mentioned in Chapter 1, previous studies have attempted to discern genetic and expression based differences that may exist between low, intermediate, and high grade PCa. The role that epigenetics may play in the identification of aggressive PCa has not been adequately addressed.

In particular, gene specific hypermethylation events have been shown to increase in tumour progression (439-441). These DNA methylation events may serve as very useful biomarkers for a number of reasons, as they are (a) positively detectable signals, (b) stable modifications to

DNA, and (c) DNA can be easily isolated and analyzed from archived paraffin embedded material with sensitive detection techniques. Accordingly, recent studies have analyzed methylated genes in PCa along with prognostic indicators, each identifying an independent set of markers (314, 430, 431). However, most studies have not investigated the PCa “methylome” in primary prostate tissue, limiting their analysis to a set of previously identified genes. In this study, we sought to analyze methylation on a genome wide scale using human CpG island microarrays to uncover novel methylated loci associated with aggressive PCa. In addition, we sought to determine novel methylated loci present specifically in ERG fusion positive or fusion

66 negative PCa. We chose to focus additionally on ERG fusion status as it is present in approximately half of all PCa cases and represents a distinctly different etiology as compared to

ERG fusion negative PCa (442, 443). In addition, a link between DNA methylation and ERG expression in PCa has been reported for the prognostic marker PITX2 and for global LINE-1 methylation status.

2.3 Materials and Methods

2.3.1 Patient Samples

50 fresh frozen PCa tissue samples obtained from prostatectomy specimens of patients with PCa diagnosed between 2001 and 2007 were collected from the tissue bank at the University Health

Network (UHN), Toronto. Patients who received therapy prior to surgery were excluded. For 10 tissue samples of GS 7 origin, both a GP 3 and GP 4 were included in the analysis. The remaining specimens from the Gleason 7 category were either GP3 or GP4.

All patients consented to the donation of removed tissue to the UHN tissue bank and samples were obtained according to protocols approved by the Research Ethics Board from Mount Sinai

Hospital (MSH) and UHN, Toronto, ON, Canada. PCa specimens were subjected to histological examination by an expert pathologist (TVDK) for independent confirmation of the Gleason grades.

2.3.2 DNA isolation

Fresh frozen archived tissues derived from pure GP3 or pure GP4 specimens were snap-frozen in liquid nitrogen, crushed, and genomic DNA was isolated using the QIAamp DNA mini kit

(Qiagen) according to the manufacturer’s recommended protocol. For GS7 specimens containing a mixture of GP3 and GP4, frozen tissue was sectioned (10 µm) onto slides and stained with hematoxylin and eosin (H&E). Areas which were distinctly GP3 or GP4 were manually dissected directly from H&E slides by TVDK using a 20 gauge needle. DNA was then extracted from all samples using the QIAamp DNA mini kit according to the manufacturer’s recommended protocol.

2.3.3 Differential Methylation Hybridization (DMH) and CpG Island Microarrays

The differential methylation hybridization technique for preparation of methylated amplicons was carried out as described previously. Briefly, genomic DNA (0.2 µg) from each case was digested with MseI. The cleaved ends were ligated with annealed H-12/H-24 linkers, followed by two successive rounds of digestion with methylation-sensitive enzymes, namely HpaII and

BstUI. Linker PCR reactions were then performed with pre-treated DNA to generate the final target amplicons for microarray hybridization. Final amplicons were purified using the QIAquick

PCR purification kit (Qiagen) according to the manufacturer’s protocol. The reference sample consisted of DNA isolated from lymphocytes of six healthy men age-matched with PCa patients.

Reference samples were similarly treated for final target generation and pooled amplicons were co-hybridized with test cases for individual arrays. Co-hybridization to Agilent human CpG island microarrays (244K) was performed by the UHN microarray facility according to Agilent protocol.

2.3.4 Statistical Analyses

Processed and normalized log ratios obtained from Agilent Feature Extraction Software for each array were first batch corrected. This was done to adjust for differences arising as a result of differences in Agilent array batches or due to specimen processing differences that could have arisen from DNA preparation performed in different batches. Statistical analyses were performed by Dr. Laurent Briollais and Ms. Nino Demetrashvili using an empirical Bayes t-test with the statistical package limma of R. The Bayes t-test has the same interpretation as an ordinary t-statistic except that the standard errors have been moderated across genes (shrunk towards a common value) using a simple Bayesian model. This has the effect of borrowing information from the ensemble of genes to make the inference about each individual gene more robust. The moderated t-statistic has an increased number of degrees of freedom compared to the ordinary t-statistic, reflecting the greater reliability associated with the smoothed standard error.

Our analyses were conducted after pre-processing performed by Agilent Feature Extraction

Software. A loess normalization procedure was then performed within arrays to remove any systematic trends which arise from the microarray technology from the methylation measures.

Data analysis was also performed using CisGenome software version 2.0 (Johns Hopkins

University). The advantage to using CisGenome software is incorporation of adjacent probe signal for determination of significant regions (peaks) of methylated DNA. Normalized log ratios were used for peak detection using a moving average window with the following parameters: (1) minimum fold change of 1.5; (2) discard peaks < 2 probes and 70 bp in length; (3) combine multiple peaks if distance between is < 300 bp and 5 probes; (4) half-window size of 2 probes

69 and 120 bp. Peaks were annotated to adjacent gene regions based on proximity (≤ 10,000 bp from transcription start site or transcription end site).

Hierarchical clustering of methylated probes was performed using Cluster 3.0 software

(University of Tokyo, Human Genome Center) and dendograms were produced using Treeview software (Eisen lab, Standford University). Probes were initially filtered using a minimum standard deviation for inclusion across the entire dataset of 1.5. This yielded a total number of

193 probes. Clustering of the filtered probeset and cases was then performed using Euclidean distance as the similarity metric and average linkage as the clustering method.

2.3.5 MassARRAY EpiTYPER Analysis

Quantitative analysis of CpG dinucleotide methylation was performed using a mass spectrometry approach as available by MassARRAY® EpiTYPER analysis (Sequenom). EpiTYPER analysis is a MALDI TOF mass spectrometry based method that provides a quantitative view of CpG dinucleotide methylation to single or multiple dinucleotide resolution. DNA was first bisulfite modified, tagged with a T7 promoter, and transcribed into RNA. RNA was then cleaved with

RNase A and cleavage products of different mass resolved by the MS instrument. Analysis was performed by the Analytical Genetics Technology Centre (AGTC), Princess Margaret Hospital,

Toronto, ON using selected frozen tissue specimen DNA that was used for CpG island microarray analysis. Regions analyzed by EpiTYPER corresponded to those that showed an enriched signal in the CpG island array results. All analyses were performed in triplicate and averages and standard errors were calculated.

2.3.6 ERG Immunohistochemistry

Immunostaining of the tissue microarrays for ERG was performed as follows: Deparaffinized 4

μm sections were dehydrated, blocked in 0.6% hydrogen peroxide in methanol for 20 minutes and processed for antigen retrieval in EDTA (pH 9.0) for 30 minutes in a microwave, followed by 30 minutes of cooling in EDTA buffer. Sections were then blocked in 1% horse serum followed by an overnight incubation with the ERG–MAb mouse monoclonal antibody (Biocare

Medical clone 9Fy, Concord, CA), diluted 1:300 at room temperature. The immunostaining was developed using the Polymer-HRP immunohistochemistry (IHC) kit (Biogenex, Fremont, CA) according to manufacturer’s instructions. Next, sections were counterstained in hematoxylin for

1 min, dehydrated, cleared and mounted. Immunostained slides were scanned using the Aperio system at objective 20X, facilitating the scoring of the individual TMA cores. ERG staining was evaluated based on percentage of epithelial cells staining positive and the intensity of staining relative to an internal control (endothelial cells with positive staining). Cores with faint or negative endothelial cell staining were excluded from analysis. Intensity was graded on a scale of

0-3 with 0 representing no staining, 1-faint positivity, 2-intensity equal to internal control and 3- intensity greater than internal control. ERG expression was then separated into binary values for positive and negative expression. Those cores with an intensity of 1 or more in greater than 10% of cells were considered positive while a score of 0 or staining in ≤10% of cells were considered negative. We considered a case positive for ERG expression if any of the arrayed cores from that case displayed positive ERG IHC as described above.

2.4 Results

2.4.1 GS6 versus GS8 analysis

Initial analyses were performed on the GS 6 and GS 8 microarray cases in order to discover methylated genes that differentiated pure GP3 from pure GP4 (444). Data was analyzed using a

Bayes t-test in R statistical computing. Using a 2-fold enrichment signal difference (positive or negative) and a p-value ≤ 0.05 between the two grades as a threshold, there was a set of 613 array probes significantly different between GS6 and GS8 cancers (top 25 genes in Table 2.1).

Intriguingly, 359 (58.6%) of these probes had decreased methylation intensity in high grade PCa versus low grade PCa. In addition, there was a significantly greater proportion of differentially methylated probes occurring in intergenic regions than expected and significantly lower proportion of differentially methylated probes occurring in intragenic regions than expected

(Figure 2.1A; χ2 p-value = 1.07 x 10-13). There was also a significant difference in regional methylation comparing the hypermethylated probes in GS6 versus GS8 and the hypomethylated

probes (Figure 2.1B; χ2 p-value = 3.4 x 10-5). Furthermore, while methylation was discovered in

genes previously identified as methylated in PCa (i.e. CDKN2A), the majority of loci (>90%)

that were significantly different between GS6 and GS8 were novel candidates. The largest fold

difference in the positive direction was within the CAP-GLY domain containing linker protein

family, member 4 (CLIP4) promoter region (5.6), while the largest fold difference in the

negative direction was in the LSM6 gene (-3.4). Interestingly, we observed differential

methylation in a number of different homeobox containing transcription factors (Table 2.2).

We chose to select two genes for validation (Chapter 3) by further filtering the GS6 versus GS8

Bayes t-test analysis using the following features: (1) Significant difference between GS6 and

GS8 (p-values ≤ 0.01; fold change > 2); (2) Methylation signal in the promoter or 5’ end of the gene, as defined by 1000 bp upstream or 500 bp downstream of the transcription start site; (3)

Characterized or putative role in carcinogenesis, cancer progression or development.

This led to the selection of homeobox D3 (HOXD3), a gene with an identified role in lung cancer and melanoma invasion/motility (432-434), and transforming growth factor beta 2 (TGFβ2), which is part of the well characterized TGFβ pathway.

Gene Probe Location Fold Change p-value CLIP4 PROMOTER 5.669052 0.006138 TBX15 INSIDE 4.86474 0.00011 chr4:041570155-041570205 Unknown 4.070236 0.002902 chr1:119345005-119345064 Unknown 4.008472 2.64E-05 COL9A1 INSIDE 3.827843 0.001167 FEZF1 PROMOTER 3.742668 0.020775 HOXD3 INSIDE 3.68626 0.000148 VAX1 INSIDE 3.678967 0.005076 VAX1 INSIDE 3.559365 0.002059 ZNF96 INSIDE 3.539512 0.0037 HOXD3 INSIDE 3.528301 0.000509 chr6:074081375-074081424 Unknown 3.491073 0.005857 DHH INSIDE 3.490936 0.000126 chr8:103682333-103682384 Unknown 3.429685 0.003012 LSM6 INSIDE 0.292252 0.000438 EBF1 PROMOTER 3.344361 5.67E-05 HIST1H1B INSIDE 3.333514 0.001096 chr1:119351272-119351316 Unknown 3.294519 0.004008 HIST1H2BD DOWNSTREAM 3.206353 0.007212 chr1:119344681-119344725 Unknown 3.179581 0.000319 chr1:119351447-119351491 Unknown 3.167829 0.007953 TBX15 INSIDE 3.155447 0.010201 C1orf32 INSIDE 3.103492 0.010649 TGFB2 INSIDE 3.073627 0.007606 KLF11 PROMOTER 0.327019 0.001211

Table 2.1. Top 25 differentially methylated probes comparing GS6 versus GS8

Figure 2.1. Genomic location of differential methylation comparing GS6 versus GS8. (A) Proportion of probe locations differentially methylated between GS6 and GS8 compared to total array representation of probe locations. (B) Proportion of probe locations for hypermethylated loci versus hypomethylated loci from GS6 versus GS8 comparison.

Gene Probe Location Fold Change p-value HOXD3 INSIDE 3.68626 0.00015 VAX1 INSIDE 3.67897 0.00508 VAX1 INSIDE 3.55937 0.00206 HOXD3 INSIDE 3.5283 0.00051 HLXB9 INSIDE 2.93647 0.0256 GBX2 INSIDE 2.89864 0.01324 VAX1 INSIDE 2.84757 0.00832 HLX1 DOWNSTREAM 2.83177 0.00088 HOXC11 PROMOTER 2.7986 0.01337 NKX2-2 DOWNSTREAM 2.74832 0.00432 NKX2-2 DOWNSTREAM 2.72834 0.00706 POU4F3 DOWNSTREAM 2.62689 0.00225 SIM1 INSIDE 2.59994 0.03971 SIM2 INSIDE 2.58873 0.00911 LHX2 INSIDE 2.56805 0.03714 NKX2-2 DOWNSTREAM 2.55788 0.10099 SIM1 INSIDE 2.50591 0.02308 ONECUT1 PROMOTER 2.48766 0.01187 HOXD3 INSIDE 2.45095 0.00055 HOXA11 PROMOTER 0.39585 0.00107 EVX2 INSIDE 0.38233 0.00067 BARHL2 PROMOTER 0.36799 0.06402 HOXA11 PROMOTER 0.35939 0.00034 EVX2 INSIDE 0.34906 0.0045 POU3F3 PROMOTER 0.34706 0.00313

Table 2.2. Top 25 differentially methylated homeobox probes comparing GS6 versus GS8

2.4.2 GS7 specimens and incorporation of recurrence data

Further analysis of microarray data was performed incorporating GS7 PCa specimens and comparing to low grade GS6 PCa, again comparing using the Bayes t-test. Overall, fold change differences between GS7 and GS6 were much less pronounced than the GS8 and GS6 comparison. For example, the top probe ranked by positive fold change in the GS7 versus GS6 comparison (TBX15) had a fold change of 2.3, compared to 5.6 (CLIP4) from the GS8 versus

GS6 comparison. Similar results were found for probes ranked by negative fold change, with a -

1.96 fold change found for the top gene (ANKRD15) in the GS7 versus GS6 comparison. The associated p-values for top ranked probes by fold change were also larger in the GS7 versus GS6 comparison, as 5/10 top ranked positive fold change probes had p-values > 0.05, compared to

0/10 in the GS8 versus GS6 dataset. Accordingly, we reduced the stringency of probes that were filtered to include fold changes of 1.5 (positive or negative direction) and p-values ≤ 0.1. This produced a list of 445 differentially methylated probes (top 25 genes in Table 2.3). Surprisingly,

393 (88.3%) of these had decreased methylation intensities, translating to hypomethylation in

GS7 versus GS6 cases. Again, the proportion of intragenic and intergenic methylation events was significantly different than expected values (χ2 p-value = 7.11 x 10-10), as were the

proportions in regional methylation comparing the hypermethylated probes in GS6 versus GS7

and the hypomethylated probes (Figure 2.2; χ2 p-value = 2.3 x 10-5). Next, we compared the

overlap between this comparison and the GS8 versus GS6 comparison. 319 probes overlapped

between the two comparisons (71.7% of the GS7 versus GS6 dataset), suggesting that a large

number of methylation changes arise in intermediate grade PCa.

Follow-up data for each patient was also obtained and used to compare two groups of patients - those who had biochemical recurrence versus those that did not. Cases that were missing follow- up data or cases without recurrence but with follow-up time < 5 years were excluded from this analysis. Significant differences in methylation for each probe were defined similarly to the GS8 versus GS6 analysis – that is, having a fold change > 2 and a p-value ≤ 0.05. Using these thresholds, 105 probes were differentially methylated comparing recurrent versus non-recurrent cases (Table 2.4). Of the 105 total probes, 38 (36.2%) were hypomethylated in recurrent cases, an opposite trend to what was observed in GS analyses. Again, there was a significantly greater proportion of differentially methylated probes occurring in intergenic regions than expected and significantly lower proportion of differentially methylated probes occurring in intragenic regions than expected (χ2 p-value = 1.8 x 10-5).

Gene Probe Location Fold Change p-value TBX15 INSIDE 2.29664 0.04777 DHH INSIDE 1.9958 0.01716 HOXD3 INSIDE 1.91942 0.01671 HOXD3 INSIDE 1.91752 0.01333 COL9A1 INSIDE 1.8919 0.09379 chr1:119345005-119345064 Unknown 1.86224 0.06945 chr6:027172564-027172610 Unknown 1.80556 0.01968 HIST1H1B INSIDE 1.75431 0.09551 chr10:000449644-000449696 Unknown 0.56962 0.04285 LIM2 PROMOTER 0.56937 0.02152 SNAI2 DOWNSTREAM 0.56922 0.04587 FOXC1 INSIDE 0.56842 0.00957 chr4:137496955-137497014 Unknown 0.56651 0.03275 ABLIM2 INSIDE 0.56181 0.01055 SEMA6B INSIDE 0.56046 0.01148 MCF2L INSIDE 0.5553 0.0028 POU3F3 PROMOTER 0.5542 0.06901 MYOM2 INSIDE 0.55211 0.00273 chr13:111943320-111943379 Unknown 0.54889 0.03727 chr14:036044192-036044251 Unknown 0.53485 0.00387 WDR86 INSIDE 0.53226 0.01343 EVX2 INSIDE 0.53169 0.01442 EVX2 INSIDE 0.53138 0.005 LSM6 INSIDE 0.51435 0.00765 ANKRD15 INSIDE 0.51108 0.00309

Table 2.3. Top 25 differentially methylated probes comparing GS6 versus GS7

Figure 2.2. Genomic location of differential methylation comparing GS6 versus GS7. (A) Proportion of probe locations differentially methylated between GS6 and GS7 compared to total array representation of probe locations. (B) Proportion of probe locations for hypermethylated loci versus hypomethylated loci from GS6 versus GS7 comparison.

Gene Probe Location Fold Change p-value HOXD8 PROMOTER 3.43020.00077 chr10:022765894-022765938 Unknown 3.10797 5.17E-05 chr6:106536153-106536197 Unknown 3.02507 0.01008 chr6:013877712-013877756 Unknown 2.9788 0.02077 chr17:044075123-044075167 Unknown 2.93245 0.00126 PCDHGB5 INSIDE 2.91706 0.01916 chr3:185714653-185714703 Unknown 2.81009 0.00455 ABCC9 PROMOTER 2.791590.04114 HOXD8 PROMOTER 2.728022.59E-05 GENE X PROMOTER 2.65941 0.02964 chr17:025952291-025952335 Unknown 2.62573 0.00021 chr8:103682333-103682384 Unknown 2.54135 0.0134 SRD5A2 INSIDE 2.5384 0.0339 chr8:071109789-071109833 Unknown 2.49161 0.0074 TBX15 INSIDE 2.46326 0.04879 chr2:019932354-019932400 Unknown 2.45417 0.03961 AATF PROMOTER 2.445560.00055 CALCA INSIDE 2.42995 0.04886 ZNF146 DOWNSTREAM 2.426050.01051 TBX15 INSIDE 2.41452 0.00966 chr5:072562288-072562336 Unknown 2.41219 0.02768 HELT PROMOTER 2.409550.03797 LARP5 INSIDE 0.41035 0.00117 SLC22A3 INSIDE 0.4087 0.024 BRUNOL4 INSIDE 0.384 0.00011

Table 2.4. Top 25 differentially methylated probes comparing recurrent versus non recurrent cases

For the selection of additional candidate genes, we chose to incorporate biochemical recurrence data to select markers which may relate to disease outcome independent of GS. The previous requirements of promoter/5’ end methylation and having a putative role in cancer development were not considered in candidate gene filtering in order to select potential progression-associated methylated loci in a more unbiased fashion. HOXD8 was the top candidates from this analysis while GENE X was significant in both GS6 versus GS7 and recurrence analysis (Table 2.4); thus, both which were chosen for validation in Chapter 3. TBX15 and CYP26A1 were validated by another member of the lab (Liyang Liu).

2.4.3 EpiTYPER analysis

The EpiTYPER analysis included a subset of frozen tumour cases that showed methylation signal enrichment of ≥ 3 or a lack of methylation signal (≤ 2 fold) on the microarrays. This was performed in order to validate methylation signal from the CpG island microarrays and to map exact locations of methylation differences to an individual CpG dinucleotide level.

Data obtained from EpiTYPER analysis confirmed the enrichment/methylation profiles for

HOXD3, HOXD8, and GENE X that were evident from the microarray results (Figure 2.3).

HOXD3 displayed a distinct pattern of increased methylation in the GS8 cases as compared to the GS6 cases. The analysis of fresh frozen DNA samples F, I, 4, and 8 confirmed increased methylation in the high grade specimens, at least with respect to the four cases analyzed (Figure

2.3A). GS8 cases F and I had an average methylation of 72% and 43% respectively, across all 27

CpG dinucleotides analyzed, while low grade samples 4 and 8 respectively had an average methylation of 19% and 35% (paired t-test p-values < 0.05). In particular, differences in local

CpG dinucleotide methylation were found in CpGs 6-11, where high grade cases F and I averaged 71% and 37%, respectively, and low grade cases 4 and 8 average 12% and 10%, respectively.

For HOXD8, there was a significant difference in methylation between GS7 case H-4 (average

76% methylation) and GS6 case 7 (average 41% methylation; Figure 2.3B, paired t-test p-value

= 3.15 x 10-10). Robust differences in methylation were observed at CpG dinucleotides 23 and

24-26 in particular, with case H-4 averaging 80% methylation and case 7 having 17%

methylation. Finally, EpiTYPER analysis of the GENE X promoter region again confirmed the

methylation profile observed from the CpG island microarrays. GS7 case H-4 had an average

methylation across CpG residues of 62% versus 4% for case C (Figure 2.3C; paired t-test p-value

= 1.81 x 10-13). Distinct differences in methylation were observed for CpG residues 2-5, 9, and

22, all of which were completely unmethylated in case C and averaged 65% methylation for case

H-4.

Figure 2.3. EpiTYPER quantitative methylation profiling. (A) HOXD3, (B) HOXD8, and (C) GENE X quantitative methylation signals in indicated frozen tumour specimens. HG – high grade; IG – intermediate grade; LG – low grade

2.4.4 Total GS analysis

To discover genes that had significantly different methylation profiles across all GSs (6, 7 and

8), data was analyzed using both a quantile regression method and CisGenome analysis to define peak regions of methylation. Using the quantile regression method, criteria for differential methylation were a fold change of 1.5 (positive or negative) and p-value ≤ 0.05. This produced a list of 232 differentially methylated probes across all GSs analyzed (Table 2.5). Of the 232 total probes, 94 (40.5%) were hypomethylated in higher grade PCa. Once more, the distribution of methylated probe locations was significantly different than expected (χ2 p-value = 9.57 x 10-9).

In addition to probe level analysis, we conducted a regional enrichment level analysis of the CpG

island microarrays across all GSs using CisGenome software. Regional analysis of microarrays

reduces the likelihood of false positive results that may arise, for example, from cross-

hybridization of DNA fragments to non-specific oligonucleotides on tiling array platforms.

Using this strategy, we discovered 156 regions hypermethylated in higher grade PCa and 135

regions hypomethylated in higher grade PCa (Table 2.6). Of these 291 peaks, 230 were

annotated to genes (within 10,000 bp upstream or downstream) and 10 of the 230 genes had multiple regions with methylation differences. Thus, a total of 220 unique genes were identified through regional analysis. Comparing with the probe level analysis list, there was an overlap of

153 genes (69.5%), indicating a large degree of concordance between the two analyses.

Gene Probe Location Beta value p-value CLIP4 PROMOTER 1.25155 0.00455 TBX15 INSIDE 1.14118 0.00121 chr4:041570155- 041570205 Unknown 1.01256 0.0065 chr1:119345005- 119345064 Unknown 1.00153 0.00069 COL9A1 INSIDE 0.96827 0.00422 FEZF1 PROMOTER 0.95203 0.02449 HOXD3 INSIDE 0.94108 6.15E-05 VAX1 INSIDE 0.93965 0.00884 VAX1 INSIDE 0.91581 0.00226 ZNF96 INSIDE 0.91178 0.00785 HOXD3 INSIDE 0.90949 0.00045 chr6:074081375- 074081424 Unknown 0.90184 0.00652 DHH INSIDE 0.90181 0.00047 chr8:103682333- 103682384 Unknown 0.88904 0.00379 EBF1 PROMOTER 0.87087 0.00019 HIST1H1B INSIDE 0.86852 0.0055 chr1:119351272- 119351316 Unknown 0.86003 0.00513 HIST1H2BD DOWNSTREAM 0.84047 0.00539 chr1:119344681- 119344725 Unknown 0.83442 0.00061 chr1:119351447- 119351491 Unknown 0.83175 0.01026 TBX15 INSIDE 0.82892 0.01128 C1orf32 INSIDE 0.81695 0.01595 TGFB2 INSIDE 0.80997 0.00508 KLF11 PROMOTER -0.8063 0.00197 LSM6 INSIDE -0.8874 9.00E-05

Table 2.5 Top 25 differentially methylated probes across all GS

Peak Maximum Chromosome Peak Start Peak End Length Gene Fold Change chr1 119350724 1193523631640 --- 3.50836 chr1 219135546 219135689144 --- 3.32832 chr17 45992057 45992417361 CACNA1G 3.32105 chr4 55218253 55218607355 KIT 3.28896 chr19 41427968 41428472505 ZNF146 3.23814 chr1 219134183 219134725543 HLX1 3.22976 chr17 41330118 41330510393 MAPT 3.21442 chr1 219131711 219132058348 HLX1 3.2007 chr15 81114286 81114480195 --- 3.18632 chr1 119328339 119328630292 TBX15 3.15383 chr12 56307577 56308112536 B4GALNT1 3.11793 chr22 49312401 49312849449 ECGF1 3.11265 chr10 118882126 118882424299 VAX1 3.08007 chr20 21437474 214387861313 NKX2-2 3.05851 chr2 200033836 200034393558 SATB2 -3.0238 chr1 25629583 25629788206 TMEM57 -3.0654 chr4 120768832 120768982151 PDE5A -3.1056 chr2 73466134 73466292159 ALMS1 -3.1252 chr7 105539352 105539464113 SYPL1 -3.2117 chr1 100778355 10077842571 GPR88 -3.2141 chr5 132389877 13238996690 ZCCHC10 -3.2263 chr21 36429965 36430078114 CBR3 -3.2986 chr17 33177425 3317749773 TCF2 -3.3149 chr2 10100090 10100579490 KLF11 -3.5886 chr19 60858812 60858961150 U2AF2 -3.6033

Table 2.6 Top 25 differentially methylated peaks across all GS

2.4.5 ERG stratification

Next, we sought to determine if there were gene specific DNA methylation differences by stratifying the microarray cases based on ERG protein expression. Using CisGenome regional analysis, we found a list of 130 differentially methylated genes (100 hypermethylated in ERG positive, 30 hypermethylated in ERG negative; Table 2.7). Similar to the GS analysis, a significantly greater than expected number of differentially methylated peaks were found in intergenic regions, while there were fewer than expected differentially methylated peaks in intragenic regions (p-value = 3.84 x 10-4).

Peak Maximum Chromosome Peak Start Peak End Length Gene Fold Change chr2 12775784 12776170 387 TRIB2 40.785 chr5 3650352 3651412 1061 IRX1 21.2279 chr8 25955233 25955446 214 EBF2 19.5302 chr3 62332155 62332285 131 FEZF2 16.4676 chr11 32412001 32412166 166 WT1 16.1898 chr6 1.01E+08 1.01E+08 404 SIM1 16.0131 chr5 72711901 72712808 908 --- 13.2932 chr11 32411433 32411691 259 WT1 12.8879 chr16 53922346 53922990 645 IRX6 11.7505 chr7 19118491 19118587 97 TWIST1 9.33735 chr1 63558855 63559063 209 FOXD3 9.19942 chr16 53528873 53528988 116 IRX5 9.15515 chr7 19112356 19112448 93 TWIST1 8.84539 chr18 53257774 53257947 174 ONECUT2 8.77873 chr16 53527927 53528551 625 IRX5 8.66535 chr5 1.54E+08 1.54E+08 140 --- 8.31492 chr15 87753299 87753483 185 --- 7.87803 chr12 1.13E+08 1.13E+08 924 TBX5 7.60735 chr6 1.01E+08 1.01E+08 87 SIM1 7.26373 chr5 54552027 54552563 537 --- 7.01921 chr1 90964901 90965322 422 BARHL2 6.99487 chr5 32749871 32750145 275 NPR3 6.93848 chr9 78818692 78819184 493 FOXB2 6.61329 chr8 10627042 10627895 854 SOX7 -8.4525 chr9 92995629 92996282 654 --- -9.4094

Table 2.7 Top 25 differentially methylated peaks comparing ERG positive versus ERG negative

2.4.6 Hierarchical clustering

All of the aforementioned analyses and results were based on stratification of cases into distinct categories based on clinicopathological features of PCa. Therefore, in order to group samples together based on methylation profiles in an unbiased fashion, we performed hierarchical clustering of cases and probes using Agilent CpG island microarray data. From this analysis, 193

(110 unique genes annotated) probes distinguished two distinct clusters of cases, with 14 cases in cluster 1 and 25 cases in cluster 2 (Figure 2.4). Interestingly, when analyzing the distribution of cases in cluster 1 versus cluster 2 by clinicopathological data (GS, stage, recurrence) and ERG expression status, only ERG expression status was significantly different (Fisher’s exact p-value

= 0.002).

Figure 2.4. Unsupervised hierarchical clustering of all cases and the 193 probes with standard deviation ≥1.5 across all PCa specimens. A – Pearson χ2 analysis of GS distribution separated by clusters 1 and 2; B – Fisher’s exact analysis separated by clusters 1 and 2

2.4.7 Functional annotation of methylated genes

To identify groups of genes belonging to different biological pathways that may be targeted by

DNA methylation, we performed DAVID functional clustering analysis using the peaks called from CisGenome analysis of GS and ERG stratification data. Using the functional annotation enrichment feature, there was strong enrichment for transcription factors in all datasets (Table

2.8 and 2.9). In particular, homeobox gene transcription factor hypermethylation was strongly enriched in higher grade PCa (Benjamini p-value = 7.8 x 10-9). For hypomethylated genes,

transcriptional repressors were the most significantly enriched annotated gene category. With

respect to the ERG stratified data, transcription factors and the homeobox gene family were

again strongly enriched (Table 2.9) in both the ERG positive hypermethylated dataset and the

ERG negative hypermethylated data set (Table 2.9).

GS hypermethylated Term Benjamini GO:0003700~transcription factor activity 6.12E-15 dna-binding 1.35E-13 Homeobox 1.26E-13 DNA-binding region:Homeobox 6.19E-12 IPR001356:Homeobox 8.10E-12 IPR017970:Homeobox, conserved site 4.40E-12 IPR012287:Homeodomain-related 5.67E-12 SM00389:HOX 6.38E-11 developmental protein 9.62E-11 GO:0030528~transcription regulator activity 6.27E-10 GS hypomethylated Term Benjamini repressor 4.66E-04 nucleus 0.001043 transcription regulation 0.002142 Transcription 0.00246 GO:0030528~transcription regulator activity 0.008888 GO:0003700~transcription factor activity 0.033818 dna-binding 0.041551

Table 2.8. Significantly enriched functional terms for differentially methylated genes according to GS (as determined by DAVID analysis). Benjamini – Benjamini p-value

ERG positive hypermethylated Term Benjamini dna-binding 1.52E-19 GO:0003700~transcription factor activity 8.12E-17 GO:0006355~regulation of transcription, DNA-dependent 1.15E-14 GO:0030528~transcription regulator activity 1.75E-15 GO:0051252~regulation of RNA metabolic process 1.31E-14 GO:0003677~DNA binding 7.88E-15 GO:0043565~sequence-specific DNA binding 5.88E-15 GO:0045449~regulation of transcription 4.15E-12 GO:0006357~regulation of transcription from RNA polymerase II promoter 6.69E-12 developmental protein 3.34E-12 transcription regulation 4.66E-11 ERG negative hypermethylated Term Benjamini GO:0003700~transcription factor activity 5.66E-06 DNA-binding region:Homeobox 8.97E-06 GO:0006357~regulation of transcription from RNA polymerase II promoter 5.76E-05 GO:0007389~pattern specification process 5.85E-05 Homeobox 2.27E-05 dna-binding 1.44E-05 IPR001356:Homeobox 3.27E-05 IPR017970:Homeobox, conserved site 1.68E-05 IPR012287:Homeodomain-related 1.42E-05 SM00389:HOX 9.83E-06

Table 2.9. Significantly enriched functional terms for differentially methylated genes according to ERG status (as determined by DAVID analysis)

2.5 Discussion

We have used human CpG island microarrays to identify novel methylated genes in aggressive

PCa ranging from low to high grade as well as recurrent versus non-recurrent tumours. We found that we were able to identify CpG islands that are both quantitatively more methylated and methylated at an increased frequency in intermediate and high grade versus low grade PCa. This may reflect an overall shift to a greater state of methylation within CpG islands as the tumour progresses towards a higher grade. While it should be noted that quantitative methylation differences from our arrays may be artifacts as the result of overall copy number changes, the majority of top differentially methylated regions, regardless of the analysis performed, are not within regions known to have frequent copy number alterations. However, SOX7 (a gene discovered as hypomethylated in the ERG analysis), is found on chromosome 8p which is a region of frequent deletions in PCa.

Most genes uncovered through our arrays have either never been shown to be methylated in PCa or in other types of cancers. Other previously described methylated genes in prostate cancer, such as CDKN2A, PTGS2, and RUNX3, all showed evidence of methylation based on fold changes and statistical significance. The stringency of the statistical analyses that we performed could have prevented the inclusion of these genes within our top genes of the progression or cancer/reference dataset. Therefore, this may not be indicative of a lack of methylation, but instead can be explained by quantitative methylation levels. It is possible that methylation of these genes may have occurred in fewer cells and/or in a fewer number of CpG dinucleotides, thus producing a less robust signal in our screen. Alternatively, grouping of cases in statistical analyses may have filtered these genes out, since methylation in a fewer number of specimens

95 would create a lower average and higher variability across these cases. We were surprised to find that the best characterized methylation event in PCa, hypermethylation of the GSTP1 promoter, was not captured in our array screen results. It is possible that the method we used for target

DNA preparation in combination with the microarray platform is responsible for the lack of detection of GSTP1 methylation signal. Sequence analysis of GSTP1 revealed that our methylated DNA enrichment method would produce a fragment of approximately 1900 bp, which may affect annealing to probes of significantly smaller length (approximately 45-60 mer) or may not remain intact following methylation sensitive digestion. Additionally, PCR amplification of this region following enzymatic digestions may have been less than optimal, and would thus lead to lack of signal.

We initially analyzed GS6 and GS8 specimens as these contain pure populations of GP3 and

GP4, respectively, and they represent two distinct ends of the prognostic spectrum with the former generally behaving in an indolent manner and the latter being faster growing and much more likely to metastasize. We chose HOXD3 and TGFβ2 for further validation (Chapter 3) from the GS6 versus GS8 analysis, as these are novel targets for methylation in PCa and the methylation signal was proximal to the promoter region of both genes. They represent a subset of genes where silencing may play a role in the development of high grade prostate cancers based on our array results, but also based on available functional information from current literature.

Therefore, these genes do not necessarily reflect the greatest statistical significance or the greatest methylation fold change of either two datasets that we produced. The genes with the greatest fold changes from the total GS analyses (CLIP4, U2AF2) will require future validation in a large series of prostate tumours. BMP7, a gene which we discovered as methylated in a

96 subset of PCa specimens (444), was not considered for validation as it did not meet any of the criteria for inclusion.

We next chose to analyze intermediate grade GS7 tumours and incorporate follow-up recurrence data into our analysis. The addition of GS7 tumours made it possible to distinguish stepwise increases and decreases in DNA methylation that occur from GS6 to GS7 to GS8. It is worthwhile to note that differerences when analyzing GS6 versus GS7 were not as pronounced as GS6 versus GS8. This is not surprising, however, given that GS7 tumours were an admixture of GP3 and GP4, while GS6 is pure GP3 and GS8 is pure GP4. Thus, heterogeneity of GPs within overall GS may be responsible for this observation. Integrating biochemical recurrence data into the analysis produced a set of genes with significant overlap with the GS data, yet may identify genes related to disease outcome independent of GS, and thus may eventually be useful in a post-treatment setting for PCa prognosis. This analysis led to the selection of HOXD8 and

GENE X for validation, which were among the top 10 differentially methylated genes according to biochemical recurrence.

Interestingly, we noted DNA methylation changes corresponding to GS that occurred at a greater than expected frequency in regions of the genome that were outside of gene promoters (intra- and intergenic). Recent studies have similarly identified tissue-specific and cancer-specific differentially methylated regions (DMRs) that occur outside of the promoter region of genes and at “shore” or “shelf” CpG regions (289). The consequence of altered DNA methylation profiles at non-promoter loci remains unknown, but is likely context dependent. For example, intragenic methylation may work to repress expression of alternative downstream first exons (294).

Intragenic methylation may also affect splice site usage and lead to alternative mRNA isoforms, as methylation has been shown to influence CTCF binding and RNA pol II elongation during

97 transcription (295). Alternatively, intergenic methylation may repress alternative upstream first exons. In addition, intergenic methylation may silence hitherto unidentified genes, such as ncRNAs (miRNAs and lincRNAs) which have not been fully characterized throughout the genome. It is also possible that intergenic methylation represses enhancer elements necessary for robust gene transcription (297, 298).

In addition, we included data obtained from ERG IHC on this cohort to assess potential relationships between ERG fusion status and DNA methylation. This produced a list of 130 gene regions, indicating an important relationship between these two variables. In fact, hierarchical clustering of the top 193 variable probes from the microarray dataset clustered 1 group with

17/24 cases positive for ERG expression together, with the other cluster containing 1/10 cases positive for expression. This suggests that ERG may be an important direct or indirect driving force in shaping the methylome, at least when compared to clinicopathological variables. The functional mechanism that leads to the relationship between methylation of these specific genes and ERG expression is unknown and needs further investigation. One possibility is that ERG drives the expression of DNA methyltransferases such as DNMT3A and DNMT3B, which subsequently leads to de novo or increased methylation at specific gene loci. Indeed, expression of epigenetic silencing enzymes HDAC1 and EZH2 have been shown to significantly co-exist with high ERG expression (in the case of HDAC1) or to be caused specifically by TMPRSS2-

ERG (in the case of EZH2) (274, 445). Alternatively, decreased histone acetylation via HDAC1 and/or increased H3K27 methylation via EZH2, and crosstalk between epigenetic modifying machinery might facilitate DNA methylation (426). Another possibility is that ERG induces expression of hitherto unidentified transcriptional co-repressor molecules that may target DNA methyltransferase activity to specific gene loci. For example, ERG has been shown to activate

98 expression of the C-MYC oncogene (442), and C-MYC can bind a MIZ-1/DNMT3A complex leading to promoter methylation and gene silencing of p21Cip1 in mice (446). Ultimately, the mechanism through which TMPRSS2-ERG may increase gene specific methylation is unknown, and thus requires further investigation.

We also show that hypermethylation associated with GS, recurrence, and ERG expression is particularly enriched within the homeobox gene family. Others have shown similar results in so- called methylation “hot spots” within HOX loci for different cancers (419, 447). Homeobox genes are a family of transcription factors that were first identified in Drosophila and characterized by their ability to cause large scale patterning abnormalities when mutated (448,

449). In humans, there are over 200 such genes containing this motif, and many of them have been identified as aberrantly expressed in cancers (450, 451). For example, overexpression of

HOXC6, PAX2, and BP1 have been described in PCa (389, 452, 453). The cause and consequence of aberrant homeobox gene DNA methylation is unknown. With respect to the latter, altered mRNA isoform expression may occur and again may be context dependent. The cause of aberrant homeobox gene methylation may be due to differential expression of epigenetic modifiers (i.e. DNMT3A/B, EZH2) and the resultant crosstalk between these modifiers (426). Localized cis-acting DNA elements may also play a role in the propensity for

DNA methylation (454). Nonetheless, there is a growing body of literature characterizing altered

DNA methylation profiles in polycomb group target genes and genes that have a “poised”

(H3K4me3/H3K27me3) promoter histone methylation profile.

In summary, we present the discovery of four novel DNA methylation targets (HOXD3, TGFβ2,

HOXD8 and GENE X) potentially associated with aggressive PCa, as well as genome-wide characterization of DNA methylation events that correlate with increased grade as well as ERG

99 expression. Future work is necessary to validate the prognostic capabilities of the novel methylation events and to ascertain the functional relationship between aberrant DNA methylation at specific loci and clinically aggressive features of PCa as well as ERG expression.

100

Chapter 3

Validation of novel methylated loci associated with PCa progression

Ken Kron1,2, Liyang Liu1,2, Vaijayanti Pethe1, Dominique Trudel3, Michael Nesbitt5, John

Trachtenberg6, Neil Fleshner6, Theodorus van der Kwast2,3, and Bharati Bapat1, 2

1. Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON, Canada

2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON,

Canada

3. Department of Pathology, University Health Network, University of Toronto, Toronto, ON,

Canada

4. Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada

5. University Health Network, University of Toronto, Toronto, ON, Canada

6. Department of Surgical Oncology, Division of Urology, University Health Network,

University of Toronto, Toronto, ON, Canada

The work in this chapter was primarily contributed by Ken Kron. VP, a research assistant, and

LL, an MSc student, assisted in performing DNA extractions. DT, a pathology resident, read the

ERG immunohistochemistry. Clinical information for patients was obtained by MN (clinical research coordinator) and JT (a surgeon), while samples were retrieved by NF (a surgeon) and

TVDK (a pathologist).

The HOXD3 data present in this chapter has been published in Laboratory Investigation: Kron

KJ, Liu L, Pethe VV, Demetrashvili N, Nesbitt ME, Trachtenberg J, et al. DNA methylation of

HOXD3 as a marker of prostate cancer progression. Lab Invest. 2010;90(7):1060-7.

101

The TGFβ2 data has been published in International Journal of Cancer: Liu L, Kron KJ, Pethe

VV, Demetrashvili N, Nesbitt ME, Trachtenberg J, et al. Association of tissue promoter methylation levels of APC, TGFbeta2, HOXD3 and RASSF1A with prostate cancer progression.

Int J Cancer. 2011;129(10):2454-62.

The relationship between HOXD3 methylation and ERG expression has been published in

Clinical Cancer Research: Kron K, Liu L, Trudel D, et al.. Correlation of ERG expression and

DNA methylation biomarkers with adverse clinicopathological features of prostate cancer. Clin

Cancer Res.

102

Chapter 3

Validation of novel methylated loci associated with PCa progression

3.1 Summary

DNA methylation in gene promoters causes gene silencing and is a common event in cancer development and progression. The ability of aberrant methylation events to serve as diagnostic and prognostic markers is being appreciated for many cancers, including prostate cancer. Using quantitative MethyLight technology, we evaluated the relationship between methylation of

HOXD3, TGFβ2, HOXD8 and GENE X and clinicopathological parameters including biochemical recurrence, pathological stage, GS (GS), and GP in a series of 219 radical prostatectomies performed between 1998 and 2001. Methylation was significantly greater in GS

7 cancers versus GS ≤6 cancers for all four genes (p-values < 0.05), as well as pT3/pT4 versus pT2 cancers for HOXD3, HOXD8, and GENE X (p-values < 0.05). There were also significant increases in methylation from GP 2 to 3 for HOXD3, HOXD8, and GENE X and from pattern 3 to 4/5 for HOXD3 and GENE X (all p-values < 0.05). Methylation of TGFβ2, HOXD8 and

GENE X was associated with biochemical recurrence in univariate analyses (p-values < 0.05), while TGFβ2 methylation was the lone methylated gene associated with biochemical recurrence in multivariate Cox regression analysis (p-value = 0.007). Furthermore, methylation of HOXD3 and HOXD8 correlated with ERG expression status (p-values < 0.01), while stratification of cases by ERG expression showed that TGFβ2 methylation was primarily predictive of recurrence in ERG negative cases. These results indicate that methylation of HOXD3, HOXD8 and GENE

103

X distinguish PCa cases with poor prognostic features from those considered more indolent, while TGFβ2 has post-surgery value in predicting disease recurrence.

3.2 Introduction

Prostate cancer (PCa) is the most common malignancy for North American men and in 2009 it was estimated to account for more than 27, 000 deaths in the United States (455). The widespread acceptance of routine PSA testing has led to dramatic increases in PCa diagnoses over the past two decades, while the benefits of such testing remain controversial (17, 26).

Therefore, predicting disease course and outcome of diagnosed cases has received attention recently as many of these cancers are of a slow growing or indolent form, while others may progress rapidly and result in metastasis and death. The best prognostic indicator for PCa, GS

(GS), characterizes the glandular architecture of the prostate and assigns a score based on “de- differentiation” of the carcinoma (35). The extent of de-differentiation is represented by GPs, which range from 1-5. Pattern 3 and pattern 4 represent an important transition from low to high grade carcinoma, and can influence patient prognosis depending on the relative proportions present in the cancer. Pure pattern 3 (GS 6) prostate cancers are low grade while pure pattern 4

(GS 8) cancers are high grade (456). GS 7 cancers (composed of glandular patterns 3 and 4 in variable amounts) are considered intermediate grade. Pretreatment identification of GS 7 cancers

(i.e. presence of GP 4) is considered key to distinguishing indolent or clinically insignificant

PCas from those that require treatment. Pathological stage is another important prognostic indicator that informs of the extent of PCa spread. pT2 cancers are confined entirely within the

104 prostate gland, pT3a PCa extends beyond the prostate into periprostatic tissue, pT3b invades semincal vesicles, and pT4 PCa extends into the bladder neck.

DNA methylation is a well characterized epigenetic event involving the addition of a methyl group to cytosine bases that precede a guanine (CpG). This modification plays an important role in promoting chromosomal stability and regulating gene expression (235, 457). In mammals, enriched DNA stretches of CpG dinucleotides known as CpG islands are common in the promoter regions of approximately 50% of genes (458). With the exception of tissue specific methylation patterns, most promoter CpG islands are hypomethylated, while approximately 70% of total CpG dinucleotides exist in a methylated state (459). Aberrant methylation patterns occur frequently in cancers, including an overall hypomethylation of genomic CpGs (460). Gene specific anomalies are also common, however, and can lead to the inactivation of tumour suppressor genes or to the activation of oncogenes (458, 461). The former arises through hypermethylation of promoters, while the latter occurs when promoter CpG islands become hypomethylated. Importantly, promoter methylation events are being recognized for their potential as both diagnostic and prognostic indicators for a variety of malignancies including bladder, prostate and colon cancers (429, 462).

We have previously identified HOXD3, TGFβ2, HOXD8, and GENE X promoter hypermethylation in PCa through a genome-wide CpG island microarray screen of 39 PCa cases spanning low to intermediate grades (444). In order to assess the contribution of methylation of these four genes as diagnostic/prognostic markers in the current study, we have now examined a larger series of PCa using a quantitative methylation approach, and we have analyzed the relationship of methylation levels with clinicopathological parameters.

105

3. 3 Materials and Methods

3.3.1 Patients and pathology

We initially sought 150 patients in each group of GS ≤6 and GS7 and 20 patients in GS≥8 based on calculations using a type I error probability of 5% and achieving 80% power. Thus, a total of

285 patients diagnosed between 1998 and 2001 were considered for specimen accrual. However, a combination of unavailable specimens in the biobank, poor quality DNA following extraction, and exclusion of patients given neoadjuvant therapy resulted in a total cohort size of 219 from the University Health Network (UHN ) in Toronto. Patient consent was obtained for accrual of removed tissue following radical prostatectomy into the UHN tissue bank. All samples and clinical and pathological follow-up information were obtained according to the protocols approved by the Research Ethics Board at Mount Sinai Hospital, Toronto and UHN, Toronto. All patients who received neo-adjuvant therapy prior to radical prostatectomy were excluded from the study.

The complete set of hematoxylin and eosin (H&E) stained slides from each prostatectomy were collected and reviewed by an expert pathologist (TVDK) to confirm GS (WHO/ISUP criteria), stage (TNM), and surgical margin status. For each case, a subset of slides was selected based on the presence of carcinoma with specific GPs representing the overall GS. Both a pattern 3 and a pattern 4 were selected for Gleason 7 cases where possible. Tumour areas representative of each

GP were marked on the H&E stained slides corresponding to an area of at least 80% neoplastic cellularity. Matched normal tissue containing at least 50% glandular content was also selected for each case where possible.

106

3.3.2 DNA extraction

Formalin-fixed, paraffin embedded tissue blocks matching the selected H&E slides were sectioned at a thickness of 10 µm. These tissue slides were then superimposed on H&E slides and each area of cancer was outlined. The circled areas of tissue slides were then scraped with a scalpel and tissue placed into 1.5 mL tubes.

DNA was extracted from tissues using a QIAamp DNA Mini Kit (Qiagen, Mississauga, ON,

Canada) with a modified protocol. Briefly, 270 µL of buffer ATL was added to tissue followed by 30 µL of proteinase K. The tissue was digested overnight at 56ºC and 20 µL of proteinase K was added the following day. The tissue was then incubated at 56ºC further for 1 hour, and an equal volume (320 µL) of buffer AL was added with an incubation of 70 ºC for 10 minutes. One volume of ethanol (320 µL) was then added, with the remainder of the steps performed according to the manufacturer’s recommended protocol for tissue DNA extraction.

3.3.3 Sodium bisulfite modification and MethyLight

A total of 400 ng of extracted DNA was converted using the EZ DNA Methylation Gold Kit

(Zymo Research Corp, Orange, CA, USA) according to the manufacturer’s protocol and eluted to a final concentration of 20 ng/µL.

The quantitative MethyLight assay was performed using 20 ng of converted DNA. The reactions were carried out in a volume of 30 µL in 96 well plates on an ABI 7500 Real-Time PCR system using the TaqMan Buffer A pack (Life Technologies) and the following reagent concentrations:

107

300 nM each primer, 100 nM probe, 200 µM each dNTP, 3.5 mM MgCl2, 0.01% Tween, 0.05% gelatin, and 0.1 U Taq polymerase. Standard curves were generated using serial dilutions of bisulfite-converted positive control supermethylated DNA (Millipore) for the gene of interest

(1:4 dilutions) and Alu-C4 repeats (1:25 dilutions). Percentage of methylated reference (PMR)

for a gene was calculated using Alu repeats as reference as follows: (gene of interest/Alu

fluorescence quantity ratio for modified specimen DNA) / (gene of interest/Alu ratio for

supermethylated DNA) X 100%. For each case, at least 1 or more (up to 6) tumour foci were

analyzed. All foci were analyzed in duplicate. Primer/probe sequences for the assayed regions

were as follows: HOXD3 (Forward) 5’-TTA AAG GTT TAT GGT TGC GC-3’; (Reverse) 5’-

TTA CGA ACA CTA AAC TAC ACC CG-3’; (Probe) 5’FAM-ACA AAA CGT TCC CGA

CGC TTC TAA AA-BHQ1-3’; TGFβ2 (Forward) 5’TTT TAG GAG AAG GCG AGT CG- 3’;

(Reverse) 5’CTC CTT AAC GTA ATA CTC TTC GTC G-3’; (Probe) 5’FAM-TCT CGC GCT

CGC AAA CGA CC-3’BHQ1; HOXD8 (Forward) 5’TAG TCG GTT TTG GTT CGT TGC-3’;

(Reverse) 5’CGT TCT AAA ACG AAA AAA AAA A CT CGC G-3’; (Probe) 5’FAM-TCC

TCG AAC AAA ACG CGA CTC CCG AAT CTC-3’BHQ1; GENE X (Forward) 5’GGA GTT

ATG AGT TGG ATT TGT TCG CG-3’; (Reverse) 5’AAA AAT CCA ACG ACT CCC ATT

CGC-3’; (Probe)5’FAM-CCC TCC GCA ACC CGA ACC TCA CCG AAA-3’BHQ1.

3.3.4 Tissue microarray construction

A range from 3 to 13 0.6 mm cores were taken from each of 253 cases diagnosed between 1998-

2001 in order to have representation of each primary, secondary, and when possible tertiary GP

present within the case as well as benign glandular tissue adjacent to the tumour. This yielded a

108 total of 1490 cores within 7 tissue microarray blocks. 5 µm serial sections of each microarray were used for H&E verification of tumour and normal tissue. The overlap between the 219 cases used for MethyLight analysis and 253 cases used for IHC of ERG protein was 204 cases.

3.3.5 ERG Immunohistochemistry

Immunostaining of the tissue microarrays for ERG was performed as follows: Deparaffinized 4

Medical clone 9Fy, Concord, CA), diluted 1:300 at room temperature. The immunostaining was developed using the Polymer-HRP immunohistochemistry (IHC) kit (Biogenex, Fremont, CA) according to the manufacturer’s instructions. Next, sections were counterstained in hematoxylin for 1 min, dehydrated, cleared and mounted. Immunostained slides were scanned using the

Aperio system at objective 20X, facilitating the scoring of the individual TMA cores. ERG staining was evaluated based on percentage of epithelial cells staining positive and the intensity of staining relative to an internal control (endothelial cells with positive staining). Cores with faint or negative endothelial cell staining were excluded from analysis. Intensity was graded on a scale of 0-3 with 0 representing no staining, 1-faint positivity, 2-intensity equal to internal control and 3-intensity greater than internal control. ERG expression was then separated into binary values for positive and negative expression. Those cores with an intensity of 1 or more in greater than 10% of cells were considered positive while a score of 0 or staining in ≤10% of cells

109 was considered negative. We considered a case positive for ERG expression if any of the arrayed cores from that case displayed positive ERG IHC as described above.

3.3.6 Statistical analysis

PMR scores for each sample analyzed were obtained from averaging duplicate runs. When multiple foci were analyzed for each PCa specimen, an individual PMR value was assigned to each case based on an average of the PMR values obtained for each focus within that case.

Receiver-Operator Curve (ROC) analysis with associated area under the curve (AUC) was performed to ascertain the ability of each methylated gene to distinguish cancer from adjacent normal tissue.

Univariate disease-free survival (biochemical recurrence) was calculated using the log-rank test and Kaplan-Meier method. Multivariate Cox proportional hazards regression analysis was used to analyze individual contributions of each variable to disease-free survival. PMR values were grouped into four quartiles for initial univariate analysis in order to determine optimal PMR thresholds. The threshold for determination of a high methylation (HM) versus low methylation

(LM) dichotomy was then chosen to be between the two quartiles which had the largest difference in overall disease-free survival. Analysis of the relationship between average PMR for each gene and GS/pattern and stage was done using non-parametric testing as the PMR data for each gene was not normally distributed. The Kruskal-Wallis H test was used for ≥ 3 groups and the Mann-Whitney U test for 2 groups. Pearson chi-square tests were used to analyze proportional differences between HM cases in each category of GS, pattern, and stage. The

110

Fisher exact method was used to replace the Pearson tests when spreadsheet cell counts were less than 5. Furthermore, we performed Wilcoxon tests for matched GP analysis.

For all described methods p-values of ≤0.05 were considered significant. Multiple testing adjustments were not performed as the four genes validated were specifically chosen due to their associations with GS and disease recurrence in Chapter 3 and adjustment doing so would increase the chances of type II errors (463). All statistics were performed using SPSS (Chicago,

IL) and R statistical software.

3.4 Results

3.4.1 Clinical and pathological variables

Table 3.1 displays the clinicopathological characteristics of the 219 patients included in the study. The median age of patients was 62 years. A total of 76 patients (34.7%) had biochemical recurrence, while the mean follow up time was 1600 days (range 63-3460).

111

Clinicopathological Characteristics Number of patients 253 Gleason Score 45 5 38 6 87 7 (3+4) 79 7 (4+3) 24 8 13 95 10 2 Pathological Stage pT2 165 pT3a 62 pT3b 21 pT4 5 Surgical Margin Status Negative 195 Positive 58 *Average pre-operative PSA (Range) 8.4 (0.1-45.8) aAverage follow-up time in years 4.38 (0.17- (Range) 9.48) aNumber of biochemical recurrences (%) 76 (34.7) aMedian Age (Range) 62 (32-75) *N = 204 aN = 219

Table 3.1. Clinicopathological characteristics of MethyLight cohort *N=204 aN=219

112

3.4.2 DNA methylation in tumour adjacent benign tissue versus PCa

Receiver-Operator curves for the four genes analyzed are shown in Figure 3.1. Each gene had an

AUC significantly greater than the chance value of 0.5 (p-values < 0.001). Methylation of GENE

X differentiated PCa from adjacent benign the best with an AUC of 0.836 (p-value = 9.86 x 10-

39), while TGFβ2 had a relatively poor AUC of 0.698 (p-value = 6.11 x 10-7). The latter result is

likely due to the overall low sensitivity of TGFβ2 (47.6% maximum at PMR of 0.05).

Nonetheless, the specificity for TGFβ2 was reasonably high at low PMRs (90.3% at PMR of

0.05). The PMR for PCa specimens was significantly greater than adjacent benign specimens for

each gene analyzed (Table 3.2, Mann-Whitney p-values < 0.001). Additionally, paired Wilcoxon

tests were performed to assess if methylation is consistently greater in PCa versus adjacent

benign tissue on a case-by-case basis. All four genes had significantly greater PMRs in PCa

versus adjacent benign tissue, with GENE X again being the most significant gene (Wilcoxon p-

value = 7.50 x 10-26).

113

Figure 3.1. Receiver-operator curve analysis. (A) HOXD3, (B) TGFβ2, (C) HOXD8, and (D) GENE X. AUC – area under the curve

114

3.4.3 DNA methylation and GS/pathological stage

PMR values for HOXD3, GENE X and HOXD8 had significant variation between GSs <7, 7, and >7 (Table 3.3; Kruskal-Wallis p-values > 0.001 for HOXD3 and GENE X, p-value = 0.032 for HOXD8), while TGFβ2 methylation did not significantly differ according to GS grouping (p- value = 0.1). The data was further tested in a pairwise fashion using groups of low (GS < 7), intermediate (GS = 7), and high grade (GS > 7). All four genes had significantly greater methylation in the GS 7 group versus the GS < 7 group, with GENE X showing the most significant results (p-value = 3.14 x 10-6). When comparing GS > 7 versus GS 7 groups, GENE

X was the only gene with significantly different methylation (p-value = 0.036). We also tested

for differences between Gleason 7 cases that were predominant pattern 3 (3+4) versus

predominant pattern 4 (4+3). Although a comparison of PMR averages revealed greater

methylation in 4+3 versus 3+4 approaching significance for HOXD3 (33.5 vs. 27.0, respectively;

p-value = 0.072), there were no other genes with significantly different PMRs (all p-values >

0.4).

The relationship of DNA methylation with pathological stage was also examined (Table 3.3).

There were significant differences between organ confined pT2 cases and locally advanced

pT3/pT4 cases for HOXD3, GENE X, and HOXD8 methylation, with HOXD3 showing the

greatest association with advanced stage (p-value = 6.93 x 10-5). Again, TGFβ2 methylation was

not significantly associated with a more advanced stage (p-value = 0.935).

115

Benign Cancer average PMR average PMR MWU test Wilcoxon test HOXD3 7.56 27.16 < 0.001 < 0.001 TGFβ2 0.28 6.83 < 0.001 < 0.001 HOXD8 8.9 26.27 < 0.001 < 0.001 GENE X 4.66 19.45 < 0.001 < 0.001

Table 3.2 Average PMR values and p-values from cancer versus benign analysis of four genes. MWU – Mann-Whitney U

GS ≤ 6 GS 7 GS ≥ 8 KW p- pT2 pT3/pT4 MWU p- PMR PMR PMR value PMR PMR value HOXD3 18.69 28.99 41.92 <0.001 21.51 32.2 <0.001 TGFβ2 3.48 7.09 14.82 0.1 6.12 6.01 0.935 HOXD8 23.27 28.9 27.9 0.032 24.42 29.4 0.018 GENE X 13.37 21.51 28.93<0.001 15.77 23.06 0.001

Table 3.3. Average PMR values for four genes according to GS and pathological stage. KW – Kruskal-Wallis; MWU – Mann-Whitney U

116

3.4.4 DNA methylation and GP

Next, DNA methylation was analyzed between each separate GP. Patterns 4 and 5 were grouped into the same high grade category due to a small number of observations for pattern 5 carcinoma

(n = 8). Initial Kruskal-Wallis analysis demonstrated significant differences between all groups for HOXD3, GENE X and HOXD8 (p-values < 0.001), while TGFβ2 methylation did not differ according to GP. Comparisons of PMR for GP 2 versus GP 3 showed significantly greater PMR in GP 3 for HOXD3, GENE X and HOXD8, but not for TGFβ2 (Figure 3.2). In addition, GP 4/5 had significantly greater DNA methylation than GP 3 for both HOXD3 and GENE X, while

HOXD8 and TGFβ2 were not significant (p-values > 0.05). Furthermore, we compared the methylation levels from identical GPs obtained from different GS. Therefore, pattern 3 from GS

7 was compared to pattern 3 from GS 6, and pattern 4 from Gleason 8 was compared to pattern 4 from Gleason 7. Interestingly, methylation of GENE X was greater in GP3 from GS7 cases compared to GP3 from GS6 cases (p-value = 0.003). Significant differences were not observed, however, for GP4 from GS > 7 versus GS 7 (p-value = 0.319). There were also no significant differences for similar comparisons made with each of the other three genes (all p-values > 0.1).

Finally, a paired Wilcoxon test was used to distinguish if methylation was consistently greater in

GP4 specimens versus GP3 specimens on a case-by-case basis (Table 3.4). There were significantly greater levels of methylation in GP4 for all four genes analyzed, with HOXD3 and

GENE X showing the greatest differences (p-values = 9.42 x 10-5 and 0.003, respectively).

117

Pattern 4 > Pattern 3 Pattern 4 < Pattern 3 Wilcoxon p-value HOXD3 77 44 <0.001 TGFβ2 40 28 0.017 HOXD8 67 47 0.027 GENE X 70 41 0.003

Table 3.4. Wilcoxon paired analysis of pattern 3 and pattern 4 specimens. Pattern 4 > Pattern 3 – number of specimens with greater PMR in pattern 4; Pattern 4 < Pattern 3 – number of specimens with greater PMR in pattern 3

Figure 3.2. Average PMR values for each GP across four genes. Circles represent outliers (between 1.5 and 3 interquartile ranges from end of box) and asterisks represent extreme cases (greater than 3 interquartile ranges from end of box). A – Mann-Whitney U p-value < 0.01; B – Mann-Whitney U p-value < 0.001

118

3.4.5 DNA methylation association with ERG

Next, we sought to determine if ERG expression is related to DNA methylation of the four genes analyzed. The monoclonal ERG antibody used in this study has previously been shown to accurately reflect the presence of TMPRSS2-ERG fusions in PCa (464), and we observed an

ERG nuclear staining pattern (Figure 3.3) similar to IHC results previously reported for similar antibodies (465).

Of the 258 cases that were represented on the TMA, 13 did not yield information due to lack of internal control endothelial cell IHC staining. Positive ERG protein expression was observed in

125/245 (51.0%) of cases, consistent with prior reports of the frequency of TMPRSS2-ERG fusions (115, 466). With regard to its association with GS, ERG expression was found more frequently in low and intermediate grade GS 5, 6 and 7 PCa, whereas GSs 8-10 had a lower frequency of positive expression (Figure 3.4). The distribution of ERG expression according to

GS was significantly different (χ2 p-value = 0.004). We also analyzed ERG expression according

to GPs 2, 3, 4, 5, and adjacent benign tissue (Figure 3.4). High grade patterns 4 and 5 were again

combined into one category as there were only 11 specimens available in the pattern 5 category

(4/11 positive; 36.4%). Of the 194 tumour adjacent normal specimens analyzed, only 1 was

positive for ERG expression. In addition, ERG expression was only present in 3/51 (5.6%) of

pattern 2 specimens, while it was highly enriched in pattern 3 (56.9%) and somewhat reduced in pattern 4 (43.0%) compared to pattern 3. Overall, the distribution of ERG expression was

significantly different across the low to high grade tumour spectrum (χ2 p-value = 1.14 x 10-10).

The difference between pattern 4 and pattern 3 expression was also statistically significant (χ2 p- value = 0.011). With respect to pathological stage, we found a significant increase in ERG

119 positivity when comparing organ confined pT2 to locally invasive pT3/pT4 PCa (Figure 3.4, χ2

p-value = 0.005).

120

Figure 3.3. Representative cores from tissue microarray ERG immunohistochemistry. (A) ERG IHC positive staining intensity of 1 in 100% of tumour epithelial cells, (B) ERG IHC negative cancer, and (C) ERG IHC negative benign glands.

Figure 3.4. Proportion of cases with positive ERG expression stratified by pathological variable. (A) GS and (B) GP categories. The proportion of ERG positivity distributed by pathological stage is shown in (C). P-values represent χ2 analyses.

121

Integrating ERG expression with DNA methylation, we found that quantitative methylation values of HOXD3 and HOXD8 were significantly greater in ERG positive versus ERG negative

PCa (Figure 3.5; p-values < 0.001 and 0.005, respectively), while TGFβ2 and GENE X were not associated with ERG expression (p-values > 0.05). The relationship for HOXD3 and HOXD8 existed regardless of GS or pathological stage, as partial correlation analysis correcting for both of these variables revealed a significant association for both genes (p-values < 0.001 and = 0.004 for HOXD3 and HOXD8, respectively).

3.4.6 Multivariate regression model for GS and pathological stage

We next tested the independent contributions of HOXD3, TGFβ2, HOXD8, GENE X, ERG, and

PSA in a logistic regression model with both pathological stage and GS considered as dependent variables (Table 3.5). In the regression model for stage, PSA, ERG expression GENE X methylation were the only significant variables (p-values = 0.014, 0.021, and 0.027 respectively), while HOXD3 methylation approached significance (p-values = 0.066). In the regression model for GS, only HOXD3 and GENE X were significant (p-values = 0.006 and 0.007, respectively).

122

Figure 3.5. Boxplots of PMR values stratified by ERG expression status. P-values obtained from Mann-Whitney U tests.

123

Stage GS p-value HR p-value HR PSA .014 1.074 .259 1.032 HOXD3 .066 1.016 .006 1.027 TGFB2 .933 .999 .186 1.020 GENE X .027 1.029 .007 1.041 HOXD8 .655 1.005 .838 1.003 ERG .021 2.255 .280 .692

Table 3.5 Multivariate regression models for pathological stage and GS

124

3.4.7 DNA methylation and biochemical progression-free survival

We next examined the relationship between biochemical recurrence and each of the four methylation markers. Univariate analysis log-rank analysis with Kaplan-Meier curves were performed for each clinicopathological variable (Figure 3.6) and each gene. Initially, PMR values were binned into quartiles to assess if there were stepwise decreases in time-to-recurrence depending on quantifiable increases in PMR and also to develop suitable threshold values (i.e. dichotomizing the PMR values) that may best determine risk of biochemical recurrence (Figure

3.7). TGFβ2 was separated by values lower than the median and further by the two upper quartiles, as most PMRs below the median were 0. For HOXD3, TGFβ2, and GENE X, the median PMR value best distinguished the number of recurrences, while for HOXD8 the third quartile PMR value was the optimal threshold as there was an increase in the proportion of recurrences at this threshold (Table 3.6). When performing a log-rank test of the PMR data separated into quartiles, only HOXD8 showed a significant difference in time-to-recurrence between the four quartiles. After dichotomizing methylation based on the aforementioned thresholds (high methylation and low methylation, HM and LM respectively), we again performed log-rank tests and Kaplan-Meier curves (Figure 3.8). TGFβ2, HOXD8, and GENE X all showed significant differences in time to recurrence (p-values = 0.036, 0.004, and 0.015, respectively), while HOXD3 approached significance (p-value = 0.080). Next, we performed univariate analysis using the three significant genes TGFβ2, HOXD8, and GENE X separated by having zero, one, two, or three of genes in the HM category. This provided the most significant separation of KM curves, with stepwise decreases in the proportion of patients free from recurrence (86%, 65.1%, 58.7%, and 35%, Figure 3.8E). It should be noted, however, that all

125 other three gene combinations were also significant in univariate analysis, suggesting an additive predictive effect that is not dependent on the genes that are included.

126

Figure 3.6. Kaplan-Meier curves and log-rank p-values with biochemical recurrence as outcome for clinicopathological variables. (A) GS, (B) pathological stage, (C) PSA and (D) surgical margin status.

127

Figure 3.7. Kaplan-Meier curves and log-rank p-values for methylated genes. (A) HOXD3, (B) TGFβ2, (C) HOXD8 and (D) GENE X with PMR values binned into four quartiles. <1 – below the first quartile; 1-<2 – between the first and second quartile; 2-<3 – between the second and third quartile; >3 – greater than third quartile

128

Figure 3.8. Kaplan-Meier curves and log-rank p-values for methylated genes stratified by high methylation and low methylation. (A) HOXD3, (B) TGFβ2, (C) HOXD8, (D) GENE X and (E) the multigene methylation signature of TGFβ2/HOXD8/GENE X. PMR values are grouped into low methylation (LM) or high methylation (HM). 0, 1, 2, 3 – number of genes methylated in multigene methylation signature

129

HOXD3 TGFβ2 HOXD8 GENE X <1 32.1 N/A 31.5 26.8 1-<2 25.9 26.8 32.1 25.9 2-<3 42.6 43.4 24.5 41.5 >3 38.2 42.6 50 44.6

Table 3.6. Percentage of patients with biochemical recurrence stratified by PMR quartile. Grayed cells represent the threshold between low methylation (LM) and high methylation (HM).

130

Furthermore, we performed multivariate Cox regression analysis of our data for each gene independently (Tables 3.7-3.10). Significant predictors of biochemical progression-free survival for all four analyses included GS, pathological stage, and surgical margin status, while age was not a significant factor in any of the four analyses. In addition, HOXD3, GENE X, and HOXD8 methylation were not significant factors, while TGFβ2 methylation was a significant predictor (p

= 0.007). The multi-gene signature of TGFβ2, HOXD8, and GENE X was also a significant predictor, although the level of significance was slightly lower than that of TGFβ2 alone (Table

3.11, p = 0.009).

131

95.0% CI for HR HR Lower Upper P-value PSA 1.031 0.994 1.069 0.103 Pathological Stage 3.007 1.817 4.978 0.000 Surgical Margins 2.301 1.401 3.777 0.001 Age 1.013 0.977 1.050 0.496 GS 3.323 1.600 6.904 0.001 HOXD3 0.973 0.583 1.624 0.918

Table 3.7. Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and HOXD3 methylation. HR – hazard ratio; CI – confidence interval

95.0% CI for HR HR Lower Upper P-value PSA 1.038 1.003 1.074 0.034 Pathological Stage 3.031 1.873 4.904 0.000 Surgical Margins 2.342 1.421 3.860 0.001 Age 1.004 0.968 1.041 0.832 GS 4.219 2.041 8.720 0.000 TGFB2 2.018 1.211 3.365 0.007

Table 3.8. Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and TGFβ2 methylation. HR – hazard ratio; CI – confidence interval

95.0% CI for HR HR Lower Upper P-value PSA 1.030 0.993 1.069 0.114 Pathological Stage 2.814 1.729 4.581 0.000 Surgical Margins 2.192 1.334 3.604 0.002 Age 1.019 0.982 1.058 0.327 GS 3.265 1.588 6.714 0.001 HOXD8 1.450 0.873 2.409 0.152

Table 3.9. Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and HOXD8 methylation. HR – hazard ratio; CI – confidence interval

132

95.0% CI for HR HR Lower Upper P-value PSA 1.031 0.994 1.069 0.106 Pathological Stage 2.907 1.776 4.759 0 Surgical Margins 2.296 1.401 3.764 0.001 Age 1.011 0.975 1.049 0.55 GS 3.173 1.527 6.596 0.002 GENE X 1.131 0.673 1.901 0.642

Table 3.10. Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and GENE X methylation. HR – hazard ratio; CI – confidence interval

95.0% CI for HR HR Lower Upper P-value PSA 1.034 0.996 1.072 0.077 Pathological Stage 2.662 1.640 4.323 0.000 Surgical Margins 2.209 1.344 3.629 0.002 Age 1.010 0.974 1.047 0.599 GS 3.241 1.589 6.610 0.001 TGFB2, HOXD8, GENE X 1.466 1.102 1.951 0.009

Table 3.11. Multivariate Cox regression model for biochemical recurrence with clinicopathological variables and three gene methylation signature. HR – hazard ratio; CI – confidence interval

133

3.4.8 ERG, DNA methylation, and biochemical progression-free survival

We next examined the relationship between ERG expression and biochemical recurrence within this patient cohort. Univariate Kaplan-Meier/log-rank analysis showed no association between

ERG expression and disease-free survival (Figure 3.9). While an interim 5 year survival trend appeard evident, further analysis limiting the cohort to a 5 year time frame showed this to be statistically non-significant (p = 0.238). We also performed multivariate Cox regression analysis including clinicopathological variables PSA, GS, pathological stage, surgical margin status, and age. ERG positive cancers had a protective effect that approached significance (p = 0.103), while increased GS, stage, and positive surgical margin status remained significant predictors (p-values

< 0.01).

Finally, we investigated the value of HOXD3, TGFβ2, HOXD8, and GENE X in predicting biochemical recurrence when stratified by ERG expression status. Interestingly, TGFβ2 methylation was predictive in ERG negative cases (Figure 3.10, p = 0.029) and not ERG positive

(p = 0.571). HOXD8 was also predictive in ERG negative cases and not ERG positives (Figure

3.10), although the p-value for ERG positive cases trended towards significant (p = 0.071). In addition, the multi-gene signature was predictive in ERG negative cases (p = 0.002) and not in

ERG positive cases (p = 0.177).

Using a multivariate Cox regression analysis for each gene plus the multi-gene signature and

ERG expression, we observed that ERG was only a significant predictor of disease recurrence in the model containing HOXD8 alone (Table 3.12; p = 0.05), while HOXD8 itself approached significance (p = 0.065).

134

Figure 3.9. Kaplan-Meier curves and log-rank p-value for ERG expression status.

135

Figure 3.10. Kaplan-Meier curves and log-rank p-values for methylated genes stratified by ERG expression status. (A,B) TGFβ2 methylation and (C,D) HOXD8 methylation stratified by (A, C) ERG negative and (B, D) ERG positive expression.

136

95.0% CI for HR HR Lower Upper P-value PSA 1.032 0.993 1.072 0.106 Pathological Stage 3.226 1.910 5.447 0.000 Surgical Margins 2.307 1.396 3.813 0.001 Age 1.020 0.983 1.059 0.297 GS 2.725 1.284 5.783 0.009 HOXD8 1.650 0.974 2.793 0.062 ERG status 0.586 0.344 1.000 0.050

Table 3.12. Multivariate Cox regression model including clinicopathological variables, HOXD8 methylation, and ERG expression status.

137

3.5 Discussion

Our results show that HOXD3, HOXD8, and GENE X methylation is elevated in intermediate grade GS 7 compared to low grade GS 6 tumours. With regard to previously characterized markers APC and RASSF1A (characterized by another student in the lab within the same cohort), both HOXD3 and GENE X have stronger associations (both p-values < 0.001 versus

0.018 for APC and 0.490 for RASSF1A). Gleason 6 to 7 represents a biologically and clinically important switch point as GS 7 cancers have an unfavourable prognosis compared to those of lower grades. Additionally, GENE X methylation increased in the GS > 7 compared to GS 7 group. Although HOXD3 was not significant for this comparison, this may be due to the relatively small number of high grade PCa specimens available (n = 19). Assigning a HOXD3,

HOXD8, and/or GENE X methylation value in conjunction with biopsy GS and clinical stage may have important clinical applications. GSs assigned on a biopsy are often inaccurate as compared to the GS assigned in the corresponding prostatectomy specimen and may lead to inappropriate treatment strategies (467, 468). Similarly, clinical staging lacks accuracy as many cancers appear to be understaged if compared with the pathological stage obtained following surgery. Therefore, markers which are indicative of a higher grade and stage of PCa could be helpful in clinical decision making. DNA methylation may also be detectable in voided urine samples or following prostatic massage as has been shown for other methylation markers (469), thus allowing a diagnosis/prognosis prior to any further invasive exams as detection of methylation would indicate the presence of high grade disease.

A unique aspect of this study was the large-scale comparison of different GPs from matched PCa specimens. We were able to show that methylation of both HOXD3 and GENE X is significantly

138 greater in high grade versus low grade GPs, either when comparing overall averages or when matching patterns to the same patient. Importantly, we observed increases in PMR for these two genes progressing from matched normal to pattern 3 and further onto high grade patterns 4 or 5.

We also observed a quantitative increase in methylation from normal to pattern 3 for TGFβ2 and

HOXD8, although there was no difference for either gene comparing GP3 to GP4/5. It is possible that methylation of HOXD3, along with other genetic and/or epigenetic events including

GENE X methylation, is responsible for progressive de-differentiation of PCa foci. For example, loss of expression of the homeobox gene PDX1 has been shown to have an inverse correlation with increasing GP (470). In addition, loss of the androgen responsive gene NKX3.1 (required for differentiating prostate epithelial cells) has been reported as has HOXC8 overexpression contributing to a loss of differentiation (191, 396, 471). Hence, it is likely that the proper expression of specific homeobox genes (including HOXD3) is crucial for the maintenance of a differentiated tissue phenotype, which is concordant with the role that homeobox genes play in the development of tissue specific architecture (472, 473). However, an alternative to this would be that HOXD3 and GENE X methylation may be an epiphenomenon of these processes instead of a cause. Future work is necessary to address this issue.

Interestingly, we did not observe any overall differences in methylation between normal tissue and GP 2 carcinomas for HOXD3 and HOXD8, while TGFβ2 and GENE X had significantly greater levels of methylation in GP 2 versus benign. One possible explanation may be that pattern 2 carcinomas often arise from the transition zone of the prostate, and transition zone carcinomas have differing gene expression profiles compared to peripheral zone cancers (474,

475). Thus, DNA methylation patterns may also differ according to different prostatic zones from which the cancers arise.

139

Univariate time-to-recurrence analysis revealed significant differences between methylated states for TGFβ2, HOXD8, and GENE X, and there was a trend towards significance for HOXD3 (p =

0.080). ). A combination of TGFβ2, HOXD8, and GENE X proved to be the most significant predictor in univariate analysis, suggesting that a multi-gene panel may have more clinical utility than any one gene alone. This has been suggested by other groups (476).

We chose to initially perform univariate analysis by binning PMR values into four quartiles for each gene. Previous studies which have employed quantitative methylation profiling have used numerous thresholds for “positive” versus “negative” methylation, including values determined from medians (314) and third quartiles (439, 440). Indeed, we have previously used the third quartile as a threshold (477). It is important to note, however, that these two thresholds are arbitrary in nature and may not be suitable for all markers. Determining a threshold from ROC analysis of cancer versus benign specimens is not well suited for our purpose either, as most cases would be classified as methylated due to the fact that there were consistently higher degrees of methylation in cancer specimens versus benign specimens. The determination of PMR threshold should be determined independently for each gene analyzed, as the spectrum of values is likely dependent on technical aspects (i.e. PCR primer efficiencies) as well as biological aspects (i.e. the amount of methylation necessary to induce gene silencing). Upon multivariate

Cox regression analysis with classical prognostic markers (GS, stage, surgical margin status), the contribution of methylation status to the model was not statistically significant for HOXD3,

HOXD8 or GENE X. There may be multiple reasons for this occurrence. First, methylation of each locus might be part of a panel of methylation markers, which together predict PCa disease course with a high degree of accuracy. Prior studies have shown the usefulness of multi-gene methylation panels for diagnosis and prognosis of PCa (439, 441).

140

The role that HOXD3 may play in PCa development and progression has not been elucidated.

Similar to PCa, this gene has been shown to be methylated in A549 lung cancer cells (447).

Functional analysis of HOXD3 has shown that overexpression drives angiogenesis, and in lung cancer A549 cells it contributes to a metastatic phenotype via coordinate expression of metastasis related genes through TGFβ dependent and independent mechanisms (432, 433).

Whether or not the paradoxical situation of methylation and increased expression leading to metastasis exists in PCa is not known. It may be that complex mechanisms exist in order to maintain proper expression of HOXD3 in tissue specific settings, and DNA methylation may only partially contribute.

In PCa PC3 cells, TGFβ2 appears to play a role in blocking apoptosis via induction of NF-κB and also increases the levels of pro-invasive genes MMP-9 and PAI-1 (478, 479). Also, TGFβ2 is preferentially secreted by prostate epithelial cells as compared to TGFβ1 and is activated by

PSA (480). Within the normal prostate, TGFβ signaling acts to promote cellular differentiation and apoptosis and inhibit cell proliferation (481), while in later stages of PCa TGFβ signaling may work to promote invasion, angiogenesis, and immune cell evasion (482, 483). It is unclear exactly what role TGFβ2 methylation may play in the pathogenesis of PCa. It is possible that

DNA methylation induced downregulation of TGFβ2 promotes proliferation and inhibits differentiation and apoptosis in primary PCa, but acts in an opposite manner in metastatic lesion consistent with functional reports in metastatic cell lines (479).

The function of HOXD8 in PCa is also unclear. Previous studies have shown similar methylation of the HOXD8 promoter region in breast cancer and mantle cell lymphoma with concomitant downregulation of HOXD8 transcript (436, 484). Furthermore, HOXD8 RNA is significantly reduced in colon cancer metastasized to the liver compared to primary colon cancer, while

141 hepatocellular carcinomas have increased levels of HOXD8 RNA (435). In lymphatic endothelial cells, HOXD8 increases the expression of PROX1 and promotes lymphangiogenesis (437).

Similar to HOXD3 and other homeobox genes, however, the role of HOXD8 is likely tissue specific and dependent on transcriptional co-factors (485).

In conclusion, we have characterized HOXD3, TGFβ2, HOXD8 and GENE X DNA methylation in a series of PCa cases and analyzed its relationship to classical clinicopathological parameters.

We have shown that methylation levels are significantly greater in intermediate grade versus low grade GSs for HOXD3, HOXD8, and GENE X, and have shown increased methylation with loss of tumour differentiation according to GPs. Furthermore, we have demonstrated a correlation between DNA methylation of HOXD3 and HOXD8 and ERG expression, suggesting a functional relationship that requires further research.

Finally we have shown that both TGFβ2 and a multi-gene panel including TGFβ2, HOXD8 and

GENE X predict biochemical recurrence, and that prognostic capabilities may in part be dependent on ERG fusion status. Future work is necessary to elucidate both the clinical utility and functional relevance of DNA methylation of these genes, including possible detection in urine/serum of PCa patients as well as the role of methylation and/or abnormal expression in PCa progression.

142

Chapter 4

Expression of novel methylated HOXD loci and putative role of HOXD8 in PCa

progression

Ken Kron1,2, Theodorus van der Kwast1,3, and Bharati Bapat1,2

1. Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON, Canada

2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON,

Canada

3. Department of Pathology, University Health Network, University of Toronto, Toronto, ON,

Canada

The data presented in this chapter was primarily contributed by Ken Kron. TVDK, a pathologist, assisted in specimen retrieval and histopathological confirmation.

143

Chapter 4

Expression of novel methylated HOXD loci and putative role of HOXD8 in PCa progression

4.1 Summary

DNA methylation silences gene transcription when present at gene promoter regions. We have discovered DNA methylation of HOXD3 and HOXD8 and shown an association with high grade disease and disease recurrence, respectively, for each gene. Here, we have used DNA demethylating treatment in vitro and qRT-PCR to assess the contribution of DNA methylation to mRNA expression of HOXD3 and HOXD8. Furthermore, we profile mRNA and DNA methylation in frozen tissue, use an in silico approach to determine the prognostic value of

HOXD3 and HOXD8 in PCa, 5’ RACE and RT-PCR to map transcription start sites of these two genes, and determine the contribution of HOXD8 to cell viability and motility. DNA demethylating treatment increased the expression of HOXD8 in LNCaP, DU145, and PC-3 cells while also increasing the expression of HOXD3 in DU145 cells. Quantitative profiling of methylation and mRNA, however, revealed no direct correlation between DNA methylation and gene expression, while mRNA expression of HOXD3 and HOXD8 in a large cohort of PCa patients showed increased expression associated with shorter time to biochemical recurrence. In

5’ RACE analysis, we found an alternative transcript of HOXD3 with transcription beginning well upstream of the canonical promoter region and within the second exon of HOXD4. We also discovered multiple HOXD8 transcripts, none of which encode for full length HOXD8 protein.

144

Finally, we find that HOXD8 does not play a role in total cell viability or cell motility. These results provide evidence of DNA methylation that does not play a direct role in silencing gene transcription in the HOXD cluster due to alternative transcription start sites.

4.2 Introduction

Homeobox genes are developmentally important transcription factors that contain a well- conserved 60 amino acid DNA binding motif referred to as the homeodomain. There are a total of 300 homeobox loci suggested in the human genome, with 235 likely functional genes and 65 pseudogenes (486). Of the 235 functional genes, 39 are in the HOX family and are located in well defined clusters along chromosomes 2, 7, 12, and 17. These are termed HOXA, HOXB,

HOXC, and HOXD, which include 11, 10, 9, and 9 paralagous genes respectively.

Management of proper HOX gene expression is maintained in large part by Polycomb and

Trithorax group of genes (487, 488). The former of these groups act to repress HOX gene transcription while the latter activates transcription. Two main Polycomb repressive complexes exist in humans, termed PRC1 and PRC2. PRC2 is composed of EZH2, SUZ12, EED, and

RbAp46 and catalyzes the repressive trimethylation of the histone H3 lysine 27 (H3K27me3) mark. MLL complexes (human counterpart to fly Trithorax proteins) act to catalyze the activating trimethylation of H3 lysine K4 (H3K4me3). Recent work has also identified important roles for miRNAs and DNA methylation in regulation of HOX gene expression. In terms of miRNAs, miR10b has been shown to regulate HOXD10 (416), while miR-196a degrades

HOXB8, HOXC8 and HOXD8 mRNA (489). With respect to DNA methylation, tissue specific patterns have been found surrounding the PAX6 gene as well as within the HOXC cluster (313,

145

490), indicating a role for tissue specific regulation of gene expression by DNA methylation.

Therefore, it is evident that controlled expression of specific HOX genes is essential and carried out under an intricate and multifaceted epigenetic system consisting of DNA methylation, histone modifications and miRNA regulation (383-386).

Deregulated homeobox gene expression has been observed in a variety of cancers including leukemia, lung and prostate carcinomas (387-389). For example, NKX3.1 (NK homeobox, family 3, A), residing on 8p21, has been identified as frequently deleted in PCa with concomitant loss of protein expression, implicating this gene as a PCa tumour suppressor (180). Other examples of abnormal HOX gene expression include NUP98-HOX gene fusion events in leukemia (392, 491) and PAX5 expression in medulloblastoma (492). Examples of downregulated HOX genes include downregulated HOXA4 corresponding with shorter overall survival in acute myeloid leukemia (400), HOXA5 inactivation via promoter methylation in breast cancer (401), and large scale methylation of HOX genes in lung cancers (493). In the prostate, HOXB13 binds the androgen receptor and mediates the response to androgens (219), and mutations of HOXB13 have been found in familial PCa patients (223). Also in PCa, it has been shown that HOXC8 mRNA is overexpressed in moderate/poorly differentiated GS7-9 carcinoma as compared to well differentiated GS3-6 (396). Follow up functional experiments revealed that a PBX1/HOXC8 complex mediates androgen independent cell growth in DU145 cells (438), and that HOXC8 blocks AR recruitment of co-activators at androgen-regulated gene enhancers while promoting the invasiveness of non-tumourigenic HPr-1 AR cells (397).

Based on results obtained in Chapters 2 and 3, we sought to determine the consequence of DNA methylation on expression of homeobox genes HOXD3 and HOXD8. Furthermore, we analyzed the role that HOXD8 plays in cell proliferation and motility in PCa cells.

146

4.3 Materials and Methods

4.3.1 5-aza-2-deoxycytidine treatment and RNA extraction of cells

LNCaP, DU145, and PC3 cells were treated with the demethylating agent 5-aza-2-deoxycytidine

(DAC) at a concentration of 2µM for 96 hours, with media/DAC replacement after 48 hours.

Control cells were treated with an equivalent volume of PBS in RPMI 1640 media supplemented with 10% FBS. All experiments were performed in triplicate. Following treatment, culture media was removed, cells were rinsed once with PBS, and 400µL Trizol (Life Technologies) reagent was added for RNA extraction. Total RNA was then extracted using the manufacturer’s recommended protocol.

4.3.2 Reverse transcription, RT-PCR, and qRT-PCR

1 µg of total RNA was reverse transcribed using the Bio-Rad iScript cDNA synthesis kit (Bio-

Rad) following the manufacturer’s recommended protocol. Following synthesis, cDNA was diluted to a concentration of 25 ng/µL.

For RT-PCR analysis of expression, 50 ng of cDNA was used in 25 µL volume reactions using

Taq Polymerase (Fermentas). All reactions were run with the following final reagent concentrations: 1x PCR buffer, 2mM MgCl2, 0.2µM of both forward and reverse primer, 0.25

units Taq polymerase, and 0.4mM each dNTP. For HOXD8 RT-PCR reactions that amplified

within the CpG island of the first full length exon, the addition of DMSO at a final concentration

of 5% was necessary. We performed 35 cycles of denaturation at 95ºC (30 seconds), annealing

147

(range from 56 ºC-62ºC depending on primer set; 30 seconds), and extension at 72ºC (1 minute for amplicons ≥ 1000 bp, 30 seconds additional for each 500 bp afterwards).

For qRT-PCR analysis, 25 ng of cDNA was used in 20 µL volume reactions using Quanta

Biosciences Perfecta SYBR Green SuperMix. All reactions were run with the following final reagent concentrations: 1x Perfecta SuperMix and 0.25µM of both forward and reverse primers.

The qPCR reaction consisted of 10 minutes initial incubation at 95ºC for enzyme activation, followed by 40 cycles of 95ºC denaturation (15 seconds) and 60ºC simultaneous annealing and extension (1 minute).

4.3.3 5’ and 3’ RACE

5’ and 3’ RACE assays were performed using total RNA from RWPE-1 and PC-3 cells and the

Ambion FirstChoice RLM RACE kit with modifications to the manufacturer’s recommended protocol. For 5’ RACE, RNA was first treated with tobacco acid pyrophosphatase to remove 7- methylguanylate caps. Oligonucleotides were then ligated to the 5’ ends of decapped mRNA using T4 RNA ligase. Reverse transcription of oligonucleotide-ligated mRNA was then performed using Superscript III (Life Technologies) with gene specific primers for HOXD3 or

HOXD8. Nested PCR was then performed using an initial outer set of primers, with the forward primer specific to the linker oligonucleotide and the reverse specific to a known region of

HOXD3 or HOXD8. Inner primers were designed in a similar fashion, with the exception that they amplified a region internal to the outer PCR reaction. Gel electrophoresis was then performed on inner PCR reactions using 1.5% agarose gels. Visible bands were excised from the gel using a clean scalpel and DNA was purified using the QIAex II purification kit (Qiagen).

148

Sequencing of purified DNA was performed by The Centre for Applied Genomics in both the forward and reverse directions using primers that were used for the inner PCR reaction.

3’ RACE was performed by first reverse transcribing RNA into cDNA using an oligonucleotide with a dT stretch and a known sequence 5’ to the dT stretch suitable for primer annealing in nested PCR reactions. Nested PCR, gel extraction, and sequencing were performed in the same manner as 5’ RACE.

4.3.4 MSKCC expression data

Normalized log2 expression values from 179 prostate specimens (29 benign, 150 PCa) were

downloaded from Gene Expression Omnibus (GEO) dataset GSE21034. Clinicopathological

information was downloaded from the Taylor et al. publication in Cancer Cell, supplementary

Table S1 (494). For statistical analyses, GS was divided into groups of <7, 7, and >7, while

pathological stage was separated into organ-confined and locally advanced (pT3/pT4). MWU,

log-rank, and Cox regression analyses were all performed using SPSS 18.

4.3.5 Generation of constructs

HOXD8 expression constructs were generated by RT-PCR using RWPE-1 mRNA as a source.

Briefly, primers were designed to amplify full length HOXD8 ORF only, truncated HOXD8

ORF only, Ensembl HOXD8 variant 002 full length mRNA, and newly discovered HOXD8

variant full length mRNA with two distinct 3’ ends. Each forward primer was designed to

contain a NheI digestion site and reverse primers were designed to contain KpnI digestion sites.

149

Amplified cDNA and pcDNA3.1 expression vector were then double digested with NheI and

KpnI and digested cDNA was ligated into pcDNA3.1. Constructs were then transformed into

DH5α and plated. Two colonies from each construct were selected, grown in 250 mL LB broth overnight, prepared by maxi-prep, and sequenced to verify the presence of the appropriate insert and insert orientation.

4.3.6 Western blots

Cells grown in 6 well dishes were lysed with 100 uL RIPA buffer, centrifuged at 4ºC at 10,000 x g for 10 minutes to remove cell debris, and the supernatant was collected and stored at -80ºC.

Protein concentrations were determined using the BCA assay (Thermo Scientific Pierce) according to the manufacturer’s instructions. We loaded 30 ug of lysate in Bio-Rad TGX precast gels, performed gel electrophoresis, and transferred to a PVDF membrane. Membranes were then blocked with 5% milk, incubated with rabbit polyclonal anti-HOXD8 antibody (dilution 1/250 in

5% milk, Imgenex IMG-6632A), which recognizes an epitope at the C-terminal of the HOXD8 protein, overnight at 4ºC. Anti-rabbit secondary antibody (Thermo Scientific Pierce #32460, dilution 1/8000) conjugated to HRP was then incubated at room temperature for 2 hours.

Chemiluminescent detection was then performed using Amersham ECL western blot detection reagent (GE Healthcare).

4.3.7 Cell viability assay

150

Cell viability was assessed using Trypan blue staining (Sigma Aldrich). Briefly, 150,000 PC3 cells were plated in 24 well plates. 24 hours after plating, cells were treated with 66 nM final concentration of siRNA in triplicate using Lipofectamine 2000 (Life Technologies) according to the manufacturer’s protocol. Fresh media was added 24 hours after treatment. 72 hours post- transfection, cells were collected by trypsinization, resuspended in Trypan blue solution, and allowed to incubate for 5 minutes at room temperature. Viable (unstained) were then counted using a hemocytometer.

4.3.8 Cell motility

Cell motility was assessed using wound scratch assays. Fully confluent PC3 cells were treated with 66 nM final concentration of siRNA in triplicate in 12 well plates. 36 hours after treatment confluent monolayers were scratched using a sterile 200 µL pipette tip and the media was replaced with RPMI 1640 supplemented with 1% FBS. Pictures were taken at the 0, 24, and 48 hours following wound induction using an Olympus CKX41 inverted light microscope.

Quantification of wound scratch closure was performed by marking scratch edges at 0 hours and counting cells that migrated into the scratch at 24 and 48 hours with Image J software.

4.4 Results

4.4.1 Methylation affects expression of HOXD3 and HOXD8

151

DNA methylation of the HOXD3 and HOXD8 promoter regions was measured in cell lines using

MethyLight. For HOXD3, neither 22Rv1 nor LNCaP cells were methylated while DU145 and

PC3 cells had similar degrees of methylation (Figure 4.1). For HOXD8, 22Rv1 was again unmethylated, while LNCaP, DU145, and PC3 cells had exhibited varying degrees of methylation. Next, we performed genome wide demethylation of LNCaP, DU145, and PC3 cells using 5-aza-2-deoxycytidine (decitabine - DAC) and measured gene expression using quantitative RT-PCR (qRT-PCR). LNCaP cells were included as they are methylated at the

HOXD8 promoter region while DU145 and PC3 cells were methylated at the promoter regions of HOXD3 and HOXD8.

HOXD3 expression increased significantly in methylated DU145 but not PC3 cells, while in unmethylated LNCaP cells there was no change in expression (Figure 4.2 A, C, E). For HOXD8, expression significantly increased in all three methylated cell lines (Figure 4.2 B, D, F).

152

Figure 4.1. Methylation of HOXD3 and HOXD8 in PCa cell lines. Percent methylation of (A) HOXD3 and (B) HOXD8 in 22RV1, LNCaP, DU145, and PC3 PCa cell lines as measured by the MethyLight assay with supermethylated DNA as a 100% reference.

153

Figure 4.2. HOXD3 and HOXD8 expression following demethylating treatment. PCa cell lines (A, B) LNCaP, (C,D) DU145, and (E, F) PC3 were treated with 5-aza-2-deoxycytidine (decitabine – DAC) and expression of (A, C, E) HOXD3 and (B, D, F) HOXD8 was assessed. * - p-value < 0.05

154

4.4.2 HOXD3 and HOXD8 expression in PCa

We next assessed the relative mRNA levels and methylation levels of HOXD3 and HOXD8 in a small panel of frozen tumours consisting of one benign, four GS6, and five GS8 specimens.

Methylation of HOXD3 displayed the expected increased PMR value in GS8 versus GS6, while the HOXD8 methylation profile was more heterogeneous and did not correlated with GS (Figure

4.3). The expression profile for both genes, however, did not show the expected inverse correlation with increased methylation (Pearson p-values > 0.1). In fact, HOXD8 mRNA was significantly increased in high grade tumours compared to low grade (MWU p-value = 0.004).

In order to verify this profile was not an artefact of the small sample size, we analyzed expression of both HOXD3 and HOXD8 using the MSKCC dataset. An initial comparison of all tumour samples versus benign revealed no clear change in expression for either HOXD3 or

HOXD8 (t-test p-values > 0.5). When analyzing expression according to GS and metastatic disease, however, there appeared to be an increase in HOXD3 and HOXD8 in high grade (GS8) and metastatic disease compared to benign tissue, although neither comparison for either gene was significantly different (Figure 4.4; p-values > 0.1). In addition, expression of HOXD3 and

HOXD8 were analyzed for their ability to predict biochemical recurrence in the MSKCC cohort.

In univariate Cox regression analysis of primary PCa specimens only, HOXD8 expression was significantly associated with shorter time to recurrence while higher HOXD3 expression trended towards significance (Figure 4.5; p-values = 0.018 and 0.064, respectively). To assess which was the stronger predictor, both variables were analyzed in a multivariate Cox regression model.

HOXD8 expression was the only significant predictor (p-value = 0.016), while the HOXD3 predictive value was completely ablated (p = 0.912).

155

Figure 4.3. Methylation and mRNA expression of HOXD3 and HOXD8 in frozen tissues. (A) Percent methylation values for HOXD3 and (B) percent methylation values for HOXD8 in panel of one benign, 4 GS6, and 5 GS8 tumours. Depicted in (C) and (D) are relative mRNA expression values for the same benign/tumour group for HOXD3 and HOXD8, respectively. There appears to be no direct link between methylation level and mRNA expression.

156

Figure 4.4. MSKCC cohort normalized log2 expression values. (A) HOXD3 and (B) HOXD8 expression values stratified according to benign tissue, GS, and metastatic tissue.

157

Figure 4.5. Kaplan-Meier curves of HOXD3 and HOXD8 expression association with biochemical recurrence. (A) MSKCC log2 mRNA expression values separated by median for HOXD3. (B) As in (A), but for HOXD8 mRNA expression.

158

4.4.3 5’ /3’RACE and RT-PCR of HOXD3 and HOXD8

One possible explanation for the lack of inverse correlation between HOXD3 and HOXD8 methylation with expression is the presence of alternative transcription start sites

(TSSs)/promoters. To address this, we performed 5’ RACE in RWPE-1 and PC-3 cells, as both cell lines had detectable levels of HOXD3 and HOXD8 mRNA. For HOXD3, only PC-3 cells yielded specific 5’ RACE products, likely due to the relatively lower levels of HOXD3 expression in RWPE-1 cells. Sequencing of the 5’ RACE product mapped the TSS of HOXD3 to the second exon of HOXD4 (Figure 4.6A). With respect to HOXD8, we obtained results from both RWPE-1 and PC3 cell lines. Sequencing of 5’ RACE products from RWPE-1 cells revealed one variant that corresponded to a previously annotated HOXD8 isoform with a TSS 46 bp upstream from full length, canonical HOXD8 (Figure 4.6B). Sequencing of the PCR product from PC3 cells, however, yielded a HOXD8 variant that hitherto was not described. This alternative form of HOXD8 had a TSS 1,016 bp downstream from full length, canonical

HOXD8. Furthermore, RT-PCR was performed of both the HOXD3 and HOXD8 genes to determine if any variants were missed by 5’ RACE analysis. For HOXD3, we observed one variant that contained the canonical HOXD3 first exon in addition to the upstream first exon

(Figure 4.6A). We also observed two different splice variants of HOXD8 that contained most of the full length, canonical HOXD8 isoform but had small introns located within the canonical first exon (Figure 4.6B). Further 3’ RACE analysis of HOXD8 mRNA yielded two different 3’ UTRs

(Figure 4.6B).

159

Figure 4.6. 5’ RACE, 3’ RACE, and RT-PCR composite map of HOXD3 and HOXD8. (A) The upper panel depicts UCSC genome browser (hg18) region for canonical HOXD3 mRNA and the upstream HOXD4 gene. Below are the results obtained from mapping sequencing of 5’ RACE and RT-PCR products, where black boxes represent exons and black lines represent introns. (B) As in panel (A) for the HOXD8 region. Three distinct isoforms are annotated in the UCSC browser for HOXD8. We verified the presence of the shorter isoform in PCa cells (top most isoform in UCSC map) and also identified a unique shorter mRNA isoform via 5’ RACE. We discovered another two novel isoforms of HOXD8 using RT-PCR and sequencing.

160

Next, we assessed the coding potential of all identified HOXD3 and HOXD8 variants using the open reading frame finder feature from NCBI. Both variants of HOXD3 were predicted to encode for canonical HOXD3 protein without any further N-terminal amino acid sequence, while all four HOXD8 variants were predicted to encode for truncated HOXD8 without any novel amino acid inclusions. All other potential ORFs did not contain an in-frame homeodomain.

The truncated form of HOXD8 lacks N-terminal amino acids 1-184 of 290 amino acids found in the full length protein, producing a protein of 106 amino acids which nonetheless contains the full homeodomain. In addition to this in silico analysis, we created expression constructs for discovered isoforms of HOXD8. Variants containing the majority of canonical HOXD8 first exon were created without a full 3’ end, while the newly identified truncated transcript was designed to contain both the long and short 3’UTRs discovered from 3’ RACE. The previously identified short HOXD8 isoform that was also identified in prostate cells only contained the short

3’UTR, as this was the only amplifiable form of this variant. Each isoform was transiently transfected in low HOXD8 expressing LNCaP and DU145 cells to assess translational ability.

Regardless of mRNA isoform transfected, a protein of approximately 17 kDa was produced along with an additional band of approximately 37 kDa, perhaps indicating dimerization that was not efficiently denatured during gel electrophoresis (Figure 4.7).

161

Figure 4.7. HOXD8 variant protein expression. Expression vectors containing variant HOXD8 isoforms were transiently transfected into (upper panel) LNCaP cells and (lower panel) DU145 cells. For both cell lines and all variants, clear bands were visible at approximately 17 kDa and 37 kDa in size. ORF only – HOXD8 truncated isoform open reading frame only; A – new variant with short 3’ UTR; B – new variant with long 3’ UTR

162

4.4.4 Expression of HOXD8 variants

As we discovered variants of HOXD8 with transcription start sites that are not proximal to observed methylated regions, we next assessed whether the expression of any isoforms varied according to methylation status. Primers were designed to amplify each discovered variant, with the exception of the two long isoforms which were amplified in total. In our internal cohort of

PCa specimens, we did not discover any appreciable differences in expression profiles that correlated well with DNA methylation results (Figure 4.8).

163

Figure 4.8. Expression of HOXD8 variants in frozen tissue. (A) qRT-PCR of Ensembl variant 002, (B) combined qRT-PCR of new long variants, and (C) qRT-PCR of new short variant revealed similar patterns of expression compared to total HOXD8 mRNA expression, again showing no correlation with methylation pattern for these cases.

164

4.4.5 Cell Viability

As all HOXD8 isoforms were overexpressed in poorly differentiated PCa compared to moderately differentiated PCa, we next assessed the consequence of HOXD8 knockdown on cell viability in PC3 cells. PC3 cells were chosen as they have robust levels of HOXD8 mRNA for endogenous knockdown and are methylated at the HOXD8 promoter region, which is representative of the paradoxical methylation and expression pattern observed in the frozen tumour cohort.

We achieved 72% and 64% knockdown of total HOXD8 mRNA using two separate siRNAs

(Figure 4.9A). Reduction of HOXD8 mRNA had no affect on the number of viable cells, indicating that HOXD8 does not play a role in proliferation/apoptosis in this in vitro model

(Figure 4.9B).

4.4.6 Cell Motility

Cell motility of siRNA treated PC3 cells was assessed using a wound scratch assay. Again, there were no significant differences in HOXD8 siRNA treated PC-3 cells versus control siRNA treated cells 24 or 48 hours post-wound induction (Figure 4.9C, D), indicating HOXD8 does not play a role in the motile behaviour of PC-3 cells.

165

Figure 4.9. HOXD8 does not affect cell viability and motility. (A) Knockdown of HOXD8 mRNA in PC-3 cells using two separate siRNAs. (B) Total number of viable PC-3 following HOXD8 siRNA treatment. (C) 0 and 48 hour wound scratch pictures of PC-3 cells treated with siRNA. (D) Quantification of wound scratch results in triplicate experiments.

166

4.5 Discussion

The above results describe several aspects of the relationship between DNA methylation and

HOXD gene expression as well as the potential implications for aberrant HOXD3 and HOXD8 gene expression patterns.

First, these results show that HOXD3 and HOXD8 gene expression is in part controlled by DNA methylation, albeit this is not the only factor involved. Demethylating and mRNA expression analysis of LNCaP cells revealed the expected pattern of DNA methylation resulting in gene silencing. In these cells, the HOXD3 promoter region was unmethylated and demethylating treatment did not correspond to increases in gene transcription. Conversely, there were detectable levels of HOXD8 promoter methylation and subsequent increases of gene transcription following demethylating treatment. For DU145 cells, the expected pattern of methylation and silencing was also observed with respect to HOXD3. Unlike LNCaP cells, the canonical HOXD3 promoter region was methylated and demethylating treatment resulted in a robust increase in gene expression. This may be attributable to methylation of an upstream CpG island, such as the one found in the second exon of HOXD4 which is proximal to the HOXD3 transcript described by 5’

RACE assays. We also observed a seemingly paradoxical pattern of DNA methylation and detectable gene expression for HOXD8 in DU145 and PC-3 cells. Nonetheless, whenever DNA methylation was present at the analyzed HOXD8 promoter region, there was a concomitant increase in gene transcription following demethylating treatment for all three cell lines analyzed

(which all have detectable levels of DNA methylation).

Demethylating treatment results suggested that DNA methylation mediated downregulation is only one factor involved in regulating HOXD3 and HOXD8 gene expression. Given the

167 prognostic role of DNA methylation in promoter regions of these genes and the demethylation results, we expected that reduced HOXD3 and HOXD8 expression would correlate with adverse clinicopathological variables and poor prognosis. In our internal frozen tumour specimens, however, HOXD3 expression was neither inversely correlated with DNA methylation nor with high grade PCa. For HOXD8, we again observed no correlation between DNA methylation and gene expression, together with a clear increase in HOXD8 mRNA that was significantly associated with high grade PCa. The poor correlation between DNA methylation and gene expression is increasingly being found through genome wide studies. These findings may be attributable, at least in part, to genes which are normally not expressed and undergo further or alternative epigenetic repression. For example, switching of epigenetic silencing marks

H3K27me3 and DNA methylation has been described in PCa cell lines, in particular within the

HOX loci (495). TSSs which are bound by RNA pol II are also protected from DNA methylation, regardless if there is active transcription occurring or not (496). Alternatively, methylation in elements distal to the TSS may not efficiently silence gene transcription, as it has been shown that methylation directly surrounding the TSS results in genes that are unable to initiate transcription (497). Using publicly available data from Taylor et al. (494) we confirmed that expression of both HOXD3 and HOXD8 does not decrease in PCa compared to normal or in higher grade PCa specimens. In fact, there is a slight increase of HOXD3 and HOXD8 expression in higher GS, and expression of both genes is significantly correlated with biochemical recurrence.

5’ RACE results revealed that the lack of correlation between methylation and expression of these two genes is most likely a result of non-canonical TSSs that exist distal to the DNA methylation regions examined. For HOXD3, we described a TSS that was hitherto unidentified.

168

This transcript begins in the second exon of HOXD4 and is slightly different than one similar previously identified transcript that encompasses most of the HOXD4 transcript and continues into HOXD3 (Ensembl variant HOXD3-005). Interestingly, a CpG island exists proximal to this

TSS with varying levels of methylation from our CpG island microarray data (Figure 4.10), suggesting that methylation of this region may better correlate with HOXD3 transcription.

For HOXD8, we observed a novel transcript in PC-3 cells that, when compared to canonical full length HOXD8, begins transcription >500 bp into the first exon. Again, this places transcription of this isoform distal to the analyzed methylated region and suggests that DNA methylation does not play a critical role in gene silencing. It should be noted, however, that the entire CpG island encompassing the region analyzed for methylation is >2000 bp and extends to the region adjacent to the novel HOXD8 transcript, leading to the potential for more proximal methylation to affect gene transcription. Based on our microarray data, this portion of the CpG island is largely devoid of methylation in primary tumours (Figure 4.10), suggesting that expression of the novel transcript is controlled by other mechanisms in PCa. This may include histone modifications or miRNAs, the latter of which is exemplified by miR-196 directed repression of

HOXD8 (238).

169

Figure 4.10. Methylation profile of HOXD8 and HOXD4/HOXD3 genomic region. Transcription start sites are marked for variant mRNA isoforms discovered through 5’ RACE/RT-PCR. Microarray specimen identifiers are listed above, while gene probes (in ascending genomic coordinate order) are listed on the side.

170

We observed expression of identical HOXD8 protein isoforms regardless of the mRNA isoform that was transiently transfected into two different PCa cell lines. Thus, it seems likely that factors exist to inhibit the expression of full length canonical HOXD8. As already mentioned, most of the first exon of full length HOXD8 is encompassed by a CpG island, and methylation of specific regions could affect HOXD8 splicing similar to other genes (295). Alternatively, binding of cell type specific trans acting factors may create open chromatin that facilitates transcription of the short HOXD8 isoform we discovered. Finally, specific trans acting factors may also bind the mRNA to regulate alternative splicing (498). Although each discovered HOXD8 mRNA isoform encodes for the same protein, the relative expression of each isoform may have important consequences. 5’ UTRs affect translational ability of mRNA as a consequence of GC content, complex RNA structure, and the presence of upstream initiation codons. For example, there are two distinct isoforms of BRCA1 mRNA with alternative 5’ UTRs, UTRa and UTRb.

UTRb is translated at an efficiency 10 times lower than UTRa as the result of stable secondary structure and upstream AUG codons (499). The western blot of alternative HOXD8 transcripts suggests differing translational abilities of each isoform, but future work is necessary to confirm this and elucidate the mechanism. In addition, we discovered two distinct 3’ UTRs of HOXD8 mRNA. Again, this may have important implications for translational ability as shorter 3’ ends may evade regulation from miRNAs or affect binding of RNA binding proteins (500).

We next assessed whether DNA methylation correlates with downregulation of specific HOXD8 transcripts. There were no discernible differences in expression patterns of alternative HOXD8 isoforms that better associated with DNA methylation pattern, suggesting that methylation of the analyzed HOXD8 region exists in an enhancer that regulates all HOXD8 isoforms (or perhaps another gene) or is a “passenger” methylation event during PCa progression. Future work

171 characterizing the HOXD8 promoter region through luciferase and chromatin conformation assays may help to clarify this.

We next assessed whether variant HOXD8 plays a role in cell viability or cell motility similar to the proliferative and invasive phenotypes observed with HOXC8 overexpression. There was no effect on either viability or motility following endogenous HOXD8 knockdown. There are several possible reasons for this. First, we used a bone metastatic in vitro model (PC-3 cells), which may not be expressing other HOXD8 co-factors, such as TALE proteins, required to elicit a response. The highly proliferative and motile behaviour of these cells may also mask subtle effects of HOXD8 with respect to these functional roles. Future work could include stable transfection of HOXD8 in low expressing cells such as androgen responsive and primary PCa derived 22Rv1 cells, androgen responsive lymph node metastatic LNCaP cells, or androgen independent brain metastatic DU145 cells. Such work would provide a more in-depth in vitro analysis and remove doubts to possible cell line specific events. Secondly, HOXD8 may not have a part in proliferation, apoptosis, or the motile behaviour of PCa cells, but instead may play a more general role in another oncogenic process, for example promoting a dedifferentiated cell type. Third, given the posterior prevalence displayed by HOX genes, it is possible that loss of

HOXD9-HOXD13 is required to observe any effect of HOXD8. Finally, overexpression of

HOXD8 may be a passenger event during PCa progression which does not have any phenotypic consequence.

We chose to examine the role of HOXD8 as it had a clear association with poor prognosis both in our internal frozen tumour cohort and in the MSKCC cohort. Furthermore, the posterior prevalence phenomenon of HOX genes would suggest that expression of HOXD8 would supercede overexpression of HOXD3. It should be noted, however, that HOXD3 may

172 nonetheless have a functional consequence, particularly in the case of null or low HOXD8 expression. Future work should examine this possibility.

Overexpression of anterior HOX genes in PCa is not surprising, given the fact that the prostate gland is a posterior organ that requires posterior HOX expression for development and differentiation (377, 379). Thus, the consequence of anterior HOXD gene overexpression, perhaps in conjunction with loss of posterior HOXD gene expression, may be the activation of pathways or gene networks necessary for anterior organ differentiation which contributes to PCa progression.

173

Chapter 5

Conclusions and future directions

DNA methylation is associated with changes in gene expression and is often deregulated in PCa.

The aims of this thesis were to further our knowledge of these changes in progression of PCa through genome-wide profiling, to validate a number of novel methylation changes as clinically useful markers, and to describe the role that DNA methylation changes of HOXD3 and HOXD8 play in expression of these two markers.

5.1 Novel DNA methylation events associated with PCa progression

Irregularities in the pattern of genome-wide methylation are common-place in the epigenome of

PCa. The work presented in this thesis expands upon prior knowledge regarding DNA methylation events, in particular identifying homeobox gene clusters as common methylation targets in the progression of PCa as well as identifying common methylation events that exist outside of gene promoter regions. Aberrant intergenic and intragenic methylation may regulate gene enhancer regions, internal promoters, or alter mRNA splicing patterns (294, 295, 297, 298).

Alternatively, many methylated loci may be considered “passenger” effects of carcinogenesis, which are not necessary for the transformative process and considered an effect rather than a cause (501). For instance, epigenetic switching from polycomb repression to DNA methylation of already silenced gene loci (495) may not drive tumourigenesis, but instead may be a result of epigenetic modifying enzymes being overexpressed (EZH2, for example). Further work is necessary to address this issue. For example, as EZH2 has been shown to have a direct effect on

174

DNA methylation, RNAi targeting EZH2 or ectopic gene expression of EZH2 can be performed followed by DNA methylation profiling in PCa cells to assess genome-wide DNA methylation consequences of EZH2 expression in this context. A similar approach with gene specific methylation assessment of MYT1 and WNT2 has been performed in U2OS cells (426).

This work also suggests that TMPRSS2:ERG fusions may play an important role in defining methylation patterns throughout the genome. Previous studies have suggested this with respect to

LINE-1 methylation and PITX2 methylation (330, 502) but have not shown it with respect to specific gene sets. This observation has important implications. First, it suggests a possible causative role behind TMPRSS2:ERG fusions and DNA methylation. As TMPRSS2:ERG is considered an early event in carcinogenesis due to its presence in PIN lesions (147, 503) and many of the methylated loci we discovered occur later in PCa progression, it suggests either a direct or indirect role for TMPRSS2:ERG in causing methylation of specific loci. As

TMPRSS2:ERG has been shown to induce EZH2 expression and is also correlated with HDAC1 expression (273, 445), it is tempting to speculate that DNA methylation of specific gene loci occurs as a secondary result of expression of these genes. Secondly, the diagnostic or prognostic capabilities of specific methylation markers may be largely dependent on TMPRSS2:ERG expression. This may be analogous to the CIMP phenotype in colorectal cancer and its strong correlation with BRAF mutations (504). Future work should involve obtaining direct proof of

TMPRSS2:ERG fusions in creating an altered methylation landscape. An initial step may be to introduce ectopic ERG in vitro followed by methylation profiling. This can also be performed in vivo using mice strains that have already been created expressing TMPRSS2:ERG fusions in the prostate (149).

175

Although our work has extended the list of methylated loci in PCa, the technology behind DNA methylome profiling has advanced to a point that allows nearly complete coverage of every CpG dinucleotide throughout the genome (505). Whole-genome bisulfite sequencing is becoming readily available as the cost of NGS continues to drop. Thus, a truly genome-wide methylation analysis, including nearly every CpG in the genome, is likely to occur for PCa specimens. This analysis will serve multiple roles, including identification of potentially diagnostic and prognostic methylation occurring outside of CpG islands or promoter regions, including shores, shelves, intergenic and intragenic locations. It will also serve to validate previously discovered diagnostic and prognostic methylation events. In fact, MethylPlex NGS by Kim et al. has shown methylation of HOXD3 in 40% of localized PCa and 75% of metastatic PCa samples, verifying results presented in this thesis (supplemental Table 10 of (330)).

In addition to determining the cause of aberrant methylation patterns and future profiling experiments, it would be worthwhile to assess the consequence of abnormal methylation patterns with respect to gene expression. One possibility would involve genome-wide methylation profiling in combination with RNA-seq experiments in tissue specimens. This would garner insight into methylation events that have putative effects on overall gene transcription, splicing pattern, and alternative promoter usage. Genes with unchanged expression but with abnormal

DNA methylation would also be identified. In vitro DNA methylation and RNA-seq analysis of cell lines treated with demethylating agents would further verify the role of DNA methylation in abnormal gene expression, splicing, or alternative promoter usage. The consequence of methylation in intergenic enhancer regions may be difficult to ascertain using these techniques, but could be further investigated using combinations of chromatin immunoprecipitations for enhancer marks (H3K4me1/2, H3K27ac) as well as chromatin conformation experiments. These

176 experiments would also assist in designing gene specific functional experiments. For example, understanding that methylation is affecting splicing and not overall gene expression could lead to experiments that identify the consequence of expression of one isoform versus another for any given gene, instead of attempting to functionally validate one specific gene product that may be taken out of context with DNA methylation. This approach, in a gene specific fashion instead of whole genome profiling, may be useful for the other novel methylation candidates TGFβ2 and

GENE X.

5.2 Validation of candidate genes

Selection of HOXD3 and TGFβ2 for methylation validation in Chapter 3 was based on statistical analysis of GS6 versus GS8 microarray cases, while HOXD8 and GENE X were selected based on statistical analysis of recurrent versus non-recurrent cases. As GS and biochemical recurrence are strongly correlated (506, 507), it is not surprising that genes selected based on GS would have a predictive value in biochemical recurrence analysis and vice versa. Interestingly, though,

TGFβ2 was not associated with GS as we expected from the microarray analysis, but instead was an independent predictor of biochemical recurrence. Thus, of all four markers analyzed, TGFβ2 appears to be the only one with potential post-operative applications, notwithstanding the possibility that methylation of the other three genes may have post-surgical adjuvant treatment response prediction capabilities.

Conversely, DNA methylation of HOXD3 and GENE X has the potential to be used in a preoperative setting based on their strong correlations with GS and pathological stage. As the clinical staging often underestimates the true extent of disease (508), one possibility consists of

177 analyzing DNA methylation in biopsy specimens to obtain a better prediction of pathological stage. Yet another possibility is non-invasive detection of methylation in serum, semen, or urine samples. If detection of DNA methylation in these samples proves to have sufficient sensitivity, it may be possible to circumvent unnecessary biopsies for patients who do not harbour PCa or do not harbour aggressive PCa. Preliminary work in biopsy specimens and urine specimens suggests that detection of methylation in these settings is feasible (Appendix Tables A.1 and A.2).

In addition to the four markers characterized in this thesis, there were dozens of other methylated loci that require validation. In fact, the methylated regions most strongly associated with GS and with ERG expression were not validated (Table 2.6 and 2.7). It is unrealistic, however, to characterize each of these genes in a large cohort given time, cost, and sample constraints. Future work may consist of multiplexing DNA methylation analysis of these genes, for example, using a microarray based platform. This could serve several important purposes.

First, it would reduce the amount of DNA necessary for analysis of multiple markers. Secondly, it would likely increase the sensitivity of detection as has previously been shown for panels of methylation markers (476). Thirdly, incorporation of different markers into one platform may allow simultaneous diagnosis and prognosis. Finally, it would reduce workload and subsequently reduce cost. Thus, multiplexing of markers has a much greater potential for translation into a clinical setting than individual gene analyses.

5.3 Role of the HOXD cluster in PCa progression

In addition to methylation of HOXD3 and HOXD8, we observed significant methylation signal throughout the HOXD cluster (Appendix Figure A.1), suggesting a broader role for DNA

178 methylation in this cluster beyond the two characterized genes. Methylation upstream of

HOXD13 and extending into EVX2, for example, was prevalent and also associated with GS based on the regional analysis performed in Chapter 2. There are a number of potential mechanisms behind widespread HOXD methylation. First, as previously discussed, EZH2 is overexpressed in aggressive PCa, is part of the PRC2 complex that targets HOX genes, and directly targets DNMTs for de novo methylation (426). Second, methylation that normally occurs in benign prostate in the HOXD cluster may undergo spreading throughout the cluster as has previously been shown in cancer (454, 509), perhaps as the result of incorrect methylation copying during DNA replication or overexpressed de novo DNMTs (510, 511). Third, and similar to the second point, a stochastic methylation event may occur within the HOXD cluster during tumourigenesis and spread as the result of mechanisms suggested above. Fourth, loss of

RNA polymerase II occupancy may predispose the HOXD cluster to aberrant methylation (496).

Finally, hitherto unidentified mechanisms behind DNA methylation targeting, such as expression of transcription factors or ncRNAs, may be responsible for the observed effect. HOXB3 and C-

MYC are two such transcription factors known to target DNA methylation (446, 512), while mir-

10a is known to transcriptionally silence HOXD4 in breast cancer via DNA methylation (417).

While DNA methylation of HOXD3 and HOXD8 influences gene expression, it is clear that it is neither the only mechanism nor is it likely the most important mechanism that controls gene expression (at least within the context of the DNA regions that we analyzed for methylation). As

HOXD genes are well characterized targets of both polycomb repressive and trithorax activating complexes, it is possible that H3K27me3 and H3K4me3 play a larger role in managing gene expression. Several studies have shown the presence of these marks within the HOXD locus in

PCa (261, 495). We have also analyzed publicly available ChIP-chip data for H3K4me3 and

179

H3K27me3 in normal PrEC cells and cancerous LNCaP cells (513) using CisGenome software for peak detection. We detected peaks for H3K27me3 at the HOXD3 region (Appendix Figure

A.2), perhaps indicating that an epigenetic switch from H3K27me3 to DNA methylation takes place in PCa as has been suggested (495). For HOXD8, the region analyzed for methylation appears to be devoid of H3K4me3 in PrEC cells, while LNCaP cells have slightly poised histone modification status (Figure 5.2). In addition, an analysis of ENCODE data using the UCSC genome browser reveals that the HOXD8 methylated region is marked by H3K4me1, H4K4me2, and H3K27ac in human umbilical vein endothelial cells (HUVECs) which are transcriptionally active at the HOXD8 locus (Appendix Figure A.3). The former two marks are often found at all gene enhancer regions while the latter is found at active gene enhancer regions specifically (514,

515). Thus, it is plausible that this region serves as an enhancer in prostate cells as well as

HUVECs, with methylation adversely affecting transcription factor binding and subsequent promoter interaction. Future work should involve chromatin immunoprecipitation experiments in vitro and in frozen tissue specimens in conjunction with DNA methylation assessment at the same genomic location to determine the overlap between epigenetic modifications. This may be done through high-density tiling platforms or targeted sequencing approaches to obtain an epigenetic map of the entire HOX locus in PCa.

It is also likely that upstream signaling including the androgen receptor pathway influences

HOXD expression as this has previously been shown for HOX genes in the rat prostate (516). As an initial experiment, we treated androgen responsive LNCaP cells with DHT for 24 hours and assessed transcription of HOXD genes (Appendix Figure A.4). Interestingly, expression of

HOXD13, HOXD10, and HOXD8 decreased approximately 3-4 fold. Thus, it is plausible that the entire HOXD cluster is downregulated in response to androgen signaling, and this may

180 contribute to the numerous phenotypes observed upon androgen stimulation such as increased cell proliferation. Cyclic AMP has also been shown to regulate HOXD genes as well CREB1/2 and NeuroD1, which all reside within the same genetic locus that has been implicated in epithelial-neuronal cell conversion (2q31-33) (517).

Despite increased DNA methylation of HOXD3 and HOXD8 in PCa, and in particular aggressive PCa, we observed slight increases of expression for these genes in higher GS. In addition, increased expression of both genes was associated with disease recurrence. Currently, there are no studies that have analysed the functional role of either gene in PCa. Previous studies have shown that HOXD3 plays an oncogenic role in both lung cancer and melanoma by increasing motility and invasion (432-434). Interestingly, increased DNA methylation of

HOXD3 has also been shown in lung cancer as well (447). Little is known regarding the function of HOXD8 beyond a role in lymphangiogenesis (437). Future work elucidating the contribution of both genes to prostate cancer progression is warranted. Initial studies may begin characterizing the coding potential of the discovered HOXD3 mRNA isoform. Also, assessment of functions that have been shown in other cell types and cancers as listed above is justified. In addition, three-dimensional cell culture assays may help in determining the role of these two genes in cell differentiation by assessing proper spheroid/lumen formation.

Altered expression of other HOXD genes may also have a functional contribution in the development and progression of PCa. HOXD13, HOXD11, and HOXD10 all have significantly reduced mRNA expression that correlates with increased GS in the MSKCC cohort (Appendix

Figure A.5; Pearson p-values < 0.001). HOXD13 is expressed in the developing and adult mouse prostate (518) which suggests a role in maintaining differentiated tissue phenotype, while

HOXD10 is a well characterized tumour suppressor in breast cancer (519). Thus, these posterior

181

HOXD genes may play a tumour suppressive role in the prostate and maintain a differentiated phenotype. Future work is necessary to address these issues.

The work presented in this thesis creates a better understanding of genome-wide DNA methylation aberrations that are associated with (a) progression of PCa and (b) the presence of a gene fusion that is implicated in causation of approximately 50% of PCa. Validation of a panel of such identified specific methylation aberrations in a large cohort suggests that these events have clinical potential, particularly in the identification of aggressive disease in a pre-treatment setting as well as a post-surgical setting. Lastly, it is shown that the consequence of methylation events in the HOXD locus is not necessarily direct transcriptional silencing but instead may be part of a broader DNA methylation phenomenon that nonetheless has implications in diagnosis, prognosis, and treatment of PCa.

182

Appendices

Case HOXD3 HOXD8 TGFB2 GENE X 1 2.6 0 0 3.41 2 12.93 0 0 1.78 3 11.34 0 0.18 1.4 4 27.91 0.56 0.35 1.27 5 7.89 16.14 0.86 2.34 6 3.29 0.44 0 2.11 7 1.68 0 0.02 1.86 8 5.87 8.93 0.2 2.77 9 7.81 19.13 19.12 38.43 10 8.1 0.17 0.52 4.95 11 6.43 4.98 0 4.78 12 2.3 0.11 0.3 1.76 13 4.25 1.64 1.09 3.08 14 16.34 0 0 3.35 15 9.33 0.23 0.45 1.76 16 23.48 0.67 0 1.75 17 3.38 0 0 1.29 18 58.14 4.81 4.1 2.89 19 13.36 1.82 0.22 3.65 20 2.57 0.4 0 2.4

Table A.1. MethyLight results from watchful waiting cohort. PMR values from 19 urine specimens with values over 10 highlighted in red.

183

HOXD3 TGFB2 RP Biopsy RP Biopsy 1 51.025 63.81 0 0 2 50.125 101.08 0 0 3 39.95 88.74 0.85 0 4 39.55833333 54.33 3.55 46.82 5 81.6 68.54 2.2 48.755 6 42.95 42.19 0 0 7 51.525 125.48 2.07 0 8 6.833333333 34.72 0.33 0 9 7.075 11.56 18.55 29.03 10 8.85 11.94 0.05 0 11 8.85 62.81 3.05 0 12 10.2 14.75 0 0 13 12.45 24.26 0.45 0 14 14.275 37.075 0 0

Table A.2. HOXD3 and TGFβ2 MethyLight results from biopsy specimens. PMR values for the corresponding radical prostatectomy (RP) are also shown.

184

Figure A.1. HOXD cluster heatmap and clustering dendogram of cases. Significant methylation exists in the HOXD3 and HOXD8 regions, but also in EVX2/HOXD13 (top of heat map) and HOXD9. Gene probes (in ascending genomic coordinate order) are listed on the side.

185

Figure A.2. Polycomb/trithorax modifications in HOXD3 and HOXD8 regions in PCa cells. (A, upper panel) H3K4me3 peaks in PrEC and LNCaP cells surrounding HOXD3 as determined by CisGenome analysis. (Lower panel) H3K27me3 peaks in PrEC and LNCaP cells. (B) As above in the HOXD8 surround region. The thick black lines in the upper panels of A and B represents the region analyzed by MethyLight in PCa. Data was obtained from GEO dataset GSE19600 (Coolen et al., Nature Cell Biology, 2010; 12(3): 235-24) and figure created by Ken Kron.

186

Figure A.3. Activating histone marks in HOXD8 region. UCSC genome browser view of ENCODE cell line ChIP-seq data (7 cell lines in total) for H3K4me3 (promoter associated), H3K4me1 (enhancer associated), and H3K27ac (active enhancer associated). RNA-seq data is also presented at the bottom (Transcription). Light blue peaks are sequencing reads from human umbilical vein endothelial cells (HUVECs). Orange and red bars are ENCODE annotated regions of strong enhancer or active promoter, respectively, in HUVECs. The thick black bar indicates the region analyzed by MethyLight in PCa samples.

187

Figure A.4. Androgen affects HOXD gene expression. Treatment of LNCaP cells with DHT for 24 hours reduced mRNA expression of posterior HOXD genes HOXD13 and HOXD10 and also reduced HOXD8 expression. Not depicted is KLK3/PSA, which was ran as a positive control and increased 5.9 fold upon DHT treatment (p-value = 0.006).

188

Figure A.5. MSKCC cohort HOXD gene expression. Log2 expression values for each gene in the HOXD cluster stratified by benign, GS, and metastatic tissue types

189

References

1. McNeal JE. Prostate cancer volume. Am J Surg Pathol. 1997;21(11):1392-3. 2. Cunha GR, Ricke W, Thomson A, Marker PC, Risbridger G, Hayward SW, et al. Hormonal, cellular, and molecular regulation of normal and neoplastic prostatic development. J Steroid Biochem Mol Biol. 2004;92(4):221-36. 3. Sugimura Y, Cunha GR, Donjacour AA. Morphogenesis of ductal networks in the mouse prostate. Biol Reprod. 1986;34(5):961-71. 4. Wang Y, Hayward S, Cao M, Thayer K, Cunha G. Cell differentiation lineage in the prostate. Differentiation. 2001;68(4-5):270-9. 5. Aumuller G, Seitz J, Lilja H, Abrahamsson PA, von der Kammer H, Scheit KH. Species- and organ-specificity of secretory proteins derived from human prostate and seminal vesicles. Prostate. 1990;17(1):31-40. 6. Takeda H, Lasnitzki I, Mizuno T. Analysis of prostatic bud induction by brief androgen treatment in the fetal rat urogenital sinus. J Endocrinol. 1986;110(3):467-70. 7. Podlasek CA, Barnett DH, Clemens JQ, Bak PM, Bushman W. Prostate development requires Sonic hedgehog expressed by the urogenital sinus epithelium. Dev Biol. 1999;209(1):28-39. 8. Sciavolino PJ, Abrams EW, Yang L, Austenberg LP, Shen MM, Abate-Shen C. Tissue- specific expression of murine Nkx3.1 in the male urogenital system. Dev Dyn. 1997;209(1):127- 38. 9. Krieger JN, Nyberg L, Jr., Nickel JC. NIH consensus definition and classification of prostatitis. Jama. 1999;282(3):236-7. 10. Thorpe A, Neal D. Benign prostatic hyperplasia. Lancet. 2003;361(9366):1359-67. 11. Schmidt LJ, Tindall DJ. Steroid 5 alpha-reductase inhibitors targeting BPH and prostate cancer. J Steroid Biochem Mol Biol.125(1-2):32-8. 12. Helpap B, Kollermann J, Oehler U. Neuroendocrine differentiation in prostatic carcinomas: histogenesis, biology, clinical relevance, and future therapeutical perspectives. Urol Int. 1999;62(3):133-8. 13. Wang W, Epstein JI. Small cell carcinoma of the prostate. A morphologic and immunohistochemical study of 95 cases. Am J Surg Pathol. 2008;32(1):65-71. 14. McNeal JE, Redwine EA, Freiha FS, Stamey TA. Zonal distribution of prostatic adenocarcinoma. Correlation with histologic pattern and direction of spread. Am J Surg Pathol. 1988;12(12):897-906. 15. Troncoso P, Babaian RJ, Ro JY, Grignon DJ, von Eschenbach AC, Ayala AG. Prostatic intraepithelial neoplasia and invasive prostatic adenocarcinoma in cystoprostatectomy specimens. Urology. 1989;34(6 Suppl):52-6. 16. Erbersdobler A, Augustin H, Schlomm T, Henke RP. Prostate cancers in the transition zone: Part 1; pathological aspects. BJU Int. 2004;94(9):1221-5. 17. Schroder FH, Hugosson J, Roobol MJ, Tammela TL, Ciatto S, Nelen V, et al. Screening and prostate-cancer mortality in a randomized European study. N Engl J Med. 2009;360(13):1320-8. 18. Johansson JE, Andren O, Andersson SO, Dickman PW, Holmberg L, Magnuson A, et al. Natural history of early, localized prostate cancer. Jama. 2004;291(22):2713-9. 19. Ward E, Jemal A, Cokkinides V, Singh GK, Cardinez C, Ghafoor A, et al. Cancer disparities by race/ethnicity and socioeconomic status. CA Cancer J Clin. 2004;54(2):78-93.

190

20. Freedman ML, Haiman CA, Patterson N, McDonald GJ, Tandon A, Waliszewska A, et al. Admixture mapping identifies 8q24 as a prostate cancer risk locus in African-American men. Proc Natl Acad Sci U S A. 2006;103(38):14068-73. 21. Kittles RA, Baffoe-Bonnie AB, Moses TY, Robbins CM, Ahaghotu C, Huusko P, et al. A common nonsense mutation in EphB2 is associated with prostate cancer risk in African American men with a positive family history. J Med Genet. 2006;43(6):507-11. 22. Magi-Galluzzi C, Tsusuki T, Elson P, Simmerman K, LaFargue C, Esgueva R, et al. TMPRSS2-ERG gene fusion prevalence and class are significantly different in prostate cancer of Caucasian, African-American and Japanese patients. Prostate.71(5):489-97. 23. Amling CL, Riffenburgh RH, Sun L, Moul JW, Lance RS, Kusuda L, et al. Pathologic variables and recurrence rates as related to obesity and race in men with prostate cancer undergoing radical prostatectomy. J Clin Oncol. 2004;22(3):439-45. 24. Do YK, Carpenter WR, Spain P, Clark JA, Hamilton RJ, Galanko JA, et al. Race, healthcare access and physician trust among prostate cancer patients. Cancer Causes Control.21(1):31-40. 25. Crawford ED, Leewansangtong S, Goktas S, Holthaus K, Baier M. Efficiency of prostate-specific antigen and digital rectal examination in screening, using 4.0 ng/ml and age- specific reference range as a cutoff for abnormal values. Prostate. 1999;38(4):296-302. 26. Andriole GL, Crawford ED, Grubb RL, 3rd, Buys SS, Chia D, Church TR, et al. Mortality results from a randomized prostate-cancer screening trial. N Engl J Med. 2009;360(13):1310-9. 27. Catalona WJ, Smith DS, Ornstein DK. Prostate cancer detection in men with serum PSA concentrations of 2.6 to 4.0 ng/mL and benign prostate examination. Enhancement of specificity with free PSA measurements. Jama. 1997;277(18):1452-5. 28. Catalona WJ, Beiser JA, Smith DS. Serum free prostate specific antigen and prostate specific antigen density measurements for predicting cancer in men with prior negative prostatic biopsies. J Urol. 1997;158(6):2162-7. 29. Catalona WJ, Southwick PC, Slawin KM, Partin AW, Brawer MK, Flanigan RC, et al. Comparison of percent free PSA, PSA density, and age-specific PSA cutoffs for prostate cancer detection and staging. Urology. 2000;56(2):255-60. 30. Carter HB, Pearson JD, Metter EJ, Brant LJ, Chan DW, Andres R, et al. Longitudinal evaluation of prostate-specific antigen levels in men with and without prostate disease. Jama. 1992;267(16):2215-20. 31. D'Amico AV, Chen MH, Roehl KA, Catalona WJ. Preoperative PSA velocity and the risk of death from prostate cancer after radical prostatectomy. N Engl J Med. 2004;351(2):125- 35. 32. Bussemakers MJ, van Bokhoven A, Verhaegh GW, Smit FP, Karthaus HF, Schalken JA, et al. DD3: a new prostate-specific gene, highly overexpressed in prostate cancer. Cancer Res. 1999;59(23):5975-9. 33. de Kok JB, Verhaegh GW, Roelofs RW, Hessels D, Kiemeney LA, Aalders TW, et al. DD3(PCA3), a very sensitive and specific marker to detect prostate tumors. Cancer Res. 2002;62(9):2695-8. 34. Hessels D, Klein Gunnewiek JM, van Oort I, Karthaus HF, van Leenders GJ, van Balken B, et al. DD3(PCA3)-based molecular urine analysis for the diagnosis of prostate cancer. Eur Urol. 2003;44(1):8-15; discussion -6.

191

35. Gleason DF. Classification of prostatic carcinomas. Cancer Chemother Rep. 1966;50(3):125-8. 36. D'Amico AV, Whittington R, Malkowicz SB, Schultz D, Blank K, Broderick GA, et al. Biochemical outcome after radical prostatectomy, external beam radiation therapy, or interstitial radiation therapy for clinically localized prostate cancer. Jama. 1998;280(11):969-74. 37. Walsh PC. Radical prostatectomy for localized prostate cancer provides durable cancer control with excellent quality of life: a structured debate. J Urol. 2000;163(6):1802-7. 38. Schally AV. Luteinizing hormone-releasing hormone analogs: their impact on the control of tumorigenesis. Peptides. 1999;20(10):1247-62. 39. Kreis W, Ahmann FR, Jordan VC, de Haan H, Scott M. Oestrogen pre-treatment abolishes luteinising hormone-releasing hormone testosterone stimulation. Br J Urol. 1988;62(4):352-4. 40. Raudrant D, Rabe T. Progestogens with antiandrogenic properties. Drugs. 2003;63(5):463-92. 41. Osoba D, Tannock IF, Ernst DS, Neville AJ. Health-related quality of life in men with metastatic prostate cancer treated with prednisone alone or mitoxantrone and prednisone. J Clin Oncol. 1999;17(6):1654-63. 42. Dimopoulos MA, Panopoulos C, Bamia C, Deliveliotis C, Alivizatos G, Pantazopoulos D, et al. Oral estramustine and oral etoposide for hormone-refractory prostate cancer. Urology. 1997;50(5):754-8. 43. Petrylak DP, Tangen CM, Hussain MH, Lara PN, Jr., Jones JA, Taplin ME, et al. Docetaxel and estramustine compared with mitoxantrone and prednisone for advanced refractory prostate cancer. N Engl J Med. 2004;351(15):1513-20. 44. Tannock IF, de Wit R, Berry WR, Horti J, Pluzanska A, Chi KN, et al. Docetaxel plus prednisone or mitoxantrone plus prednisone for advanced prostate cancer. N Engl J Med. 2004;351(15):1502-12. 45. Kantoff PW, Higano CS, Shore ND, Berger ER, Small EJ, Penson DF, et al. Sipuleucel-T immunotherapy for castration-resistant prostate cancer. N Engl J Med.363(5):411-22. 46. Higano CS. New treatment options for patients with metastatic castration-resistant prostate cancer. Cancer Treat Rev.38(5):340-5. 47. Attard G, Reid AH, Yap TA, Raynaud F, Dowsett M, Settatree S, et al. Phase I clinical trial of a selective inhibitor of CYP17, abiraterone acetate, confirms that castration-resistant prostate cancer commonly remains hormone driven. J Clin Oncol. 2008;26(28):4563-71. 48. Yap TA, Carden CP, Attard G, de Bono JS. Targeting CYP17: established and novel approaches in prostate cancer. Curr Opin Pharmacol. 2008;8(4):449-57. 49. Attard G, Reid AH, Olmos D, de Bono JS. Antitumor activity with CYP17 blockade indicates that castration-resistant prostate cancer frequently remains hormone driven. Cancer Res. 2009;69(12):4937-40. 50. Attard G, Reid AH, A'Hern R, Parker C, Oommen NB, Folkerd E, et al. Selective inhibition of CYP17 with abiraterone acetate is highly active in the treatment of castration- resistant prostate cancer. J Clin Oncol. 2009;27(23):3742-8. 51. de Bono JS, Logothetis CJ, Molina A, Fizazi K, North S, Chu L, et al. Abiraterone and increased survival in metastatic prostate cancer. N Engl J Med.364(21):1995-2005. 52. Pageau SC. Denosumab. MAbs. 2009;1(3):210-5.

192

53. McClung MR, Lewiecki EM, Cohen SB, Bolognese MA, Woodson GC, Moffett AH, et al. Denosumab in postmenopausal women with low bone mineral density. N Engl J Med. 2006;354(8):821-31. 54. Cummings SR, San Martin J, McClung MR, Siris ES, Eastell R, Reid IR, et al. Denosumab for prevention of fractures in postmenopausal women with osteoporosis. N Engl J Med. 2009;361(8):756-65. 55. Deodhar A, Dore RK, Mandel D, Schechtman J, Shergy W, Trapp R, et al. Denosumab- mediated increase in hand bone mineral density associated with decreased progression of bone erosion in rheumatoid arthritis patients. Arthritis Care Res (Hoboken).62(4):569-74. 56. Stopeck AT, Lipton A, Body JJ, Steger GG, Tonkin K, de Boer RH, et al. Denosumab compared with zoledronic acid for the treatment of bone metastases in patients with advanced breast cancer: a randomized, double-blind study. J Clin Oncol.28(35):5132-9. 57. Henry DH, Costa L, Goldwasser F, Hirsh V, Hungria V, Prausova J, et al. Randomized, double-blind study of denosumab versus zoledronic acid in the treatment of bone metastases in patients with advanced cancer (excluding breast and prostate cancer) or multiple myeloma. J Clin Oncol.29(9):1125-32. 58. Fizazi K, Carducci M, Smith M, Damiao R, Brown J, Karsh L, et al. Denosumab versus zoledronic acid for treatment of bone metastases in men with castration-resistant prostate cancer: a randomised, double-blind study. Lancet.377(9768):813-22. 59. Bergstralh DT, Ting JP. Microtubule stabilizing agents: their molecular signaling consequences and the potential for enhancement by drug combination. Cancer Treat Rev. 2006;32(3):166-79. 60. de Bono JS, Oudard S, Ozguroglu M, Hansen S, Machiels JP, Kocak I, et al. Prednisone plus cabazitaxel or mitoxantrone for metastatic castration-resistant prostate cancer progressing after docetaxel treatment: a randomised open-label trial. Lancet.376(9747):1147-54. 61. Tran C, Ouk S, Clegg NJ, Chen Y, Watson PA, Arora V, et al. Development of a second- generation antiandrogen for treatment of advanced prostate cancer. Science. 2009;324(5928):787-90. 62. Scher HI, Beer TM, Higano CS, Anand A, Taplin ME, Efstathiou E, et al. Antitumour activity of MDV3100 in castration-resistant prostate cancer: a phase 1-2 study. Lancet.375(9724):1437-46. 63. O'Mahony D, Morris JC, Quinn C, Gao W, Wilson WH, Gause B, et al. A pilot study of CTLA-4 blockade after cancer vaccine failure in patients with advanced malignancy. Clin Cancer Res. 2007;13(3):958-64. 64. Nilsson S, Franzen L, Parker C, Tyrrell C, Blom R, Tennvall J, et al. Bone-targeted radium-223 in symptomatic, hormone-refractory prostate cancer: a randomised, multicentre, placebo-controlled phase II study. Lancet Oncol. 2007;8(7):587-94. 65. Saad F, Hotte S, North S, Eigl B, Chi K, Czaykowski P, et al. Randomized phase II trial of Custirsen (OGX-011) in combination with docetaxel or mitoxantrone as second-line therapy in patients with metastatic castrate-resistant prostate cancer progressing after first-line docetaxel: CUOG trial P-06c. Clin Cancer Res.17(17):5765-73. 66. Agarwal N, Sonpavde G, Sternberg CN. Novel molecular targets for the therapy of castration-resistant prostate cancer. Eur Urol.61(5):950-60. 67. Yu EY, Wilding G, Posadas E, Gross M, Culine S, Massard C, et al. Phase II study of dasatinib in patients with metastatic castration-resistant prostate cancer. Clin Cancer Res. 2009;15(23):7421-8.

193

68. Thompson IM, Goodman PJ, Tangen CM, Lucia MS, Miller GJ, Ford LG, et al. The influence of finasteride on the development of prostate cancer. N Engl J Med. 2003;349(3):215- 24. 69. Lucia MS, Epstein JI, Goodman PJ, Darke AK, Reuter VE, Civantos F, et al. Finasteride and high-grade prostate cancer in the Prostate Cancer Prevention Trial. J Natl Cancer Inst. 2007;99(18):1375-83. 70. Andriole GL, Bostwick DG, Brawley OW, Gomella LG, Marberger M, Montorsi F, et al. Effect of dutasteride on the risk of prostate cancer. N Engl J Med.362(13):1192-202. 71. Clark LC, Combs GF, Jr., Turnbull BW, Slate EH, Chalker DK, Chow J, et al. Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial. Nutritional Prevention of Cancer Study Group. Jama. 1996;276(24):1957-63. 72. Heinonen OP, Albanes D, Virtamo J, Taylor PR, Huttunen JK, Hartman AM, et al. Prostate cancer and supplementation with alpha-tocopherol and beta-carotene: incidence and mortality in a controlled trial. J Natl Cancer Inst. 1998;90(6):440-6. 73. Lippman SM, Klein EA, Goodman PJ, Lucia MS, Thompson IM, Ford LG, et al. Effect of selenium and vitamin E on risk of prostate cancer and other cancers: the Selenium and Vitamin E Cancer Prevention Trial (SELECT). Jama. 2009;301(1):39-51. 74. Giovannucci E, Ascherio A, Rimm EB, Stampfer MJ, Colditz GA, Willett WC. Intake of carotenoids and retinol in relation to risk of prostate cancer. J Natl Cancer Inst. 1995;87(23):1767-76. 75. Kirsh VA, Mayne ST, Peters U, Chatterjee N, Leitzmann MF, Dixon LB, et al. A prospective study of lycopene and tomato product intake and risk of prostate cancer. Cancer Epidemiol Biomarkers Prev. 2006;15(1):92-8. 76. Lu QY, Hung JC, Heber D, Go VL, Reuter VE, Cordon-Cardo C, et al. Inverse associations between plasma lycopene and other carotenoids and prostate cancer. Cancer Epidemiol Biomarkers Prev. 2001;10(7):749-56. 77. Key TJ, Silcocks PB, Davey GK, Appleby PN, Bishop DT. A case-control study of diet and prostate cancer. Br J Cancer. 1997;76(5):678-87. 78. Cohen JH, Kristal AR, Stanford JL. Fruit and vegetable intakes and prostate cancer risk. J Natl Cancer Inst. 2000;92(1):61-8. 79. Freedman DM, Looker AC, Chang SC, Graubard BI. Prospective study of serum vitamin D and cancer mortality in the United States. J Natl Cancer Inst. 2007;99(21):1594-602. 80. Gissel T, Rejnmark L, Mosekilde L, Vestergaard P. Intake of vitamin D and risk of breast cancer--a meta-analysis. J Steroid Biochem Mol Biol. 2008;111(3-5):195-9. 81. Grant WB, Garland CF. The association of solar ultraviolet B (UVB) with reducing risk of cancer: multifactorial ecologic analysis of geographic variation in age-adjusted cancer mortality rates. Anticancer Res. 2006;26(4A):2687-99. 82. Hsieh TY, Ng CY, Mallouh C, Tazaki H, Wu JM. Regulation of growth, PSA/PAP and androgen receptor expression by 1 alpha,25-dihydroxyvitamin D3 in the androgen-dependent LNCaP cells. Biochem Biophys Res Commun. 1996;223(1):141-6. 83. Wang X, Chen X, Akhter J, Morris DL. The in vitro effect of vitamin D3 analogue EB- 1089 on a human prostate cancer cell line (PC-3). Br J Urol. 1997;80(2):260-2. 84. Getzenberg RH, Light BW, Lapco PE, Konety BR, Nangia AK, Acierno JS, et al. Vitamin D inhibition of prostate adenocarcinoma growth and metastasis in the Dunning rat prostate model system. Urology. 1997;50(6):999-1006.

194

85. Chan JM, Giovannucci E, Andersson SO, Yuen J, Adami HO, Wolk A. Dairy products, calcium, phosphorous, vitamin D, and risk of prostate cancer (Sweden). Cancer Causes Control. 1998;9(6):559-66. 86. Tavani A, Bertuccio P, Bosetti C, Talamini R, Negri E, Franceschi S, et al. Dietary intake of calcium, vitamin D, phosphorus and the risk of prostate cancer. Eur Urol. 2005;48(1):27-33. 87. Park SY, Murphy SP, Wilkens LR, Stram DO, Henderson BE, Kolonel LN. Calcium, vitamin D, and dairy product intake and prostate cancer risk: the Multiethnic Cohort Study. Am J Epidemiol. 2007;166(11):1259-69. 88. Ahn J, Albanes D, Peters U, Schatzkin A, Lim U, Freedman M, et al. Dairy products, calcium intake, and risk of prostate cancer in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer Epidemiol Biomarkers Prev. 2007;16(12):2623-30. 89. Ahonen MH, Tenkanen L, Teppo L, Hakama M, Tuohimaa P. Prostate cancer risk and prediagnostic serum 25-hydroxyvitamin D levels (Finland). Cancer Causes Control. 2000;11(9):847-52. 90. Nomura AM, Stemmermann GN, Lee J, Kolonel LN, Chen TC, Turner A, et al. Serum vitamin D metabolite levels and the subsequent development of prostate cancer (Hawaii, United States). Cancer Causes Control. 1998;9(4):425-32. 91. Harper CE, Cook LM, Patel BB, Wang J, Eltoum IA, Arabshahi A, et al. Genistein and resveratrol, alone and in combination, suppress prostate cancer in SV-40 tag rats. Prostate. 2009;69(15):1668-82. 92. Fleshner NE, Kapusta L, Donnelly B, Tanguay S, Chin J, Hersey K, et al. Progression from high-grade prostatic intraepithelial neoplasia to cancer: a randomized trial of combination vitamin-E, soy, and selenium. J Clin Oncol.29(17):2386-90. 93. Gupta S, Hastak K, Ahmad N, Lewin JS, Mukhtar H. Inhibition of prostate carcinogenesis in TRAMP mice by oral infusion of green tea polyphenols. Proc Natl Acad Sci U S A. 2001;98(18):10350-5. 94. Bettuzzi S, Brausi M, Rizzi F, Castagnetti G, Peracchia G, Corti A. Chemoprevention of human prostate cancer by oral administration of green tea catechins in volunteers with high- grade prostate intraepithelial neoplasia: a preliminary report from a one-year proof-of-principle study. Cancer Res. 2006;66(2):1234-40. 95. Barve A, Khor TO, Hao X, Keum YS, Yang CS, Reddy B, et al. Murine prostate cancer inhibition by dietary phytochemicals--curcumin and phenyethylisothiocyanate. Pharm Res. 2008;25(9):2181-9. 96. Harper CE, Patel BB, Wang J, Arabshahi A, Eltoum IA, Lamartiniere CA. Resveratrol suppresses prostate cancer progression in transgenic mice. Carcinogenesis. 2007;28(9):1946-53. 97. Venkateswaran V, Klotz LH. Diet and prostate cancer: mechanisms of action and implications for chemoprevention. Nat Rev Urol.7(8):442-53. 98. Cardozo CP, Michaud C, Ost MC, Fliss AE, Yang E, Patterson C, et al. C-terminal Hsp- interacting protein slows androgen receptor synthesis and reduces its rate of degradation. Arch Biochem Biophys. 2003;410(1):134-40. 99. Wang Q, Carroll JS, Brown M. Spatial and temporal recruitment of androgen receptor and its coactivators involves chromosomal looping and polymerase tracking. Mol Cell. 2005;19(5):631-42. 100. McKenna NJ, O'Malley BW. Minireview: nuclear receptor coactivators--an update. Endocrinology. 2002;143(7):2461-5.

195

101. Swinnen JV, Verhoeven G. Androgens and the control of lipid metabolism in human prostate cancer cells. J Steroid Biochem Mol Biol. 1998;65(1-6):191-8. 102. Young CY, Montgomery BT, Andrews PE, Qui SD, Bilhartz DL, Tindall DJ. Hormonal regulation of prostate-specific antigen messenger RNA in human prostatic adenocarcinoma cell line LNCaP. Cancer Res. 1991;51(14):3748-52. 103. Moilanen AM, Hakkola J, Vaarala MH, Kauppila S, Hirvikoski P, Vuoristo JT, et al. Characterization of androgen-regulated expression of CYP3A5 in human prostate. Carcinogenesis. 2007;28(5):916-21. 104. Bai G, Kasper S, Matusik RJ, Rennie PS, Moshier JA, Krongrad A. Androgen regulation of the human ornithine decarboxylase promoter in prostate cancer cells. J Androl. 1998;19(2):127-35. 105. Gao S, Lee P, Wang H, Gerald W, Adler M, Zhang L, et al. The androgen receptor directly targets the cellular Fas/FasL-associated death domain protein-like inhibitory protein gene to promote the androgen-independent growth of prostate cancer cells. Mol Endocrinol. 2005;19(7):1792-802. 106. Dehm SM, Tindall DJ. Molecular regulation of androgen action in prostate cancer. J Cell Biochem. 2006;99(2):333-44. 107. Diamandis EP. Prostate-specific Antigen: Its Usefulness in Clinical Medicine. Trends Endocrinol Metab. 1998;9(8):310-6. 108. Xi Z, Klokk TI, Korkmaz K, Kurys P, Elbi C, Risberg B, et al. Kallikrein 4 is a predominantly nuclear protein and is overexpressed in prostate cancer. Cancer Res. 2004;64(7):2365-70. 109. Klokk TI, Kilander A, Xi Z, Waehre H, Risberg B, Danielsen HE, et al. Kallikrein 4 is a proliferative factor that is overexpressed in prostate cancer. Cancer Res. 2007;67(11):5221-30. 110. Obiezu CV, Michael IP, Levesque MA, Diamandis EP. Human kallikrein 4: enzymatic activity, inhibition, and degradation of extracellular matrix proteins. Biol Chem. 2006;387(6):749-59. 111. Yousef GM, White NM, Kurlender L, Michael I, Memari N, Robb JD, et al. The kallikrein gene 5 splice variant 2 is a new biomarker for breast and ovarian cancer. Tumour Biol. 2004;25(5-6):221-7. 112. Yousef GM, Magklara A, Diamandis EP. KLK12 is a novel serine protease and a new member of the human kallikrein gene family-differential expression in breast cancer. Genomics. 2000;69(3):331-41. 113. Yousef GM, Scorilas A, Jung K, Ashworth LK, Diamandis EP. Molecular cloning of the human kallikrein 15 gene (KLK15). Up-regulation in prostate cancer. J Biol Chem. 2001;276(1):53-61. 114. Perner S, Demichelis F, Beroukhim R, Schmidt FH, Mosquera JM, Setlur S, et al. TMPRSS2:ERG fusion-associated deletions provide insight into the heterogeneity of prostate cancer. Cancer Res. 2006;66(17):8337-41. 115. Tu JJ, Rohan S, Kao J, Kitabayashi N, Mathew S, Chen YT. Gene fusions between TMPRSS2 and ETS family genes in prostate cancer: frequency and transcript variant analysis by RT-PCR and FISH on paraffin-embedded tissues. Mod Pathol. 2007;20(9):921-8. 116. Haffner MC, Aryee MJ, Toubaji A, Esopi DM, Albadine R, Gurel B, et al. Androgen- induced TOP2B-mediated double-strand breaks and prostate cancer gene rearrangements. Nat Genet.42(8):668-75.

196

117. Mani RS, Tomlins SA, Callahan K, Ghosh A, Nyati MK, Varambally S, et al. Induced chromosomal proximity and gene fusions in prostate cancer. Science. 2009;326(5957):1230. 118. Bastus NC, Boyd LK, Mao X, Stankiewicz E, Kudahetti SC, Oliver RT, et al. Androgen- induced TMPRSS2:ERG fusion in nonmalignant prostate epithelial cells. Cancer Res.70(23):9544-8. 119. Kaarbo M, Klokk TI, Saatcioglu F. Androgen signaling and its interactions with other signaling pathways in prostate cancer. Bioessays. 2007;29(12):1227-38. 120. Chan JM, Stampfer MJ, Giovannucci E, Gann PH, Ma J, Wilkinson P, et al. Plasma insulin-like growth factor-I and prostate cancer risk: a prospective study. Science. 1998;279(5350):563-6. 121. Wu JD, Haugk K, Woodke L, Nelson P, Coleman I, Plymate SR. Interaction of IGF signaling and the androgen receptor in prostate cancer progression. J Cell Biochem. 2006;99(2):392-401. 122. Toker A. mTOR and Akt signaling in cancer: SGK cycles in. Mol Cell. 2008;31(1):6-8. 123. Brunet A, Bonni A, Zigmond MJ, Lin MZ, Juo P, Hu LS, et al. Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell. 1999;96(6):857- 68. 124. Puc J, Keniry M, Li HS, Pandita TK, Choudhury AD, Memeo L, et al. Lack of PTEN sequesters CHK1 and initiates genetic instability. Cancer Cell. 2005;7(2):193-204. 125. Carnero A, Blanco-Aparicio C, Renner O, Link W, Leal JF. The PTEN/PI3K/AKT signalling pathway in cancer, therapeutic implications. Curr Cancer Drug Targets. 2008;8(3):187-98. 126. Bedolla R, Prihoda TJ, Kreisberg JI, Malik SN, Krishnegowda NK, Troyer DA, et al. Determining risk of biochemical recurrence in prostate cancer by immunohistochemical detection of PTEN expression and Akt activation. Clin Cancer Res. 2007;13(13):3860-7. 127. Rahimi RA, Leof EB. TGF-beta signaling: a tale of two responses. J Cell Biochem. 2007;102(3):593-608. 128. Feng XH, Derynck R. Specificity and versatility in tgf-beta signaling through Smads. Annu Rev Cell Dev Biol. 2005;21:659-93. 129. Guo Y, Kyprianou N. Overexpression of transforming growth factor (TGF) beta1 type II receptor restores TGF-beta1 sensitivity and signaling in human prostate cancer cells. Cell Growth Differ. 1998;9(2):185-93. 130. Massague J, Blain SW, Lo RS. TGFbeta signaling in growth control, cancer, and heritable disorders. Cell. 2000;103(2):295-309. 131. Massague J, Wotton D. Transcriptional control by the TGF-beta/Smad signaling system. Embo J. 2000;19(8):1745-54. 132. Guo Y, Kyprianou N. Restoration of transforming growth factor beta signaling pathway in human prostate cancer cells suppresses tumorigenicity via induction of caspase-1-mediated apoptosis. Cancer Res. 1999;59(6):1366-71. 133. Kyprianou N. Activation of TGF-beta signalling in human prostate cancer cells suppresses tumorigenicity via deregulation of cell cycle progression and induction of caspase-1 mediated apoptosis: significance in prostate tumorigenesis. Prostate Cancer Prostatic Dis. 1999;2(S3):S18. 134. Guo Y, Jacobs SC, Kyprianou N. Down-regulation of protein and mRNA expression for transforming growth factor-beta (TGF-beta1) type I and type II receptors in human prostate cancer. Int J Cancer. 1997;71(4):573-9.

197

135. Bello-DeOcampo D, Tindall DJ. TGF-betal/Smad signaling in prostate cancer. Curr Drug Targets. 2003;4(3):197-207. 136. Espada J, Calvo MB, Diaz-Prado S, Medina V. Wnt signalling and cancer stem cells. Clin Transl Oncol. 2009;11(7):411-27. 137. Allgeier SH, Lin TM, Vezina CM, Moore RW, Fritz WA, Chiu SY, et al. WNT5A selectively inhibits mouse ventral prostate development. Dev Biol. 2008;324(1):10-7. 138. Voeller HJ, Truica CI, Gelmann EP. Beta-catenin mutations in human prostate cancer. Cancer Res. 1998;58(12):2520-3. 139. Gao X, Zacharek A, Grignon D, Liu H, Sakr W, Porter A, et al. High-frequency of loss of expression and allelic deletion of the apc and mcc genes in human prostate-cancer. Int J Oncol. 1995;6(1):111-7. 140. Maruyama R, Toyooka S, Toyooka KO, Virmani AK, Zochbauer-Muller S, Farinas AJ, et al. Aberrant promoter methylation profile of prostate cancers and its relationship to clinicopathological features. Clin Cancer Res. 2002;8(2):514-9. 141. Liu L, Kron KJ, Pethe VV, Demetrashvili N, Nesbitt ME, Trachtenberg J, et al. Association of tissue promoter methylation levels of APC, TGFbeta2, HOXD3 and RASSF1A with prostate cancer progression. Int J Cancer. 2011;129(10):2454-62. 142. Wang Q, Symes AJ, Kane CA, Freeman A, Nariculam J, Munson P, et al. A novel role for Wnt/Ca2+ signaling in actin cytoskeleton remodeling and cell motility in prostate cancer. PLoS One.5(5):e10456. 143. Wissmann C, Wild PJ, Kaiser S, Roepcke S, Stoehr R, Woenckhaus M, et al. WIF1, a component of the Wnt pathway, is down-regulated in prostate, breast, lung, and bladder cancer. J Pathol. 2003;201(2):204-12. 144. Lodygin D, Epanchintsev A, Menssen A, Diebold J, Hermeking H. Functional epigenomics identifies genes frequently silenced in prostate cancer. Cancer Res. 2005;65(10):4218-27. 145. Tomlins SA, Rhodes DR, Perner S, Dhanasekaran SM, Mehra R, Sun XW, et al. Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science. 2005;310(5748):644-8. 146. Tomlins SA, Mehra R, Rhodes DR, Smith LR, Roulston D, Helgeson BE, et al. TMPRSS2:ETV4 gene fusions define a third molecular subtype of prostate cancer. Cancer Res. 2006;66(7):3396-400. 147. Furusato B, Gao CL, Ravindranath L, Chen Y, Cullen J, McLeod DG, et al. Mapping of TMPRSS2-ERG fusions in the context of multi-focal prostate cancer. Mod Pathol. 2008;21(2):67-75. 148. Carver BS, Tran J, Gopalan A, Chen Z, Shaikh S, Carracedo A, et al. Aberrant ERG expression cooperates with loss of PTEN to promote cancer progression in the prostate. Nat Genet. 2009;41(5):619-24. 149. King JC, Xu J, Wongvipat J, Hieronymus H, Carver BS, Leung DH, et al. Cooperativity of TMPRSS2-ERG with PI3-kinase pathway activation in prostate oncogenesis. Nat Genet. 2009;41(5):524-6. 150. Nam RK, Sugar L, Yang W, Srivastava S, Klotz LH, Yang LY, et al. Expression of the TMPRSS2:ERG fusion gene predicts cancer recurrence after surgery for localised prostate cancer. Br J Cancer. 2007;97(12):1690-5.

198

151. Saramaki OR, Harjula AE, Martikainen PM, Vessella RL, Tammela TL, Visakorpi T. TMPRSS2:ERG fusion identifies a subgroup of prostate cancers with a favorable prognosis. Clin Cancer Res. 2008;14(11):3395-400. 152. Steck PA, Pershouse MA, Jasser SA, Yung WK, Lin H, Ligon AH, et al. Identification of a candidate tumour suppressor gene, MMAC1, at chromosome 10q23.3 that is mutated in multiple advanced cancers. Nat Genet. 1997;15(4):356-62. 153. Li J, Yen C, Liaw D, Podsypanina K, Bose S, Wang SI, et al. PTEN, a putative protein tyrosine phosphatase gene mutated in human brain, breast, and prostate cancer. Science. 1997;275(5308):1943-7. 154. Li DM, Sun H. TEP1, encoded by a candidate tumor suppressor locus, is a novel protein tyrosine phosphatase regulated by transforming growth factor beta. Cancer Res. 1997;57(11):2124-9. 155. Cairns P, Okami K, Halachmi S, Halachmi N, Esteller M, Herman JG, et al. Frequent inactivation of PTEN/MMAC1 in primary prostate cancer. Cancer Res. 1997;57(22):4997-5000. 156. Wang SI, Parsons R, Ittmann M. Homozygous deletion of the PTEN tumor suppressor gene in a subset of prostate adenocarcinomas. Clin Cancer Res. 1998;4(3):811-5. 157. McMenamin ME, Soung P, Perera S, Kaplan I, Loda M, Sellers WR. Loss of PTEN expression in paraffin-embedded primary prostate cancer correlates with high Gleason score and advanced stage. Cancer Res. 1999;59(17):4291-6. 158. Verhagen PC, van Duijn PW, Hermans KG, Looijenga LH, van Gurp RJ, Stoop H, et al. The PTEN gene in locally progressive prostate cancer is preferentially inactivated by bi-allelic gene deletion. J Pathol. 2006;208(5):699-707. 159. Yoshimoto M, Cutz JC, Nuin PA, Joshua AM, Bayani J, Evans AJ, et al. Interphase FISH analysis of PTEN in histologic sections shows genomic deletions in 68% of primary prostate cancer and 23% of high-grade prostatic intra-epithelial neoplasias. Cancer Genet Cytogenet. 2006;169(2):128-37. 160. Ma X, Ziel-van der Made AC, Autar B, van der Korput HA, Vermeij M, van Duijn P, et al. Targeted biallelic inactivation of Pten in the mouse prostate leads to prostate cancer accompanied by increased epithelial cell proliferation but not by reduced apoptosis. Cancer Res. 2005;65(13):5730-9. 161. Frau M, Biasi F, Feo F, Pascale RM. Prognostic markers and putative therapeutic targets for hepatocellular carcinoma. Mol Aspects Med.31(2):179-93. 162. Chen Y, Olopade OI. MYC in breast tumor progression. Expert Rev Anticancer Ther. 2008;8(10):1689-98. 163. Bova GS, Carter BS, Bussemakers MJ, Emi M, Fujiwara Y, Kyprianou N, et al. Homozygous deletion and frequent allelic loss of chromosome 8p22 loci in human prostate cancer. Cancer Res. 1993;53(17):3869-73. 164. Van Den Berg C, Guan XY, Von Hoff D, Jenkins R, Bittner, Griffin C, et al. DNA sequence amplification in human prostate cancer identified by chromosome microdissection: potential prognostic implications. Clin Cancer Res. 1995;1(1):11-8. 165. Jenkins RB, Qian J, Lieber MM, Bostwick DG. Detection of c-myc oncogene amplification and chromosomal anomalies in metastatic prostatic carcinoma by fluorescence in situ hybridization. Cancer Res. 1997;57(3):524-31. 166. Tsuchiya N, Slezak JM, Lieber MM, Bergstralh EJ, Jenkins RB. Clinical significance of alterations of chromosome 8 detected by fluorescence in situ hybridization analysis in pathologic organ-confined prostate cancer. Genes Chromosomes Cancer. 2002;34(4):363-71.

199

167. Das K, Lau W, Sivaswaren C, Ph T, Fook-Chong S, Sl T, et al. Chromosomal changes in prostate cancer: a fluorescence in situ hybridization study. Clin Genet. 2005;68(1):40-7. 168. Kindich R, Florl AR, Kamradt J, Lehmann J, Muller M, Wullich B, et al. Relationship of NKX3.1 and MYC gene copy number ratio and DNA hypomethylation to prostate carcinoma stage. Eur Urol. 2006;49(1):169-75; discussion 75. 169. Fleming WH, Hamel A, MacDonald R, Ramsey E, Pettigrew NM, Johnston B, et al. Expression of the c-myc protooncogene in human prostatic carcinoma and benign prostatic hyperplasia. Cancer Res. 1986;46(3):1535-8. 170. Fox SB, Persad RA, Royds J, Kore RN, Silcocks PB, Collins CC. p53 and c-myc expression in stage A1 prostatic adenocarcinoma: useful prognostic determinants? J Urol. 1993;150(2 Pt 1):490-4. 171. Koh CM, Bieberich CJ, Dang CV, Nelson WG, Yegnasubramanian S, De Marzo AM. MYC and Prostate Cancer. Genes Cancer. 2010;1(6):617-28. 172. Gurel B, Iwata T, Koh CM, Jenkins RB, Lan F, Van Dang C, et al. Nuclear MYC protein overexpression is an early alteration in human prostate carcinogenesis. Mod Pathol. 2008;21(9):1156-67. 173. Kunimi K, Bergerheim US, Larsson IL, Ekman P, Collins VP. Allelotyping of human prostatic adenocarcinoma. Genomics. 1991;11(3):530-6. 174. Bergerheim US, Kunimi K, Collins VP, Ekman P. Deletion mapping of chromosomes 8, 10, and 16 in human prostatic carcinoma. Genes Chromosomes Cancer. 1991;3(3):215-20. 175. Vocke CD, Pozzatti RO, Bostwick DG, Florence CD, Jennings SB, Strup SE, et al. Analysis of 99 microdissected prostate carcinomas reveals a high frequency of allelic loss on chromosome 8p12-21. Cancer Res. 1996;56(10):2411-6. 176. Crundwell MC, Chughtai S, Knowles M, Takle L, Luscombe M, Neoptolemos JP, et al. Allelic loss on chromosomes 8p, 22q and 18q (DCC) in human prostate cancer. Int J Cancer. 1996;69(4):295-300. 177. Matsuyama H, Pan Y, Yoshihiro S, Kudren D, Naito K, Bergerheim US, et al. Clinical significance of chromosome 8p, 10q, and 16q deletions in prostate cancer. Prostate. 2003;54(2):103-11. 178. Matsuyama H, Pan Y, Oba K, Yoshihiro S, Matsuda K, Hagarth L, et al. The role of chromosome 8p22 deletion for predicting disease progression and pathological staging in prostate cancer. Aktuelle Urol. 2003;34(4):247-9. 179. El Gammal AT, Bruchmann M, Zustin J, Isbarn H, Hellwinkel OJ, Kollermann J, et al. Chromosome 8p deletions and 8q gains are associated with tumor progression and poor prognosis in prostate cancer. Clin Cancer Res.16(1):56-64. 180. He WW, Sciavolino PJ, Wing J, Augustus M, Hudson P, Meissner PS, et al. A novel human prostate-specific, androgen-regulated homeobox gene (NKX3.1) that maps to 8p21, a region frequently deleted in prostate cancer. Genomics. 1997;43(1):69-77. 181. Ornstein DK, Cinquanta M, Weiler S, Duray PH, Emmert-Buck MR, Vocke CD, et al. Expression studies and mutational analysis of the androgen regulated homeobox gene NKX3.1 in benign and malignant prostate epithelium. J Urol. 2001;165(4):1329-34. 182. Xu LL, Srikantan V, Sesterhenn IA, Augustus M, Dean R, Moul JW, et al. Expression profile of an androgen regulated prostate specific homeobox gene NKX3.1 in primary prostate cancer. J Urol. 2000;163(3):972-9.

200

183. Bowen C, Bubendorf L, Voeller HJ, Slack R, Willi N, Sauter G, et al. Loss of NKX3.1 expression in human prostate cancers correlates with tumor progression. Cancer Res. 2000;60(21):6111-5. 184. Abdulkadir SA, Magee JA, Peters TJ, Kaleem Z, Naughton CK, Humphrey PA, et al. Conditional loss of Nkx3.1 in adult mice induces prostatic intraepithelial neoplasia. Mol Cell Biol. 2002;22(5):1495-503. 185. Kim MJ, Cardiff RD, Desai N, Banach-Petrosky WA, Parsons R, Shen MM, et al. Cooperativity of Nkx3.1 and Pten loss of function in a mouse model of prostate carcinogenesis. Proc Natl Acad Sci U S A. 2002;99(5):2884-9. 186. Kim MJ, Bhatia-Gaur R, Banach-Petrosky WA, Desai N, Wang Y, Hayward SW, et al. Nkx3.1 mutant mice recapitulate early stages of prostate carcinogenesis. Cancer Res. 2002;62(11):2999-3004. 187. Lei Q, Jiao J, Xin L, Chang CJ, Wang S, Gao J, et al. NKX3.1 stabilizes p53, inhibits AKT activation, and blocks prostate cancer initiation caused by PTEN loss. Cancer Cell. 2006;9(5):367-78. 188. Bethel CR, Faith D, Li X, Guan B, Hicks JL, Lan F, et al. Decreased NKX3.1 protein expression in focal prostatic atrophy, prostatic intraepithelial neoplasia, and adenocarcinoma: association with gleason score and chromosome 8p deletion. Cancer Res. 2006;66(22):10683-90. 189. Aslan G, Irer B, Tuna B, Yorukoglu K, Saatcioglu F, Celebi I. Analysis of NKX3.1 expression in prostate cancer tissues and correlation with clinicopathologic features. Pathol Res Pract. 2006;202(2):93-8. 190. Korkmaz CG, Korkmaz KS, Manola J, Xi Z, Risberg B, Danielsen H, et al. Analysis of androgen regulated homeobox gene NKX3.1 during prostate carcinogenesis. J Urol. 2004;172(3):1134-9. 191. Asatiani E, Huang WX, Wang A, Rodriguez Ortner E, Cavalli LR, Haddad BR, et al. Deletion, methylation, and expression of the NKX3.1 suppressor gene in primary human prostate cancer. Cancer Res. 2005;65(4):1164-73. 192. Tomlins SA, Laxman B, Dhanasekaran SM, Helgeson BE, Cao X, Morris DS, et al. Distinct classes of chromosomal rearrangements create oncogenic ETS gene fusions in prostate cancer. Nature. 2007;448(7153):595-9. 193. Pflueger D, Terry S, Sboner A, Habegger L, Esgueva R, Lin PC, et al. Discovery of non- ETS gene fusions in human prostate cancer using next-generation RNA sequencing. Genome Res.21(1):56-67. 194. Barbieri CE, Baca SC, Lawrence MS, Demichelis F, Blattner M, Theurillat JP, et al. Exome sequencing identifies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nat Genet.44(6):685-9. 195. Casey G, Neville PJ, Plummer SJ, Xiang Y, Krumroy LM, Klein EA, et al. RNASEL Arg462Gln variant is implicated in up to 13% of prostate cancer cases. Nat Genet. 2002;32(4):581-3. 196. Xiang Y, Wang Z, Murakami J, Plummer S, Klein EA, Carpten JD, et al. Effects of RNase L mutations associated with prostate cancer on apoptosis induced by 2',5'- oligoadenylates. Cancer Res. 2003;63(20):6795-801. 197. Urisman A, Molinaro RJ, Fischer N, Plummer SJ, Casey G, Klein EA, et al. Identification of a novel Gammaretrovirus in prostate tumors of patients homozygous for R462Q RNASEL variant. PLoS Pathog. 2006;2(3):e25.

201

198. Fischer N, Hellwinkel O, Schulz C, Chun FK, Huland H, Aepfelbacher M, et al. Prevalence of human gammaretrovirus XMRV in sporadic prostate cancer. J Clin Virol. 2008;43(3):277-83. 199. Knouf EC, Metzger MJ, Mitchell PS, Arroyo JD, Chevillet JR, Tewari M, et al. Multiple integrated copies and high-level production of the human retrovirus XMRV (xenotropic murine leukemia virus-related virus) from 22Rv1 prostate carcinoma cells. J Virol. 2009;83(14):7353-6. 200. Robinson MJ, Tuke PW, Erlwein O, Tettmar KI, Kaye S, Naresh KN, et al. No Evidence of XMRV or MuLV Sequences in Prostate Cancer, Diffuse Large B-Cell Lymphoma, or the UK Blood Donor Population. Adv Virol.2011:782353. 201. Hong P, Li J. Lack of evidence for a role of xenotropic murine leukemia virus-related virus in the pathogenesis of prostate cancer and/or chronic fatigue syndrome. Virus Res.167(1):1-7. 202. Arnold RS, Makarova NV, Osunkoya AO, Suppiah S, Scott TA, Johnson NA, et al. XMRV infection in patients with prostate cancer: novel serologic assay and correlation with PCR and FISH. Urology.75(4):755-61. 203. Mo F, Wyatt AW, Wu C, Lapuk AV, Marra MA, Gleave ME, et al. Next-generation sequencing of prostate tumors provides independent evidence of xenotropic murine leukemia virus-related gammaretrovirus contamination. J Clin Microbiol.50(2):536-7. 204. Erlwein O, Robinson MJ, Dustan S, Weber J, Kaye S, McClure MO. DNA extraction columns contaminated with murine sequences. PLoS One.6(8):e23484. 205. Kearney MF, Spindler J, Wiegand A, Shao W, Anderson EM, Maldarelli F, et al. Multiple sources of contamination in samples from patients reported to have XMRV infection. PLoS One.7(2):e30889. 206. Carter BS, Bova GS, Beaty TH, Steinberg GD, Childs B, Isaacs WB, et al. Hereditary prostate cancer: epidemiologic and clinical features. J Urol. 1993;150(3):797-802. 207. Smith JR, Freije D, Carpten JD, Gronberg H, Xu J, Isaacs SD, et al. Major susceptibility locus for prostate cancer on chromosome 1 suggested by a genome-wide search. Science. 1996;274(5291):1371-4. 208. Cooney KA, McCarthy JD, Lange E, Huang L, Miesfeldt S, Montie JE, et al. Prostate cancer susceptibility locus on chromosome 1q: a confirmatory study. J Natl Cancer Inst. 1997;89(13):955-9. 209. Carpten J, Nupponen N, Isaacs S, Sood R, Robbins C, Xu J, et al. Germline mutations in the ribonuclease L gene in families showing linkage with HPC1. Nat Genet. 2002;30(2):181-4. 210. Rokman A, Ikonen T, Seppala EH, Nupponen N, Autio V, Mononen N, et al. Germline alterations of the RNASEL gene, a candidate HPC1 gene at 1q25, in patients and families with prostate cancer. Am J Hum Genet. 2002;70(5):1299-304. 211. Starita LM, Parvin JD. The multiple nuclear functions of BRCA1: transcription, ubiquitination and DNA repair. Curr Opin Cell Biol. 2003;15(3):345-50. 212. Tulinius H, Egilsson V, Olafsdottir GH, Sigvaldason H. Risk of prostate, ovarian, and endometrial cancer among relatives of women with breast cancer. Bmj. 1992;305(6858):855-7. 213. Sellers TA, Potter JD, Rich SS, Drinkard CR, Bostick RM, Kushi LH, et al. Familial clustering of breast and prostate cancers and risk of postmenopausal breast cancer. J Natl Cancer Inst. 1994;86(24):1860-5. 214. Scully R. Role of BRCA gene dysfunction in breast and ovarian cancer predisposition. Breast Cancer Res. 2000;2(5):324-30.

202

215. Castro E, Eeles R. The role of BRCA1 and BRCA2 in prostate cancer. Asian J Androl.14(3):409-14. 216. Tryggvadottir L, Vidarsdottir L, Thorgeirsson T, Jonasson JG, Olafsdottir EJ, Olafsdottir GH, et al. Prostate cancer progression and survival in BRCA2 mutation carriers. J Natl Cancer Inst. 2007;99(12):929-35. 217. Edwards SM, Evans DG, Hope Q, Norman AR, Barbachano Y, Bullock S, et al. Prostate cancer in BRCA2 germline mutation carriers is associated with poorer prognosis. Br J Cancer.103(6):918-24. 218. Thorne H, Willems AJ, Niedermayr E, Hoh IM, Li J, Clouston D, et al. Decreased prostate cancer-specific survival of men with BRCA2 mutations from multiple breast cancer families. Cancer Prev Res (Phila).4(7):1002-10. 219. Norris JD, Chang CY, Wittmann BM, Kunder RS, Cui H, Fan D, et al. The homeodomain protein HOXB13 regulates the cellular response to androgens. Mol Cell. 2009;36(3):405-16. 220. Kim YR, Oh KJ, Park RY, Xuan NT, Kang TW, Kwon DD, et al. HOXB13 promotes androgen independent growth of LNCaP prostate cancer cells by the activation of E2F signaling. Mol Cancer.9:124. 221. Gillanders EM, Xu J, Chang BL, Lange EM, Wiklund F, Bailey-Wilson JE, et al. Combined genome-wide scan for prostate cancer susceptibility genes. J Natl Cancer Inst. 2004;96(16):1240-7. 222. Xu J, Dimitrov L, Chang BL, Adams TS, Turner AR, Meyers DA, et al. A combined genomewide linkage scan of 1,233 families for prostate cancer-susceptibility genes conducted by the international consortium for prostate cancer genetics. Am J Hum Genet. 2005;77(2):219-29. 223. Ewing CM, Ray AM, Lange EM, Zuhlke KA, Robbins CM, Tembe WD, et al. Germline mutations in HOXB13 and prostate-cancer risk. N Engl J Med.366(2):141-9. 224. Nakagawa H, Akamatsu S, Takata R, Takahashi A, Kubo M, Nakamura Y. Prostate cancer genomics, biology, and risk assessment through genome-wide association studies. Cancer Sci.103(4):607-13. 225. Eeles RA, Kote-Jarai Z, Giles GG, Olama AA, Guy M, Jugurnauth SK, et al. Multiple newly identified loci associated with prostate cancer susceptibility. Nat Genet. 2008;40(3):316- 21. 226. Al Olama AA, Kote-Jarai Z, Giles GG, Guy M, Morrison J, Severi G, et al. Multiple loci on 8q24 associated with prostate cancer susceptibility. Nat Genet. 2009;41(10):1058-60. 227. Yeager M, Chatterjee N, Ciampa J, Jacobs KB, Gonzalez-Bosquet J, Hayes RB, et al. Identification of a new prostate cancer susceptibility locus on chromosome 8q24. Nat Genet. 2009;41(10):1055-7. 228. Zanke BW, Greenwood CM, Rangrej J, Kustra R, Tenesa A, Farrington SM, et al. Genome-wide association scan identifies a colorectal cancer susceptibility locus on chromosome 8q24. Nat Genet. 2007;39(8):989-94. 229. Easton DF, Pooley KA, Dunning AM, Pharoah PD, Thompson D, Ballinger DG, et al. Genome-wide association study identifies novel breast cancer susceptibility loci. Nature. 2007;447(7148):1087-93. 230. Kiemeney LA, Thorlacius S, Sulem P, Geller F, Aben KK, Stacey SN, et al. Sequence variant on 8q24 confers susceptibility to urinary bladder cancer. Nat Genet. 2008;40(11):1307- 12.

203

231. Jia L, Landan G, Pomerantz M, Jaschek R, Herman P, Reich D, et al. Functional enhancers at the gene-poor 8q24 cancer-linked locus. PLoS Genet. 2009;5(8):e1000597. 232. Wasserman NF, Aneas I, Nobrega MA. An 8q24 gene desert variant associated with prostate cancer risk confers differential in vivo activity to a MYC enhancer. Genome Res.20(9):1191-7. 233. Ahmadiyeh N, Pomerantz MM, Grisanzio C, Herman P, Jia L, Almendro V, et al. 8q24 prostate, breast, and colon cancer risk loci show tissue-specific long-range interaction with MYC. Proc Natl Acad Sci U S A.107(21):9742-6. 234. Pomerantz MM, Beckwith CA, Regan MM, Wyman SK, Petrovics G, Chen Y, et al. Evaluation of the 8q24 prostate cancer risk locus and MYC expression. Cancer Res. 2009;69(13):5568-74. 235. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nat Rev Genet. 2002;3(6):415-28. 236. Kim VN, Han J, Siomi MC. Biogenesis of small RNAs in animals. Nat Rev Mol Cell Biol. 2009;10(2):126-39. 237. Mendell JT, Olson EN. MicroRNAs in stress signaling and human disease. Cell.148(6):1172-87. 238. Yekta S, Shih IH, Bartel DP. MicroRNA-directed cleavage of HOXB8 mRNA. Science. 2004;304(5670):594-6. 239. He L, Hannon GJ. MicroRNAs: small RNAs with a big role in gene regulation. Nat Rev Genet. 2004;5(7):522-31. 240. Cortez MA, Ivan C, Zhou P, Wu X, Ivan M, Calin GA. microRNAs in cancer: from bench to bedside. Adv Cancer Res.108:113-57. 241. Schaefer A, Jung M, Mollenkopf HJ, Wagner I, Stephan C, Jentzmik F, et al. Diagnostic and prognostic implications of microRNA profiling in prostate carcinoma. Int J Cancer.126(5):1166-76. 242. Spahn M, Kneitz S, Scholz CJ, Stenger N, Rudiger T, Strobel P, et al. Expression of microRNA-221 is progressively reduced in aggressive prostate cancer and metastasis and predicts clinical recurrence. Int J Cancer.127(2):394-403. 243. Yaman Agaoglu F, Kovancilar M, Dizdar Y, Darendeliler E, Holdenrieder S, Dalay N, et al. Investigation of miR-21, miR-141, and miR-221 in blood circulation of patients with prostate cancer. Tumour Biol.32(3):583-8. 244. Hagman Z, Larne O, Edsjo A, Bjartell A, Ehrnstrom RA, Ulmert D, et al. miR-34c is downregulated in prostate cancer and exerts tumor suppressive functions. Int J Cancer.127(12):2768-76. 245. Pesta M, Klecka J, Kulda V, Topolcan O, Hora M, Eret V, et al. Importance of miR-20a expression in prostate cancer tissue. Anticancer Res.30(9):3579-83. 246. Peng X, Guo W, Liu T, Wang X, Tu X, Xiong D, et al. Identification of miRs-143 and - 145 that is associated with bone metastasis of prostate cancer and involved in the regulation of EMT. PLoS One.6(5):e20341. 247. Mahn R, Heukamp LC, Rogenhofer S, von Ruecker A, Muller SC, Ellinger J. Circulating microRNAs (miRNA) in serum of patients with prostate cancer. Urology.77(5):1265 e9-16. 248. Brase JC, Johannes M, Schlomm T, Falth M, Haese A, Steuber T, et al. Circulating miRNAs are correlated with tumor progression in prostate cancer. Int J Cancer.128(3):608-16. 249. Wapinski O, Chang HY. Long noncoding RNAs and human disease. Trends Cell Biol.21(6):354-61.

204

250. Rinn JL, Kertesz M, Wang JK, Squazzo SL, Xu X, Brugmann SA, et al. Functional demarcation of active and silent chromatin domains in human HOX loci by noncoding RNAs. Cell. 2007;129(7):1311-23. 251. Lin M, Pedrosa E, Shah A, Hrabovsky A, Maqbool S, Zheng D, et al. RNA-Seq of human neurons derived from iPS cells reveals candidate long non-coding RNAs involved in neurogenesis and neuropsychiatric disorders. PLoS One.6(9):e23356. 252. Wang KC, Yang YW, Liu B, Sanyal A, Corces-Zimmerman R, Chen Y, et al. A long noncoding RNA maintains active chromatin to coordinate homeotic gene expression. Nature.472(7341):120-4. 253. Srikantan V, Zou Z, Petrovics G, Xu L, Augustus M, Davis L, et al. PCGEM1, a prostate- specific gene, is overexpressed in prostate cancer. Proc Natl Acad Sci U S A. 2000;97(22):12216-21. 254. Petrovics G, Zhang W, Makarem M, Street JP, Connelly R, Sun L, et al. Elevated expression of PCGEM1, a prostate-specific gene with cell growth-promoting function, is associated with high-risk prostate cancer patients. Oncogene. 2004;23(2):605-11. 255. Fu X, Ravindranath L, Tran N, Petrovics G, Srivastava S. Regulation of apoptosis by a prostate-specific and prostate cancer-associated noncoding gene, PCGEM1. DNA Cell Biol. 2006;25(3):135-41. 256. Chung S, Nakagawa H, Uemura M, Piao L, Ashikawa K, Hosono N, et al. Association of a novel long non-coding RNA in 8q24 with prostate cancer susceptibility. Cancer Sci.102(1):245-52. 257. Prensner JR, Iyer MK, Balbin OA, Dhanasekaran SM, Cao Q, Brenner JC, et al. Transcriptome sequencing across a prostate cancer cohort identifies PCAT-1, an unannotated lincRNA implicated in disease progression. Nat Biotechnol.29(8):742-9. 258. Verdone L, Caserta M, Di Mauro E. Role of histone acetylation in the control of gene expression. Biochem Cell Biol. 2005;83(3):344-53. 259. Kouzarides T. Chromatin modifications and their function. Cell. 2007;128(4):693-705. 260. Ellinger J, Kahl P, von der Gathen J, Rogenhofer S, Heukamp LC, Gutgemann I, et al. Global levels of histone modifications predict prostate cancer recurrence. Prostate.70(1):61-9. 261. Ke XS, Qu Y, Rostad K, Li WC, Lin B, Halvorsen OJ, et al. Genome-wide profiling of histone h3 lysine 4 and lysine 27 trimethylation reveals an epigenetic signature in prostate carcinogenesis. PLoS One. 2009;4(3):e4687. 262. Yu J, Yu J, Rhodes DR, Tomlins SA, Cao X, Chen G, et al. A polycomb repression signature in metastatic prostate cancer predicts cancer outcome. Cancer Res. 2007;67(22):10657- 63. 263. Zhang Y, Fang H, Jiao J, Xu W. The structure and function of histone deacetylases: the target for anti-cancer therapy. Curr Med Chem. 2008;15(27):2840-9. 264. Patra SK, Patra A, Dahiya R. Histone deacetylase and DNA methyltransferase in human prostate cancer. Biochem Biophys Res Commun. 2001;287(3):705-13. 265. Halkidou K, Gaughan L, Cook S, Leung HY, Neal DE, Robson CN. Upregulation and nuclear recruitment of HDAC1 in hormone refractory prostate cancer. Prostate. 2004;59(2):177- 89. 266. Song Y, Shiota M, Tamiya S, Kuroiwa K, Naito S, Tsuneyoshi M. The significance of strong histone deacetylase 1 expression in the progression of prostate cancer. Histopathology.58(5):773-80.

205

267. Huffman DM, Grizzle WE, Bamman MM, Kim JS, Eltoum IA, Elgavish A, et al. SIRT1 is significantly elevated in mouse and human prostate cancer. Cancer Res. 2007;67(14):6612-8. 268. Kojima K, Ohhashi R, Fujita Y, Hamada N, Akao Y, Nozawa Y, et al. A role for SIRT1 in cell growth and chemoresistance in prostate cancer PC3 and DU145 cells. Biochem Biophys Res Commun. 2008;373(3):423-8. 269. Nakane K, Fujita Y, Terazawa R, Atsumi Y, Kato T, Nozawa Y, et al. Inhibition of cortactin and SIRT1 expression attenuates migration and invasion of prostate cancer DU145 cells. Int J Urol.19(1):71-9. 270. Simon JA, Lange CA. Roles of the EZH2 histone methyltransferase in cancer epigenetics. Mutat Res. 2008;647(1-2):21-9. 271. Varambally S, Dhanasekaran SM, Zhou M, Barrette TR, Kumar-Sinha C, Sanda MG, et al. The polycomb group protein EZH2 is involved in progression of prostate cancer. Nature. 2002;419(6907):624-9. 272. Bryant RJ, Cross NA, Eaton CL, Hamdy FC, Cunliffe VT. EZH2 promotes proliferation and invasiveness of prostate cancer cells. Prostate. 2007;67(5):547-56. 273. Yu J, Yu J, Mani RS, Cao Q, Brenner CJ, Cao X, et al. An integrated network of androgen receptor, polycomb, and TMPRSS2-ERG gene fusions in prostate cancer progression. Cancer Cell.17(5):443-54. 274. Kunderfranco P, Mello-Grand M, Cangemi R, Pellini S, Mensah A, Albertini V, et al. ETS transcription factors control transcription of EZH2 and epigenetic silencing of the tumor suppressor gene Nkx3.1 in prostate cancer. PLoS One. 2010;5(5):e10547. 275. Hoffmann MJ, Engers R, Florl AR, Otte AP, Muller M, Schulz WA. Expression changes in EZH2, but not in BMI-1, SIRT1, DNMT1 or DNMT3B are associated with DNA methylation changes in prostate cancer. Cancer Biol Ther. 2007;6(9):1403-12. 276. Chen H, Tu SW, Hsieh JT. Down-regulation of human DAB2IP gene expression mediated by polycomb Ezh2 complex and histone deacetylase in prostate cancer. J Biol Chem. 2005;280(23):22437-44. 277. Xie D, Gore C, Liu J, Pong RC, Mason R, Hao G, et al. Role of DAB2IP in modulating epithelial-to-mesenchymal transition and prostate cancer metastasis. Proc Natl Acad Sci U S A.107(6):2485-90. 278. Li E, Bestor TH, Jaenisch R. Targeted mutation of the DNA methyltransferase gene results in embryonic lethality. Cell. 1992;69(6):915-26. 279. Okano M, Bell DW, Haber DA, Li E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell. 1999;99(3):247-57. 280. Holliday R, Pugh JE. DNA modification mechanisms and gene activity during development. Science. 1975;187(4173):226-32. 281. Riggs AD. X inactivation, differentiation, and DNA methylation. Cytogenet Cell Genet. 1975;14(1):9-25. 282. Lee J, Jang SJ, Benoit N, Hoque MO, Califano JA, Trink B, et al. Presence of 5- methylcytosine in CpNpG trinucleotides in the human genome. Genomics.96(2):67-72. 283. Bird AP. CpG-rich islands and the function of DNA methylation. Nature. 1986;321(6067):209-13. 284. Kass SU, Landsberger N, Wolffe AP. DNA methylation directs a time-dependent repression of transcription initiation. Curr Biol. 1997;7(3):157-65. 285. Herman JG, Baylin SB. Gene silencing in cancer in association with promoter hypermethylation. N Engl J Med. 2003;349(21):2042-54.

206

286. Lafon-Hughes L, Di Tomaso MV, Mendez-Acuna L, Martinez-Lopez W. Chromatin- remodelling mechanisms in cancer. Mutat Res. 2008;658(3):191-214. 287. Vaissiere T, Sawan C, Herceg Z. Epigenetic interplay between histone modifications and DNA methylation in gene silencing. Mutat Res. 2008;659(1-2):40-8. 288. Ting AH, McGarvey KM, Baylin SB. The cancer epigenome--components and functional correlates. Genes Dev. 2006;20(23):3215-31. 289. Irizarry RA, Ladd-Acosta C, Wen B, Wu Z, Montano C, Onyango P, et al. The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue-specific CpG island shores. Nat Genet. 2009;41(2):178-86. 290. Clark SJ, Harrison J, Molloy PL. Sp1 binding is inhibited by (m)Cp(m)CpG methylation. Gene. 1997;195(1):67-71. 291. Wolf SF, Jolly DJ, Lunnen KD, Friedmann T, Migeon BR. Methylation of the hypoxanthine phosphoribosyltransferase locus on the human X chromosome: implications for X- chromosome inactivation. Proc Natl Acad Sci U S A. 1984;81(9):2806-10. 292. Hellman A, Chess A. Gene body-specific methylation on the active X chromosome. Science. 2007;315(5815):1141-3. 293. Rountree MR, Selker EU. DNA methylation inhibits elongation but not initiation of transcription in Neurospora crassa. Genes Dev. 1997;11(18):2383-95. 294. Maunakea AK, Nagarajan RP, Bilenky M, Ballinger TJ, D'Souza C, Fouse SD, et al. Conserved role of intragenic DNA methylation in regulating alternative promoters. Nature.466(7303):253-7. 295. Shukla S, Kavak E, Gregory M, Imashimizu M, Shutinoski B, Kashlev M, et al. CTCF- promoted RNA polymerase II pausing links DNA methylation to splicing. Nature.479(7371):74- 9. 296. Bartolomei MS, Webber AL, Brunkow ME, Tilghman SM. Epigenetic mechanisms underlying the imprinting of the mouse H19 gene. Genes Dev. 1993;7(9):1663-73. 297. Stadler MB, Murr R, Burger L, Ivanek R, Lienert F, Scholer A, et al. DNA-binding factors shape the mouse methylome at distal regulatory regions. Nature.480(7378):490-5. 298. Schmidl C, Klug M, Boeld TJ, Andreesen R, Hoffmann P, Edinger M, et al. Lineage- specific DNA methylation in T cells correlates with histone methylation and enhancer activity. Genome Res. 2009;19(7):1165-74. 299. Bell AC, Felsenfeld G. Methylation of a CTCF-dependent boundary controls imprinted expression of the Igf2 gene. Nature. 2000;405(6785):482-5. 300. Lee WH, Morton RA, Epstein JI, Brooks JD, Campbell PA, Bova GS, et al. Cytidine methylation of regulatory sequences near the pi-class glutathione S-transferase gene accompanies human prostatic carcinogenesis. Proc Natl Acad Sci U S A. 1994;91(24):11733-7. 301. Hopkins TG, Burns PA, Routledge MN. DNA methylation of GSTP1 as biomarker in diagnosis of prostate cancer. Urology. 2007;69(1):11-6. 302. Hiltunen MO, Alhonen L, Koistinaho J, Myohanen S, Paakkonen M, Marin S, et al. Hypermethylation of the APC (adenomatous polyposis coli) gene promoter region in human colorectal carcinoma. Int J Cancer. 1997;70(6):644-8. 303. Dammann R, Li C, Yoon JH, Chin PL, Bates S, Pfeifer GP. Epigenetic inactivation of a RAS association domain family protein from the lung tumour suppressor locus 3p21.3. Nat Genet. 2000;25(3):315-9.

207

304. Cote S, Sinnett D, Momparler RL. Demethylation by 5-aza-2'-deoxycytidine of specific 5-methylcytosine sites in the promoter region of the retinoic acid receptor beta gene in human colon carcinoma cells. Anticancer Drugs. 1998;9(9):743-50. 305. Kang GH, Lee S, Lee HJ, Hwang KS. Aberrant CpG island hypermethylation of multiple genes in prostate cancer and prostatic intraepithelial neoplasia. J Pathol. 2004;202(2):233-40. 306. Cho NY, Kim BH, Choi M, Yoo EJ, Moon KC, Cho YM, et al. Hypermethylation of CpG island loci and hypomethylation of LINE-1 and Alu repeats in prostate adenocarcinoma and their relationship to clinicopathological features. J Pathol. 2007;211(3):269-77. 307. Liu L, Yoon JH, Dammann R, Pfeifer GP. Frequent hypermethylation of the RASSF1A gene in prostate cancer. Oncogene. 2002;21(44):6835-40. 308. Cottrell S, Jung K, Kristiansen G, Eltze E, Semjonow A, Ittmann M, et al. Discovery and validation of 3 novel DNA methylation markers of prostate cancer prognosis. J Urol. 2007;177(5):1753-8. 309. Perry AS, Loftus B, Moroose R, Lynch TH, Hollywood D, Watson RW, et al. In silico mining identifies IGFBP3 as a novel target of methylation in prostate cancer. Br J Cancer. 2007;96(10):1587-94. 310. Toyota M, Kopecky KJ, Toyota MO, Jair KW, Willman CL, Issa JP. Methylation profiling in acute myeloid leukemia. Blood. 2001;97(9):2823-9. 311. Maier S, Nimmrich I, Koenig T, Eppenberger-Castori S, Bohlmann I, Paradiso A, et al. DNA-methylation of the homeodomain transcription factor PITX2 reliably predicts risk of distant disease recurrence in tamoxifen-treated, node-negative breast cancer patients--Technical and clinical validation in a multi-centre setting in collaboration with the European Organisation for Research and Treatment of Cancer (EORTC) PathoBiology group. Eur J Cancer. 2007;43(11):1679-86. 312. Anglim PP, Galler JS, Koss MN, Hagen JA, Turla S, Campan M, et al. Identification of a panel of sensitive and specific DNA methylation markers for squamous cell lung cancer. Mol Cancer. 2008;7:62. 313. Wu X, Rauch TA, Zhong X, Bennett WP, Latif F, Krex D, et al. CpG island hypermethylation in human astrocytomas. Cancer Res. 2010;70(7):2718-27. 314. Weiss G, Cottrell S, Distler J, Schatz P, Kristiansen G, Ittmann M, et al. DNA methylation of the PITX2 gene promoter region is a strong independent prognostic marker of biochemical recurrence in patients with prostate cancer after radical prostatectomy. J Urol. 2009;181(4):1678-85. 315. Vanaja DK, Ehrich M, Van den Boom D, Cheville JC, Karnes RJ, Tindall DJ, et al. Hypermethylation of genes for diagnosis and risk stratification of prostate cancer. Cancer Invest. 2009;27(5):549-60. 316. Banez LL, Sun L, van Leenders GJ, Wheeler TM, Bangma CH, Freedland SJ, et al. Multicenter clinical validation of PITX2 methylation as a prostate specific antigen recurrence predictor in patients with post-radical prostatectomy prostate cancer. J Urol.184(1):149-56. 317. Schatz P, Dietrich D, Koenig T, Burger M, Lukas A, Fuhrmann I, et al. Development of a diagnostic microarray assay to assess the risk of recurrence of prostate cancer based on PITX2 DNA methylation. J Mol Diagn.12(3):345-53. 318. Cho B, Lee H, Jeong S, Bang YJ, Lee HJ, Hwang KS, et al. Promoter hypomethylation of a novel cancer/testis antigen gene CAGE is correlated with its aberrant expression and is seen in premalignant stage of gastric carcinoma. Biochem Biophys Res Commun. 2003;307(1):52-63.

208

319. Ogishima T, Shiina H, Breault JE, Tabatabai L, Bassett WW, Enokida H, et al. Increased heparanase expression is caused by promoter hypomethylation and up-regulation of transcriptional factor early growth response-1 in human prostate cancer. Clin Cancer Res. 2005;11(3):1028-36. 320. Tokizane T, Shiina H, Igawa M, Enokida H, Urakami S, Kawakami T, et al. Cytochrome P450 1B1 is overexpressed and regulated by hypomethylation in prostate cancer. Clin Cancer Res. 2005;11(16):5793-801. 321. Laner T, Schulz WA, Engers R, Muller M, Florl AR. Hypomethylation of the XIST gene promoter in prostate cancer. Oncol Res. 2005;15(5):257-64. 322. Pakneshan P, Szyf M, Rabbani SA. Hypomethylation of urokinase (uPA) promoter in breast and prostate cancer: prognostic and therapeutic implications. Curr Cancer Drug Targets. 2005;5(7):471-88. 323. Bhusari S, Yang B, Kueck J, Huang W, Jarrard DF. Insulin-like growth factor-2 (IGF2) loss of imprinting marks a field defect within human prostates containing cancer. Prostate.71(15):1621-30. 324. Bedford MT, van Helden PD. Hypomethylation of DNA in pathological conditions of the human prostate. Cancer Res. 1987;47(20):5274-6. 325. Santourlidis S, Florl A, Ackermann R, Wirtz HC, Schulz WA. High frequency of alterations in DNA methylation in adenocarcinoma of the prostate. Prostate. 1999;39(3):166-74. 326. Schulz WA, Elo JP, Florl AR, Pennanen S, Santourlidis S, Engers R, et al. Genomewide DNA hypomethylation is associated with alterations on chromosome 8 in prostate carcinoma. Genes Chromosomes Cancer. 2002;35(1):58-65. 327. Yegnasubramanian S, Haffner MC, Zhang Y, Gurel B, Cornish TC, Wu Z, et al. DNA hypomethylation arises later in prostate cancer progression than CpG island hypermethylation and contributes to metastatic tumor heterogeneity. Cancer Res. 2008;68(21):8954-67. 328. Pei L, Choi JH, Liu J, Lee EJ, McCarthy B, Wilson JM, et al. Genome-wide DNA methylation analysis reveals novel epigenetic changes in chronic lymphocytic leukemia. Epigenetics.7(6). 329. Meissner A, Mikkelsen TS, Gu H, Wernig M, Hanna J, Sivachenko A, et al. Genome- scale DNA methylation maps of pluripotent and differentiated cells. Nature. 2008;454(7205):766-70. 330. Kim JH, Dhanasekaran SM, Prensner JR, Cao X, Robinson D, Kalyana-Sundaram S, et al. Deep sequencing reveals distinct patterns of DNA methylation in prostate cancer. Genome Res. 2011;21(7):1028-41. 331. Chuang LS, Ian HI, Koh TW, Ng HH, Xu G, Li BF. Human DNA-(cytosine-5) methyltransferase-PCNA complex as a target for p21WAF1. Science. 1997;277(5334):1996- 2000. 332. Bestor TH. The DNA methyltransferases of mammals. Hum Mol Genet. 2000;9(16):2395-402. 333. Esteve PO, Chin HG, Pradhan S. Human maintenance DNA (cytosine-5)- methyltransferase and p53 modulate expression of p53-repressed promoters. Proc Natl Acad Sci U S A. 2005;102(4):1000-5. 334. Walton EL, Francastel C, Velasco G. Maintenance of DNA methylation: Dnmt3b joins the dance. Epigenetics.6(11):1373-7. 335. Goll MG, Kirpekar F, Maggert KA, Yoder JA, Hsieh CL, Zhang X, et al. Methylation of tRNAAsp by the DNA methyltransferase homolog Dnmt2. Science. 2006;311(5759):395-8.

209

336. Hata K, Okano M, Lei H, Li E. Dnmt3L cooperates with the Dnmt3 family of de novo DNA methyltransferases to establish maternal imprints in mice. Development. 2002;129(8):1983-93. 337. Aapola U, Liiv I, Peterson P. Imprinting regulator DNMT3L is a transcriptional repressor associated with histone deacetylase activity. Nucleic Acids Res. 2002;30(16):3602-8. 338. Morey Kinney SR, Smiraglia DJ, James SR, Moser MT, Foster BA, Karpf AR. Stage- specific alterations of DNA methyltransferase expression, DNA hypermethylation, and DNA hypomethylation during prostate cancer progression in the transgenic adenocarcinoma of mouse prostate model. Mol Cancer Res. 2008;6(8):1365-74. 339. Zhang Q, Chen L, Helfand BT, Jang TL, Sharma V, Kozlowski J, et al. TGF-beta regulates DNA methyltransferase expression in prostate cancer, correlates with aggressive capabilities, and predicts disease recurrence. PLoS One.6(9):e25168. 340. Pritchard C, Mecham B, Dumpit R, Coleman I, Bhattacharjee M, Chen Q, et al. Conserved gene expression programs integrate mammalian prostate development and tumorigenesis. Cancer Res. 2009;69(5):1739-47. 341. Kobayashi Y, Absher DM, Gulzar ZG, Young SR, McKenney JK, Peehl DM, et al. DNA methylation profiling reveals novel biomarkers and important roles for DNA methyltransferases in prostate cancer. Genome Res. 2011;21(7):1017-27. 342. Zhu JK. Active DNA demethylation mediated by DNA glycosylases. Annu Rev Genet. 2009;43:143-66. 343. Mayer W, Niveleau A, Walter J, Fundele R, Haaf T. Demethylation of the zygotic paternal genome. Nature. 2000;403(6769):501-2. 344. Hajkova P, Erhardt S, Lane N, Haaf T, El-Maarri O, Reik W, et al. Epigenetic reprogramming in mouse primordial germ cells. Mech Dev. 2002;117(1-2):15-23. 345. Bruniquel D, Schwartz RH. Selective, stable demethylation of the interleukin-2 gene enhances transcription by an active process. Nat Immunol. 2003;4(3):235-40. 346. Bhattacharya SK, Ramchandani S, Cervoni N, Szyf M. A mammalian protein with specific demethylase activity for mCpG DNA. Nature. 1999;397(6720):579-83. 347. Fremont M, Siegmann M, Gaulis S, Matthies R, Hess D, Jost JP. Demethylation of DNA by purified chick embryo 5-methylcytosine-DNA glycosylase requires both protein and RNA. Nucleic Acids Res. 1997;25(12):2375-80. 348. Hendrich B, Hardeland U, Ng HH, Jiricny J, Bird A. The thymine glycosylase MBD4 can bind to the product of deamination at methylated CpG sites. Nature. 1999;401(6750):301-4. 349. Zhu B, Zheng Y, Hess D, Angliker H, Schwarz S, Siegmann M, et al. 5-methylcytosine- DNA glycosylase activity is present in a cloned G/T mismatch DNA glycosylase associated with the chicken embryo DNA demethylation complex. Proc Natl Acad Sci U S A. 2000;97(10):5135- 9. 350. Barreto G, Schafer A, Marhold J, Stach D, Swaminathan SK, Handa V, et al. Gadd45a promotes epigenetic gene activation by repair-mediated DNA demethylation. Nature. 2007;445(7128):671-5. 351. Engel N, Tront JS, Erinle T, Nguyen N, Latham KE, Sapienza C, et al. Conserved DNA methylation in Gadd45a(-/-) mice. Epigenetics. 2009;4(2):98-9. 352. Tahiliani M, Koh KP, Shen Y, Pastor WA, Bandukwala H, Brudno Y, et al. Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science. 2009;324(5929):930-5.

210

353. Ito S, D'Alessio AC, Taranova OV, Hong K, Sowers LC, Zhang Y. Role of Tet proteins in 5mC to 5hmC conversion, ES-cell self-renewal and inner cell mass specification. Nature.466(7310):1129-33. 354. Bocker MT, Tuorto F, Raddatz G, Musch T, Yang FC, Xu M, et al. Hydroxylation of 5- methylcytosine by TET2 maintains the active state of the mammalian HOXA cluster. Nat Commun.3:818. 355. Wu H, Zhang Y. Mechanisms and functions of Tet protein-mediated 5-methylcytosine oxidation. Genes Dev.25(23):2436-52. 356. Lorsbach RB, Moore J, Mathew S, Raimondi SC, Mukatira ST, Downing JR. TET1, a member of a novel protein family, is fused to MLL in acute myeloid leukemia containing the t(10;11)(q22;q23). Leukemia. 2003;17(3):637-41. 357. Abdel-Wahab O, Mullally A, Hedvat C, Garcia-Manero G, Patel J, Wadleigh M, et al. Genetic characterization of TET1, TET2, and TET3 alterations in myeloid malignancies. Blood. 2009;114(1):144-7. 358. Chou WC, Chou SC, Liu CY, Chen CY, Hou HA, Kuo YY, et al. TET2 mutation is an unfavorable prognostic factor in acute myeloid leukemia patients with intermediate-risk cytogenetics. Blood.118(14):3803-10. 359. Weissmann S, Alpermann T, Grossmann V, Kowarsch A, Nadarajah N, Eder C, et al. Landscape of TET2 mutations in acute myeloid leukemia. Leukemia.26(5):934-42. 360. Perez C, Martinez-Calle N, Martin-Subero JI, Segura V, Delabesse E, Fernandez- Mercado M, et al. TET2 mutations are associated with specific 5-methylcytosine and 5- hydroxymethylcytosine profiles in patients with chronic myelomonocytic leukemia. PLoS One.7(2):e31605. 361. Haffner MC, Chaux A, Meeker AK, Esopi DM, Gerber J, Pellakuru LG, et al. Global 5- hydroxymethylcytosine content is significantly reduced in tissue stem/progenitor cell compartments and in human cancers. Oncotarget.2(8):627-37. 362. Yang H, Liu Y, Bai F, Zhang JY, Ma SH, Liu J, et al. Tumor development is associated with decrease of TET gene expression and 5-methylcytosine hydroxylation. Oncogene. 363. McGinnis W, Levine MS, Hafen E, Kuroiwa A, Gehring WJ. A conserved DNA sequence in homoeotic genes of the Drosophila Antennapedia and bithorax complexes. Nature. 1984;308(5958):428-33. 364. Krumlauf R. Hox genes in vertebrate development. Cell. 1994;78(2):191-201. 365. Jacobs Y, Schnabel CA, Cleary ML. Trimeric association of Hox and TALE homeodomain proteins mediates Hoxb2 hindbrain enhancer activity. Mol Cell Biol. 1999;19(7):5134-42. 366. Schnabel CA, Jacobs Y, Cleary ML. HoxA9-mediated immortalization of myeloid progenitors requires functional interactions with TALE cofactors Pbx and Meis. Oncogene. 2000;19(5):608-16. 367. Moens CB, Selleri L. Hox cofactors in vertebrate development. Dev Biol. 2006;291(2):193-206. 368. Zhou B, Chen L, Wu X, Wang J, Yin Y, Zhu G. MH1 domain of SMAD4 binds N- terminal residues of the homeodomain of Hoxc9. Biochim Biophys Acta. 2008;1784(5):747-52. 369. Wang N, Kim HG, Cotta CV, Wan M, Tang Y, Klug CA, et al. TGFbeta/BMP inhibits the bone marrow transformation capability of Hoxa9 by repressing its DNA-binding ability. Embo J. 2006;25(7):1469-80.

211

370. Shen W, Chrobak D, Krishnan K, Lawrence HJ, Largman C. HOXB6 protein is bound to CREB-binding protein and represses globin expression in a DNA binding-dependent, PBX interaction-independent process. J Biol Chem. 2004;279(38):39895-904. 371. Brock HW, Hodgson JW, Petruk S, Mazo A. Regulatory noncoding RNAs at Hox loci. Biochem Cell Biol. 2009;87(1):27-34. 372. Favier B, Dolle P. Developmental functions of mammalian Hox genes. Mol Hum Reprod. 1997;3(2):115-31. 373. Struhl G. Role of the esc+ gene product in ensuring the selective expression of segment- specific homeotic genes in Drosophila. J Embryol Exp Morphol. 1983;76:297-331. 374. Rijli FM, Chambon P. Genetic interactions of Hox genes in limb development: learning from compound mutants. Curr Opin Genet Dev. 1997;7(4):481-7. 375. Zakany J, Gerard M, Favier B, Potter SS, Duboule D. Functional equivalence and rescue among group 11 Hox gene products in vertebral patterning. Dev Biol. 1996;176(2):325-8. 376. Mortlock DP, Innis JW. Mutation of HOXA13 in hand-foot-genital syndrome. Nat Genet. 1997;15(2):179-80. 377. Podlasek CA, Duboule D, Bushman W. Male accessory sex organ morphogenesis is altered by loss of function of Hoxd-13. Dev Dyn. 1997;208(4):454-65. 378. Warot X, Fromental-Ramain C, Fraulob V, Chambon P, Dolle P. Gene dosage-dependent effects of the Hoxa-13 and Hoxd-13 mutations on morphogenesis of the terminal parts of the digestive and urogenital tracts. Development. 1997;124(23):4781-91. 379. Podlasek CA, Clemens JQ, Bushman W. Hoxa-13 gene mutation results in abnormal seminal vesicle and prostate development. J Urol. 1999;161(5):1655-61. 380. Takahashi Y, Hamada J, Murakawa K, Takada M, Tada M, Nogami I, et al. Expression profiles of 39 HOX genes in normal human adult organs and anaplastic thyroid cancer cell lines by quantitative real-time RT-PCR system. Exp Cell Res. 2004;293(1):144-53. 381. Bijl J, van Oostveen JW, Kreike M, Rieger E, van der Raaij-Helmer LM, Walboomers JM, et al. Expression of HOXC4, HOXC5, and HOXC6 in human lymphoid cell lines, leukemias, and benign and malignant lymphoid tissue. Blood. 1996;87(5):1737-45. 382. Zhang X, Hamada J, Nishimoto A, Takahashi Y, Murai T, Tada M, et al. HOXC6 and HOXC11 increase transcription of S100beta gene in BrdU-induced in vitro differentiation of GOTO neuroblastoma cells into Schwannian cells. J Cell Mol Med. 2007;11(2):299-306. 383. Illingworth R, Kerr A, Desousa D, Jorgensen H, Ellis P, Stalker J, et al. A novel CpG island set identifies tissue-specific methylation at developmental gene loci. PLoS Biol. 2008;6(1):e22. 384. Xi S, Zhu H, Xu H, Schmidtmann A, Geiman TM, Muegge K. Lsh controls Hox gene silencing during development. Proc Natl Acad Sci U S A. 2007;104(36):14366-71. 385. Atkinson SP, Koch CM, Clelland GK, Willcox S, Fowler JC, Stewart R, et al. Epigenetic marking prepares the human HOXA cluster for activation during differentiation of pluripotent cells. Stem Cells. 2008;26(5):1174-85. 386. Yekta S, Tabin CJ, Bartel DP. MicroRNAs in the Hox network: an apparent link to posterior prevalence. Nat Rev Genet. 2008;9(10):789-96. 387. Andreeff M, Ruvolo V, Gadgil S, Zeng C, Coombes K, Chen W, et al. HOX expression patterns identify a common signature for favorable AML. Leukemia. 2008;22(11):2041-7. 388. Abe M, Hamada J, Takahashi O, Takahashi Y, Tada M, Miyamoto M, et al. Disordered expression of HOX genes in human non-small cell lung cancer. Oncol Rep. 2006;15(4):797-802.

212

389. Miller GJ, Miller HL, van Bokhoven A, Lambert JR, Werahera PN, Schirripa O, et al. Aberrant HOXC expression accompanies the malignant phenotype in human prostate. Cancer Res. 2003;63(18):5879-88. 390. Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML, Minden MD, et al. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet. 2002;30(1):41-7. 391. Faber J, Krivtsov AV, Stubbs MC, Wright R, Davis TN, van den Heuvel-Eibrink M, et al. HOXA9 is required for survival in human MLL-rearranged acute leukemias. Blood. 2009;113(11):2375-85. 392. Nakamura T, Largaespada DA, Lee MP, Johnson LA, Ohyashiki K, Toyama K, et al. Fusion of the nucleoporin gene NUP98 to HOXA9 by the chromosome translocation t(7;11)(p15;p15) in human myeloid leukaemia. Nat Genet. 1996;12(2):154-8. 393. Ghannam G, Takeda A, Camarata T, Moore MA, Viale A, Yaseen NR. The oncogene Nup98-HOXA9 induces gene transcription in myeloid cells. J Biol Chem. 2004;279(2):866-75. 394. Wu X, Chen H, Parker B, Rubin E, Zhu T, Lee JS, et al. HOXB7, a homeodomain protein, is overexpressed in breast cancer and confers epithelial-mesenchymal transition. Cancer Res. 2006;66(19):9527-34. 395. Ma XJ, Wang Z, Ryan PD, Isakoff SJ, Barmettler A, Fuller A, et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell. 2004;5(6):607-16. 396. Waltregny D, Alami Y, Clausse N, de Leval J, Castronovo V. Overexpression of the homeobox gene HOXC8 in human prostate cancer correlates with loss of tumor differentiation. Prostate. 2002;50(3):162-9. 397. Axlund SD, Lambert JR, Nordeen SK. HOXC8 inhibits androgen receptor signaling in human prostate cancer cells by inhibiting SRC-3 recruitment to direct androgen target genes. Mol Cancer Res.8(12):1643-55. 398. Jung C, Kim RS, Zhang HJ, Lee SJ, Jeng MH. HOXB13 induces growth suppression of prostate cancer cells as a repressor of hormone-activated androgen receptor signaling. Cancer Res. 2004;64(24):9185-92. 399. Jung C, Kim RS, Lee SJ, Wang C, Jeng MH. HOXB13 homeodomain protein suppresses the growth of prostate cancer cells by the negative regulation of T-cell factor 4. Cancer Res. 2004;64(9):3046-51. 400. Grubach L, Juhl-Christensen C, Rethmeier A, Olesen LH, Aggerholm A, Hokland P, et al. Gene expression profiling of Polycomb, Hox and Meis genes in patients with acute myeloid leukaemia. Eur J Haematol. 2008;81(2):112-22. 401. Piotrowski A, Benetkiewicz M, Menzel U, de Stahl TD, Mantripragada K, Grigelionis G, et al. Microarray-based survey of CpG islands identifies concurrent hyper- and hypomethylation patterns in tissues derived from patients with breast cancer. Genes Chromosomes Cancer. 2006;45(7):656-67. 402. Miao J, Wang Z, Provencher H, Muir B, Dahiya S, Carney E, et al. HOXB13 promotes ovarian cancer progression. Proc Natl Acad Sci U S A. 2007;104(43):17093-8. 403. Terranova R, Agherbi H, Boned A, Meresse S, Djabali M. Histone and DNA methylation defects at Hox genes in mice expressing a SET domain-truncated form of Mll. Proc Natl Acad Sci U S A. 2006;103(17):6629-34. 404. Hansen KH, Bracken AP, Pasini D, Dietrich N, Gehani SS, Monrad A, et al. A model for transmission of the H3K27me3 epigenetic mark. Nat Cell Biol. 2008;10(11):1291-300.

213

405. Hershko AY, Kafri T, Fainsod A, Razin A. Methylation of HoxA5 and HoxB5 and its relevance to expression during mouse development. Gene. 2003;302(1-2):65-72. 406. Chopra VS, Mishra RK. "Mir"acles in hox gene regulation. Bioessays. 2006;28(5):445-8. 407. Huret JL, Dessen P, Bernheim A. An atlas of chromosomes in hematological malignancies. Example: 11q23 and MLL partners. Leukemia. 2001;15(6):987-9. 408. Krivtsov AV, Armstrong SA. MLL translocations, histone modifications and leukaemia stem-cell development. Nat Rev Cancer. 2007;7(11):823-33. 409. Dorrance AM, Liu S, Yuan W, Becknell B, Arnoczky KJ, Guimond M, et al. Mll partial tandem duplication induces aberrant Hox expression in vivo via specific epigenetic alterations. J Clin Invest. 2006;116(10):2707-16. 410. Kleer CG, Cao Q, Varambally S, Shen R, Ota I, Tomlins SA, et al. EZH2 is a marker of aggressive breast cancer and promotes neoplastic transformation of breast epithelial cells. Proc Natl Acad Sci U S A. 2003;100(20):11606-11. 411. Raman JD, Mongan NP, Tickoo SK, Boorjian SA, Scherr DS, Gudas LJ. Increased expression of the polycomb group gene, EZH2, in transitional cell carcinoma of the bladder. Clin Cancer Res. 2005;11(24 Pt 1):8570-6. 412. Arisan S, Buyuktuncer ED, Palavan-Unsal N, Caskurlu T, Cakir OO, Ergenekon E. Increased expression of EZH2, a polycomb group protein, in bladder carcinoma. Urol Int. 2005;75(3):252-7. 413. Matsukawa Y, Semba S, Kato H, Ito A, Yanagihara K, Yokozaki H. Expression of the enhancer of zeste homolog 2 is correlated with poor prognosis in human gastric cancer. Cancer Sci. 2006;97(6):484-91. 414. Wu X, Gong Y, Yue J, Qiang B, Yuan J, Peng X. Cooperation between EZH2, NSPc1- mediated histone H2A ubiquitination and Dnmt1 in HOX gene silencing. Nucleic Acids Res. 2008;36(11):3590-9. 415. Han L, Witmer PD, Casey E, Valle D, Sukumar S. DNA methylation regulates MicroRNA expression. Cancer Biol Ther. 2007;6(8):1284-8. 416. Ma L, Teruya-Feldstein J, Weinberg RA. Tumour invasion and metastasis initiated by microRNA-10b in breast cancer. Nature. 2007;449(7163):682-8. 417. Tan Y, Zhang B, Wu T, Skogerbo G, Zhu X, Guo X, et al. Transcriptional inhibiton of Hoxd4 expression by miRNA-10a in human breast cancer cells. BMC Mol Biol. 2009;10:12. 418. Shiraishi M, Sekiguchi A, Oates AJ, Terry MJ, Miyamoto Y. HOX gene clusters are hotspots of de novo methylation in CpG islands of human lung adenocarcinomas. Oncogene. 2002;21(22):3659-62. 419. Rauch T, Wang Z, Zhang X, Zhong X, Wu X, Lau SK, et al. Homeobox gene methylation in lung cancer studied by genome-wide analysis with a microarray-based methylated CpG island recovery assay. Proc Natl Acad Sci U S A. 2007;104(13):5527-32. 420. Strathdee G, Holyoake TL, Sim A, Parker A, Oscier DG, Melo JV, et al. Inactivation of HOXA genes by hypermethylation in myeloid and lymphoid malignancy is frequent and associated with poor prognosis. Clin Cancer Res. 2007;13(17):5048-55. 421. Shu Y, Wang B, Wang J, Wang JM, Zou SQ. Identification of methylation profile of HOX genes in extrahepatic cholangiocarcinoma. World J Gastroenterol.17(29):3407-19. 422. Park JH, Park J, Choi JK, Lyu J, Bae MG, Lee YG, et al. Identification of DNA methylation changes associated with human gastric cancer. BMC Med Genomics.4:82.

214

423. Erfurth FE, Popovic R, Grembecka J, Cierpicki T, Theisler C, Xia ZB, et al. MLL protects CpG clusters from methylation within the Hoxa9 gene, maintaining transcript expression. Proc Natl Acad Sci U S A. 2008;105(21):7517-22. 424. Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L, et al. A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat Genet. 2007;39(2):237-42. 425. Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Zimmerman J, et al. Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer. Nat Genet. 2007;39(2):232-6. 426. Vire E, Brenner C, Deplus R, Blanchon L, Fraga M, Didelot C, et al. The Polycomb group protein EZH2 directly controls DNA methylation. Nature. 2006;439(7078):871-4. 427. Skacel M, Ormsby AH, Pettay JD, Tsiftsakis EK, Liou LS, Klein EA, et al. Aneusomy of chromosomes 7, 8, and 17 and amplification of HER-2/neu and epidermal growth factor receptor in Gleason score 7 prostate carcinoma: a differential fluorescent in situ hybridization study of Gleason pattern 3 and 4 using tissue microarray. Hum Pathol. 2001;32(12):1392-7. 428. van Dekken H, Paris PL, Albertson DG, Alers JC, Andaya A, Kowbel D, et al. Evaluation of genetic patterns in different tumor areas of intermediate-grade prostatic adenocarcinomas by high-resolution genomic array analysis. Genes Chromosomes Cancer. 2004;39(3):249-56. 429. Laird PW. The power and the promise of DNA methylation markers. Nat Rev Cancer. 2003;3(4):253-66. 430. Woodson K, O'Reilly KJ, Ward DE, Walter J, Hanson J, Walk EL, et al. CD44 and PTGS2 methylation are independent prognostic markers for biochemical recurrence among prostate cancer patients with clinically localized disease. Epigenetics. 2006;1(4):183-6. 431. Ellinger J, Bastian PJ, Jurgan T, Biermann K, Kahl P, Heukamp LC, et al. CpG island hypermethylation at multiple gene sites in diagnosis and prognosis of prostate cancer. Urology. 2008;71(1):161-7. 432. Hamada J, Omatsu T, Okada F, Furuuchi K, Okubo Y, Takahashi Y, et al. Overexpression of homeobox gene HOXD3 induces coordinate expression of metastasis-related genes in human lung cancer cells. Int J Cancer. 2001;93(4):516-25. 433. Miyazaki YJ, Hamada J, Tada M, Furuuchi K, Takahashi Y, Kondo S, et al. HOXD3 enhances motility and invasiveness through the TGF-beta-dependent and -independent pathways in A549 cells. Oncogene. 2002;21(5):798-808. 434. Okubo Y, Hamada J, Takahashi Y, Tada M, Tsutsumida A, Furuuchi K, et al. Transduction of HOXD3-antisense into human melanoma cells results in decreased invasive and motile activities. Clin Exp Metastasis. 2002;19(6):503-11. 435. Kanai M, Hamada J, Takada M, Asano T, Murakawa K, Takahashi Y, et al. Aberrant expressions of HOX genes in colorectal and hepatocellular carcinomas. Oncol Rep.23(3):843-51. 436. Leshchenko VV, Kuo PY, Shaknovich R, Yang DT, Gellen T, Petrich A, et al. Genomewide DNA methylation analysis reveals novel targets for drug development in mantle cell lymphoma. Blood.116(7):1025-34. 437. Harada K, Yamazaki T, Iwata C, Yoshimatsu Y, Sase H, Mishima K, et al. Identification of targets of Prox1 during in vitro vascular differentiation from embryonic stem cells: functional roles of HoxD8 in lymphangiogenesis. J Cell Sci. 2009;122(Pt 21):3923-30.

215

438. Kikugawa T, Kinugasa Y, Shiraishi K, Nanba D, Nakashiro K, Tanji N, et al. PLZF regulates Pbx1 transcription and Pbx1-HoxC8 complex leads to androgen-independent prostate cancer proliferation. Prostate. 2006;66(10):1092-9. 439. Rosenbaum E, Hoque MO, Cohen Y, Zahurak M, Eisenberger MA, Epstein JI, et al. Promoter hypermethylation as an independent prognostic factor for relapse in patients with prostate cancer following radical prostatectomy. Clin Cancer Res. 2005;11(23):8321-5. 440. Henrique R, Ribeiro FR, Fonseca D, Hoque MO, Carvalho AL, Costa VL, et al. High promoter methylation levels of APC predict poor prognosis in sextant biopsies from prostate cancer patients. Clin Cancer Res. 2007;13(20):6122-9. 441. Richiardi L, Fiano V, Vizzini L, De Marco L, Delsedime L, Akre O, et al. Promoter methylation in APC, RUNX3, and GSTP1 and mortality in prostate cancer patients. J Clin Oncol. 2009;27(19):3161-8. 442. Sun C, Dobi A, Mohamed A, Li H, Thangapazham RL, Furusato B, et al. TMPRSS2- ERG fusion, a common genomic alteration in prostate cancer activates C-MYC and abrogates prostate epithelial differentiation. Oncogene. 2008;27(40):5348-53. 443. Leshem O, Madar S, Kogan-Sakin I, Kamer I, Goldstein I, Brosh R, et al. TMPRSS2/ERG promotes epithelial to mesenchymal transition through the ZEB1/ZEB2 axis in a prostate cancer model. PLoS One. 2011;6(7):e21650. 444. Kron K, Pethe V, Briollais L, Sadikovic B, Ozcelik H, Sunderji A, et al. Discovery of novel hypermethylated genes in prostate cancer using genomic CpG island microarrays. PLoS One. 2009;4(3):e4830. 445. Iljin K, Wolf M, Edgren H, Gupta S, Kilpinen S, Skotheim RI, et al. TMPRSS2 fusions with oncogenic ETS factors in prostate cancer involve unbalanced genomic rearrangements and are associated with HDAC1 and epigenetic reprogramming. Cancer Res. 2006;66(21):10242-6. 446. Brenner C, Deplus R, Didelot C, Loriot A, Vire E, De Smet C, et al. Myc represses transcription through recruitment of DNA methyltransferase corepressor. EMBO J. 2005;24(2):336-46. 447. Rauch T, Li H, Wu X, Pfeifer GP. MIRA-assisted microarray analysis, a new technology for the determination of DNA methylation patterns, identifies frequent methylation of homeodomain-containing genes in lung cancer cells. Cancer Res. 2006;66(16):7939-47. 448. Lewis EB. A gene complex controlling segmentation in Drosophila. Nature. 1978;276(5688):565-70. 449. Scott MP, Weiner AJ. Structural relationships among genes that control development: sequence homology between the Antennapedia, Ultrabithorax, and fushi tarazu loci of Drosophila. Proc Natl Acad Sci U S A. 1984;81(13):4115-9. 450. Christensen KL, Patrick AN, McCoy EL, Ford HL. The six family of homeobox genes in development and cancer. Adv Cancer Res. 2008;101:93-126. 451. Samuel S, Naora H. Homeobox gene expression in cancer: insights from developmental regulation and deregulation. Eur J Cancer. 2005;41(16):2428-37. 452. Khoubehi B, Kessling AM, Adshead JM, Smith GL, Smith RD, Ogden CW. Expression of the developmental and oncogenic PAX2 gene in human prostate cancer. J Urol. 2001;165(6 Pt 1):2115-20. 453. Schwartz AM, Man YG, Rezaei MK, Simmens SJ, Berg PE. BP1, a homeoprotein, is significantly expressed in prostate adenocarcinoma and is concordant with prostatic intraepithelial neoplasia. Mod Pathol. 2009;22(1):1-6.

216

454. Zhang Y, Shu J, Si J, Shen L, Estecio MR, Issa JP. Repetitive elements and enforced transcriptional repression co-operate to enhance DNA methylation spreading into a promoter CpG-island. Nucleic Acids Res. 455. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun MJ. Cancer statistics, 2009. CA Cancer J Clin. 2009;59(4):225-49. 456. Lopez-Beltran A, Mikuz G, Luque RJ, Mazzucchelli R, Montironi R. Current practice of Gleason grading of prostate carcinoma. Virchows Arch. 2006;448(2):111-8. 457. Eden A, Gaudet F, Waghmare A, Jaenisch R. Chromosomal instability and tumors promoted by DNA hypomethylation. Science. 2003;300(5618):455. 458. Baylin SB. DNA methylation and gene silencing in cancer. Nat Clin Pract Oncol. 2005;2 Suppl 1:S4-11. 459. Brena RM, Costello JF. Genome-epigenome interactions in cancer. Hum Mol Genet. 2007;16 Spec No 1:R96-105. 460. Wilson AS, Power BE, Molloy PL. DNA hypomethylation and human diseases. Biochim Biophys Acta. 2007;1775(1):138-62. 461. Lee PS, Teaberry VS, Bland AE, Huang Z, Whitaker RS, Baba T, et al. Elevated MAL expression is accompanied by promoter hypomethylation and platinum resistance in epithelial ovarian cancer. Int J Cancer. 2009. 462. Kim YK, Kim WJ. Epigenetic markers as promising prognosticators for bladder cancer. Int J Urol. 2009;16(1):17-22. 463. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology. 1990;1(1):43-6. 464. Furusato B, Tan SH, Young D, Dobi A, Sun C, Mohamed AA, et al. ERG oncoprotein expression in prostate cancer: clonal progression of ERG-positive tumor cells and potential for ERG-based stratification. Prostate Cancer Prostatic Dis. 2010;13(3):228-37. 465. Park K, Tomlins SA, Mudaliar KM, Chiu YL, Esgueva R, Mehra R, et al. Antibody- based detection of ERG rearrangement-positive prostate cancer. Neoplasia. 2010;12(7):590-8. 466. Braun M, Scheble VJ, Menon R, Scharf G, Wilbertz T, Petersen K, et al. Relevance of cohort design for studying the frequency of the ERG rearrangement in prostate cancer. Histopathology. 2011;58(7):1028-36. 467. Koksal IT, Ozcan F, Kadioglu TC, Esen T, Kilicaslan I, Tunc M. Discrepancy between Gleason scores of biopsy and radical prostatectomy specimens. Eur Urol. 2000;37(6):670-4. 468. Pinthus JH, Witkos M, Fleshner NE, Sweet J, Evans A, Jewett MA, et al. Prostate cancers scored as Gleason 6 on prostate biopsy are frequently Gleason 7 tumors at radical prostatectomy: implication on outcome. J Urol. 2006;176(3):979-84; discussion 84. 469. Goessl C, Muller M, Heicappell R, Krause H, Straub B, Schrader M, et al. DNA-based detection of prostate cancer in urine after prostatic massage. Urology. 2001;58(3):335-8. 470. Jonmarker S, Glaessgen A, Culp WD, Pisa P, Lewensohn R, Ekman P, et al. Expression of PDX-1 in prostate cancer, prostatic intraepithelial neoplasia and benign prostatic tissue. Apmis. 2008;116(6):491-8. 471. Basuyaux JP, Ferreira E, Stehelin D, Buttice G. The Ets transcription factors interact with each other and with the c-Fos/c-Jun complex via distinct protein domains in a DNA-dependent and -independent manner. J Biol Chem. 1997;272(42):26188-95. 472. Abate-Shen C, Shen MM, Gelmann E. Integrating differentiation and cancer: the Nkx3.1 homeobox gene in prostate organogenesis and carcinogenesis. Differentiation. 2008;76(6):717- 27.

217

473. Duverger O, Morasso MI. Role of homeobox genes in the patterning, specification, and differentiation of ectodermal appendages in mammals. J Cell Physiol. 2008;216(2):337-46. 474. Noguchi M, Stamey TA, Neal JE, Yemoto CE. An analysis of 148 consecutive transition zone cancers: clinical and histological characteristics. J Urol. 2000;163(6):1751-5. 475. Shannon BA, McNeal JE, Cohen RJ. Transition zone carcinoma of the prostate gland: a common indolent tumour type that occasionally manifests aggressive behaviour. Pathology. 2003;35(6):467-71. 476. Phe V, Cussenot O, Roupret M. Methylated genes as potential biomarkers in prostate cancer. BJU Int.105(10):1364-70. 477. Kron KJ, Liu L, Pethe VV, Demetrashvili N, Nesbitt ME, Trachtenberg J, et al. DNA methylation of HOXD3 as a marker of prostate cancer progression. Lab Invest. 2010;90(7):1060- 7. 478. Lu T, Burdelya LG, Swiatkowski SM, Boiko AD, Howe PH, Stark GR, et al. Secreted transforming growth factor beta2 activates NF-kappaB, blocks apoptosis, and is essential for the survival of some tumor cells. Proc Natl Acad Sci U S A. 2004;101(18):7112-7. 479. Konrad L, Scheiber JA, Schwarz L, Schrader AJ, Hofmann R. TGF-beta1 and TGF-beta2 strongly enhance the secretion of plasminogen activator inhibitor-1 and matrix metalloproteinase-9 of the human prostate cancer cell line PC-3. Regul Pept. 2009;155(1-3):28- 32. 480. Dallas SL, Zhao S, Cramer SD, Chen Z, Peehl DM, Bonewald LF. Preferential production of latent transforming growth factor beta-2 by primary prostatic epithelial cells and its activation by prostate-specific antigen. J Cell Physiol. 2005;202(2):361-70. 481. Danielpour D. Functions and regulation of transforming growth factor-beta (TGF-beta) in the prostate. Eur J Cancer. 2005;41(6):846-57. 482. Reynolds AR, Kyprianou N. Growth factor signalling in prostatic growth: significance in tumour development and therapeutic targeting. Br J Pharmacol. 2006;147 Suppl 2:S144-52. 483. Matthews E, Yang T, Janulis L, Goodwin S, Kundu SD, Karpus WJ, et al. Down- regulation of TGF-beta1 production restores immunogenicity in prostate cancer cells. Br J Cancer. 2000;83(4):519-25. 484. Tommasi S, Karm DL, Wu X, Yen Y, Pfeifer GP. Methylation of homeobox genes is a frequent and early epigenetic event in breast cancer. Breast Cancer Res. 2009;11(1):R14. 485. Zappavigna V, Falciola L, Helmer-Citterich M, Mavilio F, Bianchi ME. HMG1 interacts with HOX proteins and enhances their DNA binding and transcriptional activation. Embo J. 1996;15(18):4981-91. 486. Holland PW, Booth HA, Bruford EA. Classification and nomenclature of all human homeobox genes. BMC Biol. 2007;5:47. 487. Schuettengruber B, Martinez AM, Iovino N, Cavalli G. Trithorax group proteins: switching genes on and keeping them active. Nat Rev Mol Cell Biol.12(12):799-814. 488. Richly H, Aloia L, Di Croce L. Roles of the Polycomb group proteins in stem cells and cancer. Cell Death Dis.2:e204. 489. Schimanski CC, Frerichs K, Rahman F, Berger M, Lang H, Galle PR, et al. High miR- 196a levels promote the oncogenic phenotype of colorectal cancer cells. World J Gastroenterol. 2009;15(17):2089-96. 490. Salem CE, Markl ID, Bender CM, Gonzales FA, Jones PA, Liang G. PAX6 methylation and ectopic expression in human tumor cells. Int J Cancer. 2000;87(2):179-85.

218

491. Panagopoulos I, Isaksson M, Billstrom R, Strombeck B, Mitelman F, Johansson B. Fusion of the NUP98 gene and the homeobox gene HOXC13 in acute myeloid leukemia with t(11;12)(p15;q13). Genes Chromosomes Cancer. 2003;36(1):107-12. 492. Kozmik Z, Sure U, Ruedi D, Busslinger M, Aguzzi A. Deregulated expression of PAX5 in medulloblastoma. Proc Natl Acad Sci U S A. 1995;92(12):5709-13. 493. Pfeifer GP, Rauch TA. DNA methylation patterns in lung carcinomas. Semin Cancer Biol. 2009;19(3):181-7. 494. Taylor BS, Schultz N, Hieronymus H, Gopalan A, Xiao Y, Carver BS, et al. Integrative genomic profiling of human prostate cancer. Cancer Cell.18(1):11-22. 495. Gal-Yam EN, Egger G, Iniguez L, Holster H, Einarsson S, Zhang X, et al. Frequent switching of Polycomb repressive marks and DNA hypermethylation in the PC3 prostate cancer cell line. Proc Natl Acad Sci U S A. 2008;105(35):12979-84. 496. Takeshima H, Yamashita S, Shimazu T, Niwa T, Ushijima T. The presence of RNA polymerase II, active or stalled, predicts epigenetic fate of promoter CpG islands. Genome Res. 2009;19(11):1974-82. 497. Venolia L, Gartler SM. Comparison of transformation efficiency of human active and inactive X-chromosomal DNA. Nature. 1983;302(5903):82-3. 498. Douglas AG, Wood MJ. RNA splicing: disease and therapy. Brief Funct Genomics.10(3):151-64. 499. Sobczak K, Krzyzosiak WJ. Structural determinants of BRCA1 translational regulation. J Biol Chem. 2002;277(19):17349-58. 500. Di Giammartino DC, Nishida K, Manley JL. Mechanisms and consequences of alternative polyadenylation. Mol Cell.43(6):853-66. 501. Kalari S, Pfeifer GP. Identification of driver and passenger DNA methylation in cancer by epigenomic analysis. Adv Genet.70:277-308. 502. Vinarskaja A, Schulz WA, Ingenwerth M, Hader C, Arsov C. Association of PITX2 mRNA down-regulation in prostate cancer with promoter hypermethylation and poor prognosis. Urol Oncol. 2011. 503. Clark J, Attard G, Jhavar S, Flohr P, Reid A, De-Bono J, et al. Complex patterns of ETS gene alteration arise during cancer development in the human prostate. Oncogene. 2008;27(14):1993-2003. 504. Kambara T, Simms LA, Whitehall VL, Spring KJ, Wynter CV, Walsh MD, et al. BRAF mutation is associated with DNA methylation in serrated polyps and cancers of the colorectum. Gut. 2004;53(8):1137-44. 505. Bibikova M, Fan JB. Genome-wide DNA methylation profiling. Wiley Interdiscip Rev Syst Biol Med.2(2):210-23. 506. Veltri RW, Miller MC, Partin AW, Coffey DS, Epstein JI. Ability to predict biochemical progression using Gleason score and a computer-generated quantitative nuclear grade derived from cancer cell nuclei. Urology. 1996;48(5):685-91. 507. Kupelian PA, Katcher J, Levin HS, Klein EA. Stage T1-2 prostate cancer: a multivariate analysis of factors affecting biochemical and clinical failures after radical prostatectomy. Int J Radiat Oncol Biol Phys. 1997;37(5):1043-52. 508. Ghavamian R, Blute ML, Bergstralh EJ, Slezak J, Zincke H. Comparison of clinically nonpalpable prostate-specific antigen-detected (cT1c) versus palpable (cT2) prostate cancers in patients undergoing radical retropubic prostatectomy. Urology. 1999;54(1):105-10.

219

509. Wang X, Fan J, Liu D, Fu S, Ingvarsson S, Chen H. Spreading of Alu methylation to the promoter of the MLH1 gene in gastrointestinal cancer. PLoS One.6(10):e25913. 510. Watanabe N, Okochi-Takada E, Yagi Y, Furuta JI, Ushijima T. Decreased fidelity in replicating DNA methylation patterns in cancer cells leads to dense methylation of a CpG island. Curr Top Microbiol Immunol. 2006;310:199-210. 511. Ibrahim AE, Arends MJ, Silva AL, Wyllie AH, Greger L, Ito Y, et al. Sequential DNA methylation changes are associated with DNMT3B overexpression in colorectal neoplastic progression. Gut.60(4):499-508. 512. Palakurthy RK, Wajapeyee N, Santra MK, Gazin C, Lin L, Gobeil S, et al. Epigenetic silencing of the RASSF1A tumor suppressor gene through HOXB3-mediated induction of DNMT3B expression. Mol Cell. 2009;36(2):219-30. 513. Coolen MW, Stirzaker C, Song JZ, Statham AL, Kassir Z, Moreno CS, et al. Consolidation of the cancer genome into domains of repressive chromatin by long-range epigenetic silencing (LRES) reduces transcriptional plasticity. Nat Cell Biol.12(3):235-46. 514. Heintzman ND, Stuart RK, Hon G, Fu Y, Ching CW, Hawkins RD, et al. Distinct and predictive chromatin signatures of transcriptional promoters and enhancers in the human genome. Nat Genet. 2007;39(3):311-8. 515. Creyghton MP, Cheng AW, Welstead GG, Kooistra T, Carey BW, Steine EJ, et al. Histone H3K27ac separates active from poised enhancers and predicts developmental state. Proc Natl Acad Sci U S A.107(50):21931-6. 516. Huang L, Pu Y, Hepps D, Danielpour D, Prins GS. Posterior Hox gene expression and differential androgen regulation in the developing and adult rat prostate lobes. Endocrinology. 2007;148(3):1235-45. 517. Cantile M, Kisslinger A, Cindolo L, Schiavo G, D'Anto V, Franco R, et al. cAMP induced modifications of HOX D gene expression in prostate cells allow the identification of a chromosomal area involved in vivo with neuroendocrine differentiation of human advanced prostate cancers. J Cell Physiol. 2005;205(2):202-10. 518. Oefelein M, Chin-Chance C, Bushman W. Expression of the homeotic gene Hox-d13 in the developing and adult mouse prostate. J Urol. 1996;155(1):342-6. 519. Carrio M, Arderiu G, Myers C, Boudreau NJ. Homeobox D10 induces phenotypic reversion of breast tumor cells in a three-dimensional culture model. Cancer Res. 2005;65(16):7177-85.