uveal melanoma: genetic and epigenetic …
TRANSCRIPT
UVEAL MELANOMA: GENETIC AND EPIGENETIC
CHARACTERISATION
Charlotte L. Ness MD
Institute of Clinical Medicine, Faculty of Medicine, University of Oslo
and
Department of Ophthalmology, Oslo University Hospital
Ph.D. thesis
2020
© Charlotte Larsen Ness, 2021
Series of dissertations submitted to the Faculty of Medicine, University of Oslo
ISBN 978-82-8377-853-3
All rights reserved. No part of this publication may be reproduced or transmitted, in any form or by any means, without permission.
Cover: Hanne Baadsgaard Utigard. Print production: Reprosentralen, University of Oslo.
2
TABLE OF CONTENTS
Contents
1. Acknowledgements 4
2. Abbreviations 5
3. List of papers 7
Paper I 7
Paper II 7
Paper III 7
4. Sammendrag 8
5. Introduction 10
5.1 Uveal melanoma. Disease and management 10
5.1.1 Location, epidemiology and risk factors 10
5.1.2 Symptoms and diagnosis 11
5.1.3 TNM classification and prognostic pathological parameters 12
5.1.4 Treatment 13
5.1.5 Management of metastatic disease 14
5.2 Genetic determinants in uveal melanoma 15
5.2.1 Cytogenetic features 15
5.2.2 Molecular pathways and genomic alterations in UM 15
5.2.3 Binary clustering of uveal melanomas 20
5.2.4 Genetic alterations in metastatic UM 23
5.3 Epigenetics 23
5.3.1 DNA methylation 24
5.4 In vitro and in vivo preclinical models for studying UM 27
5.4.1. Three-dimensional in vitro models 28
5.4.2 In vitro assays for studying the metastatic process of UM 29
5.4.3 In vivo assays for studying UM 30
6. Aims of the thesis 31
7. Methods and methodological considerations 32
7.1 In vitro cultivation 32
7.2 Immunohistochemistry 33
7.3 Electron microscopy 35
3
7.4 Real-time quantitative reverse transcription PCR (qRT-PCR) 36
7.5 RNAscope in situ hybridisation 38
7.6 Microarrays 39
7.6.1 cDNA microarrays 39
7.6.2 DNA methylation assay 41
7.7 Western blot 43
8. Summary of results 44
8.1 Paper I 44
8.2 Paper II 44
8.3 Paper III 45
9.1 Discussion Paper I 47
9.2 Discussion Paper II 49
9.3 Discussion Paper III 52
10. Conclusions and future perspectives 55
11. References 57
Paper I-III 76
4
1. Acknowledgements
The work leading to this thesis was performed at Center for Eye Research, Department of
Ophthalmology, Oslo University Hospital and University of Oslo.
I wish to thank everyone who has contributed to this thesis. In particular I would like to express
my deepest gratitude to co-supervisor Agate Noer without whom this thesis would not have
been completed. Your guidance, extensive knowledge and the amount of time you have put in
the project has been of immense importance. Thank you for being optimistic and encouraging
throughout these years.
I would like to extend my sincere thanks to my supervisor, Professor Morten C. Moe for
offering me the opportunity to carry out my PhD work and for constructive feedback and
support. I am also grateful to my co-supervisor Professor Emeritus Bjørn Nicolaissen for his
valuable advises.
My gratitude goes to Kirankumar Katta for his indispensable work on paper III. Special thanks
to Øystein Garred and Theresa Kumar for providing in-depth histopathological knowledge and
help in obtaining tissue samples. Further, I would like to acknowledge the contributions of all
co-authors.
Many thanks go to my colleagues at Center for Eye Research for practical support and for
creating a social work environment.
Lastly, my appreciation goes to close friends and family for their patience and encouragement.
The work was funded by the South-Eastern Norway Regional Health Authority (project
2012104), Norwegian Cancer Society (project 5808589) and supported by grants from the
Norwegian Association of the Blind and Partially Sighted, Arthur and Odd Clausons
ophthalmological fund, Aase and Knut Tønjums ophthalmological fund, Futura fund, Unifor
Frimed, Inger Holms memorial fund, “Stiftelsen for fremme av kreftforskning” at University
of Oslo and “Legat til fremme av kreftforskning”.
5
2. Abbreviations
AEC aminoethyl carbazole
AR antigen retrieval
BAP1 BRCA1 Associated Protein 1
BRAF B-RAF Proto-Oncogene
BSA bovine serum albumin
CNV copy number variation
CpG 5`-cytosine-phosphate-guanine-3
CPGI 5`-cytosine-phosphate-guanine-3` island
CSC cancer stem cell
CT computed tomography
Ct threshold cycle
CTL4 cytotoxic T lymphocyte antigen 4
Cq quantification cycle
Cx43 connexin 43
DAB 3,3`-diaminobenzidine
DMR differentially methylated region
DMP differentially methylated position
DNA deoxyribonucleic acid
DNMT DNA methyltransferase
ECM extracellular matrix
EZH2 enhancer for zeste homolog 2
FAO fatty acid oxidation
FFPE formalin-fixed paraffin embedded
FITC fluorescein isothiocyanate
FNAB fine needle aspiration biopsy
GNA11 guanine nucleotide-binding protein subunit alpha-11
GNAQ guanine nucleotide-binding protein G(q) subunit alpha
HIER heat induced epitope retrieval
HRP horseradish peroxidase
ICI immune checkpoint inhibitors
IHC immunohistochemistry
6
IPA ingenuity pathway analysis
LOA loss of adherence
M3 monosomy 3
MCTS multicellular tumour spheroids
MEK mitogen-activated protein kinase kinase
MRI magnetic resonance imaging
PCA principal component analysis
PCR polymerase chain reaction
PD1 programmed cell death protein 1
PET positron emission tomography
PRC1 polycomb repressive complex 1
PRC2 polycomb repressive complex 2
PR-DUB polycomb repressive-deubiquitinase complex
qRT-PCR quantitative reverse transcription polymerase chain reaction
RIN RNA integrity number
RNA ribonucleic acid
RPE retinal pigmented epithelium
SEM scanning electron microscope
TEM transmission electron microscope
TET ten-eleven translocation enzymes
tRNA transfer ribonucleic acid
TSS transcription start site
UM uveal melanoma
2D two-dimensional
3D three-dimensional
5-hmC 5-hydroxymethylcytosine
5-mC 5-methylcytosine
7
3. List of papers
Paper I
C. Ness, Ø. Garred, N. Eide., T. Kumar, OK. Olstad, TP. Bærland, G. Petrovski, MC. Moe, A.
Noer
Multicellular tumor spheroids of human uveal melanoma induce genes associated with
anoikis resistance, lipogenesis, and SSXs.
Exp Eye Res. 2021 Feb;203:108426. doi: 10.1016/j.exer.2020.108426. Epub 2020 Dec 30.
Paper II
C. Ness, K. Katta, Ø. Garred, T. Kumar, OK. Olstad, G. Petrovski, MC. Moe. A. Noer
Integrated differential DNA methylation and gene expression of formalin-fixed paraffin-
embedded uveal melanoma specimens identifies genes associated with early metastasis
and poor prognosis
Mol Vis. 2017 Oct 3;23:680-694. eCollection 2017.
Paper III
K. Katta*, C. Ness*, Ø. Garred, T. Kumar, OK. Olstad, N. Eide, B. Nicolaissen, G. Petrovski,
MC. Moe. A. Noer
*= co-first authors
Connexin 43 expression and subcellular distribution is dysregulated in human uveal
melanoma
Manuscript
8
4. Sammendrag
Uvealt malignt melanom (UM) er den vanligste formen for primær øyekreft. Det er en alvorlig
sykdom hvor opptil 50% av pasientene utvikler metastaser. Ved spredning har en ingen
effektive behandlingsalternativer. Avhandlingen består av 3 separate arbeider hvor en
undersøker basale mekanismer ved UM som kan ha betydning for spredning og som kan
representere angrepspunkter for fremtidig behandling.
Den første artikkelen tar for seg mekanismer som kan være assosiert med disseminering av
kreftceller og overlevelse av de disseminerte cellene. En sammenlignet i dette arbeidet
primærtumorer mot adherente cellekulturer og multicellulære-tumorsfæroider (MCTS).
Effekten av de forskjellige vekstbetingelsene ble undersøkt vha. elektronmikroskopi, DNA-
mikromatrise, qRT-PCR, RNAscope og immunohistokjemi (IHC). MCTS fremviste
egenskaper assosiert med motstand mot anoikis, som oppregulering av ANGPTL4 og økt
fettmetabolisme. MCTS viste også økt ekspresjon av Synovial sarcoma, X breakpoint proteiner
(SSXer), SSXer er kjente mål for immunterapi ved andre kreftformer.
I den andre artikkelen undersøkte en DNA-metyleringsmønstre i UM. Endringer i DNA-
metylering har betydning for utvikling og progresjon av kreft. Disse endringene er i
utgangspunktet reversible og er derfor attraktive mål for kreftbehandling. En undersøkte i denne
studien DNA-metylering i eldre formalinfikserte parafinblokker hvor en koblet funn til data fra
patologirapporter, Kreftregisteret og Dødsårsaksregisteret. DNA-metylering ble analysert vha.
Illumina Infinium HumanMethylation450 mikromatrise. En utførte videre en integrert analyse
mellom DNA-metylering og genekspresjon på et utvalg av prøver og fant her endringer
assosiert med tidlig metastasering. Genuttrykket ble undersøkt vha. DNA-mikromatrise og
qRT-PCR. Hypermetylerte gen inkluderte de antatte tumorsuppressor-genene RNF13, ZNF217
og HYAL1, mens hypometylerte gen inkluderte de antatte onkogenene TMEM200C, RGS10,
ADAM12 og PAM.
I det tredje arbeidet undersøkte en ekspresjon av connexin 43 (Cx43) i primære UM og vurderte
effekten av inhibering av Enhancer of Zeste homolog 2 (EZH2) på Cx43-ekspresjon.
Connexiner er involvert i en rekke cellulære prosesser og er ofte dysregulerte ved kreft. UM
biopsier og cellekulturer ble sammenlignet med choroidale biopsier og uveale melanocytter fra
friske donorer vha. DNA-mikromatrise og qRT-PCR. Cx43-uttrykk i primærtumorer ble
undersøkt vha. IHC og korrelert med histopatologiske data. Videre undersøkte en effekten av
EZH2-inhibitoren Tazemetostat på Cx43-ekspresjon i UM cellelinjer. Effekten ble evaluert
vha. morfologisk vurdering, ATP-analyse, qRT-PCR, immunocytokjemi (ICC) og Western
9
blot. Ekspresjon av Cx43 var redusert i UM i forhold til friske kontroller. UM fremviste også
redusert membran-innfarging. Tazemetostat medførte ikke endringer i Cx43-ekspresjon, men
en observerte en reduksjon av H3K27me3 uavhengig av BAP1 status.
10
5. Introduction
Cancer is a leading cause of death worldwide and can be looked upon as diverse collection of
diseases characterised by the dysregulation of important pathways that control normal cellular
homeostasis (1). In 2000 Hanahan and Weinberg presented six hallmarks of cancer as a
framework to explain the escape from normal control mechanisms. The six hallmarks include
sustaining proliferative signalling, evading growth suppressors, resisting cell death, enabling
replicative immortality, inducing angiogenesis and metastasis (2). In 2011 two new hallmarks
were added, namely the reprogramming of energy metabolism and the evasion of immune
destruction (3). Factors leading to these hallmarks include both genetic and epigenetic events.
Uveal melanoma (UM) is a relatively rare malignancy, thus receiving far less research attention
than cancers with a higher incidence. Compared to cutaneous melanoma, the unravelling of the
biology of UM is still in its beginning. UM has a high propensity for metastatic spread and is a
devastating disease for 50% of the patients (4). The lack of effective therapeutic options for
patients with metastatic disease urges the need for in-depth biological knowledge in order to
develop new and improved therapeutics.
5.1 Uveal melanoma. Disease and management
5.1.1 Location, epidemiology and risk factors
UM is the most common primary intraocular malignancy with an incidence of 5-8 cases per
million per year in Norway (4). The tumour arises from the pigmented cells of the posterior
uveal tract (choroid and ciliary body) or the anterior uveal tract (iris) (Figure 1). The choroid
is the most common location comprising approximately 85% of the cases. Iris melanomas
constitute 5% of the tumours, while ciliary body melanomas comprise 10% of the cases (4, 5).
The incidence is higher for those of Caucasian ethnicity, and especially for individuals with fair
skin and light iris colour. In contradistinction to skin melanoma, UM is not evidently associated
with sun exposure (6, 7). The median age at diagnosis is 60 years for patients with tumours of
the choroid and ciliary body, while the median age for iris melanomas is lower, at 43 years (8,
9). A significant predilection for gender has not been shown (4). Choroidal nevi can in some
instances undergo malignant transformation (10).
11
Approximately 40-50% of the patients develop metastatic disease. Once metastatic disease
occurs, the survival rate drops dramatically with a life expectancy of only 4 -17 months (11).
Systemic metastases are most commonly found in the liver (89%), followed by lungs (29%),
bone (17%) and skin (12%). An estimated 10% of the patients develop another primary
malignancy (12).
Figure 1: Anatomy of the eye. Horizontal section. (Courtesy of Geir A. Qvale, Oslo, Norway)
5.1.2 Symptoms and diagnosis
Initially most of the tumours are asymptomatic, though some patients can experience early
symptoms secondary to the localisation of the tumour. Iris melanomas are often detected at an
early stage due to distortion of the iris as a visible phenomenon. As the tumour enlarges the
patient can experience blurred vision (ciliary body), decreased visual acuity, floaters, photopsia,
and visual field defects. Exudative detachment of the retina can be seen, more rarely angle-
12
closure glaucoma (13). Patients with susceptible lesions should be examined in a slit lamp
followed by ultrasonography. Malignant tumours often present themselves as prominent,
grey/brown tumours that have a circular/ oval shape. Orange pigmentation, tumour exudation
and lower serous detachment can also be observed. Magnetic resonance imaging (MRI) and
fluorescein angiography can be used to characterise the lesion. About 88% of UMs show low
echogenicity on ultrasonography. Fluorescein angiography can detect risk features such as dye
leakage and irregular vessels (4). Fine needle aspiration (FNAB) is valuable tool in stratifying
lesions into malignant and benign categories, thus securing the initial diagnosis. Evaluation of
metastatic spread is assessed by clinical examination, blood samples (including assessment of
liver status), X-ray of the thorax and ultrasonography of the abdomen. Computed tomography
(CT) scan and MRI can be used to specify unclear sonographic findings. Positron emission
tomography (PET)/CT scan is not routinely performed (4, 14). Differential diagnosis of UM
include choroidal nevus, intraocular metastases, congenital hypertrophy of the retinal pigment
epithelium (RPE), haemorrhagic RPE detachment, choroidal haemangioma, age-related
macular degeneration, RPE hyperplasia, among others (14, 15).
5.1.3 TNM classification and prognostic pathological parameters
Several factors have prognostic impact on UM, including histological parameters and
extraocular extension. UM is staged according to the tumour, lymph nodes, metastasis (TNM)
classification of malignant tumours (16). T describes the size of the original tumour and whether
or not it has invaded nearby tissue. N tells of the involvement of lymph nodes, whereas M
reports on the presence of metastatic disease. Due to a relative lack of lymphatic outflow from
the eye, regional lymph node metastases are rare and can primarily be seen in cases of extra
ocular extension of the tumour (17). Intratumoural lymphatics seen in some UM with
extraocular extension are hypothesized to be recruited from conjunctival lymphatics (18). The
presence of metastases and lymphatic spread impairs prognosis significantly (16). Cell type is
assessed during routine histopathological examination and is an important prognostic indicator.
UMs are classified into three different subgroups according to cell morphology, namely
Spindle, Epithelioid and Mixed tumours (Figure 2). Spindle celled UMs are characterised by
an elongated nucleus and can be further divided into Spindle A and Spindle B cells. Spindle A
nuclei lack nucleoli, while nucleoli is a characteristic feature of Spindle B nuclei. Epithelioid
tumours resemble epithelium cells with eosinophilic cytoplasm, polygonal shape and prominent
13
nucleoli. The definition of mixed tumours is elusive, a proposed definition is the presence of
minimum 10% of Spindle or Epithelioid cells, most often the presence of >10% Epithelioid
cells (14). Additional prognostic factors include tumour size, mitotic activity, status of
extrascleral extension, mean diameter of the ten largest nucleoli, presence of mitotic figures,
presence of lymphatic infiltrates and architecture of the microcirculation (4, 19). Important
chromosomal abbreviations will be addressed more thoroughly in section “4.2 Genetic
determinants in UM”.
Figure 2: Histopathology of Uveal melanoma: (A) Epithelioid cells and (B) Spindle cells.
5.1.4 Treatment
Small tumours (height < 2mm and diameter < 4mm) detected by routine examination can often
be observed by regular fundus screening examinations. If the tumour shows signs of growth,
intervention should be considered (4). For many years, enucleation was considered the sole
method of treatment for larger UMs. Today, the advances in the field of radiotherapy have
greatly increased the possibility of preserving the eye. The Norwegian Health Council
recommends enucleation in cases where the tumour is large, has a substantial extrascleral
outgrowth or encircles more than 180 degrees of the optic nerve. Transscleral local resection is
a surgical option primarily for tumours localised anteriorly and is commonly followed by
brachytherapy. This method is used for patients who are not candidates for radiation therapy,
but are highly motivated to retain their eye. Plaque brachytherapy is the most widely used
14
radiotherapy and offers a targeted delivery of radiation to the tumour. A small plate (plaque)
containing seeds of ruthenium (Ru-106) or iodide (I-125) is attached episclerally of the lesion.
Another method, less commonly used, is proton beam therapy. This method can be used on
larger tumours and on tumours localised closer to the optic disc/fovea. Proton therapy is
currently not available in Norway, though selected patients can be treated abroad. Additional
methods of treatment exist, these are rarely used as monotherapy/ curative treatment (20).
5.1.5 Management of metastatic disease
Despite advances in treatment of primary UM, the mortality rate has remained largely
unchanged (4). Few treatment options exist for patients who develop metastatic disease. Local
resection of liver metastases either by surgery or by stereotactic radiosurgery is reported to
prolong survival. Unfortunately most patients present with diffuse involvement of the liver and
will therefore not qualify for surgical resection. Isolated liver perfusion with melphalan, and
hypothermia have been tested for patients with multiple liver metastases. Additional treatment
strategies include chemoembolization or radioembolization with Yttrium 90 microspheres (4).
Over the last decade proto-oncogene B-Raf (BRAF), mitogen-activated protein kinase kinase
(MEK) and checkpoint inhibitors have revolutionised the treatment of patients with cutaneous
melanoma (CM). Most UM are not sensitive to BRAF inihibitors since they don’t harbour
BRAF mutations (21). The use of MEK inhibitors is also questionable (22). Immunotherapy
with immune checkpoint inhibitors (ICI) such as anti-cytotoxic T lymphocyte antigen (CTLA)-
4 (ipilimumab) or anti-programmed cell death protein (PD)-1 antibodies (pembrolizumab or
nivolumab) has shown some effect. In a retrospective study, a partial response to first-line
treatment was observed in 7% of patients treated with anti-PD-1 monotherapy and in 21% of
those treated with combined anti-CTLA-4 plus anti-PD-1 therapy. The estimated one-year
overall survival rate increased from 25.0% to 41.9% and the median overall survival improved
from 7.8 months to 10.0 months (23).
15
5.2 Genetic determinants in uveal melanoma
5.2.1 Cytogenetic features
Chromosomal aberrations are important determinants for metastatic spread. The loss of
chromosome 3 is considered one of the most valuable prognostic markers. Monosomy 3 (M3)
is associated with decreased survival and the presence of risk factors such as large tumour
diameter, epithelioid cell type and extraocular extension (14, 24). Partial deletions of one copy
of chromosome 3 and isodisomy also correlate with metastatic disease (25, 26). Gain of 8q
(trisomy 8, isochromosome 8q and amplification of the c-Myc gene), in addition to M3, greatly
impairs prognosis. The five-year disease specific mortality rate for M3 tumours is 40%. The
co-existence with 8q gain increases the mortality rate to 66% (27). The loss of a part or all of
chromosome 1 is another factor contributing to poor outcome and occurs more frequently in
M3 tumours (24, 28). The loss of chromosome 6q is also associated with poor prognosis. In
contradistinction to the aforementioned chromosomal changes, the gain of chromosome 6p is a
predictor for better prognosis and is rarely seen together with M3 (estimated coexistence of 4%)
(29). Abnormalities in the q-arm of chromosome 16 are relatively common in UM, though not
associated with survival or other cytogenetic/ histopathological parameters (24).
A summary of the percentages of the main chromosomal aberrations in UM from various
studies reviewed by Dogrusöz et al are shown in Table 1 (30).
Loss of 1p Monosomy 3 Gain of 6p Gain of 8q
Range 19-34% 25-65% 18-54% 37-63%
Table 1. Frequency of common chromosome alterations with evident prognostic significance. Summary
of studies reviewed by Dogrusöz et al 2017
5.2.2 Molecular pathways and genomic alterations in UM
Despite their common embryological origin, the genetic characteristics of UM differs greatly
from those seen in cutaneous melanoma. UM lacks mutations typically associated with CM and
has a low mutational burden (31).
16
Activation of the G protein subunit alpha 11/Q (Gα11/Q) pathway is important in early UM
development and occurs in almost all primary UM via single amino acid substitution mutations
in G protein subunit alpha Q (GNAQ) (57%) and G protein subunit alpha 11 (GNA11) (41%)
(32-34). GNAQ and GNA11 encode two closely related G-alpha subunits that are components
of G protein coupled receptor proteins (GPCR). GPCR receptors encompass numerous
physiological functions and are critical in tissue homeostasis and cellular proliferation (35, 36).
Mutations in GNAQ and GNA11 are not sufficient for malignant transformation alone (33, 37).
Primary tumours that do not harbour mutations in GNAQ or GNA11 usually have mutations in
the Gα11/Q pathway associated genes Cysteinyl leukotriene receptor 2 (CYSLTR2) and
Phospholipase C β4 (PLCB4). CYSLTR2 encodes a G-protein coupled receptor and is
constitutively activated in 4% of primary UM. PLBC4 activates signalling downstream by
directly binding Gαq and is activated in 2.5%–4% of primary UM (31, 38, 39).
Mutated Gα proteins mediate the activation of the phospholipase C (PLC)/ protein kinase C
(PKC) pathway and multiple downstream signalling pathways, including the rapidly
accelerated fibrosarcoma (RAF)/MEK/ extracellular-signal-regulated kinase (ERK) pathway.
In addition to phosphoinositide 3-kinase (PI3K)/ AK strain transforming serine/threonine
kinase (AKT)/ mammalian target of rapamycin (mTOR), and triple functional domain protein
(Trio)/Ras homologous (Rho)/ Ras-related C3 botulinum toxin substrate (Rac)/yes-associated
protein 1 (YAP1) pathway (40). Activation of the mitogen-activated protein kinase (MAPK)
cascade is seen in up to 86% of primary UM (41, 42). The activation of PLC and subsequent
cleavage of phosphatidylinositol diphosphate (PIP2) into inositol triphosphate (IP3) and
diacylglycerol (DAG) results in activation of protein kinase C (PKC). PKC activates the MAPK
pathway via targets including RAF, MEK and ERK, and results in transcription of genes
involved in proliferation, differentiation and cell survival (Figure 3). The downstream
activation of MEK has stimulated the testing of MEK inhibitors in the treatment of UM (22).
The PI3K/ AKT pathway is highly activated in many cancers and has been shown to promote
proliferation and reduce apoptosis (Figure 3). AKT is a serine/ threonine kinase and is activated
by phosphorylation. The phosphorylated AKT can further inactivate proteins involved in
apoptosis and its expression correlates with poor prognosis in UM (43, 44). Phosphatase and
tensin homolog (PTEN) acts as a tumour suppressor by negatively regulating the AKT/ protein
kinase B (PKB) signalling pathway (45). Loss of heterozygosity of at least one PTEN marker
has been demonstrated in 76% of primary UM, in addition loss of cytoplasmic PTEN expression
17
is associated with cancer relapse (46). Downregulation of PTEN is suggested to be a late event
in tumour progression due to its association with increased aneuploidy (29).
Gαq/11 signalling also promotes the activation of the Trio/Rho/Rac/YAP1 pathway (Figure 3).
YAP is hypothesized to promote the transcription of transcription factors associated with cell
growth and viability and is a proposed therapeutic target (47).
In the majority of UMs, the p53 and retinoblastoma (Rb) pathways are functionally inhibited,
although mutations in the TP53 and RB1 are rare (48, 49). Both pathways serve as tumour
suppressors. The Rb protein prevents the cell from replicating damaged deoxyribonucleic acid
(DNA) and can induce growth arrest in the G1 phase (50).
Figure 3: Oncogenic signalling pathways in UM. G-protein coupled receptors (GPCR) signal through
the heterotrimeric proteins, Gα and Gβγ. Mutations in GNAQ or GNA11 lead to constitutive activation
of Gα and downstream stimulation of the mitogen-activated protein kinase (MAPK)
pathway via phospholipase C (PLCβ) and protein kinase C (PKC). The phosphotidylinositol-3 kinase
(PI3K)/AKT/mTOR and the Yes-activated protein (YAP) pathways are also activated. Adapted from
Park et al 2018 and Yang et al 2018.
18
In addition to mutations in GNAQ, GNA11, PLCB4 and CYSTLR2, UM is characterised by
mutations in three secondary driver genes; BRCA1 associated protein 1 (BAP1), Splicing factor
3B subunit 1 (SF3B1) and Eukaryotic translation initiation factor 1A, X-linked (EIF1AX).
BAP1
The strong association between M3 in UM and metastatic disease suggested that one or more
tumour suppressor genes were located on chromosome 3. In 2010 Harbour et al discovered that
BAP1, located at chromosome 3p21.1, was mutated in 47% of UM and in 85% of metastatic
UM (51). BAP1 belongs to a specific group of proteases, called deubiquitinating enzymes
(DUB). BAP1 is a catalytic component of the Polycomb repressive deubiquitinase complex
(PR-DUB) that mediates deubiquitination of histone H2A monoubiquitinated at 'Lys-119'
(H2AK119ub1), thus antagonising the activity of polycomb repressive complex 1 (PRC1) (52,
53). The exact role of BAP1 in gene regulation is still enigmatic. In healthy cells BAP1 removes
ubiquitin from H2AK119 and thereby releases repression of transcription (54). Additionally,
BAP1 binds and deubiquitinates the Transcriptional regulator host cell factor (HCF-1) (55, 56).
HCF-1 regulates gene expression by serving as a scaffold for chromatin remodelling complexes
and by binding to several transcription factors (57, 58).
Enhancer of Zeste Homolog 2 (EZH2) is the enzymatically active core subunit of the Polycomb
repressive core 2 complex (PRC2). PRC2 methylates the lysine residue at position 27 of histone
3 (H3K27), which facilitates chromatin compaction and gene silencing (59). EZH2 is opposed
by the switch/sucrose non-fermentable (SWI/SNF) multiprotein complex. The SWI/SNF family
of chromatin remodelling complexes serve to either enhance or suppress gene transcription
through mobilization of nucleosomes (60). As cells differentiate, EZH2 activity is increasingly
opposed by SWI/SNF, thus facilitating gene expression and terminal differentiation (61). EZH2
has the capacity to silence tumour suppressor genes and micro ribonucleic acid (microRNAs),
but can also function as a gene activator (62, 63). The overexpression of EZH2 due to aberrant
activation of EZH2 or loss-of-function mutations in the SWI/SNF complex is associated with
cancer aggressiveness and advanced disease (64, 65). Loss of BAP1 function has previously
been shown to increase EZH2, thus leading to EZH2 -dependent transformation (66). A recent
study assessed whether EZH2 deletion could restore expression of BAP1-regulated genes.
Deletion of EZH2 in cells already depleted of BAP1 did not impair proliferation. Of the genes
downregulated in BAP1 depleted cells, most of them remained silent in the EZH2/BAP1 double
19
knockout model, the small set of genes upregulated in the double knockout model was also
upregulated in BAP1 positive/ EZH2 negative cells. BAP1 promotes gene expression in a
manner that is largely independent of an antagonism with the PRC2 complex (53).
A summary of BAP1 functions and interacting protein partners is presented in Figure 4 (67).
Figure 4: Summary of the functional roles of BAP1. BAP1 regulates the DNA damage repair pathway
through interactions with BRCA1, BARD1 and RAD51. BAP1 interact with HCF1 in a number of
processes involved in cell-cycle control and proliferation. BAP1 binds to ASXL to form the PR-DUB
complex, responsible for regulation of chromatin through Histone H2A deubiquitination. BAP1 is
associated in a number of regulated cell death pathways including apoptosis and ferroptosis. BAP1 is
implicated in immune regulation. Courtesy of Louie et al 2020.
20
SF3B1
SF3B1 encodes a core component of the ribonucleic acid (RNA) splicing machinery, the
spliceosome processes precursor messenger RNA (mRNA) into mature transcripts and is
located on chromosome 2q33. SF3B1 ensures correct splicing by retaining pre-mRNA to define
the site for splicing, thus mutations in this gene can result in unique aberrant proteins but also
in loss of expression (68, 69). Mutations in SF3B1 are detected in approximately 15% of UM
cases (70). SF3B1 mutations are mainly restricted to tumours without M3, and are associated
with late metastatic spread (34).
EIF1AX
The EIF1AX gene is located on chromosome Xp22 and approximately 17% of UMs harbour
this mutation (31, 71, 72). EIF1AX has a role in initiating translation through a combination of
stabilisation of the ribosome and recognition of target mRNA thus preparing mRNA for
translation (73). EIF1AX encodes the eukaryotic translation factor 1A (eIF1A) that is essential
in the transfer of methionyl initiator transfer ribonucleic acid (tRNA) to the small (40s)
ribosomal unit (74). EIF1AX mutations are inversely associated with metastatic disease, most
mutations are identified in tumours with disomy 3 (D3) (48%) and are rare in M3 tumours (3%)
(71, 72).
5.2.3 Binary clustering of uveal melanomas
In 2004 Onken et al presented subclustering of UM into two distinct molecular classes based
on gene expression profile. The division into Class 1 (1a low-grade tumours, 1b low-grade with
metastatic potential) and Class 2 (high-grade tumours) was strongly correlated with cytological
severity and survival (75). A significant association between genes expressed in Class 2
tumours and those expressed in primitive ectodermal and neural stem cells has also been
demonstrated (51, 76). In 2010 the gene expression findings were commercialised as the
DecisionDx-UM GEP test. The test migrates the initial findings into a 15-gene quantitative
polymerase chain reaction (qPCR) assay with 12 discriminating genes and 3 control genes, and
is claimed to be superior to assessment of M3 and clinicopathological prognostic factors for
predicting metastasis (Table 2) (77-80).
21
Table 2: Summary of the 15 genes tested in the DecisionDx-UM GEP test.
Gene symbol Gene name
Upregulated in Class 2 uveal melanoma
CDH1 E-cadherin
ECM1 Extracellular matrix protein
HTR2B 5-Hydroxytryptamine (serotonin) receptor 2B
RAB31 RAB31, member RAS oncogene family
Downregulated in Class 2 uveal melanoma
EIF1B Eukaryotic translation initiation factor 1 B
FXR1 Fragile X mental retardation, autosomal homolog 1
ID2 Inhibitor of DNA binding 2
LMCD1 LIM and cysteine-rich domain
LTA4A Leukotriene A4 hydrolase
MTUS1 Microtubule-associated tumour suppressor 1
ROBO1 Roundabout, axon guidance receptor, 1
SATB1 SATB homeobox 1
Control genes
MRPS21 Mitochondrial ribosomal protein S21
RBM23 RNA-binding motif protein 23
SP130 Sin3A-associated protein, 130kDa
The cost-benefit of the commercial test is a subject of debate. A prospective, 5-year multi-
centre study has shown that Class 1A offers a 2% chance of the UM spreading over the next
five years. Class 1B has a 21% chance of metastasis over five years, while Class 2, high risk
UM, has 72% chance of metastasis within five years (81). Regardless of test results, long-term
follow-up is of importance since metastatic disease is often seen within the first 10 years after
diagnosis and can also be seen more than 25 years after treatment of the primary tumour (12).
It should also be noted that it is possible to receive both Class 1 and Class 2 test results in the
setting of a non-melanoma malignancy, thus histopathology should be performed in addition to
the DecisionDx-UM GEP test for correct diagnosis (82). The discovery of BAP1 loss/ mutation
in aggressive UMs has raised the question whether immunohistochemistry (IHC) could be more
cost effective since it can easily be implemented as a routine staining at Pathology Departments
22
(83). Intratumoural heterogeneity and sampling errors are possible IHC drawbacks, though this
has also been shown for the DecisionDx-UM GEP test (84).
A modified version of the DecisionDx-UM GEP test that includes preferentially expressed
antigen I melanoma (PRAME)) is also commercially available. PRAME has been shown to be
an independent biomarker for metastasis in UM. PRAME positivity is associated with an
increased risk of metastasis in Class 1 tumours and a shorter time to metastasis in Class 2
tumours. If a tumour is negative for PRAME, the prognosis indicated by the DecisionDx-UM
Class is not expected to be altered (85, 86).
More recently, Class 1 and Class 2 tumours have been further divided into the subcategories A-
D (87). The subdivision was based on data from the Cancer Genome Atlas, where primary
tumour material from 80 patients with UM have been analysed for histologic features,
chromosome copy number, genetic mutations, expression of RNA, proteins, DNA methylation
status, in addition to factors such as biochemical pathways and immune markers (39).
A summary of driver and secondary genetic alterations in UM development and progression is
shown in Figure 5, included in the figure is serine and arginine rich splicing factor 2 (SRSF2)
associated with Class 1b tumours (40).
Figure 5: Acquisition of driver and secondary genetic alterations drive uveal melanoma (UM)
development and progression. The sequential acquisition of genetic changes (highlighted within the
vertical arrows) leads to distinct genetic profiles that reflect the risk of UM metastases. Courtesy of Park
et al 2018.
23
5.2.4 Genetic alterations in metastatic UM
The emergence of metastatic disease in UM can be seen months to decades after primary
surgery and the latency period could reflect the time needed to acquire distinct oncogenic
alterations (88).
Kiilgaard et al performed DNA sequencing of 35 primary UM and matched metastases (89). In
contradistinction to several other cancers, the metastases of UM tend to have more oncogenic
mutations than primary UM (90). The study showed that copy number (CN) changes of 6p, 1q
and gains of 8q were enriched in metastases. This was in concordance with previous
publications detecting chromosome 3 monosomy (73%), 8q gain (89%), 6q loss (64%), 1p loss
(47%), 8p loss (45%), 1q gain (35%), and 16q loss (32%) in liver metastases (91, 92). The
amplitude of 8q tended to increase from primary tumour to metastases. As expected in
metastatic disease, the number of UM with SF3B1 or EIF1AX mutations was low (n=7). These
cases showed additional oncogenic alterations in e.g. CDKN2a and PTEN. In one EIF1AX
mutant case a part of the tumour had acquired a deletion of chromosome 3. Mutations in
chromatin remodelling factors were also observed, including mutations in polybromo 1
(PBRM1) and EZH2.
5.3 Epigenetics
Epigenetics is a rapidly developing field in clinical medicine and biomedical research, and is
considered to be one of the hallmarks of cancer (93-95). An epigenetic trait is the constantly-
heritable phenotype resulting from changes in a chromosome without alterations in the DNA
sequence (96). At least three types of epigenetic modifications regulate chromatin
conformation: DNA methylation, histone modifications, and non-coding RNAs. Histone
modifications are posttranslational modifications of histone proteins which includes
methylation, acetylation, citrullination, SUMOylation, phosphorylation, ADP-ribosylation and
ubiquitination (97). Histone modifications can activate or silence transcription by controlling
the accessibility of DNA to the transcriptional machinery and by protein interactions (98, 99).
Non-coding RNAs (ncRNAs) function to regulate gene expression at the transcriptional and
post-transcriptional level and can be divided into two main groups, namely short ncRNAs (<30
nucleotides) and long ncRNAs (>200 nucleotides). Micro RNAs (small (≈22 nucleotides),
single stranded, non-coding RNAs) are among the most studies ncRNAs and can repress gene
expression by binding to complementary sequences of mRNA thereby preventing their
24
translation. MiRNAs can drive tumorigenesis by overexpression of oncogenic miRNAs
(oncomirs) or by loss of tumour suppressor miRNAs (100, 101).
Differential DNA methylation in UM was the subject of interest in paper II and the process of
DNA methylation will therefore be highlighted in section 5.3.1.
5.3.1 DNA methylation
In mammalian cells, DNA methylation occurs almost exclusively at the C-5 position of cytosine
(5mC) in cytosine-phosphate-guanine (CpG) nucleotides. The majority of CpG nucleotides in
the genome are methylated and most of the methylated CpGs are located in regions with low
density of CpGs. Of the approximately 28 million CpG sites present in the human genome, 60-
80% of the cytosines are methylated as 5mC (102). Regions of the genome that are enriched in
CpG repeats are referred to as CpG islands (CGIs). A CGI is defined as a region of DNA > 200
base pairs with a GC content ≥50%, and the ratio of observed/expected CpG >0.6 (103). CGIs
are present in or near approximately 40% of gene promoters (104). Although the bulk of
genome is methylated at its CpGs, CGIs are mostly unmethylated in somatic cells (105). In
general, hypermethylation of promoters is associated with gene silencing, while methylation of
gene bodies is often a permissive mark (106, 107).
The methylation of cytosines is mediated by a class of enzymes called DNA methyltransferases
(DNMTs) and involves the transfer of a methyl group from S-adenosyl-methionine (SAM).
Five members of the DNMT family have been identified, but only three possess an inherent
enzymatic activity (DNMT1, DNMT3a, DNMT3b). DNMT1 is a maintenance
methyltransferase, ensuring the methylation status of each CpG during replication (108, 109).
DNMT3a and DNMT3b are essential for de novo methylation and mammalian development
(109). DNA methylation is assumed to interfere with transcription by either physiologically
impede the binding of transcriptional proteins to the gene or by recruiting methyl-CpG-binding
domain proteins (MBD) to methylated DNA. MBD proteins can further recruit proteins
involved in chromatin remodelling and induce conformational changes and silencing (110,
111).
In the absence of functional DNA methylation maintenance machinery, 5mC can be lost during
successive rounds of replication, thus leading to passive DNA demethylation. By contrast,
active DNA methylation refers to an enzymatic process that removes or modifies the methyl
group from 5mC (112). Active demethylation by oxidation is achieved by Ten-eleven
25
translocation (TET) -enzymes (TET1, TET2, TET3). These enzymes convert unmodified 5mC
to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC),
followed by excision of 5fC or 5caC mediated by thymine DNA glycosylase (TDG) coupled
with base excision repair (BER) (113). Another proposed mechanism for demethylation of 5mC
involves the deamination of 5mC and 5hmC by the deaminase enzymes activation-induced
cytidine deaminase (AID)/ apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like
APOBEC (114). An illustration of DNA methylation addition, maintenance and removal is
presented in Figure 6 (115).
Figure 6: DNA methylation predominantly occurs at the fifth carbon atom of cytosine bases. Its
deposition is catalysed by the de novo DNMTs DNMT3A and DNMT3B. Introduced methylation
patterns are preserved by the maintenance DNMT DNMT1 during replication. Passive DNA
demethylation is considered to be achieved across cell division in the absence of DNMT1 maintenance
activity. Active removal includes the mammalian TET1–3 proteins that are capable of converting 5-
methylcytosine to its oxidised derivative 5-hydroxymethylcytosine (5hmC) and further to 5-
formylcytosine and 5-carboxylcytosine (not indicated here). Courtesy of Ambrosi et al 2017.
Cancer cells display an aberrant methylation pattern recognised by global hypomethylation and
hypermethylation of promoter associated CpGs (116, 117). The global DNA hypomethylation
in cancer is mainly due to hypomethylation of highly repetitive DNA sequences e.g. short
interspersed nucleotide elements and long interspersed nucleotide elements (118-120). The
dense methylation of these regions as seen in normal tissue presumably maintains genomic
integrity by preventing translocations, genomic disruptions and genomic instability (117, 121-
124). The aberrant hypermethylation of promoters seen in cancer, is often associated with the
26
silencing of tumour suppressor genes such as BRCA1 and Von Hippel-Lindau tumour
suppressor (VHL) (106, 125, 126). Rarely, promoter methylation can also serve as a permissive
mark (127). The exact role of promoter methylation in gene silencing is a subject of discussion
and could be a late event in gene silencing, secondary to nucleosome positioning (128-130).
Even though the understanding of promoter methylation in gene silencing is in its beginning,
the link between cancer and methylation is undeniable. The lifetime risk of cancer is correlated
to the degree of abnormal methylation changes that occur during the ageing of normal tissue.
Tissue where the cells have a relatively high degree of abnormal methylation (e.g. colon) has a
higher propensity for developing cancer than cells characterised by a lower degree of aberrant
methylation (131).
Despite controversy over the role of promoter methylation in gene silencing, DNMT inhibitors
have shown promise in the treatment of various cancers, especially haematological cancers
(132-134). The differential effect of DNMT inhibitors in diverse cancer could be due to their
heterogeneous methylation pattern (135). It should be noted that DNMT inhibitors might exert
their effect by other mechanisms than promoter demethylation (107, 136). Intriguingly a
DNMT inhibitor prodrug has been shown to have the ability to up-regulate HLA class 1
antigens, thus indicating a potential in increasing immunogenicity and immune recognition of
neoplastic cells (137).
Relatively few studies have investigated the methylome of UM, though hypermethylation has
been shown in areas associated with promoters for genes regulating the cell cycle, and
extracellular matrix degradation, e.g. p16INK4a, RASSF1a, RASEF, Embryonal fyn-associated
substrate (EFS) and Metalloproteinase inhibitor 3 (TIMP3) (138-142).
The global methylation profile of UM has been shown to coincide with clustering into Class 1
and Class 2 tumours (39, 143). Robertson et al showed that EIF1AX-mutant tumours were
restricted to DNA methylation cluster 1, while UM in DNA methylation clusters 2 and 3 were
highly enriched with tumours harbouring SF3B1/SRFR2 mutations. Thus, D3 UM with EIF1AX
versus SF3B1/SRFR2 mutations possess distinct DNA methylation patterns. Monosomy 3
(M3)/BAP1-aberrant UM tumours showed a single global DNA methylation profile (39).
Interestingly partial deletion of chromosome 3 is associated with low- risk Class 1 UMs (144),
thus raising the question why complete loss of chromosome 3 is required for Class 2 GEP. This
was further investigated by Harbour et al who showed that the most significant and densely
clustered hypermethylated/ downregulated gene loci in Class 2 UMs were located on
chromosome 3, which contained many of the axon guidance cues, neural crest specification,
27
and melanocyte differentiation genes (e.g., Roundabout homolog 1 (ROBO1), Plexin B1
(PLXNB1), Semaphorin-3B (SEMA3B), Cell adhesion molecule L1 like (CHL1), Special AT-
rich sequence-binding protein-1 (SATB1), Microphthalmia-associated transcription factor
MITF, Dishevelled Segment Polarity Protein 3 (DVL3), and Rapidly accelerated fibrosarcoma
-1 Proto-Oncogene, Serine/Threonine Kinase (RAF1). Since these genes undergo repressive
methylation changes on the sole remaining copy of chromosome 3, it could explain why the
other copy of chromosome 3 must be lost to acquire the metastasising Class 2 UM phenotype.
Additionally a novel hypermethylated site within the BAP1 locus was found in all Class 2
tumours, suggesting that BAP1 itself is epigenetically regulated (145).
Characterisation of the methylome provides mechanistic insight into the development and
progression of UM and could lay the foundation for the development of new therapeutics.
Methylation profiles can also serve as diagnostic and prognostic markers in addition to
predicting responsiveness to therapy and monitoring of response (146-152).
5.3.1.1 DNA hydroxymethylation
The conversion of 5mC to 5hmC by TET enzymes has gained considerable attention as 5hmC
has been shown to be a relatively stable epigenetic mark whose role in transcription regulation
is linked to its genomic location (153). The 5hmC levels vary between different cell types and
tissues and are highest in neurons, while cancer cells have lower levels compared to
corresponding normal tissue (154, 155). Once cancer is formed, a lower level of 5-hmC
correlates with poor prognosis (156, 157). 5hmC has a greater relative abundance in gene bodies
compared to gene promoters, where 5hmC modified CpGs are generally depleted. An
enrichment of 5hmC CpGs over enhancer elements and some transcriptional start sites is
associated with silenced genes, while gene body methylation is often associated with active
genes (158-162).
5.4 In vitro and in vivo preclinical models for studying UM
In vitro and in vivo models play a pivotal role in basic and translational cancer research and are
important tools to investigate the pathogenesis of metastatic UMs and for drug testing.
Two-dimensional (2D) monolayer cultures of primary tumour cells and cell lines are
indispensable in UM research and allow for expansion of cells and drug testing under direct
28
visualisation. At the same time, these cultures represent non-physiological conditions that
might not be representative for cancer cells residing in the complex microenvironment of
primary tumours and metastatic niches.
Three-dimensional (3D) cultures are hypothesised to recapitulate in vivo growth including cell
connectivity, polarity, tissue architecture and gene expression (163). That said, 3D cell culture
models are often more time consuming, difficult to implement in standard workflows and often
pose a challenge in imaging and quantitative analyses (164).
5.4.1. Three-dimensional in vitro models
Broadly, 3D cell cultures are classified as Scaffold-based (cells grown in presence of a support)
and Scaffold-free techniques.
5.4.1.1 Scaffold based techniques
Scaffolds used for 3D cell cultures range from extra cellular matrix (ECM)-like matrices to
simple mechanical structures and can be further divided into hydrogels and solid-state scaffolds.
Hydrogels are water swollen polymeric material and include natural hydrogels (e.g. agarose,
laminin, collagen, hyaluronic acid) and synthetic hydrogels. Matrigel is an example of a natural
hydrogel and is a solubilised basement membrane preparation extracted from the Engelbreth-
Holm-Swarm mouse sarcoma and provides ECM protein such as laminin, collagen IV,
proteoglycans and a number of growth factors (165). Solid state scaffolds have the ability to
organise positioning of cells in a reproducible and controllable manner (166).
5.4.1.2 Scaffold-free techniques
Scaffold-free 3D cultures facilitate the formation of spheroids (multicellular aggregates) (167).
The formation of spheroids relies on either forced or self-assembled clustering of cells.
Hanging drop cultures involves the culturing of cells in a drop of media suspended in the lid of
a cell culture dish, meaning that the drop has to be small enough to adhere to the lid under
manipulation. Aggregation can also be promoted using plates with low attachment coating (low
adhesion plates), these plates have a higher volume capacity than the hanging drop method and
often result in the formation of one spheroid per well. An additional technique is magnetic
29
levitation where cells are preloaded with magnetic nanoparticles and further aggregated using
an externally applied magnetic field. Spheroids can also be generated with the aid of bioreactors
e.g. spinner flasks and rotational bioreactors. Bioreactors provides greater reproducibility and
can produce a larger number of uniform spheroids. Another way to introduce flow to cell culture
systems is through the use of microfluidic devices. These devices contain micro-channels, thus
allowing for continuous delivery of nutrients and the creation of gradient concentrations of
biochemical signals (168, 169).
5.4.2 In vitro assays for studying the metastatic process of UM
The dismal outcome of metastatic UM urges the need for validated models to study the
mechanisms controlling metastasis. The sequential steps of metastasis include the degradation
of ECM, intravasation into blood vessels, circulation within the bloodstream, attachment to the
endothelium of a target organ and the extravasation into connective tissue before proliferation
(170). The complex process of metastasis renders a uniform model unlikely since a thorough
understanding of every step is needed. Several in vitro models for studying metastasis have
been developed, each with its strengths and limitations.
Migratory and invasive capacity are prerequisite skills for metastatic spread. Boyden chamber
assay and its modifications can be used to study invasion, chemotaxis (migration towards a
chemical concentration gradient) and haptotaxis (ECM protein gradient) of tumour cells. The
standard Boyden chamber assay involves the seeding of cancer cells on top of a transwell
membrane suspended over a larger well which contain medium/ chemoattractants. Cells are
allowed to migrate through the porous membrane before migratory cells are stained and
counted, modifications involve the addition of e.g. Matrigel on top of the membranes and
addition of feeder layers. Migratory cells can be detected and quantifies by both colorimetric
and fluorometric methods (171).
A simple and well-developed method to assess cell migration is the scratch assay. This method
introduces a “scratch” in a monolayer cell culture and images are captured at the beginning and
at regular intervals. Images are then compared to quantify the migration rate of the cells. The
Ring assay uses the same concept as the scratch assay, cells are allowed to grow to confluency
before a central ring is removed, allowing cells to migrate into this area. Cell migration can also
be evaluated by microcarrier bead assay, where cells are grown on microcarriers before being
30
transferred to plastic wells, migration to the plastic wells is assessed after removal of the beads
(171, 172).
The use of the previous mentioned microfluidic devices is emerging as these small chambers
enables control over local gradients, fluid flow, tissue mechanics, and composition of the local
environment. Additionally, these chambers can be made optically accessible for live
observation. Microfluidic devices could represent a valuable substitute for animal models in
preclinical trials (173).
5.4.3 In vivo assays for studying UM
Great advancements in medical research have been attributed to the use of animal models and
these models are still valuable tools in cancer research. The use of animal models raises several
ethical questions and the three R`s (Replacement, Reduction and Refinement) should always
be taken into consideration (174). In cancer research, a well-designed animal model can provide
insight into basic pathobiology and the process of metastatic spread. Testing of novel
therapeutics also rely in these models, as animal models represent a bridge between in vitro
research and clinical trials.
Animal models can be divided into spontaneous models, transgenic models and induced
models. The relatively low incidence of UM, even in animals, limits the use of spontaneous
models.
The genetic engineering of transgenic animal models allows oncogenes to be constitutively or
conditionally expressed and tumour suppressor genes to be silenced (175). In cutaneous
melanoma, numerous transgenic mouse models have been successfully established. Attempts
in developing transgenic models in UM have been undertaken, including the development of a
GNAQ mutant mouse strain, unfortunately these models have failed to develop liver metastases,
thus no transgenic models are currently available for UM (176, 177).
Induced animal models involves the artificial introduction of disease by radiation, chemical
agents, viruses, cells or tissues. Several induced animal models exist, including intraocular,
intrasplenic, intravenous and intrahepatic injections of tumour cells in addition to patient
derived xenografts. It should be noted that the ability to grow metastases is tumour dependent,
and that these models rely on immunocompromised animals (171, 176, 177).
31
6. Aims of the thesis
The overall aim of the thesis is to shed light upon underlying mechanisms in the development
and progression of UM, thus unravelling potential treatment strategies and improve prognostic
assessment.
More specifically:
1) In the first paper, our aim was to compare the differential gene expression of
multicellular tumour spheroids (MCTS) of UM to primary tumour tissue and adherent
cultures, with a special emphasis on unravelling the pathways and survival mechanisms
pathognomonic for disseminated and circulating cancer cells.
2) In the second paper, we sought to delineate biologically relevant groups and genes in
FFPE derived UM specimens by correlating histopathological data and survival data of
the patients with methylation profiles and gene expression.
3) The aim of the third paper was to investigate the differential expression of Cx43 in
primary UM biopsies and cultures vs healthy choroidal tissue and choroidal
melanocytes and explore potential regulatory mechanisms.
32
7. Methods and methodological considerations
All experiments were performed in accordance with the Declaration of Helsinki. Both tissue
harvesting and the use of archived paraffin embedded tissue blocks were approved by the local
Committees for Medical Research Ethics. Fresh tissue samples were obtained after informed
written consent.
7.1 In vitro cultivation
Various procedures for preparation of single cell suspensions from tumour tissue exist. The rate
of success is determined by dissection procedure, tissue quality, method of separation
(enzymatic or filtration). The use of enzymatic digestion is dependent on enzyme used,
concentration, temperature, and length of incubation. The most widely enzymes used include
trypsin, collagenase, dispase, hyaluronidase, papain and elastase. Other commercially available
solutions include Accutase (Innovative Cell Technologies, Inc., San Diego, US) and TrypLE
(Thermo Fisher Scientific, Waltham, US), enzymes that allegedly cause less damage than
trypsin. Our method of choice was based on personal experience and published literature. Other
groups have had success with non-enzymatic separation of tumour tissue (filter, cloth, mincing)
due to small sample size we preferred an enzymatic approach (178). In our experience
dissociation in 0.25% Trypsin digestion often resulted in an overgrowth of fibroblasts or cell
death in adherent cultures (179). In the first paper, UM tissue was obtained from patients
undergoing enucleation of the eye. After surgery, the eye was transferred to a 0.9% NaCl and
transported to the Pathology Department where an Ophthalmological Pathologist excised a
small portion of the tumour for research purposes. The tissue was minced into small pieces in
a mixture of 1mg/ml of collagenase I and IV, before incubation for 1 h at 37C. The pellet was
resuspended in RPMI 1640, 10% FBS, 0.5% Penicillin/Streptomycin and 0.25% Amphotericin
B. Gentamycin 75µg/ml was added to ensure removal of fibroblasts from the cell culture (180).
This protocol was kindly provided by Tina Maria Ludowika Jehs from the University of
Copenhagen and was originally intended for isolation of uveal melanocytes. In our experience,
the protocol results in a homogenous cell culture viable for 1-3 passages if the quality of the
starting material is satisfactory (data not shown). In addition we tested a neuronal dissociation
kit from as described by Tura et al, our results indicated increased cell yield using this method
33
(181). Unfortunately the miniscule samples we obtained for research purpose did not allow for
parallel testing of the isolation protocols.
RPMI 1640 was chosen over DMEM/F12, as the latter was seemingly more favourable for
fibroblasts (182). Alpha-MEM is proposed to be superior to RPMI1640, implementation of this
media to our protocol could thus optimise culture conditions further (183).
For establishment of spheroid cultures, cells were grown for 7 days as adherent cultures before
trypsinisation in a 0.25% solution under careful supervision. The cells were resuspended in
hESC +MEF, 0.5% Penicillin/Streptomycin and 0.25% Amphotericin B and transferred to
ultra-low attachment plates (Corning, Sigma-Aldrich, St. Louis, Missouri, United States). This
media was chosen based on studies on skin melanomas (184). The use of cone shaped wells
and the fact that the UM cells retained their pigmentation enabled us to change ¼ of the media
every second/third day without disturbing the cells.
In the third paper we used a modified protocol developed by co-author Kirankumar Katta. This
protocol implemented elements from isolation protocols of choroidal melanocytes. Briefly, the
samples were treated with Dispase II, filtered and cultured in Ham`s F12 with 10% FBS and
antibiotics.
7.2 Immunohistochemistry
IHC is the demonstration of antigens in tissue sections by the use of labelled antibodies as
specific reagents through antigen-antibody interactions that are visualised by a marker. IHC
does not only allow visualisation of proteins, but also allows the user to determine the
subcellular location and/or co-localisation of them (185). Successful detection of antigens by
IHC depends on a variety of factors, starting with tissue sample and fixation. Tissue should be
rapidly preserved to avoid the breakdown of cellular proteins and tissue architecture. Formalin
is the most commonly used fixative and was used in all experiments included in this thesis.
Fixation time should be standardised for all tissues. Prolonged fixation in formalin can result in
excessive cross-linking, thus making the antigen unrecognisable for the antibody. Fortunately
most antigens can be demasked if a proper method of antigen retrieval (AR) is used (186).
Under-fixation of samples is considered a more serious problem than over-fixation, since the
core of the sample will only be fixated by alcohol before immersion in paraffin, thereby creating
a heterogeneous fixation throughout the sample (185). Spheroid derived cells from paper I were
fixated at 4oC overnight. Formalin-fixed, paraffin-embedded (FFPE) tissue from the diagnostic
34
biobank was processed according to their standardised protocols. After fixation in formalin, the
tissue is dehydrated in a series of xylene and alcohol dehydration before embedding in paraffin.
The effect of long-term storage of FFPE tissue is debatable, FFPE blocks can be stored for >25
years if stored at a cool place, FFPE slides should be limited for 7 days, though loss of
antigenicity is also suggested to be antigen dependent (187). Paper II and III included the use
of archived FFPE tissue for IHC. In the context of UM research, assessment of BAP1 status is
advisable. Staining of BAP1 was performed. Unfortunately the staining was inconclusive for
several of the samples due to negative innate control (data not shown in publications), reduced
BAP1 antigenicity in old FFPE tissue has also been encountered by other groups (188).
The tissue used in our experiments was sectioned at 3.5-4μm, since thicker sections can lead to
difficulties in the interpretation of the staining due to multi–layering of the cells. After
sectioning, the tissue was dried before further processing. This should be done at temperatures
less than 60 degrees to avoid loss of antigenicity. Our standard protocol included drying for 1h
at 59 degrees and overnight at 38 degrees. To prevent detachment of sections during AR, super
frost slides were used (Thermo Fisher Scientific).
AR is an essential step in order to reverse the changes induced by fixation. The choice of AR
depends on the targeted antigen and the type of antibody (185). Length of treatment,
temperature, pH, and chemical composition of the AR solution are major factors that influence
the effect of AR. To subgroups of AR exist, namely heat induced epitope retrieval (HIER) and
proteolytic induced epitope retrieval (PIER). HIER includes water bath (PT-link), pressure
cooker heating and microwave heating. PIER consists of various methods of enzymatic
degradation. Our experiments were conducted using PT-link HIER. We also tested microwave
heating and enzymatic digestion with trypsin, in our experience these were inferior to the PT-
link for AR for the chosen epitopes. After AR the sections were treated with a blocking solution.
The time period of blocking is critical since prolonged treatment can result in masking of
antigen and too short incubation time can result in non-specific binding of the secondary
antibody. Blocking solutions include commercial blocking buffers, milk and serum. The choice
of serum should be the same species as the secondary antibody is generated in (189). Our
samples showed considerable less fluorescent background staining if the samples were treated
with goat or donkey serum compared to bovine serum albumin (BSA) or milk. For fluorescent
staining we used a concentration of 10% serum in phosphate-buffered saline (PBS). The
dilution of primary antibodies was chosen based on testing of multiple dilutions, dilutions used
in previous publications and recommended dilution from the supplier. Monoclonal antibodies
35
recognise one epitope, while polyclonal antibodies detect several epitopes on the same antigen,
both of them have their advantages and disadvantages (185).
The direct method of IHC staining is rarely used as the indirect method of staining allows for
greater signal amplifications. The indirect method involves the use of a primary antibody that
detects the epitope(s) of the antigen and a labelled secondary antibody that react with the
primary antibody. The secondary antibody can be conjugated with a fluorescent label; e.g.
fluorescein isothiocyanate (FITC), Rhodamine or Texas Red or by an enzyme e.g. alkaline
phosphatase, peroxidase or glucose oxidase. For enzymatic labels a third layer is added, e.g.
the avidin-biotin complex (ABC) method where the third layer includes peroxidase that can be
developed to different colorimetric end- products when it reacts to 3,3`-diaminobenzidine
(DAB) or other substrates (189). Staining with chromogens such as DAB and amino ethyl
carbazole (AEC) are often used on whole tissue as fluorescent probes can produce too much
background staining, thus making it difficult to distinguish different cell types. Chromogenic
staining has also the advantage that it can be stored, compared to fluorescent probes that often
fade. Heavy pigmentation can be seen in melanoma specimens, hence AEC can be favourable
as it produces a red staining that is distinguishable from melanin. Fluorescent probes have a
high sensitivity and are often preferred for double staining (185) (189).
7.3 Electron microscopy
Electron microscopy allows for visualisation of ultrastructural cellular elements by using a
beam of electrons to create an image of the specimen. The path of the electrons is controlled by
electromagnetic and/or electrostatic lenses. By focusing this beam onto a sample, a resolution
of 0.05nm can be achieved. A resolution at this level enables the user to study subcellular
structures such as membrane structures and organelles. The resolution of the microscope is
increased if the accelerating voltage of the electron beam is increased (190). There are two types
of electron microscopes; scanning electron microscope (SEM) and transmission electron
microscope (TEM). SEM produces an image by detecting secondary electrons that are emitted
from the primary electron beam, the detection of the scattered electrons generates a 3D image
of the surface. TEM produces 2D images of the specimen, though at a higher resolution than
SEM. TEM images are generated by focusing the beam of electrons through a thin specimen
followed by detection by a sensor or film. The preparation of samples varies between TEM and
SEM. TEM samples are cut into ultrathin sections, before treatment with heavy metals (e.g.
36
lead and uranium) to give contrast between different structures. SEM samples on the other hand
are not sectioned. Following fixation and dehydration, they are coated with a thin layer of metal
e.g. gold pallidum to obtain conductivity for imaging.
Samples for EM are either fixated by cryofixation or chemical fixation. We used 0.1M
glutaraldehyde as a fixative for our tissue. The fixation was carried out at 4°C overnight
followed by washing in 0.2M cacodylatebuffer. The tissue was then postfixated in a mixture of
1% osmium tetroxide and 0.2M cacodylatebuffer for 60 minutes before dehydrated through a
graded series of ethanol up to 100%. Further, the tissue was immersed in propylene oxide for
2x5 minutes and a mixture of epon and propylene oxide, before embedment in epon. Ultra-thin
sections (60–70 nm thick) were cut on a Leica Ultracut Ultramicrotome UCT (Leica, Wetzlar,
Germany), stained with uranyl acetate and lead citrate and examined using a Tecnai12
transmission electron microscope (Phillips, Amsterdam, Netherlands).
Important aspects of preparation of samples for TEM include quick fixation of samples to avoid
ultrastructural changes, e.g. oxygen deprivation and mitochondrial alterations. Caution must
also be taken to avoid artefacts when processing the samples.
7.4 Real-time quantitative reverse transcription PCR (qRT-PCR)
Polymerase chain reaction (PCR) is an amplification technique for cloning the specific or
targeted parts of a DNA sequence to generate numerous copies. The method is based on the
ability of DNA polymerase to synthesize a new strand of DNA complementary to the offered
template strand. The PCR cycle consists of three steps; denaturation, annealing and extension.
Denaturation by heating creates two separate strands. The single stranded DNA then anneals to
a given DNA primer that marks the starts of the DNA fragment of interest. The new strand,
marked by the primer, is elongated by the addition of nucleotides by DNA polymerase, thereby
creating a new strand. This is generally repeated in 30-40 cycles (191).
The PCR run is divided into three phases: 1) The exponential phase: where there is an exact
doubling in each cycle. 2) The linear phase: where the reaction components are being
consumed, the reaction slows down and products start to degrade. 3) The plateau or endpoint:
the reaction has stopped and PCR products will degrade if left long enough. The traditional
PCR measures the end-point using agarose gel, yielding low sensitivity and resolution.
Measuring end-product on agarose gel also makes it difficult to quantify starting material,
especially due to sample variation in the end point. qRT-PCR overcomes these obstacles by
37
measuring PCR product in the exponential phase and has become the method of choice for
quantifying RNA. PCR product is quantified relative to an external standard curve or to one or
more co-amplified control mRNAs. Except for time and method for measurement of product,
qRT-PCR also differs from PCR with the addition of the preliminary step, namely the initial
conversion of RNA into a DNA template by an RNA-dependent DNA polymerase (reverse
transcriptase).
Detection and quantification is achieved by fluorescent dyes, such as TaqMan probes or SYBR
green. TaqMan uses a sequence specific probe that is designed with a high energy dye, Reporter,
at the 5`end and a low energy molecule, Quencher, at the 3`end. The Quencher inhibits the
Reporters dye emission when they are in close proximity of each other. When the probe is
cleaved by the 5`endonuclease activity of the polymerase the Reporter is no longer inhibited of
the Quencher and a signal is generated. The specificity of TaqMan is conferred at three levels;
through two PCR primers and the probe. Some suppliers also include a minor groove binder
for extra specificity. TaqMan is more specific than SYBR green and was our method of choice
for detection of specific genes to verify our microarray results. SYBR green was used in paper
II as a quality control of DNA before bisulfite conversion according to the manufacturers’
protocol (See section 6.6.2 DNA methylation assay).
SYBR green emits a signal when it is bound to nascent double stranded (ds) DNA. The PCR
product can be verified by plotting fluorescence as a function of temperature to generate a
melting curve of the amplicon. Synthesis of several PCR products can be seen as peaks in the
melting curve, indicating unspecific primer binding and unspecific results. SYBR green has the
advantage that only one pair of primers is needed, making it less expensive than TaqMan,
though only one target sequence can be monitored in one tube (192). We used pre-made
TaqMan primer for qRT-PCR in all papers, meaning that the manufacturer guaranteed high
amplification efficiency. For genes that should be verified after microarray analysis, we
checked that the probes spanned the correct exon(s).
The Ct (cycle threshold) is defined as the number of cycles required for the fluorescent signal
to cross the threshold, where the threshold is a fluorescent signal that is significantly above the
background fluorescence. The threshold cycle is inversely proportional to the original relative
expression level of the gene of interest. There are various methods for presenting RT-PCR data,
including presentation at absolute or relative expression levels. Absolute expression is
dependent of transformation of data via a standard curve. Relative quantification uses an
internal control for relative presentation of the data. As an endogenous control for our TaqMan
38
RT-PCR we used 18S. 18S was chosen after testing of several housekeeping genes, including
GAPDH and CDKNA1, showing that 18S was more stably expressed across UM samples. The
most widely used method for analysing RT-PCR data is by the comparative CT method, though
other methods exist (193).
A successful PCR run is dependent on several factors. RNA isolation is critical since RNA is
easily degraded and there is a risk of co-purifying inhibitors of the RT or PCR, thus generating
inconsistent results. For RNA isolation from fresh frozen UM samples in paper I and III we
used the Qiagen RNeasy kit, adding an extra step of centrifugation for 10 minutes to remove
unsolvable material after Qiazol treatment to avoid clogging of the columns. Samples can be
further purified using the Zymo PCR inhibitor removal kit (Paper I). DNase was added to avoid
genomic DNA contamination. The RNA concentration and quality was measured using
Nanodrop (Wilmington, DE) and Qubit fluorometer (Thermo Fisher Scientific). A 260/280
ratio of 1.8 on Nanodrop is generally accepted as pure for DNA and a ratio of 2 for RNA. The
260/230 ratio is used as a secondary measure for nucleic acid purity and is ideally in the range
of 2.0-2.2. The RNA quality of samples intended for additional microarray analyses was
assessed by Agilent 2100 Bioanalyser. Samples with a RNA Integrity Number (RIN) values
above 8 are considered to be of high quality.
7.5 RNAscope in situ hybridisation
RNAscope permits direct visualization of RNA in FFPE tissue with single molecule sensitivity
and single cell resolution (194). The in situ analysis of RNA is a supplement and alternative to
RT-PCR. Available samples for research purpose in UM are often minuscule, thus isolation of
high quality RNA can be difficult. RNA isolated from FFPE tissue can be partially degraded
and not suitable for qRT-PCR. RNAscope enables the use of FFPE tissue for RNA analyses,
thereby increasing the amount of tissue available by including stored diagnostic paraffin blocks
from Pathology departments. The RNAscope technique also preserves the architecture of the
tissue, making it possible to map observed signal to individual cells. Whole tissue and
microdissection of tissue both carry the risk of including unwanted cells into RNA extraction,
e.g. including RNA from immune cells in a RT-PCR run. The RNA probes in the RNAscope
technique consist of 28-25 bases complementary to the RNA, a spacer sequence and a 14-base
tail sequence. A pair of target probes, each possessing an individual tail sequence, hybridize
contiguously to a target region of approximately 50 bases. The double-probe design strategy
39
ensures superior background control (194). Probes are hybridized to a cascade of signal
amplification molecules culminating in binding of horseradish peroxidase (HRP)- or alkaline
phosphatase (AP)- labelled probes. DAB or Fast Red substrates are used to detect target RNA
(194). Duplex array and a multiplex fluorescent assay are also available. The duplex assay uses
AP-based Fast Red and HRP-based green. The multiplex assay allows detection of three RNA
targets. Signal detection is performed using dyes with excitation and emission properties
equivalent to those of FITC, Cu3 and Cy5 dyes. In paper I we used Fast red due to the
pigmentation of the UM samples. Fast red produces red fluorescence in addition to the red
reaction product, thus providing a greater level of sensitivity (195). RNA staining signal was
identified as red punctate dots. Each sample was quality controlled for RNA integrity with a
probe specific to the housekeeping gene peptidylprolyl isomerase B (PPIB) mRNA. Negative
control background staining was evaluated using a probe specific to bacterial
dihydrodipicolinate reductase DapB gene.
7.6 Microarrays
7.6.1 cDNA microarrays
DNA microarray technology is used for parallel gene expression analysis of a number of genes
of known and unknown functions. Additionally, the technology can be used for detecting
polymorphisms and mutations (196). The principle of DNA microarrays is the hybridisation of
oligonucleotide probes on a chip to a complementary DNA sequence. This sequence can
represent a known gene or another DNA element.
Microarrays are cost-effective and offer well defined analysis pipelines and standardized
approaches for data submission compared to RNA-sequencing protocols. A disadvantage is that
the array only detects designated sequences. Several factors can contribute to output errors,
among them RNA quality, labelling, hybridisation and detection of the fluorescent signal (196).
There are several platforms available, the ones that are most commonly used are Agilent,
Affymetrix and Illumina. Illumina’s HumanHT-12 v4 Expression BeadChip (Illumina) was
used for the microarray in paper. The array targets 43 770 RefSeq transcripts.
A single gene often has more than one transcript and there are usually several probes for a given
gene. The RNA sample is reverse transcribed into cDNA, followed by an
amplification/labelling step (in vitro transcription) to synthesise biotin-labelled cRNA. The
40
quantity of the labelled cRNA was measured using the NanoDrop Spectrophotometer, and the
quality and size distribution of the labelled cRNA assessed using the 2100 Bioanalyzer. This
was done to be able to hybridize equal amounts of successfully labelled cRNA to the array. For
each sample, 750 ng of biotin labelled cRNA was hybridized to Illumina’s HumanHT-12 v4
Expression BeadChip. After hybridisation to the complementary probe, the array is washed and
scanned. Binding of oligonucleotides is measured by a fluorescent signal.
J-express (http://www.molmine.com/magma/analysis/fss.htm) and Rank Product analysis were
used to identify genes with a two fold up- or downregulation and q-values less than 0.05
between the different groups (197). 1000 permutations were run for each comparison. The rank
product assumes that under the null hypothesis, the order of all items is random, and the
probability of finding a specific item among the top r of n items in a list is p = r/n. Multiplying
these probabilities leads to the definition of the rank product.
In paper II, samples were run on the Affymetrix Human Clariom ™ D Array (Thermo Fisher
Science), targeting more than 540,000 transcripts, thus including rare and low expressing
transcripts. This array is especially suitable for detection of biomarkers due to its
comprehensive coverage of the transcriptome. For each sample a total of 50ng of RNA
extracted from FFPE derived UM was subjected to the GeneChip™ WT Pico Reagent Kit and
WT Labelling Kit (Thermo Fisher Science). After hybridization, washing and staining, the array
was scanned and the Robust Multichip Analysis (RMA) algorithm was applied for generation
of signal values and normalization. Gene transcript with maximal signal values of less than 5
(log2) across all arrays were removed to filter for low and non-expressed genes.
The differential gene expression of the two groups (“Subset Early metastasis” vs “Subset No
metastasis”) was analysed using a one way ANOVA model. The results were expressed as fold
changes (FC). Genes with FC ≥ |±1.5| and a P-value < 0.05 were regarded as significantly
regulated.
In paper III, we used the Affymetrix GeneChip™ Human Gene 2.0 ST Array (Thermo Fisher
Science). This array covers >30 000 coding transcripts and >11 000 intergenic non-coding
transcripts. 150 ng of total RNA was subjected to the GeneChip WT PLUS Reagent Kit and
analysed according to the procedure described in paper II.
41
7.6.2 DNA methylation assay
Analysis of DNA methylation in FFPE derived samples was carried out using the Illumina 450k
bead chip array. This method was chosen due to Illuminas validated protocol for FFPE derived
samples (198, 199). The array covers 99% of RefSeq genes with an average of 17CpG sites per
gene distributed across the promoter, 5`UTR, first exon, gene body, and 3`UTR. The array
covers 96% of CpG islands and interrogates more than 485.000 methylation sites per sample.
As with other arrays, the Illumina 450k bead chip has some limitations compared to sequencing
techniques. The total coverage of CpG sites in the genome is around 2% which means that some
features (e.g. enhancers) are barely covered. Additionally the design does not allow for allele-
specific methylated DNA detection (200).
After DNA isolation, DNA was measured and quality checked using Qubit fluorometer and
Nanodrop. Quality control was performed using Illumina's Infinium HD FFPE QC assay. This
assay is a qRT-PCR assay for a single marker that selects samples suitable for subsequent
restoration based on a difference in quantification cycle (Cq) value (delta Cq less than 5)
between a standard proprietary template and the sample. If the quality of DNA is satisfactory,
the DNA is bisulphite converted using the EZ-96 DNA Methylation-Gold™ Kit according to
the manufacturer's protocol (Zymo Research, Orange, CA). Successful bisulphite conversion
of DNA is essential for detection of methylated CpG islands. Bisulphite conversion is used to
deaminate unmethylated cytosine to produce uracil in DNA. Methylated cytosines are protected
from the conversion to uracil, thus allowing direct detection to determine the location of
unmethylated cytosines and 5-methylcytosines. After bisulphite conversion, samples
underwent restoration using the Illumina Infinium HD Restoration protocol, and 4 μl of
bisulphite-converted restored DNA was used for hybridization on the Infinium
HumanMethylation450 BeadChip. The 450k bead chip array uses two different means
(chemistries) to detect methylated and unmethylated CpG sites. The Infinium 1 assay design
employs two bead types per CpG locus, one each for the methylated and unmethylated state.
The Infinium II design uses 1 bead type, where the methylated state is determined at the single
base extension step after hybridisation.
Hybridisation and scanning was performed at the Core facility, Radiumhospitalet, Oslo
University Hospital. The intensities of the images were extracted using the GenomeStudio
(v.2011.1) Methylation module (1.9.0) software, which normalises within-sample data using
different internal controls that are present on the HumanMethylation450 BeadChip and internal
42
background probes. The methylation score for each CpG was represented as a β-value
according to the fluorescent intensity ratio representing any value between 0 (unmethylated)
and 1 (completely methylated).
The processing of raw data was conducted with help from GeneVia Technologies, FL. The
minfi Bioconductor package was used for data analyses (201). This package uses the R
statistical programming language. Probe filtering was performed to remove probes that have
failed to hybridise (low p-value) , probes that overlap with SNPs, that cross hybridise with
multiple genomic locations and probes on sex chromosomes (200).
Limma Bioconductor package was used to identify single CpG (differentially methylated
positions, DMPs) that were significantly methylated between the two groups (202). Limma uses
an empirical Bayes method to moderate the standard errors of the estimated log-fold changes.
This leads to more stable inference and improved power because there is borrowing of strength
from the body of probes when making inference about each individual probe. The statistics
used is called the moderated t-statistic, which us computed for each probe and then adjusted for
multiple testing using the Benjamini-Hochberg method (203).
Differentially methylated regions (DMRs) were analysed using R package DMRcate, v. 1.18.0
that is based on limma (204). DMRs that were constituted by at least two consecutive significant
CpGs separated by a maximum of 1000 nucleotide gaps were included.
Genomic location, relation to CGI and gene association was supplied for the DMRs and DMPs.
Relation to CGI was classified as flanking CGI shores and shelves, and open sea. Shores are
regions up to 2 kb and shelves 2–4 kb from CGIs. Shores were annotated according to their
chromosome orientation from the p- to q-arms as in N- and S-shores, respectively. The open
sea regions represent CpGs not associated with a CGI. CPGs were annotated depending on the
gene specific orientation in TSS1500 (−1500 to −200), TSS200 (−200 to TSS), 5′-UTR, 1st
exon, gene body and 3′-UTR.
Copy number profiles were generated using the Conumee R package in Bioconductor (205).
To define chromosomal gains and losses, an absolute segment mean threshold ≥0.3 was applied
(206).
The processing of raw data from the 450k array is complex and could not have been performed
without the help of experienced bioinformaticians from the GeneVia team in Finland. The
choice of methods was based on previous publications, methodological review papers and the
experience of the GeneVia team.
43
7.7 Western blot
Western blot is a semi-quantitative technique that utilises antibodies for the detection of
proteins in serum, cell lysates, cell culture supernatants or tissue samples. The technique
involves separation of proteins according to their size by electrophoresis and transfer of the
separated protein onto a polymer membrane using an electrical current. The protein is then
subjected to immunological detection using antibodies. Unbound primary antibody is washed
away before a labelled secondary antibody is added. Positive and negative controls are added
to confirm antibody specificity and to detect potential unspecific staining. (207).
Sample preparation should be performed at cold temperature with protease inhibitors to avoid
denaturation of the proteins. Accurate determination of protein concentration is important to
avoid gel overload. Several colorimetric, reagent-based protein assay techniques have been
developed. Protein is added to the reagent, producing a colour change in proportion to the
amount added. In paper III we used Micro BCA™ Protein Assay Kit (Thermo Fisher
Scientific). The BCA assay is based on the fact that proteins reduces copper that reacts with
BCA to form a coloured complex whose absorption is proportional to the amount of protein
present. The product absorbs at 562 nm. The concentration is determined by reference to a
standard curve consisting of known concentrations of a purified reference protein, most
commonly bovine serum albumin (BSA) (207). The sample is further denatured and separated
using gel electrophoresis. This is achieved by the addition of Sodium dodecyl sulphate (SDS)
that denatures proteins and confers negative charge. The proteins are separated according to
their weight by their migration to the positively charged anode. The concentration of
polyacrylamide in the gel is also of importance, as a lower acrylamide concentration increases
the resolution of higher molecular weight protein. After separation, the proteins are transferred
to a polyvinylidene difluoride (PVDF) membrane by an electrical current (electroblotting). The
proteins remain their organisation within the gel. Before using antibodies, a blocking solution
is added to prevent non-specific binding of antibodies. After addition of the primary antibody,
a secondary HRP conjugated antibody was added before chemiluminescent detection. The light
emission is caused by the oxidation between HRP and the enhanced chemiluminescent solution
(ECL). Potential problems performing Western blot includes high background signal, the
detection of an additional band, weak or no signal detection. The detection of an additional
band could be due to too much protein per lane, multimeric protein assembly, non-specific
antigen and antibody binding, degradation of protein, protein variants and contamination of
reagents (208).
44
8. Summary of results
8.1 Paper I
The first paper sheds light on potential adaptive mechanisms in UM during cancer
dissemination and highlights the metabolic flexibility cancer cells need to possess in order to
survive. Dissemination of cancer cells from the primary tumour is believed to be an early event
as circulating malignant cells can be seen in up to 88% of UM patients at the time of diagnosis.
Additionally, micrometastases to the bone marrow have been found in 29% of the patients (209,
210). Curiously these findings don`t correlate with overall survival. Cancer cells disseminating
from the primary tumour have to adapt to a changing micro-milieu in order to generate
metastatic disease (211-213). The generation of multicellular tumour spheroids (MCTS) by
anchorage- independent growth is associated with enrichment of an aggressive phenotype
characterised by chemoresistance, invasiveness and expression of undifferentiated markers
(214-216). By comparing the differential gene expression of tumoursphere cultures of UM
(n=4) to primary tumour tissue (n=4) and adherent cultures (n=4) we sought to unravel potential
pathways and survival mechanisms pathognomonic for disseminated and circulating cancer
cells.
The differential gene expression of tumour biopsies, adherent cell- and tumoursphere- cultures
from four patients with uveal melanoma (n=4) was examined. The different conditions were
evaluated by microarray analysis, qRT-PCR, RNA-scope, IHC and TEM followed by gene
expression bioinformatics. The multicellular spheroid tumoursphere cultures displayed traits
associated with anoikis resistance demonstrated by ANGPTL4 upregulation, and a shift towards
a lipogenic profile. Additionally, the multicellular spheres showed a marked upregulation of
synovial sarcoma X breakpoint proteins (SSXs), known targets for immunotherapy in several
cancers (217, 218).
8.2 Paper II
The second paper investigates the methylation profile and expression profile of FFPE derived
UM specimen. Data was coupled to histopathological classification and data from the Cancer
Registry of Norway and the Norwegian Cause of Death Registry.
FFPE samples from 23 UM patients who underwent enucleation of the eye in the period 1976-
1989 were included. Samples were divided into 3 subgroups; 1) Death within 5 years (Early
45
metastasis), 2) Death later than 9 years after diagnosis (Late metastasis) and 3) Alive or other
cause of death more than 18 years after primary diagnosis (No metastasis). All subgroups
contained samples with different histological profile, thus including tumours classified as
epithelioid, mixed and spindle shaped. DNA was extracted from the FFPE tissue, bisulphite
converted and analysed by the 450k DNA methylation microarray. Principal component
analysis (PCA) and consensus clustering showed clustering into two groups according to
chromosome 3 status (M3 and D3). Relative methylation contrasting survival groups and
histology groups was analysed without the detection of significant differentially methylated
probes (DMPs) or differentially methylated regions (DMRs). This was anticipated after
evaluating the PCA plot since the samples failed to show clear clustering properties at group
level. Samples with spindle cell histology and late relapse showed the most uniform cluster in
the PCA plot. A subset of eight samples was selected based on preliminary MDS plots,
histopathological classification, chromosome 3 status, survival status and clustering properties.
The comparison “Subset Early metastasis” (n=4) vs “Subset No metastasis” detected 348 DMPs
and 36 DMRs. The DMPs and DMRs from the subset comparison were cross linked to gene
expression data from the same donors, thus revealing a potential mechanistic role of DNA
methylation in the regulation of 26 genes for the DMPs and 4 genes for the 3 detected DMRs.
RNF13, ZNF217 and HYAL1 are candidate tumour suppressors and TMEM200C, RGS10,
ADAM12 and PAM are candidate oncogenes linked to early metastasis.
8.3 Paper III
In the third paper, we explored the differential expression of Connexin 43 (Cx43) in healthy
choroidal tissue (n=6) and cultured uveal melanocytes (n=6) vs UM biopsies (n=6) and cultured
UM (n=6). Aberrant Cx43 expression has been observed in several cancers, however there is
controversy over its role in carcinogenesis and its implication in patient survival. Cx43 is
proposed to function both an oncogene and as a tumour suppressor.
Briefly; primary cell cultures of UM and uveal melanocytes were established. RNA was
extracted from the cell cultures in addition to healthy choroidal tissue and UM biopsies. The
differential gene expression of the 4 sample groups was investigated by microarray and qRT-
PCR showing decreased expression of Cx43 in UM compared to healthy choroidal controls.
46
Protein expression of Cx43 was further examined by IHC. Since the UM specimens included
in the immunohistochemical evaluation (n=9) all harboured histopathological traits associated
with aggressive disease, a panel of less aggressive UM FFPE samples were added for
comparison (n=5). The latter FFPE samples were coupled to data from the Cancer Registry of
Norway (https://www.kreftregisteret.no/en/) and the Norwegian Cause of Death Registry
(https://www.fhi.no/en/hn/health-registries/cause-of-death-registry/), thus providing
information about time and cause of death in addition to information about metastatic spread.
In general UMs displayed diffuse cytoplasmic staining that was comprehensively weaker than
the staining in uveal melanocytes. Most of the tumours showed some degree of heterogeneous
staining. This heterogeneity was independent of histology, proximity to vessels and localisation
within the tumour. Few cells displayed membranous staining pattern. The less aggressive FFPE
samples (long-term survivors) were mostly negative for Cx43, though one of them showed a
staining pattern similar to the more aggressive FFPE samples. A clear relationship between
Cx43 expression in UM and risk of metastatic disease could not be established, though we did
demonstrate changes in the cellular distribution of Cx43 in UM vs healthy uveal melanocytes.
An inverse correlation in EZH2 and Cx43 expression was evident in choroidal tissue and UM,
suggesting an EZH2-dependent mechanism in the regulation of Cx43. This was further assessed
by testing the effect of the EZH2 inhibitor Tazemetostat (EPZ-6438) in UM cell lines (n=3).
The use of Tazemetostat did not induce changes in the expression of Cx43, however
methylation at lysine residue at position 27 of histone 3 (H3K27) was evident for all donors
regardless of BAP1 status.
47
9. Discussion
9.1 Discussion Paper I
The first paper compares the differential gene expression of multicellular tumour spheroids
(MCTS) cultures of UM to primary tumour tissue and adherent cell cultures. The culturing of
UM as MCTS was used as a model to mimic the anchorage independent growth needed for
cancer dissemination. The dissemination of cancer cells is an early event in UM as elusive
metastases can be seen years after treatment of the primary tumour.
The characterisation and identification of these disseminated cells can lay the foundation for
adjuvant treatment strategies that could reduce cancer relapse significantly.
Cell culturing of UM is often hampered by the amount of starting material and growth
properties of the tumour. The propagation of new UM cell lines is known to be challenging
(180). The study was successful in generating primary adherent cultures and first-passage
tumoursphere cultures, allowing us to compare them with tumour tissue from the same donor.
The study included 3 donors with epithelioid histology. Generalisation of the findings could be
limited due to histological homogeneity and few samples, though the results could reflect
adaptive survival mechanism in aggressive tumours as all donors developed metastases.
Ultra-low attachment cone-shaped wells and changes in the composition of media were used to
induce loss of adherence (LOA) in MCTS. The hESC-MEF media has previously been used for
the culturing of putative cancer stem cells (CSCs), there are also reports of enrichment of CSCs
as spheroids without prior cell sorting (184, 219, 220). Such enrichment of stemness-associated
genes could not be inferred from our microarray data. The significance of culture conditions
upon gene expression was demonstrated by principal component analysis (PCA) and
hierarchical clustering, showing clustering according to culture conditions. Primary tumour
biopsies had a higher expression of genes associated with macrophages (CD68), endothelial
cells (von Willebrand factor), and T-cells (CD3D, CD8A, and CD2), in addition to increased
expression of human leukocyte antigen (HLA), thus mirroring cellular heterogeneity within the
primary tumour. The melanoma profile of the MCTS was verified by staining for a-melanoma,
a marker that recognizes HMB-45, MART-1, and Tyrosinase.
Alterations in cellular metabolism and energetics are hallmarks of cancer (3). One of the earliest
observations of altered tumour metabolism was the tendency of cancer cells to favour glycolysis
rather than the more efficient oxidative phosphorylation pathway under aerobic conditions, a
process termed the Warburg effect (221). Changes in lipid metabolism have emerged as a key
48
feature of cancer cells to thrive under challenging conditions. The metabolic flexibility of
cancer cells is demonstrated by their ability to switch between different pathways of FA
acquisition, e.g. elevation of de novo synthesis of fatty acids and the ability to increase the
uptake of exogenous fatty acids (222, 223). A notable feature of the MCTS was the metabolic
shift towards a lipogenic profile. Pathways and genes associated with lipid metabolism,
biosynthesis of unsaturated fatty acids, cholesterol biosynthesis and lipid storage were
upregulated in the MCTS. One of these upregulated genes was Perilipin 2 (PLIN 2), a protein
that coats intracellular lipid storage droplets. The expression of PLIN2 was verified by IHC in
all three MCTS samples. These findings stimulated us to investigate primary tumour tissue and
MCTS at a morphological level by performing TEM. Unfortunately the MCTS of the donors
included in the array had not been processed for TEM since the material was scarce and RNA
extraction and IHC were the priority. TEM of MCTS was therefore performed on a
supplementary donor with epithelioid histology that underwent the same culture conditions as
the primary donors. The TEM images of the supplementary donor showed numerous lipid
droplets in addition to mitochondria. Lipid droplets were also abundant in primary tumour
biopsies. The presence of lipid droplets in UM has been described in the literature earlier, both
as a response to radiation and in untreated tumour tissue (224, 225). Balloon cells are especially
rich in lipids, the abundance of lipids has previously been looked upon as a sign of degradation,
recently these cells are shown to have a high metastatic potential and are found in less
differentiated tumours (225, 226).
The metabolic shift in the MCTS could be a consequence of LOA. LOA is shown to inhibit
uptake of glucose and glycolysis. This leads to diminished levels of ATP and NADPH,
secondary metabolic stress and the generation of ROS that induces cell death by anoikis (212).
The induction of fatty acid oxidation (FAO) restores ATP production and increases NADPH,
thus preventing anoikis (212, 227). Upregulation of FAO in MCTS was indicated by
upregulation of Enoyl-CoA hydratase 1 (ECH1), Peroxisomal D3,D2-enoyl-CoA isomerase
(PECI) and Acyl-CoA thioesterase 1 (ACOT1). FAO is considered an advantageous metabolic
trait for cancer cells and is linked to anoikis resistance (228). The top upregulated gene in
MCTS was angiopoietin like 4 (ANGPTL4), ANGPTL4 is known to stimulate intracellular
lipolysis, thereby supplying substrate for FAO (229).
Our understanding of lipid metabolism in cancer is evolving, thus offering a rationale to develop
a new generation of therapeutic agents targeting the lipogenic profile of UM (230)
49
Another key finding in the paper was the upregulation of the synovial sarcoma X breakpoint
proteins SSX2 and SSX4. SSX expression is confined to the testis, placenta, at low levels in
the thyroid, and in a wide range of tumours (including synovial sarcoma) (231). SSX2 has gene-
regulating properties as has been shown to activate repressed genes by indirectly antagonize the
polycomb repressive group members EZH2 and BMI1. SSX2 has also been shown to have
DNA-binding properties and to negatively regulate the distribution of histone mark H3K27me3,
implying that SSX2 plays a role in the regulation of chromatin structure and function (232).
The IHC staining of primary tumour biopsies and MCTS displayed minimal expression of
SSX4 in the biopsies, while a higher percentage of the cells in the MCTS were positive. If this
is valid for a larger sample set is yet to be assessed. The expression of SSXs in cutaneous
melanoma tissue and cell lines ranges from 21-40% (231, 233). Our results indicate that
upregulation of SSXs is a process related to increased cellular stress e.g. LOA, meaning that
SSXs could be used for detection and targeting of disseminated cancer cells in UM. The tissue
restricted expression of SSXs, make them attractive targets for immunotherapy. Potential
methods include the use of tumour vaccines; the introduction of additional SSXs to promote
recognition and enhance the UM immune response. Another is to directly enhance the
immunological recognition of T- lymphocytes against SSXs, known as adoptive T-cell therapy
(234).
9.2 Discussion Paper II
Paper II investigates the differential methylation pattern of FFPE derived UM specimens and
correlates DNA methylation profiles to survival. Access to UM tissue is hampered by tumours
size, the rarity of the disease and the emergence of brachytherapy in UM treatment. The use of
FFPE UMs greatly expands the selection of available samples and can be linked to long term
survival data and comprehensive medical records. Our publication demonstrates the feasibility
of using FFPE UMs in methylation studies. As outlined in the methodological considerations
section the use of FFPE derived materials presents some pitfalls. Our samples yielded a high
amount of DNA that passed the PCR-based quality control supplied by Illumina and the results
are likely to be comparable to the usage of fresh frozen samples (199). The use of RNA derived
from FFPE tissue is a subject of debate since formalin can induce crosslinking and degradation
of RNA, hence RIN values are generally low (235). Several factors determine the quality of
FFPE derived RNA (236). Formalin fixation is likely to have started immediately after
50
isolation, though the length of fixation is not known and could be prolonged for some of the
samples. The RIN values of our FFPE derived RNA were within the ranges recommended by
Qiagen, though there is a possibility that some relevant genes are not detected due to the quality
of the RNA.
Although DNA methylation is the best characterised epigenetic modification in most cancers,
the characterisation of the methylome of UM is still in its initial phase.
In 2009 Landreville et al stated that clustering of global methylation profile coincided with
clustering into Class 1 and Class 2 UM (referring unpublished data) (143). Our preliminary
clustering analysis running 30 000 probes with the greatest variation in terms of standard
deviation did not demonstrate sub-clustering into 2 groups based on survival properties, nor did
we observe clustering based on histopathological classification (spindle/epithelioid/mixed)
(237). Shortly after our preliminary analyses, Robertson et al published a distinguished paper
demonstrating clustering into 4 subgroups that coincided with Class 1 and Class 2 tumours by
performing unsupervised consensus clustering on the most variable 1% of CpG probes (39).
Our data was reanalysed with the help from GeneVia (Helsinki, Finland). Performing consensus
clustering of the 1% most variable CpG probes showed a stable 4 cluster solution, however
these four clusters didn`t coincide with neither chromosome status nor survival. A potential
explanation could be small sample size and the selection of samples included in our study.
While Robertson et al had a representative selection of 80 sample including all 4 GEP, our
selection was based on survival data and pathology reports only. Gene expression profile was
not available for all 23 samples, meaning that the selection of GEP included could be skewed.
In 2019 Field et al performed an unsupervised PCA on the 20% most variable probes on the
same dataset as Robertson et al and on a set of 12 samples, demonstrating clustering according
to chromosome 3 status (145, 238). We were able to reproduce this result by performing
consensus clustering of the 1% most variable CpG probes and PCA on the top 500 differentially
methylated probes, showing a two cluster solutions according to chromosome 3 status. This
two cluster solution did not coincide with survival nor histological classification (as expected
since loss of chromosome 3 is not restricted to e.g. epithelioid tumours nor does it accurately
predict UM metastatic disease). As inferred from clustering analyses, we did not detect any
significant DMPs or DMRs comparing predefined survival groups, nor histological groups.
This could be due to heterogeneity within sample groups as all survival groups contained
samples classified as spindle, epithelioid and mixed, while all histological subgroups contained
51
samples classified as “Early”, “Late” and “No” metastasis. An additional explanation could be
intratumoural heterogeneity. A study on lung adenocarcinomas showed that intratumoural
heterogeneity increases with larger tumour size, advanced age and postsurgical recurrence
(239). Our series was composed of relatively large tumours. Large UM tumours are known to
host a greater frequency of genomic alterations (240). Differential DNA methylation within
tumours has been demonstrated in lung carcinomas where on average 25% of all differentially
methylated probes were shared by all regions from the same tumours (239). Further, our
predefined survival groups could be subject to confounding factors since cancer relapse could
be undetected or present itself at a later stage.
Based on our preliminary clustering analyses and the clustering analysis performed by
GeneVia, we extracted 2 subsets of samples for comparison. Four samples (n=4) defined as
“Subset Early metastasis” was compared to four samples (n=4) defined as “Subset No
metastasis”. The 2 subgroups included samples of approximately the same size; “Subset Early
metastasis” included patients with spindle and mixed histology, while “Subset No metastasis”
included spindle cell histology only. The age at diagnosis ranged from 35 to 85 years in the
“Subset Early metastasis”, while the range was from 43 to 59 years in “Subset No metastasis”.
This comparison yielded significant DMPs and DMRs. RNA was extracted from the same
samples (n=8) and analysed by microarray. Gene expression analysis by qRT-PCR of the same
samples coincided with the top three up and downregulated genes from the array. Out of the
348 differentially methylated CpGs in “Subset Early metastasis” vs “Subset No metastasis”, 26
DMPs corresponded with changes in gene expression. Out of the 36 DMRs, the overlapping
promoters of 3 of these DMRs corresponded with significant changes in gene expression of 4
genes from the one-way ANOVA. Several of these DMPs displayed inverse relationship
between promoter methylation and gene expression as anticipated (e.g. decreased methylation
in CpG shore and islands located in TSS and a corresponding increase in gene expression).
A potential problem that is rarely discussed in other publications is the inability of the 450 k
array to discriminate between 5mC and 5hmC. This could be overcome by performing oxidative
bisulfite sequencing (Ox-BS). The process includes a selective oxidative step that deprotects
hydroxymethylation and converts 5hmC to 5fC, which, after bisulfite treatment, becomes a
uracil. The main drawbacks of Ox-BS are the oxidative degradation of DNA and longer bisulfite
treatment required for complete 5fC deamination (241). If detection of 5hmc represent a
52
potential bias in 450k analyses is debatable. In general the presence of 5hmc is low in cancer
samples and shows a low abundancy in promoter areas (155, 242, 243).
The strengths of the study could be enhanced by supplying expression data for all donors and
performing bisulphite sequencing (117). The preservation of FFPE samples for future research
has been a priority as protocols for FFPE derived tissue are rapidly evolving and would enable
us to perform more extensive studies with less sample input.
The study indicates that subclustering according to global methylation profile is hampered by
donor heterogeneity when comparing few samples and is mainly associated with chromosome
3 status. By subtracting a subset of samples we were able to detect significant DMPs and DMRs
correlated with survival that could be relevant for the progression and aggressiveness of UM,
as demonstrated by changes in the expression of genes related to these probes.
9.3 Discussion Paper III
The third paper assesses the expression of Cx43 in primary UM and explores a potential
regulatory role of EZH2 in Cx43 expression and distribution.
Connexins are a family of transmembrane proteins capable of forming gap junctions, thus
providing communication between adjacent cells (244). In addition to intercellular
communication (GJIC), connexins are proposed to exert effects through hemichannel signalling
and gap junctional independent pathways e.g. via their C-terminal tail (245). To the best of our
knowledge, only one paper on Cx43 expression in UM has previously been published. This
paper showed increased Cx43 staining in primary UM compared to dermal nevi. Moreover,
they noted increased staining in tumours harbouring scleral invasion and observed membranous
staining of Cx43 in relation to blood vessels.
We compared our samples to healthy choroidal tissue and cultured uveal melanocytes, showing
reduced cytoplasmic and membranous staining in UM. We did detect increased staining around
some vessels, though most of the staining was attributed to expression of Cx43 in macrophages
and endothelial cells. In the aforementioned paper, scleral invasion was accounted for as a
measure of tumour aggressiveness. Except from UMB6, UMB7 and UMB13 (and the older
archived specimens), the samples included in our study showed scleral invasion. UMB6 and
UMB7 were both BAP1 negative and associated with poor prognosis, while UMB 13 was BAP1
53
positive. We did not detect any difference in staining associated with BAP1 status nor scleral
invasion.
Since all of the recent tumours included in our study had a more aggressive profile, we chose
to include archived specimens known to be long-term survivors. The archived FFPE tissue had
positive internal control cells, though potential biases are long-term storage and variation in
time and length of formalin fixation. In general most of these tumours displayed a weaker
staining, however UMB15 showed similar staining pattern as more aggressive tumours. Hence,
staining intensity of Cx43 should be interpreted with caution as a measure of tumour
aggressiveness. Interestingly, intratumoural heterogeneity of Cx43 was seen in several
specimens, this could be indicative of subgroups of cells with different properties and/or innate
flexibility of UM to express Cx43. ICC of UM cell lines showed perinuclear expression of Cx43
while this was not observed by IHC in primary UM. This could be due to technical issues or
e.g. degradation of Cx43, though it could also reflect a potential role of Cx43 in downstream
transcription processes, such as RNA splicing, processing, export and translation (246, 247).
In cutaneous melanoma cell lines, overexpression of Cx43 resulted in suppression of
anchorage-independent growth and a reduction in proliferative and metastatic capacity (248,
249).
While several studies support the notion that connexins are tumour suppressors, there are
examples of studies indicative of a tumour promoting role of connexins, especially in metastatic
lesions (250). Cutaneous melanomas have shown an increased expression of Cx43 in metastatic
lesions, mostly in their intracellular compartments. Rarely Cx43 assembled into functional gap
junctions (251). Increased Cx43 expression and associated upregulation of GJIC has shown
enhanced therapeutic effect of cisplatin (244, 252). Considering that UM is a cancer recognised
by its resistance to chemotherapeutics, improved chemosensitivity through GJIC restoration
could have a vast potential in future treatment protocols. The implementation of such strategies
relies on characterisation of Cx43 expression in all sequential steps of metastasis (3).
In paper III, we explored a potential role of EZH2 in regulating Cx43 expression. The
overexpression of EZH2 due to aberrant activation of EZH2 or loss-of-function mutations in
the SWI/SNF complex is associated with cancer aggressiveness and advanced disease (64, 65).
As discussed in paper III, the differential effect of EZH2 inhibition related to BAP1 status has
previously been studied for mesothelioma in a mouse model. This study showed that EZH2
silencing abrogated in vivo tumour formation of BAP1-mutant, but not of wild-type, cell lines.
Additionally EZH2 reduced the metastatic potential in BAP1- mutant cell lines (66). The effect
54
of EZH2 inhibition has been tested on uveal melanoma cell lines, though the paper is no longer
accessible at the journal site, the authors claim that uveal melanoma cells are resistant to EZH2
inhibition regardless of BAP1 status (253). The aforementioned paper has been subject for
criticism (254). Firstly, cell lines that lacked BAP1 tended to grow more slowly, and therapeutic
efficacy should be evaluated at a later time point (255). Secondly, since previous functional
work had shown that loss of BAP1 did not promote proliferation, colony formation or in vivo
tumour growth of uveal melanoma cells, migration and invasion studies could be better means
in the evaluation of therapeutic effect. Thirdly, they questioned the target inhibition of the
compound used since reduced polycomb activity (like the depletion of H3K27me3) had not
been determined.
Among the 2 primary cell lines used for testing, one was depleted of BAP1 (MP38). In line
with previous observations, MP38 grew slower than the BAP1 positive cell lines (253).
Estimated doubling time of MP38 was 80 hours, while doubling time for MP41 and MP46 was
closer to 40 hours. Thus, it is likely that 10 days of treatment with Tazemetostat is enough to
detect effect upon Cx43 expression of MP38, though it could be too short to detect resistance
to treatment as observed in MP46. Regardless of effect on Cx43 expression and variability in
the induction of cell death, Tazemetostat resulted in a reduction of H3K27me3 in all cell lines
independently of BAP1 status.
55
10. Conclusions and future perspectives
The treatment of primary UM is invariably successful, meaning that tumour cell dissemination
is an early event. This is reflected by the detection of circulating tumour cells and tumour cells
in the bone marrow after primary treatment, regardless of estimated metastatic risk. The aim of
the present thesis was to shed light upon mechanisms and traits of UM that enables cells to
metastasise, thus highlighting potential treatment strategies since drug resistance to targeted
therapy remains the major challenge in UM treatment.
Paper I investigates the metabolic flexibility in cultured UM cells in the context of cancer
dissemination and LOA. The cells displayed a shift towards a more lipogenic profile upon LOA
in addition to increased expression of SSXs. Targeting SSXs could be a means of selectively
targeting disseminated cancers cells and needs to be explored further, e.g. by the assessment of
SSX expression in circulating cancer cells. Fatty acid metabolism has been shown to be more
than a survival strategy for striving cancer cells, as lipids can serve as oncogenic signalling
molecules (256). Targeting of lipid metabolism has shown promise in preclinical and clinical
trials, especially after the development of new compounds with less off-target effects.
Considering the importance of lipid metabolism in normal whole-body metabolic
homoeostasis, targeting is more likely to be a part of combinatorial treatment.
In paper II we demonstrated significant changes in DNA methylation and corresponding gene
expression between subgroups of UM (early metastasis vs no metastasis). Previous work has
shown that loss of BAP1 leads to the to the methylomic repatterning profile characteristic of
Class 2 UMs (145). Our study features genes that could be directly regulated by DNA
methylation, hence these genes could be important in the progression of metastatic disease and
potential secondary targets for DNMT-inhibitors. Our knowledge of DNA methylation in the
evolvement of metastatic disease could be enhanced by assessing DNA methylation patterns in
UM metastases or by studying differential DNA methylation in preclinical metastases models.
The development of selective and reversible DNMT-inhibitors brings hope for safe and
efficient targeting of aberrant DNA methylation patterns in UM (257).
In paper III we investigated the expression of Cx43 in UM, demonstrating a downregulation of
Cx43 in tumours vs healthy choroidal tissue. Tumours expressing Cx43 showed predominately
cytoplasmic localisation of the protein. Restoration of functional gap junctions by increasing
Cx43 expression might seem like an appealing treatment strategy to increase chemosensitivity,
especially for UM. It should be taken into consideration that several of our tumours showed
heterogeneous expression of Cx43. In addition, studies on other cancers have shown an
56
upregulation of Cx43 in metastases, highlighting a potential risk of fuelling metastatic disease
by restoring Cx43 expression. Hence, it would be of great value to explore Cx43 expression in
disseminated tumour cells and in metastases. The proposed role of connexins in oncogenic
signalling is also intriguing and widens the potential application of connexin-targeted therapy.
Paper III also explored the effect of EZH2 inhibition upon Cx43 expression. Though alterations
in Cx43 expression were not seen, EZH2 inhibition resulted in a reduction of H3K27me3 in all
cell lines independently of BAP1 status, furthermore, partial cell death was observed.
Tazemetostat has shown relatively few side effects in clinical trials and could thus be relevant
in combinatorial therapy (258).
In conclusion the present thesis has elucidated genetic and epigenetic traits of UM that can be
of importance in the development of future treatment strategies.
57
11. References
1. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al.
Cancer incidence and mortality worldwide: sources, methods and major patterns
in GLOBOCAN 2012. International journal of cancer Journal international du
cancer. 2015;136(5):E359-86.
2. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell. 2000;100(1):57-
70.
3. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell.
2011;144(5):646-74.
4. Helsedirektoratet. Nasjonalt handlingsprogram med retningslinjer for
diagnostikk, behandling og oppfølging av maligne melanomer. 9. ed:
Helsedirektoratet; 2020. p. 87-96.
5. Shields CL, Kaliki S, Furuta M, Mashayekhi A, Shields JA. Clinical
spectrum and prognosis of uveal melanoma based on age at presentation in 8,033
cases. Retina (Philadelphia, Pa). 2012;32(7):1363-72.
6. Weis E, Shah CP, Lajous M, Shields JA, Shields CL. The association
between host susceptibility factors and uveal melanoma: a meta-analysis.
Archives of ophthalmology (Chicago, Ill : 1960). 2006;124(1):54-60.
7. de Lange MJ, Razzaq L, Versluis M, Verlinde S, Dogrusoz M, Bohringer
S, et al. Distribution of GNAQ and GNA11 Mutation Signatures in Uveal
Melanoma Points to a Light Dependent Mutation Mechanism. PLoS One.
2015;10(9):e0138002.
8. Andreoli MT, Mieler WF, Leiderman YI. Epidemiological trends in uveal
melanoma. The British journal of ophthalmology. 2015.
9. Krohn J, Dahl O. Incidence of iris melanoma in western Norway. Acta
ophthalmologica. 2008;86(1):116-7.
10. Singh AD, Kalyani P, Topham A. Estimating the risk of malignant
transformation of a choroidal nevus. Ophthalmology. 2005;112(10):1784-9.
11. Kivela TT, Piperno-Neumann S, Desjardins L, Schmittel A, Bechrakis N,
Midena E, et al. Validation of a Prognostic Staging for Metastatic Uveal
Melanoma: A Collaborative Study of the European Ophthalmic Oncology Group.
American journal of ophthalmology. 2016.
12. Diener-West M, Reynolds SM, Agugliaro DJ, Caldwell R, Cumming K,
Earle JD, et al. Development of metastatic disease after enrollment in the COMS
trials for treatment of choroidal melanoma: Collaborative Ocular Melanoma
Study Group Report No. 26. Archives of ophthalmology (Chicago, Ill : 1960).
2005;123(12):1639-43.
13. Eskelin S, Kivela T. Mode of presentation and time to treatment of uveal
melanoma in Finland. The British journal of ophthalmology. 2002;86(3):333-8.
58
14. van Beek JGM, Koopmans AE, Verdijk RM, Naus NC, de Klein A, Kilic
E. Diagnosis, Histopathologic and Genetic Classification of Uveal Melanoma. In:
Duc GHT, editor. Melanoma - From Early Detection to Treatment: InTech; 2012.
p. 137-69.
15. Shields CL, Manalac J, Das C, Ferguson K, Shields JA. Choroidal
melanoma: clinical features, classification, and top 10 pseudomelanomas. Current
opinion in ophthalmology. 2014;25(3):177-85.
16. Mellen PL, Morton SJ, Shields CL. American joint committee on cancer
staging of uveal melanoma. Oman journal of ophthalmology. 2013;6(2):116-8.
17. Nakao S, Hafezi-Moghadam A, Ishibashi T. Lymphatics and
lymphangiogenesis in the eye. J Ophthalmol. 2012;2012:783163.
18. van Beek JGM, van den Bosch QCC, Naus N, Paridaens D, de Klein A,
Kiliç E, et al. Absence of Intraocular Lymphatic Vessels in Uveal Melanomas
with Extrascleral Growth. Cancers (Basel). 2019;11(2).
19. Eagle RC, Jr. The pathology of ocular cancer. Eye (London, England).
2013;27(2):128-36.
20. Shields JA, Shields CL. Management of posterior uveal melanoma: past,
present, and future: the 2014 charles L. Schepens lecture. Ophthalmology.
2015;122(2):414-28.
21. Cruz F, 3rd, Rubin BP, Wilson D, Town A, Schroeder A, Haley A, et al.
Absence of BRAF and NRAS mutations in uveal melanoma. Cancer research.
2003;63(18):5761-6.
22. Steeb T, Wessely A, Ruzicka T, Heppt MV, Berking C. How to MEK the
best of uveal melanoma: A systematic review on the efficacy and safety of MEK
inhibitors in metastatic or unresectable uveal melanoma. Eur J Cancer.
2018;103:41-51.
23. Bol KF, Ellebaek E, Hoejberg L, Bagger MM, Larsen MS, Klausen TW, et
al. Real-World Impact of Immune Checkpoint Inhibitors in Metastatic Uveal
Melanoma. Cancers (Basel). 2019;11(10).
24. Kilic E, van Gils W, Lodder E, Beverloo HB, van Til ME, Mooy CM, et al.
Clinical and cytogenetic analyses in uveal melanoma. Investigative
ophthalmology & visual science. 2006;47(9):3703-7.
25. Tschentscher F, Prescher G, Horsman DE, White VA, Rieder H,
Anastassiou G, et al. Partial deletions of the long and short arm of chromosome 3
point to two tumor suppressor genes in uveal melanoma. Cancer research.
2001;61(8):3439-42.
26. White VA, McNeil BK, Thiberville L, Horsman DE. Acquired
homozygosity (isodisomy) of chromosome 3 during clonal evolution of a uveal
melanoma: association with morphologic heterogeneity. Genes Chromosomes
Cancer. 1996;15(2):138-43.
27. Scholes AG, Damato BE, Nunn J, Hiscott P, Grierson I, Field JK.
Monosomy 3 in uveal melanoma: correlation with clinical and histologic
59
predictors of survival. Investigative ophthalmology & visual science.
2003;44(3):1008-11.
28. Kilic E, Naus NC, van Gils W, Klaver CC, van Til ME, Verbiest MM, et
al. Concurrent loss of chromosome arm 1p and chromosome 3 predicts a
decreased disease-free survival in uveal melanoma patients. Investigative
ophthalmology & visual science. 2005;46(7):2253-7.
29. Ehlers JP, Worley L, Onken MD, Harbour JW. Integrative genomic
analysis of aneuploidy in uveal melanoma. Clinical cancer research : an official
journal of the American Association for Cancer Research. 2008;14(1):115-22.
30. Dogrusoz M, Jager MJ. Genetic prognostication in uveal melanoma. Acta
ophthalmologica. 2017.
31. Johansson P, Aoude LG, Wadt K, Glasson WJ, Warrier SK, Hewitt AW, et
al. Deep sequencing of uveal melanoma identifies a recurrent mutation in PLCB4.
Oncotarget. 2016;7(4):4624-31.
32. Van Raamsdonk CD, Griewank KG, Crosby MB, Garrido MC, Vemula S,
Wiesner T, et al. Mutations in GNA11 in uveal melanoma. N Engl J Med.
2010;363(23):2191-9.
33. Van Raamsdonk CD, Bezrookove V, Green G, Bauer J, Gaugler L, O'Brien
JM, et al. Frequent somatic mutations of GNAQ in uveal melanoma and blue
naevi. Nature. 2009;457(7229):599-602.
34. Yavuzyigitoglu S, Koopmans AE, Verdijk RM, Vaarwater J, Eussen B, van
Bodegom A, et al. Uveal Melanomas with SF3B1 Mutations: A Distinct Subclass
Associated with Late-Onset Metastases. Ophthalmology. 2016;123(5):1118-28.
35. Shoushtari AN, Carvajal RD. GNAQ and GNA11 mutations in uveal
melanoma. Melanoma research. 2014;24(6):525-34.
36. O'Hayre M, Vazquez-Prado J, Kufareva I, Stawiski EW, Handel TM,
Seshagiri S, et al. The emerging mutational landscape of G proteins and G-
protein-coupled receptors in cancer. Nat Rev Cancer. 2013;13(6):412-24.
37. Harbour JW. The genetics of uveal melanoma: an emerging framework for
targeted therapy. Pigment cell & melanoma research. 2012;25(2):171-81.
38. Moore AR, Ceraudo E, Sher JJ, Guan Y, Shoushtari AN, Chang MT, et al.
Recurrent activating mutations of G-protein-coupled receptor CYSLTR2 in uveal
melanoma. Nat Genet. 2016;48(6):675-80.
39. Robertson AG, Shih J, Yau C, Gibb EA, Oba J, Mungall KL, et al.
Integrative Analysis Identifies Four Molecular and Clinical Subsets in Uveal
Melanoma. Cancer Cell. 2018;33(1):151.
40. Park JJ, Diefenbach RJ, Joshua AM, Kefford RF, Carlino MS, Rizos H.
Oncogenic signaling in uveal melanoma. Pigment cell & melanoma research.
2018;31(6):661-72.
41. Populo H, Vinagre J, Lopes JM, Soares P. Analysis of GNAQ mutations,
proliferation and MAPK pathway activation in uveal melanomas. The British
journal of ophthalmology. 2011;95(5):715-9.
60
42. Zuidervaart W, van Nieuwpoort F, Stark M, Dijkman R, Packer L,
Borgstein AM, et al. Activation of the MAPK pathway is a common event in uveal
melanomas although it rarely occurs through mutation of BRAF or RAS. British
journal of cancer. 2005;92(11):2032-8.
43. Vasudevan KM, Garraway LA. AKT signaling in physiology and disease.
Curr Top Microbiol Immunol. 2010;347:105-33.
44. Saraiva VS, Caissie AL, Segal L, Edelstein C, Burnier MN, Jr.
Immunohistochemical expression of phospho-Akt in uveal melanoma. Melanoma
research. 2005;15(4):245-50.
45. Carnero A, Paramio JM. The PTEN/PI3K/AKT Pathway in vivo, Cancer
Mouse Models. Front Oncol. 2014;4:252.
46. Abdel-Rahman MH, Yang Y, Zhou XP, Craig EL, Davidorf FH, Eng C.
High frequency of submicroscopic hemizygous deletion is a major mechanism of
loss of expression of PTEN in uveal melanoma. J Clin Oncol. 2006;24(2):288-95.
47. Feng X, Degese MS, Iglesias-Bartolome R, Vaque JP, Molinolo AA,
Rodrigues M, et al. Hippo-independent activation of YAP by the GNAQ uveal
melanoma oncogene through a trio-regulated rho GTPase signaling circuitry.
Cancer Cell. 2014;25(6):831-45.
48. Brantley MA, Jr., Harbour JW. Deregulation of the Rb and p53 pathways
in uveal melanoma. Am J Pathol. 2000;157(6):1795-801.
49. Chana JS, Wilson GD, Cree IA, Alexander RA, Myatt N, Neale M, et al. c-
myc, p53, and Bcl-2 expression and clinical outcome in uveal melanoma. The
British journal of ophthalmology. 1999;83(1):110-4.
50. Giacinti C, Giordano A. RB and cell cycle progression. Oncogene.
2006;25(38):5220-7.
51. Harbour JW, Onken MD, Roberson ED, Duan S, Cao L, Worley LA, et al.
Frequent mutation of BAP1 in metastasizing uveal melanomas. Science.
2010;330(6009):1410-3.
52. Scheuermann JC, de Ayala Alonso AG, Oktaba K, Ly-Hartig N, McGinty
RK, Fraterman S, et al. Histone H2A deubiquitinase activity of the Polycomb
repressive complex PR-DUB. Nature. 2010;465(7295):243-7.
53. Campagne A, Lee MK, Zielinski D, Michaud A, Le Corre S, Dingli F, et
al. BAP1 complex promotes transcription by opposing PRC1-mediated H2A
ubiquitylation. Nature communications. 2019;10(1):348.
54. Ismail IH, Davidson R, Gagne JP, Xu ZZ, Poirier GG, Hendzel MJ.
Germline mutations in BAP1 impair its function in DNA double-strand break
repair. Cancer research. 2014;74(16):4282-94.
55. Kristie TM, Liang Y, Vogel JL. Control of alpha-herpesvirus IE gene
expression by HCF-1 coupled chromatin modification activities. Biochimica et
biophysica acta. 2010;1799(3-4):257-65.
56. Misaghi S, Ottosen S, Izrael-Tomasevic A, Arnott D, Lamkanfi M, Lee J,
et al. Association of C-terminal ubiquitin hydrolase BRCA1-associated protein 1
61
with cell cycle regulator host cell factor 1. Molecular and cellular biology.
2009;29(8):2181-92.
57. Knez J, Piluso D, Bilan P, Capone JP. Host cell factor-1 and E2F4 interact
via multiple determinants in each protein. Molecular and cellular biochemistry.
2006;288(1-2):79-90.
58. Tyagi S, Chabes AL, Wysocka J, Herr W. E2F activation of S phase
promoters via association with HCF-1 and the MLL family of histone H3K4
methyltransferases. Molecular cell. 2007;27(1):107-19.
59. Di Croce L, Helin K. Transcriptional regulation by Polycomb group
proteins. Nat Struct Mol Biol. 2013;20(10):1147-55.
60. Weissman B, Knudsen KE. Hijacking the chromatin remodeling
machinery: impact of SWI/SNF perturbations in cancer. Cancer research.
2009;69(21):8223-30.
61. Italiano A, Soria JC, Toulmonde M, Michot JM, Lucchesi C, Varga A, et
al. Tazemetostat, an EZH2 inhibitor, in relapsed or refractory B-cell non-Hodgkin
lymphoma and advanced solid tumours: a first-in-human, open-label, phase 1
study. Lancet Oncol. 2018;19(5):649-59.
62. Kondo Y, Shen L, Cheng AS, Ahmed S, Boumber Y, Charo C, et al. Gene
silencing in cancer by histone H3 lysine 27 trimethylation independent of
promoter DNA methylation. Nat Genet. 2008;40(6):741-50.
63. Xu K, Wu ZJ, Groner AC, He HH, Cai C, Lis RT, et al. EZH2 oncogenic
activity in castration-resistant prostate cancer cells is Polycomb-independent.
Science. 2012;338(6113):1465-9.
64. Bachmann IM, Halvorsen OJ, Collett K, Stefansson IM, Straume O,
Haukaas SA, et al. EZH2 expression is associated with high proliferation rate and
aggressive tumor subgroups in cutaneous melanoma and cancers of the
endometrium, prostate, and breast. J Clin Oncol. 2006;24(2):268-73.
65. 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.
66. LaFave LM, Beguelin W, Koche R, Teater M, Spitzer B, Chramiec A, et
al. Loss of BAP1 function leads to EZH2-dependent transformation. Nature
medicine. 2015;21(11):1344-9.
67. Louie BH, Kurzrock R. BAP1: Not just a BRCA1-associated protein.
Cancer Treat Rev. 2020;90:102091.
68. Alsafadi S, Houy A, Battistella A, Popova T, Wassef M, Henry E, et al.
Cancer-associated SF3B1 mutations affect alternative splicing by promoting
alternative branchpoint usage. Nature communications. 2016;7:10615.
69. Paolella BR, Gibson WJ, Urbanski LM, Alberta JA, Zack TI,
Bandopadhayay P, et al. Copy-number and gene dependency analysis reveals
partial copy loss of wild-type SF3B1 as a novel cancer vulnerability. eLife.
2017;6.
62
70. Furney SJ, Pedersen M, Gentien D, Dumont AG, Rapinat A, Desjardins L,
et al. SF3B1 mutations are associated with alternative splicing in uveal melanoma.
Cancer Discov. 2013;3(10):1122-9.
71. Martin M, Masshofer L, Temming P, Rahmann S, Metz C, Bornfeld N, et
al. Exome sequencing identifies recurrent somatic mutations in EIF1AX and
SF3B1 in uveal melanoma with disomy 3. Nat Genet. 2013;45(8):933-6.
72. Decatur CL, Ong E, Garg N, Anbunathan H, Bowcock AM, Field MG, et
al. Driver Mutations in Uveal Melanoma: Associations With Gene Expression
Profile and Patient Outcomes. JAMA ophthalmology. 2016;134(7):728-33.
73. Kusters-Vandevelde HV, Creytens D, van Engen-van Grunsven AC,
Jeunink M, Winnepenninckx V, Groenen PJ, et al. SF3B1 and EIF1AX mutations
occur in primary leptomeningeal melanocytic neoplasms; yet another similarity
to uveal melanomas. Acta Neuropathol Commun. 2016;4:5.
74. Chaudhuri J, Si K, Maitra U. Function of eukaryotic translation initiation
factor 1A (eIF1A) (formerly called eIF-4C) in initiation of protein synthesis. The
Journal of biological chemistry. 1997;272(12):7883-91.
75. Onken MD, Worley LA, Ehlers JP, Harbour JW. Gene expression profiling
in uveal melanoma reveals two molecular classes and predicts metastatic death.
Cancer research. 2004;64(20):7205-9.
76. Chang SH, Worley LA, Onken MD, Harbour JW. Prognostic biomarkers
in uveal melanoma: evidence for a stem cell-like phenotype associated with
metastasis. Melanoma research. 2008;18(3):191-200.
77. Onken MD, Worley LA, Tuscan MD, Harbour JW. An accurate, clinically
feasible multi-gene expression assay for predicting metastasis in uveal melanoma.
The Journal of molecular diagnostics : JMD. 2010;12(4):461-8.
78. Worley LA, Onken MD, Person E, Robirds D, Branson J, Char DH, et al.
Transcriptomic versus chromosomal prognostic markers and clinical outcome in
uveal melanoma. Clinical cancer research : an official journal of the American
Association for Cancer Research. 2007;13(5):1466-71.
79. Petrausch U, Martus P, Tonnies H, Bechrakis NE, Lenze D, Wansel S, et
al. Significance of gene expression analysis in uveal melanoma in comparison to
standard risk factors for risk assessment of subsequent metastases. Eye (London,
England). 2008;22(8):997-1007.
80. van Gils W, Lodder EM, Mensink HW, Kilic E, Naus NC, Bruggenwirth
HT, et al. Gene expression profiling in uveal melanoma: two regions on 3p related
to prognosis. Investigative ophthalmology & visual science. 2008;49(10):4254-
62.
81. Onken MD, Worley LA, Char DH, Augsburger JJ, Correa ZM, Nudleman
E, et al. Collaborative Ocular Oncology Group report number 1: prospective
validation of a multi-gene prognostic assay in uveal melanoma. Ophthalmology.
2012;119(8):1596-603.
63
82. Klufas MA, Itty S, McCannel CA, Glasgow BJ, Moreno C, McCannel TA.
Variable Results for Uveal Melanoma-Specific Gene Expression Profile
Prognostic Test in Choroidal Metastasis. JAMA ophthalmology.
2015;133(9):1073-6.
83. Shah AA, Bourne TD, Murali R. BAP1 protein loss by
immunohistochemistry: a potentially useful tool for prognostic prediction in
patients with uveal melanoma. Pathology. 2013;45(7):651-6.
84. Augsburger JJ, Correa ZM, Augsburger BD. Frequency and implications
of discordant gene expression profile class in posterior uveal melanomas sampled
by fine needle aspiration biopsy. American journal of ophthalmology.
2015;159(2):248-56.
85. Field MG, Decatur CL, Kurtenbach S, Gezgin G, van der Velden PA, Jager
MJ, et al. PRAME as an Independent Biomarker for Metastasis in Uveal
Melanoma. Clinical cancer research : an official journal of the American
Association for Cancer Research. 2016;22(5):1234-42.
86. Field MG, Durante MA, Decatur CL, Tarlan B, Oelschlager KM, Stone JF,
et al. Epigenetic reprogramming and aberrant expression of PRAME are
associated with increased metastatic risk in Class 1 and Class 2 uveal melanomas.
Oncotarget. 2016;7(37):59209-19.
87. Jager MJ, Brouwer NJ, Esmaeli B. The Cancer Genome Atlas Project: An
Integrated Molecular View of Uveal Melanoma. Ophthalmology.
2018;125(8):1139-42.
88. Kujala E, Makitie T, Kivela T. Very long-term prognosis of patients with
malignant uveal melanoma. Investigative ophthalmology & visual science.
2003;44(11):4651-9.
89. Shain AH, Bagger MM, Yu R, Chang D, Liu S, Vemula S, et al. The genetic
evolution of metastatic uveal melanoma. Nat Genet. 2019;51(7):1123-30.
90. Reiter JG, Makohon-Moore AP, Gerold JM, Heyde A, Attiyeh MA,
Kohutek ZA, et al. Minimal functional driver gene heterogeneity among untreated
metastases. Science. 2018;361(6406):1033-7.
91. Royer-Bertrand B, Torsello M, Rimoldi D, El Zaoui I, Cisarova K, Pescini-
Gobert R, et al. Comprehensive Genetic Landscape of Uveal Melanoma by
Whole-Genome Sequencing. Am J Hum Genet. 2016;99(5):1190-8.
92. Trolet J, Hupe P, Huon I, Lebigot I, Decraene C, Delattre O, et al. Genomic
profiling and identification of high-risk uveal melanoma by array CGH analysis
of primary tumors and liver metastases. Investigative ophthalmology & visual
science. 2009;50(6):2572-80.
93. Sandoval J, Esteller M. Cancer epigenomics: beyond genomics. Current
opinion in genetics & development. 2012;22(1):50-5.
94. Waldmann T, Schneider R. Targeting histone modifications--epigenetics in
cancer. Current opinion in cell biology. 2013;25(2):184-9.
64
95. Wolffe AP, Matzke MA. Epigenetics: regulation through repression.
Science. 1999;286(5439):481-6.
96. Berger SL, Kouzarides T, Shiekhattar R, Shilatifard A. An operational
definition of epigenetics. Genes & development. 2009;23(7):781-3.
97. Zhao Y, Garcia BA. Comprehensive Catalog of Currently Documented
Histone Modifications. Cold Spring Harbor perspectives in biology. 2015;7(9).
98. Kouzarides T. Chromatin modifications and their function. Cell.
2007;128(4):693-705.
99. Bannister AJ, Kouzarides T. Regulation of chromatin by histone
modifications. Cell research. 2011;21(3):381-95.
100. Lin S, Gregory RI. MicroRNA biogenesis pathways in cancer. Nat Rev
Cancer. 2015;15(6):321-33.
101. Filipowicz W, Bhattacharyya SN, Sonenberg N. Mechanisms of post-
transcriptional regulation by microRNAs: are the answers in sight? Nature
reviews Genetics. 2008;9(2):102-14.
102. Smith ZD, Meissner A. DNA methylation: roles in mammalian
development. Nature reviews Genetics. 2013;14(3):204-20.
103. Takai D, Jones PA. Comprehensive analysis of CpG islands in human
chromosomes 21 and 22. Proceedings of the National Academy of Sciences of the
United States of America. 2002;99(6):3740-5.
104. Fatemi M, Pao MM, Jeong S, Gal-Yam EN, Egger G, Weisenberger DJ, et
al. Footprinting of mammalian promoters: use of a CpG DNA methyltransferase
revealing nucleosome positions at a single molecule level. Nucleic acids research.
2005;33(20):e176.
105. Fazzari MJ, Greally JM. Epigenomics: beyond CpG islands. Nature
reviews Genetics. 2004;5(6):446-55.
106. Bird AP, Wolffe AP. Methylation-induced repression--belts, braces, and
chromatin. Cell. 1999;99(5):451-4.
107. Yang X, Han H, De Carvalho DD, Lay FD, Jones PA, Liang G. Gene body
methylation can alter gene expression and is a therapeutic target in cancer. Cancer
Cell. 2014;26(4):577-90.
108. Robertson KD, Uzvolgyi E, Liang G, Talmadge C, Sumegi J, Gonzales FA,
et al. The human DNA methyltransferases (DNMTs) 1, 3a and 3b: coordinate
mRNA expression in normal tissues and overexpression in tumors. Nucleic acids
research. 1999;27(11):2291-8.
109. Probst AV, Dunleavy E, Almouzni G. Epigenetic inheritance during the
cell cycle. Nature reviews Molecular cell biology. 2009;10(3):192-206.
110. Du Q, Luu PL, Stirzaker C, Clark SJ. Methyl-CpG-binding domain
proteins: readers of the epigenome. Epigenomics. 2015:1-23.
111. Deaton AM, Bird A. CpG islands and the regulation of transcription. Genes
& development. 2011;25(10):1010-22.
65
112. Sadakierska-Chudy A, Kostrzewa RM, Filip M. A comprehensive view of
the epigenetic landscape part I: DNA methylation, passive and active DNA
demethylation pathways and histone variants. Neurotoxicity research.
2015;27(1):84-97.
113. Wu X, Zhang Y. TET-mediated active DNA demethylation: mechanism,
function and beyond. Nature reviews Genetics. 2017;18(9):517-34.
114. Schutsky EK, Nabel CS, Davis AKF, DeNizio JE, Kohli RM. APOBEC3A
efficiently deaminates methylated, but not TET-oxidized, cytosine bases in DNA.
Nucleic acids research. 2017;45(13):7655-65.
115. Ambrosi C, Manzo M, Baubec T. Dynamics and Context-Dependent Roles
of DNA Methylation. J Mol Biol. 2017;429(10):1459-75.
116. Brocato J, Costa M. Basic mechanics of DNA methylation and the unique
landscape of the DNA methylome in metal-induced carcinogenesis. Critical
reviews in toxicology. 2013;43(6):493-514.
117. Torano EG, Petrus S, Fernandez AF, Fraga MF. Global DNA
hypomethylation in cancer: review of validated methods and clinical significance.
Clinical chemistry and laboratory medicine : CCLM / FESCC. 2012;50(10):1733-
42.
118. Qu G, Dubeau L, Narayan A, Yu MC, Ehrlich M. Satellite DNA
hypomethylation vs. overall genomic hypomethylation in ovarian epithelial
tumors of different malignant potential. Mutation research. 1999;423(1-2):91-
101.
119. Xiao-Jie L, Hui-Ying X, Qi X, Jiang X, Shi-Jie M. LINE-1 in cancer:
multifaceted functions and potential clinical implications. Genetics in medicine :
official journal of the American College of Medical Genetics. 2015.
120. Rodriguez J, Vives L, Jorda M, Morales C, Munoz M, Vendrell E, et al.
Genome-wide tracking of unmethylated DNA Alu repeats in normal and cancer
cells. Nucleic acids research. 2008;36(3):770-84.
121. Chen RZ, Pettersson U, Beard C, Jackson-Grusby L, Jaenisch R. DNA
hypomethylation leads to elevated mutation rates. Nature. 1998;395(6697):89-93.
122. Eden A, Gaudet F, Waghmare A, Jaenisch R. Chromosomal instability and
tumors promoted by DNA hypomethylation. Science. 2003;300(5618):455.
123. Yagi S, Hirosawa M, Shiota K. DNA methylation profile: a composer-,
conductor-, and player-orchestrated Mammalian genome consisting of genes and
transposable genetic elements. The Journal of reproduction and development.
2012;58(3):265-73.
124. Anwar SL, Krech T, Hasemeier B, Schipper E, Schweitzer N, Vogel A, et
al. Loss of imprinting and allelic switching at the DLK1-MEG3 locus in human
hepatocellular carcinoma. PLoS One. 2012;7(11):e49462.
125. Yamashita N, Tokunaga E, Kitao H, Hitchins M, Inoue Y, Tanaka K, et al.
Epigenetic Inactivation of BRCA1 Through Promoter Hypermethylation and Its
66
Clinical Importance in Triple-Negative Breast Cancer. Clinical breast cancer.
2015;15(6):498-504.
126. Herman JG, Latif F, Weng Y, Lerman MI, Zbar B, Liu S, et al. Silencing
of the VHL tumor-suppressor gene by DNA methylation in renal carcinoma.
Proceedings of the National Academy of Sciences of the United States of
America. 1994;91(21):9700-4.
127. Rishi V, Bhattacharya P, Chatterjee R, Rozenberg J, Zhao J, Glass K, et al.
CpG methylation of half-CRE sequences creates C/EBPalpha binding sites that
activate some tissue-specific genes. Proceedings of the National Academy of
Sciences of the United States of America. 2010;107(47):20311-6.
128. Hesson LB, Sloane MA, Wong JW, Nunez AC, Srivastava S, Ng B, et al.
Altered promoter nucleosome positioning is an early event in gene silencing.
Epigenetics. 2014;9(10):1422-30.
129. Sproul D, Nestor C, Culley J, Dickson JH, Dixon JM, Harrison DJ, et al.
Transcriptionally repressed genes become aberrantly methylated and distinguish
tumors of different lineages in breast cancer. Proceedings of the National
Academy of Sciences of the United States of America. 2011;108(11):4364-9.
130. Moarii M, Boeva V, Vert JP, Reyal F. Changes in correlation between
promoter methylation and gene expression in cancer. BMC genomics.
2015;16(1):873.
131. Klutstein M, Moss J, Kaplan T, Cedar H. Contribution of epigenetic
mechanisms to variation in cancer risk among tissues. Proceedings of the National
Academy of Sciences of the United States of America. 2017;114(9):2230-4.
132. Derissen EJ, Beijnen JH, Schellens JH. Concise drug review: azacitidine
and decitabine. The oncologist. 2013;18(5):619-24.
133. Khan C, Pathe N, Fazal S, Lister J, Rossetti JM. Azacitidine in the
management of patients with myelodysplastic syndromes. Therapeutic advances
in hematology. 2012;3(6):355-73.
134. Fu S, Hu W, Iyer R, Kavanagh JJ, Coleman RL, Levenback CF, et al. Phase
1b-2a study to reverse platinum resistance through use of a hypomethylating
agent, azacitidine, in patients with platinum-resistant or platinum-refractory
epithelial ovarian cancer. Cancer. 2011;117(8):1661-9.
135. Stefanska B, Huang J, Bhattacharyya B, Suderman M, Hallett M, Han ZG,
et al. Definition of the landscape of promoter DNA hypomethylation in liver
cancer. Cancer research. 2011;71(17):5891-903.
136. Nguyen CT, Weisenberger DJ, Velicescu M, Gonzales FA, Lin JC, Liang
G, et al. Histone H3-lysine 9 methylation is associated with aberrant gene
silencing in cancer cells and is rapidly reversed by 5-aza-2'-deoxycytidine. Cancer
research. 2002;62(22):6456-61.
137. Coral S, Parisi G, Nicolay HJ, Colizzi F, Danielli R, Fratta E, et al.
Immunomodulatory activity of SGI-110, a 5-aza-2'-deoxycytidine-containing
67
demethylating dinucleotide. Cancer immunology, immunotherapy : CII.
2013;62(3):605-14.
138. van der Velden PA, Metzelaar-Blok JA, Bergman W, Monique H, Hurks
H, Frants RR, et al. Promoter hypermethylation: a common cause of reduced
p16(INK4a) expression in uveal melanoma. Cancer research. 2001;61(13):5303-
6.
139. Maat W, van der Velden PA, Out-Luiting C, Plug M, Dirks-Mulder A,
Jager MJ, et al. Epigenetic inactivation of RASSF1a in uveal melanoma.
Investigative ophthalmology & visual science. 2007;48(2):486-90.
140. van der Velden PA, Zuidervaart W, Hurks MH, Pavey S, Ksander BR,
Krijgsman E, et al. Expression profiling reveals that methylation of TIMP3 is
involved in uveal melanoma development. International journal of cancer Journal
international du cancer. 2003;106(4):472-9.
141. Maat W, Beiboer SH, Jager MJ, Luyten GP, Gruis NA, van der Velden PA.
Epigenetic regulation identifies RASEF as a tumor-suppressor gene in uveal
melanoma. Investigative ophthalmology & visual science. 2008;49(4):1291-8.
142. Neumann LC, Weinhausel A, Thomas S, Horsthemke B, Lohmann DR,
Zeschnigk M. EFS shows biallelic methylation in uveal melanoma with poor
prognosis as well as tissue-specific methylation. BMC cancer. 2011;11:380.
143. Landreville S, Agapova OA, Harbour JW. Emerging insights into the
molecular pathogenesis of uveal melanoma. Future oncology (London, England).
2008;4(5):629-36.
144. Field MG, Durante MA, Anbunathan H, Cai LZ, Decatur CL, Bowcock
AM, et al. Punctuated evolution of canonical genomic aberrations in uveal
melanoma. Nature communications. 2018;9(1):116.
145. Field MG, Kuznetsov JN, Bussies PL, Cai LZ, Alawa KA, Decatur CL, et
al. BAP1 Loss Is Associated with DNA Methylomic Repatterning in Highly
Aggressive Class 2 Uveal Melanomas. Clinical cancer research : an official
journal of the American Association for Cancer Research. 2019;25(18):5663-73.
146. Eissa S, Zohny SF, Shehata HH, Hegazy MG, Salem AM, Esmat M.
Urinary retinoic acid receptor-beta2 gene promoter methylation and
hyaluronidase activity as noninvasive tests for diagnosis of bladder cancer.
Clinical biochemistry. 2012;45(6):402-7.
147. Kawai Y, Sakano S, Suehiro Y, Okada T, Korenaga Y, Hara T, et al.
Methylation level of the RASSF1A promoter is an independent prognostic factor
for clear-cell renal cell carcinoma. Annals of oncology : official journal of the
European Society for Medical Oncology / ESMO. 2010;21(8):1612-7.
148. Avraham A, Uhlmann R, Shperber A, Birnbaum M, Sandbank J, Sella A,
et al. Serum DNA methylation for monitoring response to neoadjuvant
chemotherapy in breast cancer patients. International journal of cancer Journal
international du cancer. 2012;131(7):E1166-72.
68
149. Stone A, Zotenko E, Locke WJ, Korbie D, Millar EK, Pidsley R, et al. DNA
methylation of oestrogen-regulated enhancers defines endocrine sensitivity in
breast cancer. Nature communications. 2015;6:7758.
150. Pernia O, Belda-Iniesta C, Pulido V, Cortes-Sempere M, Rodriguez C,
Vera O, et al. Methylation status of IGFBP-3 as a useful clinical tool for deciding
on a concomitant radiotherapy. Epigenetics. 2014;9(11):1446-53.
151. Dullea A, Marignol L. MGMT testing allows for personalised therapy in
the temozolomide era. Tumour biology : the journal of the International Society
for Oncodevelopmental Biology and Medicine. 2016;37(1):87-96.
152. Li Y, Yang Y, Lu Y, Herman JG, Brock MV, Zhao P, et al. Predictive value
of CHFR and MLH1 methylation in human gastric cancer. Gastric cancer : official
journal of the International Gastric Cancer Association and the Japanese Gastric
Cancer Association. 2015;18(2):280-7.
153. Bachman M, Uribe-Lewis S, Yang X, Williams M, Murrell A,
Balasubramanian S. 5-Hydroxymethylcytosine is a predominantly stable DNA
modification. Nature chemistry. 2014;6(12):1049-55.
154. Lian CG, Xu Y, Ceol C, Wu F, Larson A, Dresser K, et al. Loss of 5-
hydroxymethylcytosine is an epigenetic hallmark of melanoma. Cell.
2012;150(6):1135-46.
155. 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.
2011;2(8):627-37.
156. Zhang Y, Wu K, Shao Y, Sui F, Yang Q, Shi B, et al. Decreased 5-
Hydroxymethylcytosine (5-hmC) predicts poor prognosis in early-stage laryngeal
squamous cell carcinoma. American journal of cancer research. 2016;6(5):1089-
98.
157. Liao Y, Gu J, Wu Y, Long X, Ge DI, Xu J, et al. Low level of 5-
Hydroxymethylcytosine predicts poor prognosis in non-small cell lung cancer.
Oncology letters. 2016;11(6):3753-60.
158. Song CX, Szulwach KE, Fu Y, Dai Q, Yi C, Li X, et al. Selective chemical
labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine. Nat
Biotechnol. 2011;29(1):68-72.
159. Song CX, Yi C, He C. Mapping recently identified nucleotide variants in
the genome and transcriptome. Nat Biotechnol. 2012;30(11):1107-16.
160. Thomson JP, Hunter JM, Nestor CE, Dunican DS, Terranova R, Moggs JG,
et al. Comparative analysis of affinity-based 5-hydroxymethylation enrichment
techniques. Nucleic Acids Res. 2013;41(22):e206.
161. Robertson J, Robertson AB, Klungland A. The presence of 5-
hydroxymethylcytosine at the gene promoter and not in the gene body negatively
regulates gene expression. Biochem Biophys Res Commun. 2011;411(1):40-3.
69
162. Lin IH, Chen YF, Hsu MT. Correlated 5-Hydroxymethylcytosine (5hmC)
and Gene Expression Profiles Underpin Gene and Organ-Specific Epigenetic
Regulation in Adult Mouse Brain and Liver. PLoS One. 2017;12(1):e0170779.
163. Langhans SA. Three-Dimensional in Vitro Cell Culture Models in Drug
Discovery and Drug Repositioning. Front Pharmacol. 2018;9:6.
164. Cui X, Hartanto Y, Zhang H. Advances in multicellular spheroids
formation. J R Soc Interface. 2017;14(127).
165. Gjorevski N, Sachs N, Manfrin A, Giger S, Bragina ME, Ordonez-Moran
P, et al. Designer matrices for intestinal stem cell and organoid culture. Nature.
2016;539(7630):560-4.
166. Jitcy Saji Joseph STM, Monde Ntwasa Two-Dimensional (2D) and Three-
Dimensional (3D) Cell Culturing in Drug Discovery, Cell Culture, 2018
[November 28th 2018:[Available from: https://www.intechopen.com/books/cell-
culture/two-dimensional-2d-and-three-dimensional-3d-cell-culturing-in-drug-
discovery.
167. Ness C, Garred O, Eide NA, Kumar T, Olstad OK, Baerland TP, et al.
Multicellular tumor spheroids of human uveal melanoma induce genes associated
with anoikis resistance, lipogenesis, and SSXs. Molecular vision. 2017;23:680-
94.
168. Lv D, Hu Z, Lu L, Lu H, Xu X. Three-dimensional cell culture: A powerful
tool in tumor research and drug discovery. Oncol Lett. 2017;14(6):6999-7010.
169. Coluccio ML, Perozziello G, Malara N, Parrotta E, Zhang P, Gentile F, et
al. Microfluidic platforms for cell cultures and investigations. Microelectronic
Engineering. 2019;208:14-28.
170. Valastyan S, Weinberg RA. Tumor metastasis: molecular insights and
evolving paradigms. Cell. 2011;147(2):275-92.
171. Yang H, Cao J, Grossniklaus HE. Uveal Melanoma Metastasis Models.
Ocul Oncol Pathol. 2015;1(3):151-60.
172. Kramer N, Walzl A, Unger C, Rosner M, Krupitza G, Hengstschlager M,
et al. In vitro cell migration and invasion assays. Mutation research.
2013;752(1):10-24.
173. Sleeboom JJF, Eslami Amirabadi H, Nair P, Sahlgren CM, den Toonder
JMJ. Metastasis in context: modeling the tumor microenvironment with cancer-
on-a-chip approaches. Dis Model Mech. 2018;11(3).
174. National Centre for the Replacement RaRoAiR. The 3Rs 2016 [Available
from: https://www.nc3rs.org.uk/the-3rs.
175. Lampreht Tratar U, Horvat S, Cemazar M. Transgenic Mouse Models in
Cancer Research. Front Oncol. 2018;8:268.
176. Cao J, Jager MJ. Animal Eye Models for Uveal Melanoma. Ocul Oncol
Pathol. 2015;1(3):141-50.
70
177. Stei MM, Loeffler KU, Holz FG, Herwig MC. Animal Models of Uveal
Melanoma: Methods, Applicability, and Limitations. BioMed research
international. 2016;2016:4521807.
178. Luyten GP, Naus NC, Mooy CM, Hagemeijer A, Kan-Mitchell J, Van
Drunen E, et al. Establishment and characterization of primary and metastatic
uveal melanoma cell lines. International journal of cancer Journal international du
cancer. 1996;66(3):380-7.
179. Hu DN, McCormick SA, Ritch R, Pelton-Henrion K. Studies of human
uveal melanocytes in vitro: isolation, purification and cultivation of human uveal
melanocytes. Investigative ophthalmology & visual science. 1993;34(7):2210-9.
180. Nareyeck G, Zeschnigk M, Bornfeld N, Anastassiou G. Novel cell lines
derived by long-term culture of primary uveal melanomas. Ophthalmologica
Journal international d'ophtalmologie International journal of ophthalmology
Zeitschrift fur Augenheilkunde. 2009;223(3):196-201.
181. Tura A, Luke J, Merz H, Reinsberg M, Luke M, Jager MJ, et al.
Identification of circulating melanoma cells in uveal melanoma patients by dual-
marker immunoenrichment. Investigative ophthalmology & visual science.
2014;55(7):4395-404.
182. Haastert K, Mauritz C, Chaturvedi S, Grothe C. Human and rat adult
Schwann cell cultures: fast and efficient enrichment and highly effective non-viral
transfection protocol. Nature protocols. 2007;2(1):99-104.
183. Angi M, Versluis M, Kalirai H. Culturing Uveal Melanoma Cells. Ocular
Oncology and Pathology. 2015;1:126-32.
184. Fang D, Nguyen TK, Leishear K, Finko R, Kulp AN, Hotz S, et al. A
tumorigenic subpopulation with stem cell properties in melanomas. Cancer
research. 2005;65(20):9328-37.
185. Ramos-Vara JA, Miller MA. When tissue antigens and antibodies get
along: revisiting the technical aspects of immunohistochemistry--the red, brown,
and blue technique. Veterinary pathology. 2014;51(1):42-87.
186. Arber DA. Effect of prolonged formalin fixation on the
immunohistochemical reactivity of breast markers. Applied
immunohistochemistry & molecular morphology : AIMM / official publication of
the Society for Applied Immunohistochemistry. 2002;10(2):183-6.
187. Engel KB, Moore HM. Effects of preanalytical variables on the detection
of proteins by immunohistochemistry in formalin-fixed, paraffin-embedded
tissue. Archives of pathology & laboratory medicine. 2011;135(5):537-43.
188. Herwig-Carl MC, Sharma A, Moulin A, Strack C, Loeffler KU. BAP1
Immunostaining in Uveal Melanoma: Potentials and Pitfalls. Ocul Oncol Pathol.
2018;4(5):297.
189. Rudbeck CTL. Immunohistochemical Staining Methods. Sixth edition ed.
DAKO AT, editor2013.
71
190. Gordon RE. Electron microscopy: a brief history and review of current
clinical application. Methods in molecular biology (Clifton, NJ). 2014;1180:119-
35.
191. Techonologies ABL. Real-time PCR handbook 2014 [Available from:
https://www.thermofisher.com/content/dam/LifeTech/global/Forms/PDF/real-
time-pcr-handbook.pdf.
192. Peirson SN, Butler JN. Quantitative polymerase chain reaction. Methods in
molecular biology (Clifton, NJ). 2007;362:349-62.
193. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the
comparative C(T) method. Nature protocols. 2008;3(6):1101-8.
194. Wang F, Flanagan J, Su N, Wang LC, Bui S, Nielson A, et al. RNAscope:
a novel in situ RNA analysis platform for formalin-fixed, paraffin-embedded
tissues. The Journal of molecular diagnostics : JMD. 2012;14(1):22-9.
195. Murdoch A, Jenkinson EJ, Johnson GD, Owen JJ. Alkaline phosphatase-
fast red, a new fluorescent label. Application in double labelling for cell cycle
analysis. Journal of immunological methods. 1990;132(1):45-9.
196. Bumgarner R. Overview of DNA microarrays: types, applications, and
their future. Current protocols in molecular biology / edited by Frederick M
Ausubel [et al]. 2013;Chapter 22:Unit 22.1.
197. Breitling R, Armengaud P, Amtmann A, Herzyk P. Rank products: a
simple, yet powerful, new method to detect differentially regulated genes in
replicated microarray experiments. FEBS letters. 2004;573(1-3):83-92.
198. de Ruijter TC, de Hoon JP, Slaats J, de Vries B, Janssen MJ, van Wezel T,
et al. Formalin-fixed, paraffin-embedded (FFPE) tissue epigenomics using
Infinium HumanMethylation450 BeadChip assays. Laboratory investigation; a
journal of technical methods and pathology. 2015;95(7):833-42.
199. Moran S, Vizoso M, Martinez-Cardus A, Gomez A, Matias-Guiu X,
Chiavenna SM, et al. Validation of DNA methylation profiling in formalin-fixed
paraffin-embedded samples using the Infinium HumanMethylation450
Microarray. Epigenetics. 2014;9(6):829-33.
200. Morris TJ, Beck S. Analysis pipelines and packages for Infinium
HumanMethylation450 BeadChip (450k) data. Methods (San Diego, Calif).
2015;72:3-8.
201. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP,
Hansen KD, et al. Minfi: a flexible and comprehensive Bioconductor package for
the analysis of Infinium DNA methylation microarrays. Bioinformatics (Oxford,
England). 2014;30(10):1363-9.
202. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers
differential expression analyses for RNA-sequencing and microarray studies.
Nucleic acids research. 2015;43(7):e47.
72
203. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate - a
Practical and Powerful Approach to Multiple Testing. J R Stat Soc B.
1995;57(1):289-300.
204. Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K, R VL, et al. De
novo identification of differentially methylated regions in the human genome.
Epigenetics Chromatin. 2015;8:6.
205. Hovestadt V, Zapatka M. Enhanced copy-number variation analysis using
Illumina DNA methylation arrays. R package version 1.9.0. 2017.
206. Souren NY, Gerdes LA, Lutsik P, Gasparoni G, Beltran E, Salhab A, et al.
DNA methylation signatures of monozygotic twins clinically discordant for
multiple sclerosis. Nature communications. 2019;10(1):2094.
207. Hnasko TS, Hnasko RM. The Western Blot. Methods in molecular biology
(Clifton, NJ). 2015;1318:87-96.
208. Mahmood T, Yang PC. Western blot: technique, theory, and trouble
shooting. N Am J Med Sci. 2012;4(9):429-34.
209. Callejo SA, Antecka E, Blanco PL, Edelstein C, Burnier MN, Jr.
Identification of circulating malignant cells and its correlation with prognostic
factors and treatment in uveal melanoma. A prospective longitudinal study. Eye
(London, England). 2007;21(6):752-9.
210. Eide N, Faye RS, Hoifodt HK, Sandstad B, Qvale G, Faber R, et al.
Immunomagnetic detection of micrometastatic cells in bone marrow of uveal
melanoma patients: a paradox. Acta ophthalmologica. 2015;93(1):59-66.
211. Buchheit CL, Weigel KJ, Schafer ZT. Cancer cell survival during
detachment from the ECM: multiple barriers to tumour progression. Nat Rev
Cancer. 2014;14(9):632-41.
212. Schafer ZT, Grassian AR, Song L, Jiang Z, Gerhart-Hines Z, Irie HY, et al.
Antioxidant and oncogene rescue of metabolic defects caused by loss of matrix
attachment. Nature. 2009;461(7260):109-13.
213. Paoli P, Giannoni E, Chiarugi P. Anoikis molecular pathways and its role
in cancer progression. Biochimica et biophysica acta. 2013;1833(12):3481-98.
214. Weiswald LB, Richon S, Validire P, Briffod M, Lai-Kuen R, Cordelieres
FP, et al. Newly characterised ex vivo colospheres as a three-dimensional colon
cancer cell model of tumour aggressiveness. British journal of cancer.
2009;101(3):473-82.
215. Emmink BL, Van Houdt WJ, Vries RG, Hoogwater FJ, Govaert KM,
Verheem A, et al. Differentiated human colorectal cancer cells protect tumor-
initiating cells from irinotecan. Gastroenterology. 2011;141(1):269-78.
216. Calvet CY, Andre FM, Mir LM. The culture of cancer cell lines as
tumorspheres does not systematically result in cancer stem cell enrichment. PLoS
One. 2014;9(2):e89644.
73
217. Smith HA, McNeel DG. The SSX family of cancer-testis antigens as target
proteins for tumor therapy. Clinical & developmental immunology.
2010;2010:150591.
218. D'Arcy P, Maruwge W, Wolahan B, Ma L, Brodin B. Oncogenic functions
of the cancer-testis antigen SSX on the proliferation, survival, and signaling
pathways of cancer cells. PLoS One. 2014;9(4):e95136.
219. Tirino V, Desiderio V, Paino F, Papaccio G, De Rosa M. Methods for
cancer stem cell detection and isolation. Methods in molecular biology (Clifton,
NJ). 2012;879:513-29.
220. Dontu G, Abdallah WM, Foley JM, Jackson KW, Clarke MF, Kawamura
MJ, et al. In vitro propagation and transcriptional profiling of human mammary
stem/progenitor cells. Genes & development. 2003;17(10):1253-70.
221. Warburg O. On the origin of cancer cells. Science. 1956;123(3191):309-
14.
222. Bensaad K, Favaro E, Lewis CA, Peck B, Lord S, Collins JM, et al. Fatty
acid uptake and lipid storage induced by HIF-1alpha contribute to cell growth and
survival after hypoxia-reoxygenation. Cell reports. 2014;9(1):349-65.
223. Rohrig F, Schulze A. The multifaceted roles of fatty acid synthesis in
cancer. Nat Rev Cancer. 2016;16(11):732-49.
224. Seddon JM, Gragoudas ES, Albert DM. Ciliary body and choroidal
melanomas treated by proton beam irradiation. Histopathologic study of eyes.
Archives of ophthalmology (Chicago, Ill : 1960). 1983;101(9):1402-8.
225. Driot JY, Rault J, Bonnin P, Liotet S. Electron microscopy of three cases
of choroid malignant melanomas. International ophthalmology. 1987;10(4):193-
202.
226. Khalil MK. Balloon cell malignant melanoma of the choroid:
ultrastructural studies. The British journal of ophthalmology. 1983;67(9):579-84.
227. Pike LS, Smift AL, Croteau NJ, Ferrick DA, Wu M. Inhibition of fatty acid
oxidation by etomoxir impairs NADPH production and increases reactive oxygen
species resulting in ATP depletion and cell death in human glioblastoma cells.
Biochimica et biophysica acta. 2011;1807(6):726-34.
228. Carracedo A, Cantley LC, Pandolfi PP. Cancer metabolism: fatty acid
oxidation in the limelight. Nat Rev Cancer. 2013;13(4):227-32.
229. Gray NE, Lam LN, Yang K, Zhou AY, Koliwad S, Wang JC. Angiopoietin-
like 4 (Angptl4) protein is a physiological mediator of intracellular lipolysis in
murine adipocytes. The Journal of biological chemistry. 2012;287(11):8444-56.
230. Chen M, Huang J. The expanded role of fatty acid metabolism in cancer:
new aspects and targets. Precis Clin Med. 2019;2(3):183-91.
231. Tureci O, Chen YT, Sahin U, Gure AO, Zwick C, Villena C, et al.
Expression of SSX genes in human tumors. International journal of cancer Journal
international du cancer. 1998;77(1):19-23.
74
232. Gjerstorff MF, Relster MM, Greve KB, Moeller JB, Elias D, Lindgreen JN,
et al. SSX2 is a novel DNA-binding protein that antagonizes polycomb group
body formation and gene repression. Nucleic acids research. 2014;42(18):11433-
46.
233. dos Santos NR, Torensma R, de Vries TJ, Schreurs MW, de Bruijn DR,
Kater-Baats E, et al. Heterogeneous expression of the SSX cancer/testis antigens
in human melanoma lesions and cell lines. Cancer research. 2000;60(6):1654-62.
234. Bowen WS, Svrivastava AK, Batra L, Barsoumian H, Shirwan H. Current
challenges for cancer vaccine adjuvant development. Expert review of vaccines.
2018;17(3):207-15.
235. von Ahlfen S, Missel A, Bendrat K, Schlumpberger M. Determinants of
RNA quality from FFPE samples. PLoS One. 2007;2(12):e1261.
236. Wimmer I, Troscher AR, Brunner F, Rubino SJ, Bien CG, Weiner HL, et
al. Systematic evaluation of RNA quality, microarray data reliability and pathway
analysis in fresh, fresh frozen and formalin-fixed paraffin-embedded tissue
samples. Scientific reports. 2018;8(1):6351.
237. Rohrich M, Koelsche C, Schrimpf D, Capper D, Sahm F, Kratz A, et al.
Methylation-based classification of benign and malignant peripheral nerve sheath
tumors. Acta neuropathologica. 2016;131(6):877-87.
238. Senbabaoglu Y, Michailidis G, Li JZ. Critical limitations of consensus
clustering in class discovery. Scientific reports. 2014;4:6207.
239. Quek K, Li J, Estecio M, Zhang J, Fujimoto J, Roarty E, et al. DNA
methylation intratumor heterogeneity in localized lung adenocarcinomas.
Oncotarget. 2017;8(13):21994-2002.
240. Shields CL, Say EAT, Hasanreisoglu M, Saktanasate J, Lawson BM, Landy
JE, et al. Cytogenetic Abnormalities in Uveal Melanoma Based on Tumor
Features and Size in 1059 Patients: The 2016 W. Richard Green Lecture.
Ophthalmology. 2017;124(5):609-18.
241. Barros-Silva D, Marques CJ, Henrique R, Jeronimo C. Profiling DNA
Methylation Based on Next-Generation Sequencing Approaches: New Insights
and Clinical Applications. Genes (Basel). 2018;9(9).
242. Hlady RA, Sathyanarayan A, Thompson JJ, Zhou D, Wu Q, Pham K, et al.
Integrating the Epigenome to Identify Drivers of Hepatocellular Carcinoma.
Hepatology. 2019;69(2):639-52.
243. Skvortsova K, Zotenko E, Luu PL, Gould CM, Nair SS, Clark SJ, et al.
Comprehensive evaluation of genome-wide 5-hydroxymethylcytosine profiling
approaches in human DNA. Epigenetics Chromatin. 2017;10:16.
244. Aasen T, Mesnil M, Naus CC, Lampe PD, Laird DW. Gap junctions and
cancer: communicating for 50 years. Nat Rev Cancer. 2016;16(12):775-88.
245. Bates DC, Sin WC, Aftab Q, Naus CC. Connexin43 enhances glioma
invasion by a mechanism involving the carboxy terminus. Glia.
2007;55(15):1554-64.
75
246. Gago-Fuentes R, Fernández-Puente P, Megias D, Carpintero-Fernández P,
Mateos J, Acea B, et al. Proteomic Analysis of Connexin 43 Reveals Novel
Interactors Related to Osteoarthritis. Mol Cell Proteomics. 2015;14(7):1831-45.
247. Leithe E, Mesnil M, Aasen T. The connexin 43 C-terminus: A tail of many
tales. Biochim Biophys Acta Biomembr. 2018;1860(1):48-64.
248. Su YA, Bittner ML, Chen Y, Tao L, Jiang Y, Zhang Y, et al. Identification
of tumor-suppressor genes using human melanoma cell lines UACC903,
UACC903(+6), and SRS3 by comparison of expression profiles. Mol Carcinog.
2000;28(2):119-27.
249. Tittarelli A, Guerrero I, Tempio F, Gleisner MA, Avalos I, Sabanegh S, et
al. Overexpression of connexin 43 reduces melanoma proliferative and metastatic
capacity. British journal of cancer. 2015;113(2):259-67.
250. Aasen T, Leithe E, Graham SV, Kameritsch P, Mayan MD, Mesnil M, et
al. Connexins in cancer: bridging the gap to the clinic. Oncogene.
2019;38(23):4429-51.
251. Alaga KC, Crawford M, Dagnino L, Laird DW. Aberrant Cx43 Expression
and Mislocalization in Metastatic Human Melanomas. J Cancer. 2017;8(7):1123-
8.
252. Cheng YJ, Chang MY, Chang WW, Wang WK, Liu CF, Lin ST, et al.
Resveratrol Enhances Chemosensitivity in Mouse Melanoma Model Through
Connexin 43 Upregulation. Environ Toxicol. 2015;30(8):877-86.
253. Schoumacher M, Le Corre S, Houy A, Mulugeta E, Stern MH, Roman-
Roman S, et al. Uveal melanoma cells are resistant to EZH2 inhibition regardless
of BAP1 status. Nature medicine. 2016;22(6):577-8.
254. LaFave LM, Beguelin W, Koche R, Teater M, Spitzer B, Chramiec A, et
al. Reply to "Uveal melanoma cells are resistant to EZH2 inhibition regardless of
BAP1 status". Nature medicine. 2016;22(6):578-9.
255. Matatall KA, Agapova OA, Onken MD, Worley LA, Bowcock AM,
Harbour JW. BAP1 deficiency causes loss of melanocytic cell identity in uveal
melanoma. BMC cancer. 2013;13:371.
256. Koundouros N, Poulogiannis G. Reprogramming of fatty acid metabolism
in cancer. British journal of cancer. 2020;122(1):4-22.
257. Wang L, Lee JY, Gao L, Yin J, Duan Y, Jimenez LA, et al. A DNA aptamer
for binding and inhibition of DNA methyltransferase 1. Nucleic acids research.
2019;47(22):11527-37.
258. Gounder M, Schöffski P, Jones RL, Agulnik M, Cote GM, Villalobos VM,
et al. Tazemetostat in advanced epithelioid sarcoma with loss of INI1/SMARCB1:
an international, open-label, phase 2 basket study. Lancet Oncol.
2020;21(11):1423-32.
76
Paper I-III
I
Uveal melanoma (UM) is the most common primary intraocular malignancy in adults with an incidence of approximately 5.1 per million per year in the United States [1], while in Europe, the incidence varies from less than 2 per million per year in Spain and southern Italy to more than 8 per million per year in Scandinavia [2,3]. Despite advances in the diagnosis and treatment of the disease, the prognosis has remained largely unchanged [1,4]. UM has a high propensity for metastatic spread. Relapse can be seen several years after treatment, and 40–50% of patients will eventually die of metastatic disease [4-7]. Dissemination of cancer cells from the primary tumor is believed to be an early event in UM. Circulating malignant cells (CMCs) have been detected in up to 88% of patients with UM and can be found at the time of diagnosis but also years after the primary tumor has been
removed [8]. Micrometastatic cells have also been found in the bone marrow of patients with UM in 29% of cases [9]. Intriguingly, the presence of disseminated cells in bone marrow and the bloodstream does not correlate with overall survival [8,10]. Cancer cells disseminating from the primary tumor have to adapt to a changing micromilieu to generate metastatic disease. The various tissues of the metastatic route provide a different nutritional supply, pH, and oxygen concen-tration; thus, the malignant cells have to exhibit metabolic flexibility to sustain growth and survival [11-13].
Anchorage-independent growth and resistance to anoikis (cell death induced by loss of extracellular matrix attachment as in circulating metastatic cells) are essential features of disseminated cancer cells and metastatic progression [14-16]. The generation of multicellular tumor spheroids (MCTS) by anchorage-independent growth is associated with enrichment of an aggressive phenotype characterized by chemoresistance, invasiveness, and expression of undifferentiated markers [17-21]. The present study aims to compare the differential
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680>Received 10 April 2017 | Accepted 1 October 2017 | Published 3 October 2017
© 2017 Molecular Vision
680
Multicellular tumor spheroids of human uveal melanoma induce genes associated with anoikis resistance, lipogenesis, and SSXs
Charlotte Ness,1,2 Øystein Garred,3 Nils A. Eide,1 Theresa Kumar,3 Ole K. Olstad,4 Thomas P. Bærland,1 Goran Petrovski,1,2 Morten C. Moe,1,2 Agate Noer1,2
(The last two authors co-senior authors for this study)
1Center for Eye Research, Department of Ophthalmology, Oslo University Hospital and University of Oslo, Norway; 2Norwegian Center for Stem Cell Research, Oslo University Hospital, Norway; 3Department of Pathology, Oslo University Hospital, Norway; 4Department of Medical Biochemistry, Oslo University Hospital, Norway
Purpose: Uveal melanoma (UM) has a high propensity for metastatic spread, and approximately 40–50% of patients die of metastatic disease. Metastases can be found at the time of diagnosis but also several years after the primary tumor has been removed. The survival of disseminated cancer cells is known to be linked to anchorage independence, anoikis resistance, and an adaptive cellular metabolism. The cultivation of cancer cells as multicellular tumor spheroids (MCTS) by anchorage-independent growth enriches for a more aggressive phenotype. The present study examines the differential gene expression of adherent cell cultures, non-adherent MCTS cultures, and uncultured tumor biopsies from three patients with UM. We elucidate the biochemical differences between the culture conditions to find whether the culture of UM as non-adherent MCTS could be linked to an anchorage-independent and more aggressive phenotype, thus unravelling potential targets for treatment of UM dissemination.Methods: The various culture conditions were evaluated with microarray analysis, quantitative reverse-transcription polymerase chain reaction (qRT-PCR), RNAscope, immunohistochemistry (IHC), and transmission electron microscopy (TEM) followed by gene expression bioinformatics.Results: The MCTS cultures displayed traits associated with anoikis resistance demonstrated by ANGPTL4 upregulation, and a shift toward a lipogenic profile by upregulation of ACOT1 (lipid metabolism), FADS1 (biosynthesis of unsaturated fatty acids), SC4MOL, DHCR7, LSS (cholesterol biosynthesis), OSBPL9 (intracellular lipid receptor), and PLIN2 (lipid storage). Additionally, the present study shows marked upregulation of synovial sarcoma X breakpoint proteins (SSXs), transcriptional repressors related to the Polycomb group (PcG) proteins that modulate epigenetic silencing of genes.Conclusions: The MCTS cultures displayed traits associated with anoikis resistance, a metabolic shift toward a lipogenic profile, and upregulation of SSXs, related to the PcG proteins.
Correspondence to: Agate Noer, Center for Eye Research, Department of Ophthalmology, Oslo University Hospital, Postboks 4956 Nydalen 0424 Oslo, Norway; Phone: +47 23 01 61 98; FAX: +47 22 11 80 00; email: [email protected]
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
681
gene expression of MCTS of UM to primary tumor tissue and adherent cultures, with a special emphasis on unravelling the pathways and survival mechanisms pathognomonic for disseminated and circulating cancer cells.
METHODS
All experiments were conducted in accordance with the Declaration of Helsinki (2013), and all tissue harvesting was approved by the Local Committees for Medical Research Ethics (REK Ref. 2009/1973 and REK Ref. 2013/803–1). The study is adhered to the ARVO statement on human subjects. Informed written consent was obtained from patients before tissue harvesting. All reagents used in the present study were from Sigma-Aldrich (St. Louis, MO) unless otherwise stated.
Biopsies and cell cultures: UM biopsies from patients under-going enucleation of the eye were included in this study. After enucleation, the ophthalmic pathologist excised fresh tumor tissue for use in research before formalin fixation for routine histopathological examination. The UM of the three donors (D1, D2, and D3) was classified as mixed (D1) and epithelioid (D2 and D3) types with a routine histopathological examination. Retrospectively, donors D1, D2, and D3 all had confirmed liver metastases. A fourth supplementary donor was added to the study after data were obtained. The UM of this donor, D(S), was classified as epithelioid, and the donor tissue underwent the same culture conditions as the tissue from donors D1, D2, and D3.
A fraction of the tissue was snap-frozen and stored at −80 °C. The remaining sample was minced with scissors in collagenase I and IV (1 mg/ml), before being incubated for 1 h at 37 °C. After dissociation, the tissue was cultured adherently for 7 days in RPMI 1640 (Invitrogen, Carlsbad, CA), 10% fetal bovine serum (FBS), penicillin/streptomycin (100 U/ml, P4333), and amphotericin B (2.5 µg/ml, A2942) in addition to gentamycin (75 µg/ml; Sanofi-Aventis, Gentilly, France) to ensure the removal of fibroblasts [22]. After 7 days of adherent culturing, the cells were trypsinized using Trypsin-EDTA (0.25%, T4049) and pelleted into three frac-tions of 100,000 cells. The first fraction of the cells was collected for RNA analyses, the second fraction for further adherent growth, and the third for non-adherent growth as MCTS. The term MCTS is used for this non-adherent culture of tumor cells, in accordance with the nomenclature, and is considered aggregation and compaction of tumor cells [21]. The cell fraction for MCTS culture was plated at a density of 500–1,000 cells per well on Corning Costar ultra-low attach-ment, polystyrene, round-bottom 96-well plates (CLS7007) in melanoma stem cell medium (MSCM) (1) and (2): (1) 30% human embryonic stem cell medium (hESC); (78% KnockOut
DMEM/F12 (Cat. no. 12660–012, Thermo Fisher Scientific Inc., Waltham, MA), 20% KnockOut serum replacer (Cat. no. 10828–028, Thermo Fisher Scientific Inc.), 1% MEM non-essential amino acids (Cat. no. 11140–050, Thermo Fisher Scientific Inc.), 4 ng/ml basic fibroblast growth factor (b-FGF; Cat. no. 13,256-029, Thermo Fisher Scientific Inc.), 1% GlutaMAX (35,050-061, Thermo Fisher Scientific Inc.), and 1.4‰ 2-mercaptoethanol (M7522) and (2) 70% mouse embryonic fibroblast (MEF) conditioned medium (AR005, R&D Systems/Bio-Techne, Minneapolis, MN) [23] with peni-cillin/streptomycin (100 U/ml) and amphotericin B (2.5 µg/ml). The cells were collected after 12 days of cell culture and further embedded in paraffin for immunohistochemistry (IHC) or pelleted and stored at −80 °C for RNA analyses.
RNA isolation: RNA from fresh frozen primary tumors (D1, D2, and D3) was isolated using the Qiagen RNeasy kit (Qiagen, Hilden, Germany). Briefly, the tissue was placed in a 4.5 ml cryotube, and 500 µl of QIAzol (Qiagen) was added before the sample was disrupted using Qiagen TissueRuptor (Qiagen), according to the manufacturer’s recommenda-tions. The sample was centrifuged at 18 400 ×g for 10 min to remove insoluble material before being processed with the Qiagen RNeasy kit with DNase. Samples were purified using the Zymo PCR inhibitor removal kit (Zymo, Irvine, CA). RNA from the pelleted samples (adherent and cultured spheres from D1, D2, and D3) was isolated as described above, except the disruption step using the Qiagen Tissu-eRuptor. RNA concentration and purity were determined using NanoDrop (Wilmington, DE) and Bioanalyzer (Agilent 2100, Agilent, Santa Clara, CA). All nine samples had RNA integrity number (RIN) values above 8 before being analyzed with microarray and PCR [24].
Immunohistochemistry: The growth media in the 96-well plates was diluted by gently adding Hanks’ Balanced Salt solution (Thermo Fisher Scientific Inc.). Then the MCTS were allowed to make sediment before the media was care-fully removed. A mixture of human plasma and thrombin (Sigma–Aldrich) was used to clot the MCTS together before fixation in 4% paraformaldehyde (PFA) and embedment in paraffin. Then 3.5 μm sections were cut and stained [25]. Ki-67 staining was performed using the Envision + Dual Link HRP (K4065, Dako, Glostrup, Denmark) and AEC + Substrate chromogen ready-to-use (k3461, Dako). Briefly, the K4065 kit protocol was followed until the addition of 3,3′-diaminobenzidine (DAB). After polymer horseradish peroxidase (HRP), 3-amino-9-ethylcarbazole (AEC) chro-mogen from the kit k3461 was added, and the sections were washed and counterstained with hematoxylin according to the k3461 protocol. Negative controls without primary antibody
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
682
were included for all stainings. The following primary antibodies and dilutions were used (rabbit:rb, mouse:ms): Ki-67 (rb, 1:200; Thermo Fisher Scientific Inc.), SSX4 (rb, 1;50; Acris), anti-melanoma (a-melanoma) [HMB45 + MART1 (DT101 + BC199) + tyrosinase (T311)] (ms, 1:50; ab733, Abcam, Cambridge, UK), Perilipin (rb, 1:100; Santa Cruz Biotechnology Inc., Dallas, TX), and ANGPTL4 (rb, 1:500; Abcam). The secondary antibodies had the fluorescent marker Alexa Fluor 488 (1:500; Invitrogen). Hoechst (1:500; Invitrogen) was used for nuclear staining. The sections were analyzed using a Zeiss Axio Observer.Z1 fluorescence micro-scope (Zeiss, Oberkochen, Germany). Sections were also stained with hematoxylin and eosin (H&E) for morphological examination.
Microarray: Microarray analysis was performed at the Genomics Core Facility, Oslo University Hospital and Helse Sør-Øst. HumanHT-12 v4 Expression BeadChip (Illumina, San Diego, CA) was used for the analysis. It targets more than 31,000 annotated genes with 47,000 probes mainly derived from the National Center for Biotechnology Information Reference Sequence (NCBI) RefSeq Release 38 (November 7, 2009). For each sample, 440 ng of total RNA was amplified and labeled using the Illumina TotalPrep-96 RNA Amplifica-tion Kit protocol. The quantity of labeled copy RNA (cRNA) was measured using the NanoDrop spectrophotometer (Wilmington, DE). The quality and size distribution of the labeled cRNA were assessed using the 2100 Bioanalyzer. This was done to be able to hybridize equal amounts of successfully labeled cRNA to the arrays. For each sample, 750 ng of biotin-labeled cRNA was hybridized to the Illumina HumanHT-12 v4 Expression BeadChip. J-Express and rank product (RP) analysis were used to further identify differ-ently expressed genes with ≥2 fold up- or downregulation and q values ≤0.05 between the different groups. One thousand permutations (1,000*) were run for each RP analysis [26].
Quantitative reverse-transcription PCR: RNA concentration and purity were measured using NanoDrop. Reverse tran-scription (RT) was performed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Abingdon, UK) with 50 ng total RNA per 20 μl RT reaction. Copy DNA (cDNA) was diluted to a volume of 50 µl (1 ng/µl) after cDNA synthesis. Quantitative PCR (qPCR) was performed using the StepOnePlus RT–PCR system (Applied Biosystems) and Taqman Gene Expression assays following the manufac-turer’s protocols (Applied Biosystems). The TaqMan Gene Expression Assays used include ANGPTL4 (Hs01101127_m1) and 18S (Hs03003631_g1). The thermal cycling conditions were 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. All samples were run in duplicate (each
reaction: 2.5 μl/2.5 ng cDNA in a total volume of 12.5 μl). The data were analyzed using the 2−ΔΔCt method to find the relative changes in gene expression as a fold change between the samples. The uncultured tumor sample was chosen as the calibrator and equaled one, while the other samples had fold changes related to the uncultured tumor calibrator sample. The 18S probe, primers, and assay (Hs03003631_g1) were used as a loading control to quantify the differences in cDNA input between the samples.
RNAscope in situ hybridization: RNA in situ hybridization was performed using the RNAscope® 2.5 High Definition (HD)- Red assay (Advanced Cell Diagnostics, Hayward, CA) according to the manufacturer’s instructions using the standard pretreatment protocol. Sections were mounted using Prolong Gold with 4',6-diamidine-2'-phenylindole dihydro-chloride (DAPI). RNAscope permits direct visualization of RNA in formalin-fixed, paraffin-embedded (FFPE) tissue with single molecule sensitivity and single cell resolution [27]. RNAscope Probe-Hs-SSX4–01 (Cat. no. 468,641, Advanced Cell Diagnostics) was used. Hybridization signals were detected with chromogenic reactions using Fast Red. Fast Red produces red fluorescence in addition to the red reaction product, thus providing a greater level of sensitivity [28]. The RNA staining signal was identified as red punctate dots. Each sample was quality controlled for RNA integrity with a probe specific to peptidyl-prolyl cis-trans isomerase B (PPIB) mRNA. Negative control background staining was evaluated using a probe (Cat.no. 3100439, Advanced Cell Diagnostics, Newark, NJ) specific to the bacterial dihydro-dipicolinate reductase (DapB) gene (Gene ID EF191515). The sections were analyzed with a Zeiss Axio Observer.Z1 fluorescence microscope.
Pathway and gene ontology analysis: Data from the micro-array analysis were imported into Ingenuity Pathway Analysis (IPA) software in the search for biologic pathways and Gene Ontology to identify potential networks. Principal component analysis (PCA) and unsupervised hierarchical clustering were performed using the Partek Genomics Suite software (Partek, Inc., Chesterfield, MO).
Transmission electron microscopy: Primary tissue from uncultured tumor D1 and the donor D(S) cultured as MCTS were fixed at 4 °C overnight in glutaraldehyde (0.1 M). The tissue was washed four times in cacodylate buffer (0.2 M) before post-fixation in a mixture of 1% osmium tetroxide and cacodylate buffer (0.2 M) for 60 min. The tissue was further rinsed in cacodylate buffer (0.2 M) before being dehydrated through a graded series of ethanol up to 100%. The tissue was then immersed in propylene oxide for 2 ×5 min and a mixture of Epon and propylene oxide before embedment in
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
683
Epon. Ultrathin sections (60–70 nm thick) were cut on a Leica Ultracut Ultramicrotome UCT (Leica, Wetzlar, Germany), stained with uranyl acetate and lead citrate, and examined using a Tecnai12 transmission electron microscope (Phillips, Amsterdam, the Netherlands).
RESULTS
Cultivation of uveal melanoma: The cells cultured as MCTS grew as large aggregations involving the majority of the cells in the well (Figure 1A and insets). The MCTS were mitoti-cally active, as seen with the positive Ki67 staining with a score of 1%, 2%, and 4% for donors D1, D2, and D3, respec-tively (Figure 1C). The melanoma profile of the MCTS was verified by staining for a-melanoma, a marker that recognizes HMB-45, MART-1, and tyrosinase. More than 90% of the
cells in the MCTS-derived paraffin sections stained positive for this marker (Figure 1D).
Genetic clustering is determined by the culture conditions: The gene expression profiles of the UMs (D1, D2, and D3), uncultured, cultured as MCTS, or cultured as adherent primary cells, were comprehensively analyzed with micro-array analysis. PCA was performed on raw data from the microarray with a false discovery rate (FDR) of 10%. This type of analysis clusters the samples and represents them on a three-dimensional space based on the differential relative gene expression. The PCA plot shows that the clustering was mainly determined by the culture conditions (Figure 2A).
The relative gene expression of UMs (uncultured, cultured as MCTS, or cultured as adherent primary tumor cells) was further investigated by performing an unsupervised
Figure 1. Multicellular tumor spheroid culture of primary uveal melanoma cells. A: Single cells (upper inset) after primary tumor isolation, during cultivation small pigmented tumor spheres formed (lower inset), and further developing resulting into large spheroid structures if not passaged. B: Adherent cell culture of primary uveal melanoma (UM) cells. C: Ki67 staining (*) of UM multicellular tumor spheroid (MCTS). D: Immunohistochemical staining of antimelanoma (green) and Hoechst staining of the nucleus (blue; right panel) with the corresponding light-microscopic image (left panel) of UM MCTS.
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
684
hierarchical clustering with an FDR of 10% presented as a heat map (Figure 2B). The heat map shows significant downregulation of the surface markers in the cultured cells compared to the uncultured primary tumor biopsy. These markers reflect the cellular heterogeneity of the primary tumor and the loss of macrophages (CD68), endothelial cells (von Willebrand factor), and T-cells (CD3D, CD8A, and CD2) in the cell cultures (Table 1). Additionally, there is a marked downregulation of human leukocyte antigen (HLA) expres-sion in MCTS (Table 1) and in the adherent primary tumor cells (Appendix 1). This finding is in accordance with the work of van Essen et al. who showed downregulation of HLA expression upon loss of tumor-infiltrating leukocytes [29].
The genes found to be upregulated in the unsupervised hierarchical clustering (Figure 2B) were in concordance with many of the genes found in the RP analysis (Table 1 and supplementary data). The RP analysis (q≤0.05) resulted in 206 genes ≥2 fold upregulated and 373 genes ≥2 fold down-regulated in MCTS versus uncultured tumor biopsies. Two
hundred eighteen genes were found to be ≥2 fold upregulated, and 552 genes were ≥2 fold downregulated in adherent cell cultures versus the uncultured tumor biopsies. Sixty-four genes were found to be ≥2 fold upregulated, and 71 genes were ≥2 fold downregulated in adherent cell cultures versus the MCTS.
The genes from the RP analysis were further analyzed with Ingenuity IPA software. The differences in molecular and cellular functions between the various culture conditions are shown in Figure 3.
There was a noticeable increase in the cellular strain in the MCTS compared to the uncultured tumor biopsies, indicated by increased free radical scavenging, enhanced drug metabolism, and the increase in lipid metabolism in the MCTS versus adherent cells and uncultured tumor biopsies. Associated pathways and molecules in lipid metabolism in the MCTS versus uncultured tumor biopsies are shown in Figure 3. Alterations in the lipid metabolism include seven networks:
Figure 2. Gene expression in uveal melanoma donors (D1, D2, and D3) cultured as primary adherent cells (red), multicellular tumor spheroids (blue), and uncultured tumor biopsies (green). A: Prin-cipal component analysis (PCA) plot of gene expression in uveal melanoma donors (D1, D2, and D3) cultured as primary adherent cells (red), multicellular tumor spher-oids (blue), and uncultured tumor biopsies (green). B: Hierarchical clustering of gene expression in uveal melanoma donors (D1, D2, and D3), where each row represents the single sample tested: adherent cultures (D1, D2, and D3; red), multicellular tumor spheroids (MCTS; D1, D2, and D3; blue), and uncultured tumors (D1, D2, and D3; green), while each column represents a single probe set (gene symbol or Illumina ID number) analyzed. Relative gene expression
is presented in color: Red is higher-level expression relative to the sample mean, blue is relatively lower level expression, and gray is no change in expression.
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
685
synthesis of lipids, steroid metabolism, metabolism of choles-terol, metabolism of lipid membrane derivatives, synthesis of cholesterol, and conversion of lipid and fatty acid metabolism.
MCTS display a genetic profile indicating EMT and anoikis resistance: Anoikis is a form of apoptosis induced by loss or inappropriate cell adhesion [30]. The process of epithelial-to-mesenchymal-transition (EMT) is considered an important feature of anoikis [31]. Rank product data revealed 3.5-fold upregulation of snail family transcriptional repressor 2 (SNAI2; Gene ID: 6591, OMIM 602150) and 0.6-fold downregulation of cadherin 1 (CDH1; Gene ID: 999, OMIM 192090; E-cadherin) in MCTS (Table 1). Anoikis resistance is also supported by the upregulation of pyruvate
dehydrogenase kinase 4 (PDK4; Gene ID: 5166, OMIM 602527), an enzyme that inactivates pyruvate dehydrogenase (PDH), which is required for the conversion of pyruvate to acetyl-CoA. PDK4 is upregulated in response to loss of adher-ence (LOA) and reduces reactive oxygen species (ROS) strain [32]. Noticeably, there was strong upregulation of angiopoi-etin like 4 (ANGPTL4; Gene ID: 51129, OMIM 605910) in the MCTS (Table 1, Figure 4). ANGPTL4 has recently been shown to be associated with an angiogenic phenotype of UM, and thus being involved in metastatic spread [33]. ANGPTL4 is thought to contribute to anoikis resistance by inducing conformational changes that enable resistance to inducers of apoptosis [34,35]. ANGPTL4 is further known to stimulate
Table 1. lisT of selecTed genes, including The Ten mosT up- and downregulaTed, from The microarray rank product (rp) analysis (≥ 2fold up- or down- regulated, q≤0.05) in multicellular tumour spheroids (mcts) versus
uncultured tumours and mcts versus adherent cultures (see supplementary data for the complete list).
Up in MCTS vs. uncultured tumours
Down in MCTS vs. uncultured tumours
Up in MCTS vs. adherent cultures
Down in MCTS vs. adherent cultures
Gene symbol
Fold change
Gene symbol
Fold change
Gene symbol
Fold change
Gene symbol
Fold change
ANGPTL4 27.1 HLA-DRA -32.1 ANGPTL4 21.1 VGF -11SSX4 6.4 CD74 -22.9 SSX4 6.2 ID3 -10.7ASPA 4.7 C1QB -17.8 ASPA 4.6 MIR1974 -5.9SSX2 4.5 VWF -14 SSX2 4.3 ILMN_1881909 -3.9LDLR 4.4 CD14 -12.2 APOD 3.6 ID2 -4.6MT1X 4.3 C1QC -11.7 IL17D 3.2 CTGF -4.1HTR2B 4.3 HLA-DMB -11.1 NRXN2 3.2 ID1 -3.9FCRLA 4.1 HLA-DRB1 -11.7 MT1X 3.1 SRGN -3.1SQLE 4 HLA-DMB -11.1 COL16A1 3 NPTX1 -2.8PRUNE2 4 HLA-DPA1 -10.4 MAL 2.8 PENK -2.7SLC2A10 3.8 ARHGDIB -9.7 BMF 2.7 CAPS -2.4SNAI2 3.5 TYROBP -8.7 SSX5 2.7 ODC1 -2.3FADS1 2.9 SLC15A3 -6.4 MT1G 2.5 RNU1A3 -2.3ECH1 2.7 HBA2 -6.3 AEBP1 2.5 CYR61 -2.3PLIN2 2.5 HBB -5.5 CDH19 2.4 LOC389342 -2.3DHCR7 2.5 ITGB2 -5.2 CLCNKA 2.4 MAL2 -2.1OSBPL9 2.5 IL18BP -4.9 PKNOX2 2.4 CDCA7 -2.1BMF 2.5 SNORD3A -4.5 PDK4 2.2 WFDC1 -2.1PDK4 2.5 CD68 -3.7 MT2A 2.2 HSP90B1 -2.2LSS 2.4 CXCL16 -3.5 SLC2A10 2.2 IFI6 -2.2MT2A 2.4 CD8A -3.3 GPR125 2.2 LAMA1 -2.2SC4MOL 2.3 CD3D -2.7 LSS 2.1 THBS2 -2.2PECI 2.1 CDH1 -2.6 CREB1 2.1 CTSL1 -2.2MT1E 2.1 VCAM1 -2.1 MT1E 2.1 EIF5A -2ACOT1 2 CD2 -2.1 FADS1 2 QPCT -2
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
686
intracellular lipolysis, thus supplying substrate for fatty acid oxidation (FAO) [35]. Upregulation of FAO in MCTS is indicated by upregulation of enoyl-CoA hydratase 1 (ECH1; Gene ID: 1891, OMIM 600696), peroxisomal D3,D2-enoyl-CoA isomerase (PECI; Gene ID: 10455, OMIM 608024), and acyl-CoA thioesterase 1 (ACOT1, Gene ID: 641371, OMIM 614313; Table 1). FAO has been proven to be an advantageous metabolic trait for cancer cells and is linked to anoikis resis-tance [36].
MCTS culture conditions induce a metabolic shift toward a lipogenic profile: Microarray results indicated a metabolic shift toward a lipogenic profile in the MCTS. A high content of lipid droplets (LDs) and stored-cholesterol ester is strongly associated with tumor aggressiveness [37,38]. As shown in
Table 1, ACOT1 (lipid metabolism), fatty acid desaturase 1 (FADS1, Gene ID: 3992, OMIM 606148; biosynthesis of unsaturated fatty acids), sterol-C4-mehtyl oxidase-like (SC4MOL, Gene ID: 6307, OMIM 607545), 7-dehydrocho-lesterol reductase (DHCR7, Gene ID: 1717, OMIM 602858), lanosterol synthase (LSS, Gene ID: 4047, OMIM 600909; cholesterol biosynthesis), and oxysterol binding protein like 9 (OSBPL9, Gene ID: 114883, OMIM 606737; intracel-lular lipid receptor) all showed marked upregulation in the MCTS cultures compared to primary tumors. SC4MOL, LSS, and FAD1 were also found to be upregulated in the MCTS cultures compared to the adherent cultures. The microarray results also demonstrated increased lipid storage by upregula-tion of perilipin 2 (PLIN2, Gene ID: 123, OMIM 103195). PLIN2 belongs to the perilipin family, members of which
Figure 3. Molecular and cellular functions being upregulated in tumor biopsies versus multicellular tumor spheroids (upper left panel), multicellular tumor spheroids versus tumor biopsies (upper right panel), multicellular tumor spheroids versus adherent cultures (lower left panel), and adherent cultures versus multicellular tumor spheroids (lower right panel). The number of molecules upregulated is shown in brackets. MCTS = tumors cultivated as multicellular tumor spheroids; adherent cultures = adherent cultivated tumors; tumor biopsies = uncultured primary tumor tissue.
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
687
coat intracellular lipid storage droplets [39]. The presence of PLIN2 was verified with IHC of donors D1, D2, and D3 (Figure 4). Morphological examination with transmission electron microscopy (TEM) revealed numerous lipid droplets in the supplementary donor D(S) cultured as MCTS (Figure 5). The TEM images also showed numerous mitochondria.
MCTS cultures increase the expression of cancer and testis antigens: The synovial sarcoma X breakpoint (SSX, Gene ID: 6759, OMIM 300326) gene family consists of nine highly homologous members (SSX1–9) [40]. SSX expression is confined to the testis, placenta, at low levels in the thyroid, and in a wide range of tumors (including synovial sarcoma), thus making them interesting targets for cancer therapy [41]. SSXs have been linked to EMT and anoikis resistance [42].
The microarray results showed an increase in the expres-sion of SSX4 in MCTS versus primary tumors and adherent cultures (Table 1). This presence of SSX4 mRNA was verified with RNAscope, while the SSX4 protein was verified with IHC staining. The proportion of cells expressing SSX4 in
primary tumors and MCTS was found (Figure 6). Notice-ably, the SSX protein was minimally expressed in the tumor biopsies.
DISCUSSION
By comparing the UM MCTS to biopsies and adherent cell cultures, the present study revealed a metabolic shift in the MCTS. The latter display traits associated with anoikis resis-tance, including a shift toward a lipogenic profile, as well as marked upregulation of SSXs, transcriptional repressors capable of humoral and cellular immune responses in cancer patients and putative targets for immunotherapy in cancers.
To disseminate, cancer cells have to undergo loss of adherence. Loss of adherence inhibits uptake of glucose and glycolysis which results in diminished levels of ATP and NADPH leading to metabolic stress and generation of ROS that induces anoikis [15]. The induction of FAO restores ATP production and increases NADPH, thus preventing anoikis [15,43]. This metabolic shift is also indicated in the MCTS
Figure 4. Lipogenic profile of uveal melanoma multicellular tumor spheroids. Angiopoietin like 4 (ANGPTL4; green) staining of multicel-lular tumor spheroids (MCTS), Hoechst staining of nucleus (blue; A) with corresponding light-microscopic image (B). Perilipin 2 (PLIN2) staining (green) of MCTS and Hoechst staining of the nucleus (blue; C) with the corresponding light microscopic image (D). E: Quantitative reverse-transcription PCR (qRT-PCR) of ANGPTL4 in support of the microarray finding. F: Ingenuity Pathway Analysis (IPA) based on rank product (q≤0.05) in MCTS versus the tumor, showing important molecules and pathways, including seven networks and their associated upregulated molecules in lipid metabolism. Deep red indicates more pronounced expression, and numbers below the gene symbols reflect the fold change (number on top) and q value/significance (number below).
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
688
in the present study. Malignant cells have been shown to provide and utilize fatty acids [36]. The lipogenic profile of the MCTS-derived cells reveals an increase in the synthesis of cholesterol, a trait associated with cancer aggressiveness [38,44,45]. Depletion of cholesterol has been shown to result in anoikis-like cell death [46]. Whether the lipogenic switch seen in the MCTS in the present study is valid for in vivo disseminated UM cells remains to be revealed. Lipogenic
targeting could be advantageous for solid tumors. The present study showed abundant LDs in the MCTS and in the primary tumor. The presence of LDs in UM has been described in the literature previously, as a response to radiation and in the untreated tumor tissue [47,48]. UM is characterized by its poor response to chemotherapeutics, and FAO has been shown to fuel chemoresistant cancer cells [49]. Several FAO inhibi-tors have shown promising results in mice models, although
Figure 5. Transmission electron microscopy of uveal melanoma. A: Uveal melanoma biopsy with nucleus (n), lipid droplets (li), pigment (p), mitocondria (m) and interdigitations (*) between cells. B–D: In the multicellular tumor spheroids (MCTS), the cells were less packed but contained abundant lipid droplets, pigment, interdigitations, and a dense concentration of mitochondria. D: Adherence-like junctions (***) between cells were also evident (inset). Scale bars: A, 5 μm; B, 10 μm; C, 1 μm; D, 1 μm.
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
689
Figure 6. Immunohistochemistry analysis of uveal melanoma multicellular tumor spheroids shows positive staining for SSX4 (green), Hoechst staining of nucleus (blue; A) with corresponding light-microscopic image (B). The presence of synovial sarcoma X breakpoint protein 4 (SSX4) was verified with RNAscope staining (red), Hoechst staining of nucleus (blue; C) where SSX4 RNA transcripts are shown as red chromogenic dots, and with the corresponding light-microscopic image (D). E: Percentage of SSX4-positive cells in multicellular tumor spheroids (MCTS) (D1, D2, and D3) versus uncultured primary tumors (D1, D2, and D3) analyzed with immunohistochemistry (IHC).
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
690
chemosensitization by FAO and cholesterol synthesis inhibi-tors might be even more favorable [50-52]. The present results imply that ANGPTL4 might be a key player in orchestrating lipid metabolism in MCTS. ANGPTL4 has previously been shown to play a role in anoikis resistance and in angiogenesis and oncogenesis of several cancers, including UM [34,53-55]. The link between ANGPTL4 and EMT has recently been highlighted as a decrease in EMT markers and aggressiveness after silencing of ANGPTL4 in non-small cell lung cancer [56]. Although most publications indicate an oncogenic func-tion of ANGPTL4, the opposite has been shown in gastric cancer, where it is a proposed tumor suppressor [57]. These conflicting findings suggest that further characterization of ANGPTL4 in UM is needed. The present results suggest that ANGPTL4 could be an attractive target in UM and possibly a way to target disseminated cancer cells.
Another compelling finding in the present study is the marked upregulation of SSX4. SSXs show tissue-restricted expression and are therefore regarded as attractive targets for cancer therapy [40,58]. The proteins are implied to be involved in proliferation and survival in cancer cells and formed a transient complex with beta-catenin thus altering the expression of genes involved in EMT [59]. SSXs are localized to the nucleus and contains two different repressor domains: a Krüppel associated box (KRAB) domain and a potent repressor domain (RD) [60,61]. SSXs have a close connection with the Polycomb repressive group of proteins [62,63]. SSX2 (a homologous SSX group member) has been shown to antagonize BMI1 and EZH2 through an indirect mechanism, thus activating repressed genes. Additionally, SSX2 has been shown to have DNA-binding properties and negatively regulate the distribution of histone mark H3K27me3, implying that SSX2 plays a role in the regula-tion of chromatin structure and function [64]. The exact function of SSXs in UM is not known, although the link between EMT and SSXs highlights a potential role in anoikis resistance. Disseminated cancer cells are likely to have an altered metabolic state as a survival strategy, and SSXs with their gene-regulating properties might be essential for these alterations. The synovial sarcoma fusion protein SS18-SSX2 has been associated with induction of cholesterol synthesis [65]. Whether there is a direct link between lipid metabo-lism and SSXs in UM is yet to be unveiled. SSX4 has been shown to be expressed in 21% of skin melanomas; however, SSX4 expression in UM has not yet been assessed [41]. If SSXs are highly expressed in disseminated cancer cells, it would make them valuable targets for immunotherapy. The restricted tissue expression of SSXs might lead to less severe side effects than targeting molecules and pathways involved in normal cellular homeostasis.
Cell culturing of UM is often hampered by tumor size and growth properties. A limitation of the present study is the low number and histological homogeneity of the donors included. In our experience, spindle cell tumors are more challenging to cultivate, thus making it difficult to run extensive genomic analyses on this cell type. Tumor size is an important aspect in UM research as the relative size of the tumors is small compared to other cancers, such as colon and breast. The diagnostic assessment should always be priori-tized, meaning that miniscule amounts of tissue are available for research if the primary tumor is small. Unfortunately, small tissue samples (as often seen in spindle cell UM) also show greater clonal homogeneity upon expansion provided that the same number of cells is needed for downstream analyses. By using samples from larger tumors and early cell culture passages, we hope to better reflect the innate proper-ties of the primary tumor. Epitheloid and mixed tumors are more prone to metastasis. The donors D1, D2, and D3 all had confirmed liver metastases. The selection of tumors analyzed in this study therefore is highly representative of aggressive UMs. Whether these results are valid for all UMs or solely the aggressive UMs is yet to be revealed, although there are indications that tumors with a low metastatic risk profile are more difficult to cultivate using the present protocol. The optimization of culture conditions would enable us to conduct further experiments for extensive verification of results and unravelling of epigenetic pathways.
In conclusion, we found that UM MCTS cultures undergo a metabolic shift. The MCTS display traits asso-ciated with anoikis resistance, including a shift toward a lipogenic profile. Targeting of lipid metabolism as a method to kill disseminated cancer cells could be a compelling new therapy in UM and needs further investigation. Additionally, the present study showed marked upregulation of SSXs, tran-scriptional repressors related to the PcG proteins that modu-late epigenetic silencing of genes. SSXs have been implied in the process of EMT, and their expression could be increased in cells that have conferred anoikis resistance, thus serving as a potential target for disseminated cancer cells. UM MCTS could be a suitable model to reveal novel candidate targets for treatment of UM dissemination.
APPENDIX 1. J-EXPRESS AND RANK PRODUCT (RP) ANALYSIS OF DIFFERENTIALLY EXPRESSED GENES (≥2 FOLD UP/DOWN, Q≤ 0.05) IN MULTICELLULAR TUMOR SPHEROIDS (MCTS) VERSUS UNCULTURED TUMOR BIOPSIES VS ADHERENT PRIMARY CULTURES.
To access the data, click or select the words “Appendix 1.”
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
691
ACKNOWLEDGMENTS
We would like to acknowledge all personnel at the Center for Eye Research and at the Dep. of Ophthalmology OUS that contributed in this project. We thank the personnel at the Genomics Core Facility, OUS and Helse Sør-Øst for performing microarray analysis and helping out with J-express analysis. We would also like to thank Sverre-Henning Brorson at the Dep. of Pathology, OUS for helping with TEM imaging. The work was funded by the South-Eastern Norway Regional Health Authority (Helse Sør-Øst) project 2012104, Norwegian Cancer Society project 5808589 and supported by grants from Arthur and Odd Clausons ophthalmological fund, Aase and Knut Tønjums ophthal-mological fund, Futura fund, Unifor Frimed, Norwegian Association of the Blind and Partially Sighted, Inger Holms memorial fund, Stiftelsen for fremme av kreftforskning at University of Oslo and Legat til fremme av kreftforskning. All authors contributing to the study have read and approved the manuscript. There are no conflicts of interest for any of the authors.
REFERENCES
1. Singh AD, Turell ME, Topham AK. Uveal Melanoma: Trends in Incidence, Treatment, and Survival. Ophthalmology 2011; 118:1881-5. [PMID: 21704381].
2. Virgili G, Gatta G, Ciccolallo L, Capocaccia R, Biggeri A, Crocetti E, Lutz JM, Paci E. Incidence of uveal melanoma in Europe. Ophthalmology 2007; 114:2309-15. [PMID: 17498805].
3. Bergman L, Seregard S, Nilsson B, Lundell G, Ringborg U, Ragnarsson-Olding B. Uveal melanoma survival in Sweden from 1960 to 1998. Invest Ophthalmol Vis Sci 2003; 44:3282-7. [PMID: 12882771].
4. Kujala E, Makitie T, Kivela T. Very long-term prognosis of patients with malignant uveal melanoma. Invest Ophthalmol Vis Sci 2003; 44:4651-9. [PMID: 14578381].
5. Diener-West M, Reynolds SM, Agugliaro DJ, Caldwell R, Cumming K, Earle JD, Hawkins BS, Hayman JA, Jaiyesimi I, Jampol LM, Kirkwood JM, Koh WJ, Robertson DM, Shaw JM, Straatsma BR, Thoma J. Development of metastatic disease after enrollment in the COMS trials for treatment of choroidal melanoma: Collaborative Ocular Melanoma Study Group Report No. 26. Arch Ophthalmol 2005; 123(12):1639–43.
6. Tura A, Luke J, Merz H, Reinsberg M, Luke M, Jager MJ, Grisanti S. Identification of circulating melanoma cells in uveal melanoma patients by dual-marker immunoenrich-ment. Invest Ophthalmol Vis Sci 2014; 55:4395-404. [PMID: 24970258].
7. Kaliki S, Shields CL, Shields JA. Uveal melanoma: Estimating prognosis. Indian J Ophthalmol 2015; 63:93-102. [PMID: 25827538].
8. Callejo SA, Antecka E, Blanco PL, Edelstein C, Burnier MN Jr. Identification of circulating malignant cells and its correlation with prognostic factors and treatment in uveal melanoma. A prospective longitudinal study. Eye (Lond) 2007; 21:752-9. [PMID: 16575415].
9. Eide N, Faye RS, Hoifodt HK, Overgaard R, Jebsen P, Kval-heim G, Fodstad O. Immunomagnetic detection of micro-metastatic cells in bone marrow in uveal melanoma patients. Acta ophthalmologica 2009; 87:830-6. [PMID: 19055657].
10. Eide N, Faye RS, Hoifodt HK, Sandstad B, Qvale G, Faber R, Jebsen P, Kvalheim G, Fodstad O. Immunomagnetic detection of micrometastatic cells in bone marrow of uveal melanoma patients: a paradox. Acta ophthalmologica 2015; 93:59-66. [PMID: 25613126].
11. Sullivan WJ, Christofk HR. The metabolic milieu of metas-tases. Cell 2015; 160:363-4. [PMID: 25635452].
12. Nguyen DX, Bos PD, Massague J. Metastasis: from dissemina-tion to organ-specific colonization. Nat Rev Cancer 2009; 9:274-84. [PMID: 19308067].
13. Kondo T. Stem cell-like cancer cells in cancer cell lines. Cancer biomarkers: section Dis Markers 2007; 3:245-50. .
14. Buchheit CL, Weigel KJ, Schafer ZT. Cancer cell survival during detachment from the ECM: multiple barriers to tumour progression. Nat Rev Cancer 2014; 14:632-41. [PMID: 25098270].
15. Schafer ZT, Grassian AR, Song L, Jiang Z, Gerhart-Hines Z, Irie HY, Gao S, Puigserver P, Brugge JS. Antioxidant and oncogene rescue of metabolic defects caused by loss of matrix attachment. Nature 2009; 461:109-13. [PMID: 19693011].
16. Paoli P, Giannoni E, Chiarugi P. Anoikis molecular pathways and its role in cancer progression. Biochim Biophys Acta 2013; 1833:3481-98. [PMID: 23830918].
17. Weiswald LB, Richon S, Validire P, Briffod M, Lai-Kuen R, Cordelieres FP, Bertrand F, Dargere D, Massonnet G, Maran-goni E, Gayet B, Pocard M, Bieche I, Poupon MF, Bellet D, Dangles-Marie V. Newly characterised ex vivo colospheres as a three-dimensional colon cancer cell model of tumour aggressiveness. Br J Cancer 2009; 101:473-82. [PMID: 19603013].
18. Emmink BL, Van Houdt WJ, Vries RG, Hoogwater FJ, Govaert KM, Verheem A, Nijkamp MW, Steller EJ, Jimenez CR, Clevers H, Borel Rinkes IH, Kranenburg O. Differentiated human colorectal cancer cells protect tumor-initiating cells from irinotecan. Gastroenterology 2011; 141:269-78. [PMID: 21459094].
19. Calvet CY, Andre FM, Mir LM. The culture of cancer cell lines as tumorspheres does not systematically result in cancer stem cell enrichment. PLoS One 2014; 9:e89644-[PMID: 24586931].
20. Kunjithapatham R, Karthikeyan S, Geschwind JF, Kieserman E, Lin M, Fu DX, Ganapathy-Kanniappan S. Reversal of
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
692
anchorage-independent multicellular spheroid into a mono-layer mimics a metastatic model. Sci Rep 2014; 4:6816-[PMID: 25351825].
21. Weiswald LB, Bellet D, Dangles-Marie V. Spherical cancer models in tumor biology. Neoplasia 2015; 17:1-15. [PMID: 25622895].
22. Nareyeck G, Zeschnigk M, Bornfeld N, Anastassiou G. Novel cell lines derived by long-term culture of primary uveal melanomas. Ophthalmologica Journal international d’ophtalmologie International journal of ophthalmology Z Augenheilkd 2009; 223:196-201. .
23. Fang D, Nguyen TK, Leishear K, Finko R, Kulp AN, Hotz S, Van Belle PA, Xu X, Elder DE, Herlyn M. A tumorigenic subpopulation with stem cell properties in melanomas. Cancer Res 2005; 65:9328-37. [PMID: 16230395].
24. Schroeder A, Mueller O, Stocker S, Salowsky R, Leiber M, Gassmann M, Lightfoot S, Menzel W, Granzow M, Ragg T. The RIN: an RNA integrity number for assigning integ-rity values to RNA measurements. BMC Mol Biol 2006; 7:3-[PMID: 16448564].
25. Johnsen EO, Froen RC, Albert R, Omdal BK, Sarang Z, Berta A, Nicolaissen B, Petrovski G, Moe MC. Activation of neural progenitor cells in human eyes with proliferative vitreoreti-nopathy. Exp Eye Res 2012; 98:28-36. [PMID: 22465407].
26. Breitling R, Armengaud P, Amtmann A, Herzyk P. Rank prod-ucts: a simple, yet powerful, new method to detect differen-tially regulated genes in replicated microarray experiments. FEBS Lett 2004; 573:83-92. [PMID: 15327980].
27. Wang F, Flanagan J, Su N, Wang LC, Bui S, Nielson A, Wu X, Vo HT, Ma XJ, Luo Y. RNAscope: a novel in situ RNA analysis platform for formalin-fixed, paraffin-embedded tissues. JMD 2012; 14:22-9. [PMID: 22166544].
28. Murdoch A, Jenkinson EJ, Johnson GD, Owen JJ. Alkaline phosphatase-fast red, a new fluorescent label. Application in double labelling for cell cycle analysis. J Immunol Methods 1990; 132:45-9. [PMID: 2118160].
29. van Essen TH, van Pelt SI, Bronkhorst IH, Versluis M, Nemati F, Laurent C, Luyten GP, van Hall T, van den Elsen PJ, van der Velden PA, Decaudin D, Jager MJ. Upregulation of HLA Expression in Primary Uveal Melanoma by Infil-trating Leukocytes. PLoS One 2016; 11:e0164292-[PMID: 27764126].
30. Frisch SM, Francis H. Disruption of epithelial cell-matrix interactions induces apoptosis. J Cell Biol 1994; 124:619-26. [PMID: 8106557].
31. Onder TT, Gupta PB, Mani SA, Yang J, Lander ES, Weinberg RA. Loss of E-cadherin promotes metastasis via multiple downstream transcriptional pathways. Cancer Res 2008; 68:3645-54. [PMID: 18483246].
32. Kamarajugadda S, Stemboroski L, Cai Q, Simpson NE, Nayak S, Tan M, Lu J. Glucose oxidation modulates anoikis and tumor metastasis. Mol Cell Biol 2012; 32:1893-907. [PMID: 22431524].
33. Hu K, Babapoor-Farrokhran S, Rodrigues M, Deshpande M, Puchner B, Kashiwabuchi F, Hassan SJ, Asnaghi L, Handa JT, Merbs S, Eberhart CG, Semenza GL, Montaner S, Sodhi A. Hypoxia-inducible factor 1 upregulation of both VEGF and ANGPTL4 is required to promote the angiogenic phenotype in uveal melanoma. Oncotarget 2016; [PMID: 26761211].
34. Zhang Z, Cao L, Li J, Liang X, Liu Y, Liu H, Du J, Qu Z, Cui M, Liu S, Gao L, Ma C, Zhang L, Han L, Sun W. Acquisition of anoikis resistance reveals a synoikis-like survival style in BEL7402 hepatoma cells. Cancer Lett 2008; 267:106-15. [PMID: 18433990].
35. Gray NE, Lam LN, Yang K, Zhou AY, Koliwad S, Wang JC. Angiopoietin-like 4 (Angptl4) protein is a physiological mediator of intracellular lipolysis in murine adipocytes. J Biol Chem 2012; 287:8444-56. [PMID: 22267746].
36. Carracedo A, Cantley LC, Pandolfi PP. Cancer metabolism: fatty acid oxidation in the limelight. Nat Rev Cancer 2013; 13:227-32. [PMID: 23446547].
37. Yue S, Li J, Lee SY, Lee HJ, Shao T, Song B, Cheng L, Masterson TA, Liu X, Ratliff TL, Cheng JX. Cholesteryl ester accumulation induced by PTEN loss and PI3K/AKT activation underlies human prostate cancer aggressiveness. Cell Metab 2014; 19:393-406. [PMID: 24606897].
38. Beloribi-Djefaflia S, Vasseur S, Guillaumond F. Lipid meta-bolic reprogramming in cancer cells. Oncogenesis 2016; 5:e189-[PMID: 26807644].
39. Kimmel AR, Brasaemle DL, McAndrews-Hill M, Sztalryd C, Londos C. Adoption of PERILIPIN as a unifying nomen-clature for the mammalian PAT-family of intracellular lipid storage droplet proteins. J Lipid Res 2010; 51:468-71. [PMID: 19638644].
40. Smith HA, McNeel DG. The SSX family of cancer-testis anti-gens as target proteins for tumor therapy. Clin Dev Immunol 2010; 2010:150591-[PMID: 20981248].
41. Tureci O, Chen YT, Sahin U, Gure AO, Zwick C, Villena C, Tsang S, Seitz G, Old LJ, Pfreundschuh M. Expression of SSX genes in human tumors. Int J Cancer 1998; 77:19-23. [PMID: 9639388].
42. Frisch SM, Schaller M, Cieply B. Mechanisms that link the oncogenic epithelial-mesenchymal transition to suppression of anoikis. J Cell Sci 2013; 126:21-9. [PMID: 23516327].
43. Pike LS, Smift AL, Croteau NJ, Ferrick DA, Wu M. Inhibi-tion of fatty acid oxidation by etomoxir impairs NADPH production and increases reactive oxygen species resulting in ATP depletion and cell death in human glioblastoma cells. Biochim Biophys Acta 2011; 1807:726-34. [PMID: 21692241].
44. Kuzu OF, Noory MA, Robertson GP. The Role of Cholesterol in Cancer. Cancer Res 2016; 76:2063-70. [PMID: 27197250].
45. Mollinedo F, Gajate C. Lipid rafts as major platforms for signaling regulation in cancer. Adv Biol Regul 2015; 57:130-46. [PMID: 25465296].
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
693
46. Park EK, Park MJ, Lee SH, Li YC, Kim J, Lee JS, Lee JW, Ye SK, Park JW, Kim CW, Park BK, Kim YN. Cholesterol depletion induces anoikis-like apoptosis via FAK down-regulation and caveolae internalization. J Pathol 2009; 218:337-49. [PMID: 19288501].
47. Seddon JM, Gragoudas ES, Albert DM. Ciliary body and choroidal melanomas treated by proton beam irradiation. Histopathologic study of eyes. Arch Ophthalmol 1983; 101:1402-8. [PMID: 6311146].
48. Driot JY, Rault J, Bonnin P, Liotet S. Electron microscopy of three cases of choroid malignant melanomas. Int Ophthalmol 1987; 10:193-202. [PMID: 3654058].
49. Harper ME, Antoniou A, Villalobos-Menuey E, Russo A, Trauger R, Vendemelio M, George A, Bartholomew R, Carlo D, Shaikh A, Kupperman J, Newell EW, Bespalov IA, Wallace SS, Liu Y, Rogers JR, Gibbs GL, Leahy JL, Camley RE, Melamede R, Newell MK. Characterization of a novel metabolic strategy used by drug-resistant tumor cells. FASEB J 2002; 16:1550-7. [PMID: 12374777].
50. Schlaepfer IR, Rider L, Rodrigues LU, Gijon MA, Pac CT, Romero L, Cimic A, Sirintrapun SJ, Glode LM, Eckel RH, Cramer SD. Lipid catabolism via CPT1 as a therapeutic target for prostate cancer. Mol Cancer Ther 2014; 13:2361-71. [PMID: 25122071].
51. Hossain F, Al-Khami AA, Wyczechowska D, Hernandez C, Zheng L, Reiss K, Valle LD, Trillo-Tinoco J, Maj T, Zou W, Rodriguez PC, Ochoa AC. Inhibition of Fatty Acid Oxida-tion Modulates Immunosuppressive Functions of Myeloid-Derived Suppressor Cells and Enhances Cancer Therapies. Cancer Immunol Res 2015; 3:1236-47. [PMID: 26025381].
52. Kodach LL, Jacobs RJ, Voorneveld PW, Wildenberg ME, Verspaget HW, van Wezel T, Morreau H, Hommes DW, Peppelenbosch MP, van den Brink GR, Hardwick JC. Statins augment the chemosensitivity of colorectal cancer cells inducing epigenetic reprogramming and reducing colorectal cancer cell ‘stemness’ via the bone morphogenetic protein pathway. Gut 2011; 60:1544-53. [PMID: 21551187].
53. Tan MJ, Teo Z, Sng MK, Zhu P, Tan NS. Emerging roles of angiopoietin-like 4 in human cancer. Molecular cancer research MCR. 2012; 10:677-88. [PMID: 22661548].
54. Zhu P, Tan MJ, Huang RL, Tan CK, Chong HC, Pal M, Lam CR, Boukamp P, Pan JY, Tan SH, Kersten S, Li HY, Ding JL, Tan NS. Angiopoietin-like 4 protein elevates the prosurvival intracellular O2(-):H2O2 ratio and confers anoikis resistance to tumors. Cancer Cell 2011; 19:401-15. [PMID: 21397862].
55. Tanaka J, Irie T, Yamamoto G, Yasuhara R, Isobe T, Hokazono C, Tachikawa T, Kohno Y, Mishima K. ANGPTL4 regulates the metastatic potential of oral squamous cell carcinoma. J Oral Pathol Med 2015; 44:126-33. [PMID: 25060575].
56. Zhu X, Guo X, Wu S, Wei L. ANGPTL4 Correlates with NSCLC Progression and Regulates Epithelial-Mesenchymal Transition via ERK Pathway. Lung 2016; [PMID: 27166634].
57. Okochi-Takada E, Hattori N, Tsukamoto T, Miyamoto K, Ando T, Ito S, Yamamura Y, Wakabayashi M, Nobeyama Y, Ushijima T. ANGPTL4 is a secreted tumor suppressor that inhibits angiogenesis. Oncogene 2014; 33:2273-8. [PMID: 23686315].
58. Abate-Daga D, Speiser DE, Chinnasamy N, Zheng Z, Xu H, Feldman SA, Rosenberg SA, Morgan RA. Development of a T cell receptor targeting an HLA-A*0201 restricted epitope from the cancer-testis antigen SSX2 for adoptive immu-notherapy of cancer. PLoS One 2014; 9:e93321-[PMID: 24681846].
59. D’Arcy P, Maruwge W, Wolahan B, Ma L, Brodin B. Onco-genic functions of the cancer-testis antigen SSX on the proliferation, survival, and signaling pathways of cancer cells. PLoS One 2014; 9:e95136-[PMID: 24787708].
60. dos Santos NR, de Bruijn DR, Balemans M, Janssen B, Gartner F, Lopes JM, de Leeuw B, Geurts van Kessel A. Nuclear localization of SYT, SSX and the synovial sarcoma-associated SYT-SSX fusion proteins. Hum Mol Genet 1997; 6:1549-58. [PMID: 9285793].
61. Lim FL, Soulez M, Koczan D, Thiesen HJ, Knight JC. A KRAB-related domain and a novel transcription repression domain in proteins encoded by SSX genes that are disrupted in human sarcomas. Oncogene 1998; 17:2013-8. [PMID: 9788446].
62. Kato H, Tjernberg A, Zhang W, Krutchinsky AN, An W, Takeuchi T, Ohtsuki Y, Sugano S, de Bruijn DR, Chait BT, Roeder RG. SYT associates with human SNF/SWI complexes and the C-terminal region of its fusion partner SSX1 targets histones. J Biol Chem 2002; 277:5498-505. [PMID: 11734557].
63. Soulez M, Saurin AJ, Freemont PS, Knight JC. SSX and the synovial-sarcoma-specific chimaeric protein SYT-SSX co-localize with the human Polycomb group complex. Onco-gene 1999; 18:2739-46. [PMID: 10348348].
64. Gjerstorff MF, Relster MM, Greve KB, Moeller JB, Elias D, Lindgreen JN, Schmidt S, Mollenhauer J, Voldborg B, Pedersen CB, Bruckmann NH, Mollegaard NE, Ditzel HJ. SSX2 is a novel DNA-binding protein that antagonizes poly-comb group body formation and gene repression. Nucleic Acids Res 2014; 42:11433-46. [PMID: 25249625].
65. de Bruijn DR, Allander SV, van Dijk AH, Willemse MP, Thijssen J, van Groningen JJ, Meltzer PS, van Kessel AG. The synovial-sarcoma-associated SS18–SSX2 fusion protein induces epigenetic gene (de)regulation. Cancer Res 2006; 66:9474-82. [PMID: 17018603].
Molecular Vision 2017; 23:680-694 <http://www.molvis.org/molvis/v23/680> © 2017 Molecular Vision
694
Articles are provided courtesy of Emory University and the Zhongshan Ophthalmic Center, Sun Yat-sen University, P.R. China. The print version of this article was created on 3 October 2017. This reflects all typographical corrections and errata to the article through that date. Details of any changes may be found in the online version of the article.
II
1
Integrated differential DNA methylation and gene expression of formalin-
fixed paraffin-embedded uveal melanoma specimens identifies genes
associated with early metastasis and poor prognosis
Charlotte Ness1,2
, Kirankumar Katta1, Øystein Garred
3, Theresa Kumar
3, Ole Kristoffer
Olstad4, Goran Petrovski
1,2, Morten C. Moe
1,2 and Agate Noer
1.
1Center for Eye Research, Department of Ophthalmology, Oslo University Hospital, Oslo, Norway
2Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway
3Department of Pathology, Oslo University Hospital, Norway
4Department of Medical Biochemistry, Oslo University Hospital, Norway
The authors have no conflicts-of-interest to declare.
Corresponding author:
Agate Noer, PhD, Senior Researcher, Center for Eye Research, Department of Ophthalmology, Oslo
University Hospital, Pb 4956 Nydalen, 0424 Oslo, Norway; Phone: +47 23 01 61 98
E-mail: [email protected]
2
Abstract
PURPOSE:
Uveal melanoma (UM) is an aggressive malignancy, in which nearly 50% of the patients die
from metastatic disease. Aberrant DNA methylation is recognized as an important epigenomic
event in carcinogenesis. Formalin-fixed paraffin-embedded (FFPE) samples represent a
valuable source of tumor tissue, and recent technology has enabled the use of these samples in
genome-wide DNA methylation analyses. Our aim was to investigate differential DNA
methylation in relation to histopathological classification and survival data. In addition we
sought to identify aberrant DNA methylation of genes that could be associated with metastatic
disease and poor survival.
METHODS:
FFPE samples from UM patients (n=23) who underwent enucleation of the eye in the period
1976-1989 were included. DNA methylation was assessed using the Illumina Infinium
HumanMethylation450 array and coupled to histopathological data, Cancer Registry of
Norway- (registered UM metastasis) and Norwegian Cause of Death Registry- (time and
cause of death) data. Differential DNA methylation patterns contrasting histological
classification, survival data and clustering properties were investigated. Survival groups were
defined as “Early metastasis” (metastases and death within 2-5 years after enucleation, n=8),
“Late metastasis” (metastases and death within 9-21 years after enucleation, n=7) and “No
metastasis” (no detected metastases ≥18 years after enucleation, n=8). A subset of samples
were selected based on preliminary multi-dimensional scaling (MDS) plots, histopathological
classification, chromosome 3 status, survival status and clustering properties; “Subset Early
metastasis” (n=4) vs “Subset No metastasis” (n=4). Bioinformatics analyses were conducted
3
in the R statistical software. Differentially methylated positions (DMPs) and differentially
methylated regions (DMRs) in various comparisons were assessed. Gene expression of
relevant subgroups was determined by microarray analysis and quantitative reverse-
transcription polymerase chain reaction (qRT-PCR).
RESULTS:
DNA methylation analyses identified 2 clusters that separated the samples according to
chromosome 3 status. Cluster 1 consisted of samples (n=5) with chromosome 3 disomy (D3),
while Cluster 2 was comprised of samples (n=15) with chromosome 3 monosomy (M3). 1212
DMRs and 9386 DMPs were identified in M3 vs D3. No clear clusters were formed based on
our predefined survival groups (“Early”, “Late”, “No”) nor histopathological classification
(Epithelioid, Mixed, Spindle). We identified significant changes in DNA methylation (beta
FC ≥0.2, adjusted p<0.05) between two sample subsets (n=8). “Subset Early metastasis”
(n=4) vs “Subset No metastasis” (n=4) identified 348 DMPs and 36 DMRs, and their
differential gene expression by microarray showed that 14 DMPs and 2 DMRs corresponded
to changes in gene expression (FC≥1.5, p<0.05). RNF13, ZNF217 and HYAL1 were
hypermethylated and downregulated in “Subset Early metastasis” vs “Subset No metastasis”
and could be potential tumor suppressors. TMEM200C, RGS10, ADAM12 and PAM were
hypomethylated and upregulated in “Subset Early metastasis vs “Subset No metastasis” and
could be potential oncogenes and thus markers of early metastasis and poor prognosis in UM.
4
CONCLUSIONS:
DNA methylation profiling showed differential clustering of samples according to
chromosome 3 status: Cluster 1 (D3) and Cluster 2 (M3). Integrated differential DNA
methylation and gene expression of two subsets of samples identified genes associated with
early metastasis and poor prognosis. RNF13, ZNF217 and HYAL1 are hypermethylated and
candidate tumor suppressors, while TMEM200C, RGS10, ADAM12 and PAM are
hypomethylated and candidate oncogenes linked to early metastasis. UM FFPE samples
represent a valuable source for methylome studies and enable long-time follow-up.
Key words: Uveal melanoma; DNA methylation; 450k; FFPE; Epigenetics.
Short running title: Differential DNA methylation profiles of FFPE UM
5
1. Introduction
Aberrant changes of the epigenetic landscape are common denominators during cancer
development. DNA methylation is the best-characterized epigenetic modification and is
recognized as an important player in regulating gene expression and chromatin architecture
(Brocato and Costa 2013, Shen and Laird 2013). In general, cancer cells display global DNA
hypomethylation (loss of methylation) and promoter hypermethylation (gain of methylation)
of promoter associated CpGs (Brocato and Costa 2013). Hypermethylation in cancer often
occurs at promoters associated with tumor suppressors, thus leading to inactivation of the
corresponding gene. Hypomethylation of gene promoters is considered a permissive mark and
can lead to subsequent activation of oncogenes (Herman, Latif et al. 1994, Yamashita,
Tokunaga et al. 2015). The methylation profile of a given cancer can elucidate therapeutic
targets, reveal biomarkers for early detection or identify high risk tumors (Arshad, Ye et al.
2013, Baylin and Jones 2016).
Uveal melanoma (UM) is the most common primary intraocular malignancy in adults, and has
a high propensity for metastatic spread (Kaliki, Shields et al. 2015). Numerous studies have
demonstrated a series of molecular alterations associated with and involved in UM
pathogenesis. Activating mutations in GNAQ/GNA11 are considered early driver mutations
in UM, while metastatic disease is often correlated with Monosomy 3 (M3) and loss of the
deubiquitinating enzyme BAP1 (Prescher, Bornfeld et al. 1990, Van Raamsdonk, Bezrookove
et al. 2009, Harbour, Onken et al. 2010, Van Raamsdonk, Griewank et al. 2010). Recent
studies also suggest an epigenetic contribution to the underlying molecular pathology in UM.
Clustering of UMs according to their global methylation profile has been shown to coincide
6
with clustering into Class 1 (low-risk) and Class 2 (high-risk) tumors with respect to risk of
metastasis (Landreville, Agapova et al. 2008, Robertson, Shih et al. 2018, Field, Kuznetsov et
al. 2019). BAP1 itself is hypothesized to be epigenetically regulated since a novel
hypermethylated site within the BAP1 locus has been found in all Class 2 tumors (Field,
Durante et al. 2018). Additionally, gene silencing by promoter methylation has been
demonstrated for loci involved in extracellular matrix degradation, cell cycle regulation, axon
guidance, melanogenesis and development (van der Velden, Metzelaar-Blok et al. 2001, van
der Velden, Zuidervaart et al. 2003, Maat, van der Velden et al. 2007, Maat, Beiboer et al.
2008, Neumann, Weinhausel et al. 2011, Field, Kuznetsov et al. 2019). Although
advancements in the characterization of the UM methylome have been made, this unveiling is
still in its beginning. In order to elucidate the methylome in UM further, we investigated DNA
methylation in formalin-fixed paraffin-embedded (FFPE) UM samples. FFPE samples
represent an extensive source of material and offer the possibility of long-term follow-up.
Recently, optimized protocols for DNA restoration have been developed, thus enabling the
use of FFPE-derived DNA in genome-wide DNA methylation analyses (Dumenil, Wockner et
al. 2014, Moran, Vizoso et al. 2014, de Ruijter, de Hoon et al. 2015). Further, the Illumina
Infinium HumanMethylation450 BeadChip array (HM-450K) has proven to be a robust
platform for investigating methylation in restored FFPE samples (Moran, Vizoso et al. 2014,
de Ruijter, de Hoon et al. 2015).
Our aim was to investigate differential DNA methylation in relation to UM histopathological
classification and survival data. In addition we sought to identify aberrant DNA methylation
of genes that could be linked to metastatic disease and poor survival.
7
2. Materials and methods
2.1 Samples
All experiments were conducted in accordance with the Declaration of Helsinki (2013), and
tissue harvesting was approved by the Local Committees for Medical Research Ethics (REK
Ref. 2009/1973 and REK Ref. 2013/803–1). The study adhered to the ARVO statement on
human subjects. Tumor tissue was obtained from the archive of the Department of Pathology,
Oslo University Hospital. FFPE samples were coupled to data from the Cancer Registry of
Norway (https://www.kreftregisteret.no/en/) and the Norwegian Cause of Death Registry
(https://www.fhi.no/en/hn/health-registries/cause-of-death-registry/ ), thus providing
information about time and cause of death in addition to information about metastatic spread.
FFPE samples from 23 UM patients undertaking enucleation of the eye in the period 1976-
1989 were included (Supplementary table 1). All specimens were diagnosed as UMs at the
time of initial diagnosis and re-evaluated by ophthalmic pathologists. FFPE tissue was
sectioned and stained by hematoxylin and eosin in addition to staining with leukocyte
common antigen, CD45 (mouse, 1:100; Abcam, Cambridge, UK) as previously described
(Ness, Garred et al. 2017). The samples were divided into 3 subgroups based on metastasis
and survival; 1. Metastases and death within 2-4 (mean 2.75) years (“Early
metastasis”=“Early”, n=8), 2. Metastases and death 9-21 (mean 12.7) years after diagnosis
(“Late metastasis”=“Late”, n=7) and 3. Alive or dead of other cause ≥18 (mean >24) years
after primary diagnosis (“No metastasis”=“No”, n=8). Patients with metastatic disease 5-8
years after primary diagnosis were excluded to reduce overlap between our predefined
survival groups in order to detect differential methylated patterns associated with time of
cancer relapse. All subgroups contained samples with different histological profiles, thus
including tumors classified as epithelioid (n=6), mixed (n=6) and spindle shaped (n=11) (van
Beek, Koopmans et al. 2012).
8
Based on preliminary multi-dimensional scaling (MDS plots), histopathological classification,
chromosome 3 status, survival status and clustering properties, we subtracted two subsets of
samples for comparison. These subsets were named “Subset Early metastasis” (n=4) and
“Subset No metastasis” (n=4). “Subset Early metastasis” (sample 2, 15, 16 and 17) consisted
of M3 samples and cancer relapse within 4 year. “Subset No metastasis” (sample 7, 11, 12
and 13) consisted of D3 samples and survival ≥18 years after primary diagnosis.
2.2 DNA extraction, bisulphite conversion and Illumina 450k array
DNA extraction was carried out using the QIAamp DNA FFPE Tissue Kit (Qiagen, Venlo,
NL) according to the manufacturers’ recommendations. DNA was purified using the Zymo
PCR inhibitor removal kit (Zymo, Irvine, CA, US). DNA concentration was determined using
Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, US). Quality control of
DNA samples was performed using the Infinium HD FFPE QC Assay (Illumina, San Diego,
CA, US) protocol according to Illumina’s recommendations. Samples were kept at -20 C˚
until bisulphite conversion using Zymo EZ DNA Methylation kit (Zymo). 500 ng of genomic
DNA from each sample was converted. Samples underwent restoration using the Illumina
Infinium HD Restore protocol (Illumina), and 4 μl of bisulphite-converted restored DNA was
used for hybridization on the Infinium Human Methylation 450 BeadChip (Illumina),
following the Illumina Infinium HD Methylation protocol (Illumina). Hybridization,
scanning, and raw data processing were performed at the Genomics Core Facility
(http://oslo.genomics.no), Oslo University Hospital, South-Eastern Norway Regional Health
Authority and University of Oslo. The intensities of the images were extracted using the
GenomeStudio (v.2011.1) Methylation module (1.9.0) software, which normalizes within-
sample data using different internal controls that are present on the HumanMethylation450
9
BeadChip and internal background probes. The methylation score for each CpG was
represented as a beta value according to the fluorescent intensity ratio representing any value
between 0 (unmethylated) and 1 (completely methylated). The Illumina 450k array DNA
methylation data have been deposited in NCBI's Gene Expression Omnibus (Edgar,
Domrachev et al. 2002) and are accessible through GEO Series accession number GSE156876
(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE156876).
2.3 Quality control and pre-processing
All subsequent analysis steps were performed using R, v. 3.5.1 (R Core Team 2018). The
minfi package, v.1.28.3 was used for importing data into R, for quality control and for
preprocessing of the data (Aryee, Jaffe et al. 2014). The data were quantile normalized.
Probes on loci with single nucleotide polymorphisms (SNPs), on sex chromosomes and
probes previously shown to be non-specific were removed, in total 55 568 probes (Chen,
Lemire et al. 2013). Data were visualized as MDS plots using R package limma, v. 3.83.3
plotting function (Ritchie, Phipson et al. 2015). The minfi function plotSeX was utilized for
evaluating the correctness of the gender annotations of the samples.
2.4 Copy-number variation analysis
Copy-number profiles of all 23 samples were generated using the ‘conumee’ R package in
Bioconductor as previously described (Hovestadt and Zapatka 2017). Eight healthy retina
samples obtained from ArrayExpress (https://www.ebi.ac.uk/arrayexpress, accession number
EMTAB-5535) were used as reference samples in the analysis. A list of 29 genes associated
with central nervous system tumors from the conumee package was used for in detail region
10
analysis. The copy-number ratios were plotted in a graph according to chromosomal location
and assessed manually.
2.5 Principal component analysis and consensus clustering
Probe M-value variances across all samples were calculated. The top 500 most variably
methylated probes were used for unsupervised principal component analysis using promp
function of the stats base package in R. Three-dimensional PCA plot was generated using R
package pca3d, v. 0.10.2 (Weiner 2020). Ellipsoids, if shown, indicated confidence of each
group at 95% level. The samples were coloured either by their chromosome 3 M3/D3 status,
“Early”, “Late” or “No” relapse status or by histopathological classification Spindle,
Epithelioid or Mixed tumors.
Unsupervised consensus clustering on the most variable 1% of CpG probes (4299 probes) was
carried out using R package ConsensusClusterPlus, with Euclidean distance and partitioning
around medoids (PAM) (Wilkerson and Hayes 2010). Solutions between 2-5 clusters were
evaluated for cluster stability, and for associations with clinical and chromosomal covariates.
2.6 Probe-wise differential methylation analysis
To identify differentially methylated positions (DMPs) between the samples, each individual
CpG probe was examined using limmav (Ritchie, Phipson et al. 2015). DMPs were analyzed
between: M3 vs D3, the three histology groups, the three pre-defined survival groups and the
subsets “Subset Early metastasis” (n=4) vs “Subset No metastasis” (n=4).. The latter subsets
of samples were selected based on preliminary MDS plots, histopathological classification,
11
chromosome 3 status, survival status and clustering properties. P-values were obtained using
the Benjamini-Hochberg procedure (Benjamini and Hochberg 1995).
2.7 Differential methylation analysis of regions
Differentially methylated regions (DMRs) were analyzed using R package DMRcate, v.
1.18.0 that is based on limma (Peters, Buckley et al. 2015). DMRs were analyzed between:
M3 vs D3, the three histology groups, the three pre-defined survival groups and the subsets
“Subset Early metastasis” (n=4) vs “Subset No metastasis” (n=4). DMRs that were
constituted by at least two consecutive significant CpGs separated by a maximum of 1000
nucleotide gaps were included. The overall significance of the DMRs was assessed based on
Stouffer-transformed p-values. DMRs with Stouffer p-value < 0.05 were considered
statistically significant and were visualized within their chromosomal context using DMR.plot
function of DMRcate. Annotations of individual CpGs constituting the DMRs were added
(Lawrence, Huber et al. 2013, Hansen 2016).
2.8 RNA isolation
RNA was isolated from 8 samples; donor 2, 7, 11, 12, 13, 15, 16, 17, and. The samples were
selected based on preliminary MDS plots, their histology, chromosome 3 status, survival
status and clustering properties. Extraction was carried out using the Qiagen MiRNEASY
FFPE kit (Qiagen) according to the manufacturers’ recommendations. RNA was purified
using the Zymo PCR inhibitor removal kit (Zymo). RNA concentration and purity were
determined using NanoDrop (Thermo Fisher Scientific) and Bioanalyzer (Agilent 2100,
12
Agilent, Santa Clara, CA, US). The RNA integrity number (RIN) values and 260/280 ratio
were within the ranges recommended by Qiagen.
2.9 Microarray of RNA samples
Microarray analyses were performed at the Affymetrix Core Facility, Ullevål, Oslo University
Hospital, South-Eastern Norway Regional Health Authority. Affymetrix Human Clariom ™
D Array (Affymetrix, Santa Clara, CA, US) was used for the analyses, targeting 540,000
transcripts. Total RNA (50 ng) was subjected to the GeneChip™ WT Pico Reagent Kit and
WT Labeling Kit (Affymetrix). A total of 6 cycles pre-IVT (in vitro transcription)
amplification was run according to the manufacturer`s protocol. Biotinylated and fragmented
single-stranded complementary DNAs (cDNAs) were hybridized to the arrays. The arrays
were washed and stained using an FS-450 fluidics station (Affymetrix, fluidics protocol
FS450_0001). Signal intensities were detected by a Hewlett Packard (HP, US) 30007G gene
array scanner. The scanned images were processed using the AGCC Affymetrix GeneChip
Command Console) software, and the CEL files were imported into Partek ® Genomics Suite
™ software (Partek, St. Louis, MO, US) for statistical analysis. The Robust Multichip
Analysis (RMA) algorithm was applied for generation of signal values and normalization.
Transcripts containing accession numbers that begin with the prefixes “NM_” (protein-coding
transcripts) and “NR_” (non-protein-coding transcripts) in the NCBI Reference Sequence
Database (RefSeq) were filtered out for further statistical analysis. For expression
comparisons of “Subset Early metastasis” vs “Subset No metastasis”, profiles were compared
using a one-way ANOVA method. The results were expressed as fold changes (FC). Genes
with FC ≥1.5 and a p-value < 0.05 were regarded as significantly regulated.
13
The Affymetrix gene expression data have been deposited in NCBI's Gene Expression
Omnibus (Edgar, Domrachev et al. 2002) and are accessible through GEO Series accession
number GSE156877 for (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE156877).
2.10 Quantitative reverse-transcription PCR
RNA concentration and purity were measured using NanoDrop (Thermo Fisher Scientific).
Reverse transcription (RT) was performed using the High Capacity cDNA Reverse
Transcription Kit (Applied Biosystems, Waltham, MA, US) with 1µg total RNA per 20 μl RT
reaction. Complementary DNA (cDNA) was diluted to a volume of 200 µl (5ng/µl) after
cDNA synthesis. Quantitative PCR (qPCR) was performed using the StepOnePlus Real-Time
PCR system (Applied Biosystems, Thermo Fisher Scientific) and Taqman Gene Expression
assays following the manufacturer’s protocols (Applied Biosystems, Thermo Fisher
Scientific). The TaqMan Gene Expression Assays used include adhesion G protein-coupled
receptor G1: ADGRG1 (Hs00938474_m1), 5-hydroxytrymptamine receptor 2B: HTR2B
(Hs01118766_m1), roundabout guidance receptor 1: ROBO1 (Hs00268049_m1), contactin 3:
CNTN3 (Hs00968399_m1), ADAM Metallopeptidase domain 23: ADAM23
(Hs00187022_m1), palmdelphin: PALMD (Hs00927401_m1). The thermal cycling conditions
were 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. All
samples were run in duplicates (each reaction: 2.5 μl/12.5 ng cDNA in a total volume of
12.5 μl). The data were analyzed using the 2−ΔΔCt
method to find the relative changes in gene
expression as a FC between the samples. The “No metastasis” samples were chosen as the
calibrator and equaled one. The 18S assay (Hs03003631_g1) was used as a loading control to
quantify the differences in cDNA input between the samples.
14
2.11 Pathway and gene ontology analysis
Data from the microarray analyses were imported into Ingenuity Pathway Analysis (IPA)
software (Qiagen) to identify enriched biological pathways and molecular functions.
Unsupervised hierarchical clustering (p<0.05, FC ≥1.5) was performed using Partek ®
Genomics Suite ™ software (Partek).
3. Results
3.1 PCA and consensus clustering of DNA methylation data clustered samples according
to their chromosome 3 status
Sample 21, 22 and 23 (from the original dataset of 23 samples) showed partial deletion of
chromosome 3 by CNV analysis. PCA of the 500 most differentially methylated probes of the
23 samples are presented in Supplementary figure 1. In order to reduce the number of
variables and run a pure D3 vs M3 comparison, the 3 samples harbouring partial deletion of
chromosome 3 were excluded from further analyses. In accordance with previous publications
(Robertson et al and Field et al), PCA of the top 500 most differentially methylated probes
clustered the 20 samples into two groups based on chromosome 3 status (Figure 1). No clear
clusters were formed based on our predefined survival groups (“Early”, “Late”, “No”)
(Supplementary figure 2) or histopathological classification (Epithelioid, Mixed, Spindle)
(Supplementary figure 3). The D3 cluster consists of four “Early”/ “Spindle” samples (7, 11,
12 and 13) and one “Late”/ “Mixed” sample (19), while the M3 cluster includes two “No”
15
samples (6 and 10), seven “Early” samples (1, 2, 3, 14, 15, 16 and 17) and six “Late” samples
(4, 5, 8, 9, 18 and 20). Further, the M3 cluster consists of four “Epithelioid” (3, 6, 8 and 14),
seven “Spindle” (1, 2, 4, 7, 9, 17 and 20) and five “Mixed” (5, 10, 15, 16 and 18) samples.
Figure 1: Unsupervised three-dimensional principal component analysis (PCA) based on methylation profiling
of the 20 samples (1-20) showing differential clustering of samples into two clusters according to chromosome 3
status. Blue (Cluster 1): No deletion of chromosome 3= disomy 3 (D3). Red (Cluster 2): deletion of chromosome
3 = monosomy 3 (M3).
The most variable 1% CpG probes were used for unsupervised consensus clustering, yielding
k2- k5 cluster solutions. The k2 cluster solution divided samples into 2 clusters based on
chromosome 3 status (Figure 2). None of the tested clusters were homogenous regarding
16
neither histology nor survival. However, a subgroup of samples that showed the same
clustering properties in all k2- k5 clusters could be identified. Sample 2, 15, 16 and 17
(“Subset Early metastasis”) were in Cluster 1, while sample 7, 11, 12 and 13 (“Subset No
metastasis”) were in Cluster 2. All samplesin”Subset Early metastasis” had loss of
chromosome 3 (M3) (Cluster 2) and died of metastatic disease, while all samples in “Subset
No metastasis” had D3 (Cluster 1) and were long term survivors.
Figure 2: Heat map visualizing the consensus clustering result of the top 1% most variable probes
demonstrating 2 clusters (on top). Samples (1-20) have been ordered by their DNA methylation cluster
classification. Scale annotation (1=hypermethylated -0 hypomethylated) to the right. Sample histology and time
of metastasis are also shown above heatmap and as colored bars to the right. Spi=Spindle. Epi= Epitheloid. Mix=
17
Mixed= Spindle and Epithelioid. Copy-number variation (CNV) annotation for chromosome 3 below heatmap; -
D3= chromosome 3 disomy and M3= chromosome 3 monosomy.
3.2 Analysis of differentially methylated positions and regions
The DMP analysis of M3 samples vs D3 samples showed 9386 DMPs (adjusted p-value
0.05). The DMR analyses of M3 vs D3 displayed 1212 DMRs (adjusted p-value 0.05).
No significant DMPs and DMRs were detected when comparing the various histopathological
groups (spindle, epithelioid, mixed) or the predefined survival groups (“Early”, “Late”, “No”)
(adjusted p-value 0.05).
The DMP analysis of ”Subset Early metastasis” vs “Subset No metastasis” displayed 348
DMPs (adjusted p-value 0.05) (Appendix A, showing 348 DMPs). The DMR analysis of
”Subset Early metastasis” vs “Subset No metastasis” showed 36 DMRs (adjusted p-value of
0.05). These DMRs contained 200 CpG sites (Supplementary table 1, showing the 36
significant DMRs).
3.3 Gene expression, clustering and canonical pathways of “Subset Early metastasis” vs
“Subset No metastasis”
Differential gene expression analysis (One-Way ANOVA) of “Subset Early metastasis” vs
“Subset No metastasis” identified 1536 transcripts (1394 up- and 142 down-regulated genes)
(p<0.05, FC ≥1.5). Unsupervised hierarchical clustering was performed and presented as a
heatmap (Supplementary Figure 4). The ten most up-and downregulated genes from the One-
Way ANOVA are presented in Supplementary table 3. qRT-PCR of HTR2B, ADGRG1,
18
ADAM23, ROBO1, CNTN3 and PALMD was consistent with the microarray gene expression
data (Supplementary Figure 5).
The top 20 canonical pathways as per p-value displayed upregulation of several cancer
associated pathways in “Subset Early metastasis” vs “Subset No metastasis” (Figure 3).
Figure 3: Top 20 Canonical pathways as per p-value. #: Number of molecules. Green: Downregulated in
“Subset Early metastasis” vs “Subset No metastasis”. Red: Upregulated in “Subset Early metastasis” vs “Subset
No metastasis”.
19
3.4 Integrated DNA methylation and gene expression of “Subset Early metastasis” vs
“Subset No metastasis” shows aberrant DNA methylation and gene expression of
potential oncogenes and tumor suppressors
Genes related to the 348 significant DMPs and the 36 DMRs from the comparison “Subset
Early metastasis” vs “Subset No metastasis” (beta FC ≥0.2, adjusted p<0.05) were matched to
relative changes in gene expression from the RNA microarray (FC ≥1.5, p<0.05,). The
methylation and gene expression data have been deposited in NCBI's Gene Expression
Omnibus (Edgar et al., 2002) and are accessible through GEO Series accession number
GSE160645 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE160645). Out of the
significant DMPs and DMRs, 14 DMPs displayed corresponding changes in gene expression
(Table 1), while this was shown for 2 DMRs (Table 2). Noticeably we detected
hypermethylation and downregulation of the proposed tumor suppressors RNF13, ZNF217
and HYAL1 (Frost, Mohapatra et al. 2000, Arshad, Ye et al. 2013, Cohen, Donini et al. 2015).
Hypomethylation and upregulation was shown TMEM200C, RGS10, ADAM12 and PAM,
several of them being dysregulated in cancer and proposed oncogenes (Cacan, Ali et al. 2014,
Shao, Li et al. 2014, Soni, Bode et al. 2020)
Table 1: DMPs vs gene expression in “Subset Early metastasis” vs “Subset No metastasis”. Gene. FC: relative
change in gene expression between “Subset Early metastasis” vs “Subset No metastasis” represented by fold
change. . P-value for gene expression FC. Genes in bold are hypermethylated and downregulated or
hypomethylated and upregulated. Gene region from the 450k array. Relation of CpG probe to CpG Island.
betaFC: Beta fold change; negative value: differentially less methylated in “Subset Early metastasis” vs “Subset
No metastasis” and positive value: differentially more methylated in “Subset Early metastasis” vs” Subset No
metastasis”. Adjusted P-value for betaFC in DNA methylation. Chromosome. CpG site: CpG probe from the
450k array.
Gene FC P-value Gene region Relation to
Island
betaFC Adj. P-
value
Chr CpG site
RNF13 -2.03 0.040 TSS1500 N-Shore 0.34 0.044 chr3 cg15108553
ZNF217 -1.87 0.035 TSS200 S-Shore 0.65 0.031 chr20 cg22164891
HYAL1 -1.70 0.005 TSS1500 Open Sea 0.42 0.048 chr3 cg12930727
AMN1 1.51 0.032 TSS1500 Island 0.53 0.031 chr12 cg16014770
CD47 1.54 0.017 TSS1500 Island 0.75 0.031 chr3 cg17216759
MTCH1 1.60 0.024 TSS1500 S-Shore 0.49 0.045 chr6 cg25254170
20
TUBA1B 1.66 0.018 TSS1500 S-Shore 0.53 0.031 chr12 cg13709639
SHARPIN 1.67 0.015 TSS15005UTR Island 0.41 0.045 chr8 cg11883258
TMEM200C 1.94 0.0006 1exon;TSS200 S-Shore -0.49 0.042 chr18 cg26438105
TMEM200C 1.94 0.0006 1exon:TSS200 S-Shore -0.73 0.03 chr18 cg1608637
RGS10 1.98 0,033 Body Island -0.51 0.040 chr10 cg04041960
ADAM12 2.07 0.028 Body N-Shore -0.46 0.047 chr10 cg16018302
PAM 2.21 0.031 Body Open Sea -0.42 0.044 chr5 cg22911687
EIF2AK2 2.30 0.006 TSS200 Island 0.53 0.048 chr2 cg01617117
Table 2: DMRs vs gene expression in “Subset Early metastasis” vs “Subset No metastasis”. Gene. FC: relative
change in gene expression between “Subset Early metastasis” vs “Subset No metastasis” represented by fold
change. P-value for gene expression FC. Coordinates: Chromosome number, start and end. No.CpGs: Number of
CpGs. betaFC: Mean beta fold change; negative value: differentially less methylated in “Subset Early
metastasis” vs “Subset No metastasis” and positive value: differentially more methylated in “Subset Early
metastasis” vs” Subset No metastasis”. Adjusted P-value for betaFC in DNA methylation.
Gene FC P-value Coordinates No.CpGs betaFC Adj.P-value
TMEM200C 1.94 0.0006 chr18:5890293- 5892245 9 -0.53 0.00054
ZNF217 -1.87 0.035 chr20: 52199520- 52199778 5 0.52 0.00274
3.5 Functional pathway enrichment of integrated DNA methylation and gene expression
of “Subset Early metastasis” vs “Subset No metastasis”
Gene Ontology (GO) analysis was performed to study biological functions related to
negatively and positively correlated gene expression associated with corresponding changes
in DMPs.
DMPs were calculated using a cut off betaFC ≥0.2 and a less stringent p-value (unadjusted
p<0.05) allowing us to examine a larger set of differentially methylated genes. Candidate
genes were detected using a cut off FC≥1.5 and p<0.05 for gene expression. Integrative
21
analysis resulted in a list of 110 candidate genes being upregulated and hypomethylated,
while 22 candidate genes were downregulated and hypermethylated (Supplementary table 3).
Carcinoma was identified as the top disease linked to these in total 132 hypo/upregulated and
hyper/downregulated genes (Figure 4). Axonal guidance signaling was identified as a top
pathway for the hypermethylated/downregulated genes and also a pathway present for the
hypomethylated/upregulated genes- both associated with carcinoma annotation (Figure 4).
Axonal guidance signaling genes were hypomethylated/upregulated and
hypermethylated/downregulated in “Subset Early metastasis” vs “Subset No metastasis”
(Figure 5 and 6).
22
Figure 4. Diseases or Functions Annotations for “Subset Early metastasis” vs “Subset No metastasis” in (A)
hypermethylated and upregulated genes and (B) hypomethylated and upregulated genes.
23
Figure 5. Canonical pathway (CP) Axonal Guidance Signaling is deregulated in “Subset Early metastasis” vs
“Subset No metastasis”. Lines arepointing towards the genes involved. Genes being hypomethylated and
upregulated are marked with pink boxes, while genes that are hypermethylated and downregulated are marked
with green boxes. P-values are right below the boxes and gene expression fold change (FC) below the p-values.
Figure 6. Ingenuity pathway analysis (IPA) showing 20 of 22 hypermethylated and downregulated genes in
“Subset Early metastasis” vs “Subset No metastasis” annotated to Carcinoma (p=2,64E-06). Connections
24
between genes are shown by purple arrows. Canonical pathway (CP) Axonal Guidance Signaling is shown by
lines pointing towards the genes involved.
3.6 Immunohistochemical evaluation of CD45 expression of “Subset Early metastasis”
vs “Subset No metastasis”
CD45 expression was assessed for “Subset Early metastasis” (n=4) and “Subset No
metastasis” (n=4) showing a low number of leukocytes in the sections evaluated
(Supplementary figure 6).
4 Discussion
Metastatic spread in UM can be seen several years after primary diagnosis and treatment
(Kujala, Makitie et al. 2003). The use of archived FFPE tissue provides the opportunity to
conduct retrospective studies and has a vast potential in characterization of aberrant DNA
methylation associated with cancer relapse. The possibility of comparing DNA methylation
patterns in the primary tumor to metastases of the same patient years after the primary
diagnosis should also be emphasized. The definitive endpoint in patients with UM is overall
survival. One of the strengths of this study is the implementation of the unique data material
available through The Cancer Registry of Norway and the Norwegian Cause of Death
Registry. The Norwegian Cause of Death registry collects data on deaths by age, sex, cause,
place of death, and place of residence for Norway. It contains digitized cause of death data
dating back to 1951. The Cancer Registry of Norway collects data on all new/ suspected cases
of cancers and cancer relapse, and medical doctors in the country are instructed by law to
notify this registry. The possibility of human errors in the registration of data is greatly
reduced due to the unique personal identification number in Norway. To reduce registration
bias, data from the two registries were cross-checked to ensure that the cause of death
25
coincided with cancer diagnosis. However, data from the registries cannot account for
undetected metastatic disease.
UM tumors that metastasize are known to have increased inflammatory cell populations, thus
tumor infiltration of other cell types, especially infiltration of leukocytes, could be a
confounding factor when assessing the UM methylome. The degree of leukocyte infiltration
was investigated by HE staining and by IHC of CD45 for donors in “Subset Early metastasis”
and “Subset No metastasis” showing a low degree of leukocyte infiltration in both subsets.
Yet, though the sections used for immunohistological evaluation were taken from the same
blocks and in close proximity of the sections used for DNA methylation analyses, there is a
potential risk of contamination by inflammatory cell populations in our DNA methylation
analyses.
Previous studies have demonstrated a relationship between the global methylation profiles of
UM and risk of metastatic disease assessed by gene expression profile (Robertson, Shih et al.
2018, Field, Kuznetsov et al. 2019) . We were able to reproduce a PCA plot that positioned
samples into two clusters based on chromosome 3 status. The same was demonstrated by
consensus clustering into 2 clusters. Our study was unable to detect a strong correlation
between DNA methylation clustering and risk of metastasis as shown by Robertson et al by
consensus clustering of eighty samples into 4 clusters. This could be due to our small sample
size and high level of heterogeneity within our predefined sample groups, e.g. the groups
classified by time of detected metastasis (“Early”, “Late” and “No”) all contained samples
with various histopathological classification, generating additional variability between the
samples. Importantly, samples were included based on survival properties, not genetic profile.
EIF1AX, SF3B1 and BAP1 status was unknown for our FFPE-derived samples. Three samples
with partial monosomy 3 were excluded in order to analyze pure M3 vs D3, these samples
clustered in between M3 and D3. Cases with partial M3 have previously been shown to
26
cluster near the transition point between Class 1 and Class 2 tumors and are associated with
Class 1 gene expression profile (Field, Durante et al. 2018, Field, Kuznetsov et al. 2019).
Donor 22 and 23 were both long-term survivors, while donor 21 died of metastatic disease
within 5 years. Recent work suggests that partial deletion of chromosome 3 encompassing the
BAP1 locus is associated with poor prognosis (Rodrigues, Ait Rais et al. 2020).
Significant DMPs and DMRs were detected by selecting more homogenous sample groups.
The comparison “Subset Early metastasis” vs “Subset No metastasis” yielded 348 DMPs and
36 DMRs. These DMPs and DMRs were cross-linked to gene expression data from the same
donors, thus revealing a potential mechanistic role of DNA methylation in the regulation of 14
genes for the DMPs and 2 genes for the 2detected DMRs. There is a general consensus that
methylation in the close proximity of the transcription start site (TSS) is associated with
silencing of gene expression (Jones 2012). The effect of methylation in the gene body on the
other hand is enigmatic; methylation of the 1st Exon is tightly linked to gene silencing, while
gene body methylation is associated with increased expression (Brenet, Moh et al. 2011,
Jones 2012, Anastasiadi, Esteve-Codina et al. 2018). Several of the DMPs associated with
gene expression were located in promoter areas and support the general assumption that
increased/ decreased methylation within promoter areas decrease/ increase the gene
expression of the associated gene respectively.
RNF13, ZNF217 and HYAL1 were among the genes that were hypermethylated and
downregulated in “Subset Early metastasis” vs “Subset No metastasis”. These genes could
thus be potential tumor suppressors and markers for early metastasis and poor prognosis in
UM. RNF13 (Ring finger protein 13) knockdown cells are reported to be resistant to apoptosis
and JNK activation triggered by ER stress, indicating that hypermethylation and
downregulation of RNF13 could make the cells more resistant to apoptosis (Arshad, Ye et al.
2013). ZNF217 (Zink finger protein 217) functions as an oncogene in several tumors and is
27
mostly upregulated (Cohen, Donini et al. 2015). However, this transcription factor is also
known for its dual role as a transcription activator and repressor, thus it might function as a
tumor suppressor in UM by facilitating tumor growth when it is hypermethylated and
repressed. As tumors grow they often outgrow their blood supply, leading to hypoxic
conditions in parts of the microenvironment. Hypoxia is a proposed trigger for methylation of
ZNF217 (Yuen, Chen et al. 2013). HYAL1 (Hyaluronidase 1) encodes a hyaluronidase that
degrades hyaluronan in the extracellular matrix. HYAL1 is known to be dysregulated in a
variety of cancers and both elevated and reduced depending on cancer type (Frost, Mohapatra
et al. 2000, Hautmann, Lokeshwar et al. 2001, Wang, Grigorieva et al. 2008, Tan, Wang et al.
2011).
TMEM200C, RGS10, ADAM12 and PAM were among the genes that were hypomethylated
and upregulated in “Subset Early metastasis” vs “Subset No metastasis”. ADAM12
(Disintegrin And Metalloproteinase Domain-Containing Protein 12) is a disintegrin and
metalloproteinase that can perform a proteolytic "shedding" of membrane-associated proteins
ectodomain and hence the rapid modulation of key cell signaling pathways in tissue
microenvironment. A variety of cytokines, chemokines and growth factors are activated by
these sheddase activities. ADAM12 is upregulated in breast cancer and has been reported to be
a diagnostic marker for the proliferation, migration and invasion in patients with small cell
lung cancer and promotes a stem cell-like phenotype in claudin-low breast cancer (Nariţa,
Anghel et al. 2010, Shao, Li et al. 2014, Duhachek-Muggy, Qi et al. 2017) ADAM12 was
hypomethylated and upregulated in “Subset Early metastasis” vs “Subset No metastasis” and
could function as an oncogene in UM, hence being a possible marker for early metastasis and
poor prognosis. PAM (Peptidylglycine Alpha-Amidating Monooxygenase) encodes a protein
that catalyzes the biosynthesis of many signaling peptides in humans. Reduction of PAM
expression increased survival of mice in a glioblastoma model and reduced the formation of
28
blood vessels in vitro, suggesting PAM is a potential target for antiangiogenic therapy in
glioblastoma (Soni, Bode et al. 2020). In line with the findings by Soni et al, reduced
expression of PAM might reduce neovascularization of UM metastasis, thus being a possible
treatment strategy. TMEM200C is a gene with relatively little experimental and functional
information. A differential DNA methylation study has identified it as a candidate gene
related to psychiatric illness, though its function remains elusive (Esposito, Jones et al. 2016).
A potential role for TMEM200C in UM pathogenesis and metastasis is yet to be investigated.
RGS10 (Regulator of G protein signaling 10) suppresses proinflammatory macrophage
responses and enhances survival. Studies in ovarian cancer cells suggest that RGS10 is
transcriptionally regulated by DNA and histone-targeted epigenetic mechanisms (Ali, Cacan
et al. 2013, Cacan, Ali et al. 2014). Hypomethylated and upregulated RGS10 could function as
an oncogene in UM by protecting cancer cells from proinflammatory macrophage responses
and enhance their survival.
Several of our findings are in support of previous publications (Field, Durante et al. 2018,
Robertson, Shih et al. 2018) particularly hypermethylated probes in “Subset Early metastasis”
vs “Subset No metastasis” were enriched in “shore” regions up to 1500 bp upstream of TSS
and methylation was inversely correlated to gene expression. In addition, hypomethylated
probes in “Subset Early metastasis” vs “Subset No metastasis” were mostly enriched in “open
sea” regions and gene body regions.
Significant pathways and biological functions of this comparison included genes associated
with carcinoma and malignant transformation. Furthermore, dysregulation of axonal guidance
signalling was implicated, showing hypermethylation of 7 of 11 genes and hypomethylation
of 13 of 25 genes previously described in Class 2 UM (Field, Kuznetsov et al. 2019). We also
found several hypermethylated sites in the gene body of ROBO1 in our early metastatic
samples.
29
The assessment of a mechanistic role o DNA methylation in gene regulation is outside the
boundaries of the present study, we havehighlighted a set of differentially methylated genes
discriminating poor and good prognosis (“Subset Early metastasis” vs “Subset No
metastasis”). Metastatic UM is a devastating disease, urging for new and improved therapy,
hence restoration of aberrant DNA methylation should be explored as a potential therapeutic
target. DNA hypomethylating agents have shown great promise in the treatment of
hematological malignancies (Khan, Pathe et al. 2012, Derissen, Beijnen et al. 2013). In the
setting of solid tumors, the use of epidrugs to restore sensitivity to cytotoxic or hormonal
drugs is a major goal (Fu, Hu et al. 2011, Dullea and Marignol 2016). Restoration of
chemosensitivity is especially appealing for UM, a malignancy recognized by its
chemoresistance (Buder, Gesierich et al. 2013).
In conclusion, we present differential DNA methylation profiles between subgroups
correlated to early vs no metastasis and ultimately cancer survival. The present work
accentuates factors involved in differential DNA methylation in UM and features changes in
DNA methylation correlated with gene expression in patients who develop metastatic UM.
Acknowledgments
We would like to acknowledge all personnel at the Center for Eye Research and at the
Department of Ophthalmology, Oslo University Hospital (OUH) that contributed to this
project. We thank the Genomics and Affimetrix Core Facilities at Radiumhospitalet and
Ullevål, OUH for performing the DNA methylation and RNA microarrays, with special
thanks to Berit Sletbakk Brusletto .We would also like to thank Borghild Roald at the
Department of Pathology, OUH for help with granting access to the archived FPPE tissue.
Additionally, we thank GeneVia Technologies, Finland for running the differential DNA
30
methylation analyses, and Clara-Cecilie Günther at Norwegian Computing Center (NR) for
valuable help running the preliminary DNA methylation analysis.
The work was funded by the South-Eastern Norway Regional Health Authority (project
2012104), Norwegian Cancer Society (project 5808589) and supported by grants from
Norwegian Association of the Blind and Partially Sighted, Arthur and Odd Clausons
ophthalmological fund, Aase and Knut Tønjums ophthalmological fund, Futura fund, Unifor
Frimed, Inger Holms memorial fund, “Stiftelsen for fremme av kreftforskning” at University
of Oslo and “Legat til fremme av kreftforskning”. All authors contributing to the study have
read and approved the manuscript. There are no conflicts of interest for any of the authors.
References
Ali, M. W., E. Cacan, Y. Liu, J. Y. Pierce, W. T. Creasman, M. M. Murph, R. Govindarajan, S. T. Eblen, S. F. Greer and S. B. Hooks (2013). "Transcriptional suppression, DNA methylation, and histone deacetylation of the regulator of G-protein signaling 10 (RGS10) gene in ovarian cancer cells." PLoS One 8(3): e60185.10.1371/journal.pone.0060185
Anastasiadi, D., A. Esteve-Codina and F. Piferrer (2018). "Consistent inverse correlation between DNA methylation of the first intron and gene expression across tissues and species." Epigenetics Chromatin 11(1): 37.10.1186/s13072-018-0205-1
Arshad, M., Z. Ye, X. Gu, C. K. Wong, Y. Liu, D. Li, L. Zhou, Y. Zhang, W. P. Bay, V. C. Yu and P. Li (2013). "RNF13, a RING finger protein, mediates endoplasmic reticulum stress-induced apoptosis through the inositol-requiring enzyme (IRE1α)/c-Jun NH2-terminal kinase pathway." J Biol Chem 288(12): 8726-8736.10.1074/jbc.M112.368829
Aryee, M. J., A. E. Jaffe, H. Corrada-Bravo, C. Ladd-Acosta, A. P. Feinberg, K. D. Hansen and R. A. Irizarry (2014). "Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays." Bioinformatics 30(10): 1363-1369.10.1093/bioinformatics/btu049
Baylin, S. B. and P. A. Jones (2016). "Epigenetic Determinants of Cancer." Cold Spring Harb Perspect Biol 8(9).10.1101/cshperspect.a019505
Benjamini, Y. and Y. Hochberg (1995). "Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing." Journal of the Royal Statistical Society Series B-Statistical Methodology 57(1): 289-300.DOI 10.1111/j.2517-6161.1995.tb02031.x
31
Brenet, F., M. Moh, P. Funk, E. Feierstein, A. J. Viale, N. D. Socci and J. M. Scandura (2011). "DNA methylation of the first exon is tightly linked to transcriptional silencing." PLoS One 6(1): e14524.10.1371/journal.pone.0014524
Brocato, J. and M. Costa (2013). "Basic mechanics of DNA methylation and the unique landscape of the DNA methylome in metal-induced carcinogenesis." Crit Rev Toxicol 43(6): 493-514.10.3109/10408444.2013.794769
Buder, K., A. Gesierich, G. Gelbrich and M. Goebeler (2013). "Systemic treatment of metastatic uveal melanoma: review of literature and future perspectives." Cancer Med 2(5): 674-686.10.1002/cam4.133
Cacan, E., M. W. Ali, N. H. Boyd, S. B. Hooks and S. F. Greer (2014). "Inhibition of HDAC1 and DNMT1 modulate RGS10 expression and decrease ovarian cancer chemoresistance." PLoS One 9(1): e87455.10.1371/journal.pone.0087455
Chen, Y. A., M. Lemire, S. Choufani, D. T. Butcher, D. Grafodatskaya, B. W. Zanke, S. Gallinger, T. J. Hudson and R. Weksberg (2013). "Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray." Epigenetics 8(2): 203-209.10.4161/epi.23470
Cohen, P. A., C. F. Donini, N. T. Nguyen, H. Lincet and J. A. Vendrell (2015). "The dark side of ZNF217, a key regulator of tumorigenesis with powerful biomarker value." Oncotarget 6(39): 41566-41581.10.18632/oncotarget.5893
de Ruijter, T. C., J. P. de Hoon, J. Slaats, B. de Vries, M. J. Janssen, T. van Wezel, M. J. Aarts, M. van Engeland, V. C. Tjan-Heijnen, L. Van Neste and J. Veeck (2015). "Formalin-fixed, paraffin-embedded (FFPE) tissue epigenomics using Infinium HumanMethylation450 BeadChip assays." Lab Invest 95(7): 833-842.10.1038/labinvest.2015.53
Derissen, E. J., J. H. Beijnen and J. H. Schellens (2013). "Concise drug review: azacitidine and decitabine." Oncologist 18(5): 619-624.10.1634/theoncologist.2012-0465
Duhachek-Muggy, S., Y. Qi, R. Wise, L. Alyahya, H. Li, J. Hodge and A. Zolkiewska (2017). "Metalloprotease-disintegrin ADAM12 actively promotes the stem cell-like phenotype in claudin-low breast cancer." Mol Cancer 16(1): 32.10.1186/s12943-017-0599-6
Dullea, A. and L. Marignol (2016). "MGMT testing allows for personalised therapy in the temozolomide era." Tumour Biol 37(1): 87-96.10.1007/s13277-015-4240-2
Dumenil, T. D., L. F. Wockner, M. Bettington, D. M. McKeone, K. Klein, L. M. Bowdler, G. W. Montgomery, B. A. Leggett and V. L. Whitehall (2014). "Genome-wide DNA methylation analysis of formalin-fixed paraffin embedded colorectal cancer tissue." Genes Chromosomes Cancer 53(7): 537-548.10.1002/gcc.22164
Edgar, R., M. Domrachev and A. E. Lash (2002). "Gene Expression Omnibus: NCBI gene expression and hybridization array data repository." Nucleic Acids Res 30(1): 207-210.10.1093/nar/30.1.207
Esposito, E. A., M. J. Jones, J. R. Doom, J. L. MacIsaac, M. R. Gunnar and M. S. Kobor (2016). "Differential DNA methylation in peripheral blood mononuclear cells in adolescents exposed to significant early but not later childhood adversity." Dev Psychopathol 28(4pt2): 1385-1399.10.1017/s0954579416000055
32
Field, M. G., M. A. Durante, H. Anbunathan, L. Z. Cai, C. L. Decatur, A. M. Bowcock, S. Kurtenbach and J. W. Harbour (2018). "Punctuated evolution of canonical genomic aberrations in uveal melanoma." Nat Commun 9(1): 116.10.1038/s41467-017-02428-w
Field, M. G., J. N. Kuznetsov, P. L. Bussies, L. Z. Cai, K. A. Alawa, C. L. Decatur, S. Kurtenbach and J. W. Harbour (2019). "BAP1 Loss Is Associated with DNA Methylomic Repatterning in Highly Aggressive Class 2 Uveal Melanomas." Clin Cancer Res 25(18): 5663-5673.10.1158/1078-0432.Ccr-19-0366
Frost, G. I., G. Mohapatra, T. M. Wong, A. B. Csóka, J. W. Gray and R. Stern (2000). "HYAL1LUCA-1, a candidate tumor suppressor gene on chromosome 3p21.3, is inactivated in head and neck squamous cell carcinomas by aberrant splicing of pre-mRNA." Oncogene 19(7): 870-877.10.1038/sj.onc.1203317
Fu, S., W. Hu, R. Iyer, J. J. Kavanagh, R. L. Coleman, C. F. Levenback, A. K. Sood, J. K. Wolf, D. M. Gershenson, M. Markman, B. T. Hennessy, R. Kurzrock and R. C. Bast, Jr. (2011). "Phase 1b-2a study to reverse platinum resistance through use of a hypomethylating agent, azacitidine, in patients with platinum-resistant or platinum-refractory epithelial ovarian cancer." Cancer 117(8): 1661-1669.10.1002/cncr.25701
Hansen, K. D. (2016). "IlluminaHumanMethylation450kanno.ilmn12.hg19: Annotation for Illumina's 450k methylation arrays. R package version 0.6.0. ."
Harbour, J. W., M. D. Onken, E. D. Roberson, S. Duan, L. Cao, L. A. Worley, M. L. Council, K. A. Matatall, C. Helms and A. M. Bowcock (2010). "Frequent mutation of BAP1 in metastasizing uveal melanomas." Science 330(6009): 1410-1413.10.1126/science.1194472
Hautmann, S. H., V. B. Lokeshwar, G. L. Schroeder, F. Civantos, R. C. Duncan, R. Gnann, M. G. Friedrich and M. S. Soloway (2001). "Elevated tissue expression of hyaluronic acid and hyaluronidase validates the HA-HAase urine test for bladder cancer." J Urol 165(6 Pt 1): 2068-2074.10.1097/00005392-200106000-00072
Herman, J. G., F. Latif, Y. Weng, M. I. Lerman, B. Zbar, S. Liu, D. Samid, D. S. Duan, J. R. Gnarra, W. M. Linehan and et al. (1994). "Silencing of the VHL tumor-suppressor gene by DNA methylation in renal carcinoma." Proc Natl Acad Sci U S A 91(21): 9700-9704.10.1073/pnas.91.21.9700
Hovestadt, V. and M. Zapatka (2017). "Enhanced copy-number variation analysis using Illumina DNA methylation arrays. R package version 1.9.0."
Jones, P. A. (2012). "Functions of DNA methylation: islands, start sites, gene bodies and beyond." Nat Rev Genet 13(7): 484-492.10.1038/nrg3230
Kaliki, S., C. L. Shields and J. A. Shields (2015). "Uveal melanoma: estimating prognosis." Indian J Ophthalmol 63(2): 93-102.10.4103/0301-4738.154367
Khan, C., N. Pathe, S. Fazal, J. Lister and J. M. Rossetti (2012). "Azacitidine in the management of patients with myelodysplastic syndromes." Ther Adv Hematol 3(6): 355-373.10.1177/2040620712464882
Kujala, E., T. Makitie and T. Kivela (2003). "Very long-term prognosis of patients with malignant uveal melanoma." Invest Ophthalmol Vis Sci 44(11): 4651-4659.10.1167/iovs.03-0538
Landreville, S., O. A. Agapova and J. W. Harbour (2008). "Emerging insights into the molecular pathogenesis of uveal melanoma." Future Oncol 4(5): 629-636.10.2217/14796694.4.5.629
33
Lawrence, M., W. Huber, H. Pages, P. Aboyoun, M. Carlson, R. Gentleman, M. T. Morgan and V. J. Carey (2013). "Software for computing and annotating genomic ranges." PLoS Comput Biol 9(8): e1003118.10.1371/journal.pcbi.1003118
Maat, W., S. H. Beiboer, M. J. Jager, G. P. Luyten, N. A. Gruis and P. A. van der Velden (2008). "Epigenetic regulation identifies RASEF as a tumor-suppressor gene in uveal melanoma." Invest Ophthalmol Vis Sci 49(4): 1291-1298.10.1167/iovs.07-1135
Maat, W., P. A. van der Velden, C. Out-Luiting, M. Plug, A. Dirks-Mulder, M. J. Jager and N. A. Gruis (2007). "Epigenetic inactivation of RASSF1a in uveal melanoma." Invest Ophthalmol Vis Sci 48(2): 486-490.10.1167/iovs.06-0781
Moran, S., M. Vizoso, A. Martinez-Cardus, A. Gomez, X. Matias-Guiu, S. M. Chiavenna, A. G. Fernandez and M. Esteller (2014). "Validation of DNA methylation profiling in formalin-fixed paraffin-embedded samples using the Infinium HumanMethylation450 Microarray." Epigenetics 9(6): 829-833.10.4161/epi.28790
Nariţa, D., A. Anghel, E. Seclaman, R. Ilina, N. Cireap and S. Ursoniu (2010). "Molecular profiling of ADAM12 gene in breast cancers." Rom J Morphol Embryol 51(4): 669-676
Ness, C., O. Garred, N. A. Eide, T. Kumar, O. K. Olstad, T. P. Baerland, G. Petrovski, M. C. Moe and A. Noer (2017). "Multicellular tumor spheroids of human uveal melanoma induce genes associated with anoikis resistance, lipogenesis, and SSXs." Mol Vis 23: 680-694
Neumann, L. C., A. Weinhausel, S. Thomas, B. Horsthemke, D. R. Lohmann and M. Zeschnigk (2011). "EFS shows biallelic methylation in uveal melanoma with poor prognosis as well as tissue-specific methylation." BMC Cancer 11: 380.10.1186/1471-2407-11-380
Peters, T. J., M. J. Buckley, A. L. Statham, R. Pidsley, K. Samaras, V. L. R, S. J. Clark and P. L. Molloy (2015). "De novo identification of differentially methylated regions in the human genome." Epigenetics Chromatin 8: 6.10.1186/1756-8935-8-6
Prescher, G., N. Bornfeld and R. Becher (1990). "Nonrandom chromosomal abnormalities in primary uveal melanoma." J Natl Cancer Inst 82(22): 1765-1769.10.1093/jnci/82.22.1765
R Core Team. (2018). "R: A Language and Environment for Statistical Computing." from https://www.R-project.org/.
Ritchie, M. E., B. Phipson, D. Wu, Y. Hu, C. W. Law, W. Shi and G. K. Smyth (2015). "limma powers differential expression analyses for RNA-sequencing and microarray studies." Nucleic Acids Res 43(7): e47.10.1093/nar/gkv007
Robertson, A. G., J. Shih, C. Yau, E. A. Gibb, J. Oba, K. L. Mungall, J. M. Hess, V. Uzunangelov, V. Walter, L. Danilova, T. M. Lichtenberg, M. Kucherlapati, P. K. Kimes, M. Tang, A. Penson, O. Babur, R. Akbani, C. A. Bristow, K. A. Hoadley, L. Iype, M. T. Chang, T. R. Network, A. D. Cherniack, C. Benz, G. B. Mills, R. G. W. Verhaak, K. G. Griewank, I. Felau, J. C. Zenklusen, J. E. Gershenwald, L. Schoenfield, A. J. Lazar, M. H. Abdel-Rahman, S. Roman-Roman, M. H. Stern, C. M. Cebulla, M. D. Williams, M. J. Jager, S. E. Coupland, B. Esmaeli, C. Kandoth and S. E. Woodman (2018). "Integrative Analysis Identifies Four Molecular and Clinical Subsets in Uveal Melanoma." Cancer Cell 33(1): 151.10.1016/j.ccell.2017.12.013
Rodrigues, M., K. Ait Rais, F. Salviat, N. Algret, F. Simaga, R. Barnhill, S. Gardrat, V. Servois, P. Mariani, S. Piperno-Neumann, S. Roman-Roman, O. Delattre, N. Cassoux, A. Savignoni, M. H. Stern and G.
34
Pierron (2020). "Association of Partial Chromosome 3 Deletion in Uveal Melanomas With Metastasis-Free Survival." JAMA Ophthalmol 138(2): 182-188.10.1001/jamaophthalmol.2019.5403
Shao, S., Z. Li, W. Gao, G. Yu, D. Liu and F. Pan (2014). "ADAM-12 as a diagnostic marker for the proliferation, migration and invasion in patients with small cell lung cancer." PLoS One 9(1): e85936.10.1371/journal.pone.0085936
Shen, H. and P. W. Laird (2013). "Interplay between the cancer genome and epigenome." Cell 153(1): 38-55.10.1016/j.cell.2013.03.008
Soni, H., J. Bode, C. D. L. Nguyen, L. Puccio, M. Neßling, R. M. Piro, J. Bub, E. Phillips, R. Ahrends, B. A. Eipper, B. Tews and V. Goidts (2020). "PERK-mediated expression of peptidylglycine α-amidating monooxygenase supports angiogenesis in glioblastoma." Oncogenesis 9(2): 18.10.1038/s41389-020-0201-8
Tan, J. X., X. Y. Wang, H. Y. Li, X. L. Su, L. Wang, L. Ran, K. Zheng and G. S. Ren (2011). "HYAL1 overexpression is correlated with the malignant behavior of human breast cancer." Int J Cancer 128(6): 1303-1315.10.1002/ijc.25460
van Beek, J. G. M., A. E. Koopmans, R. M. Verdijk, N. C. Naus, A. de Klein and E. Kilic (2012). Diagnosis, Histopathologic and Genetic Classification of Uveal Melanoma. Melanoma - From Early Detection to Treatment. G. H. T. Duc, InTech: 137-169.
van der Velden, P. A., J. A. Metzelaar-Blok, W. Bergman, H. Monique, H. Hurks, R. R. Frants, N. A. Gruis and M. J. Jager (2001). "Promoter hypermethylation: a common cause of reduced p16(INK4a) expression in uveal melanoma." Cancer Res 61(13): 5303-5306
van der Velden, P. A., W. Zuidervaart, M. H. Hurks, S. Pavey, B. R. Ksander, E. Krijgsman, R. R. Frants, C. P. Tensen, R. Willemze, M. J. Jager and N. A. Gruis (2003). "Expression profiling reveals that methylation of TIMP3 is involved in uveal melanoma development." Int J Cancer 106(4): 472-479.10.1002/ijc.11262
Van Raamsdonk, C. D., V. Bezrookove, G. Green, J. Bauer, L. Gaugler, J. M. O'Brien, E. M. Simpson, G. S. Barsh and B. C. Bastian (2009). "Frequent somatic mutations of GNAQ in uveal melanoma and blue naevi." Nature 457(7229): 599-602.10.1038/nature07586
Van Raamsdonk, C. D., K. G. Griewank, M. B. Crosby, M. C. Garrido, S. Vemula, T. Wiesner, A. C. Obenauf, W. Wackernagel, G. Green, N. Bouvier, M. M. Sozen, G. Baimukanova, R. Roy, A. Heguy, I. Dolgalev, R. Khanin, K. Busam, M. R. Speicher, J. O'Brien and B. C. Bastian (2010). "Mutations in GNA11 in uveal melanoma." N Engl J Med 363(23): 2191-2199.10.1056/NEJMoa1000584
Wang, F., E. V. Grigorieva, J. Li, V. N. Senchenko, T. V. Pavlova, E. A. Anedchenko, A. V. Kudryavtseva, A. Tsimanis, D. Angeloni, M. I. Lerman, V. I. Kashuba, G. Klein and E. R. Zabarovsky (2008). "HYAL1 and HYAL2 inhibit tumour growth in vivo but not in vitro." PLoS One 3(8): e3031.10.1371/journal.pone.0003031
Weiner, J. (2020). "pca3D: Three Dimensional PCA Plots. R package version 0.10.2." from https://CRAN.R-project.org/package=pca3d.
Wilkerson, M. D. and D. N. Hayes (2010). "ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking." Bioinformatics 26(12): 1572-1573.10.1093/bioinformatics/btq170
35
Yamashita, N., E. Tokunaga, H. Kitao, M. Hitchins, Y. Inoue, K. Tanaka, Y. Hisamatsu, K. Taketani, S. Akiyoshi, S. Okada, Y. Oda, H. Saeki, E. Oki and Y. Maehara (2015). "Epigenetic Inactivation of BRCA1 Through Promoter Hypermethylation and Its Clinical Importance in Triple-Negative Breast Cancer." Clin Breast Cancer 15(6): 498-504.10.1016/j.clbc.2015.06.009
Yuen, R. K., B. Chen, J. D. Blair, W. P. Robinson and D. M. Nelson (2013). "Hypoxia alters the epigenetic profile in cultured human placental trophoblasts." Epigenetics 8(2): 192-202.10.4161/epi.23400
Supplementary data
Supplementary Figure 1: Unsupervised three-dimensional principal component analysis (PCA) based on
methylation profiling of 23 samples (1-23) showing differential clustering of samples according to chromosome
3 status. Blue (D3): No deletion of chromosome 3= disomy 3. Red (M3): deletion of chromosome 3 =
monosomy 3. Green (D3/M3): Partial loss of chromosome 3.
36
Supplementary Figure 2: Unsupervised principal component analysis (PCA) based on methylation profiling
showing differential clustering of the 20 samples (sample 1-20). Samples have been colored according to
histopathological classification “Epithelioid” (blue) “Spindle cell” (green) and “Mixed” (red).
Supplementary Figure 3: Unsupervised principal component analysis (PCA) based on methylation profiling
showing differential clustering of the 20 samples (sample 1-20). Samples have been colored according to
histopathological classification “Epithelioid” (blue) “Spindle cell” (green) and “Mixed” (red).
37
Supplementary Figure 4: Hierarchical clustering of up- and downregulated genes in uveal melanoma sample 2,
15, 16 and 17 (“Subset Early metastasis”) vs sample 7, 11,12and 13 (“Subset No metastasis”), where each row
represents the single sample tested, while each column represents a single probe set (gene name) analyzed.
Relative gene expression is presented in color: Red is higher-level expression relative to the sample mean, blue
is relatively lower level expression, and gray is no change in expression.
38
Supplementary Figure 5: Quantitative reverse-transcription PCR (qRT-PCR) of HTR2B, ADGRG1, ADAM23,
ROBO1, CNTN3 and PALMD in support of the microarray findings in “Subset Early metastasis” (sample 2, 15,
16 and 17) vs “Subset No metastasis” (7, 11, 12 and 13).
39
Supplementary figure 6: CD45 expression in “Subset Early metastasis” (sample 3, 17, 18 and 19, n=4) and
“Subset No metastasis” (sample 8,13, 14 and 15, n=4) was assessed by immunohistochemistry using CD45
antibody (1:100). CD45 positive cells (leukocytes) were identified by red staining.
40
Supplementary table 1: Patient and tumor characteristics. #: Sample number. Group: E=”Early metastasis”,
L=”Late metastasis” and N= “No metastasis”. Histo=Histology: Ep=Epithelioid. S= Spindle. M= Mixed. Chr3:
chromosome 3 status assessed by copy number variation analysis: monosomy 3 (M3) or disomy 3 (D3). Size:
Size of the primary tumor from the pathology reports in millimeter. Age: Age at diagnosis. Sex: M=male.
F=female. Enucleation: Year of enucleation. Met: Year of registered metastatic disease. Death: Year of death.
All patients with registered metastatic disease died from UM. Patients without registered metastatic disease died
from other causes. SA: Still alive when data was collected. Subset: SEM= “Subset Early metastasis”. SNM=
“Subset No metastasis”. RNA: Samples included in RNA analyses. TNM: classification at diagnosis (T=tumor
size, N= nodes involved, M= metastasis).
# Group Histo Chr3 Size Age Sex Enucleation Met Death Subset RNA TNM
1 E Spi M3 11x11
x8
39 M 1988 1992 1992 T2 Nx M0
2 E Spi M3 12x12
x3
57 M 1985 1988 1988 SEM RNA T2 Nx M0
3 E Epi M3 8x12 70 M 1989 1992 1994 T3 Nx M0
4 L Spi M3 10x6 72 M 1979 1990 1990 T2 Nx M0
5 L Mix M3 15x7 34 F 1980 2000 2001 T3 Nx M0
6 N Epi M3 11x7x
5
72 F 1981 - 1998 RNA T2 Nx M0
7 N Spi D3 10 43 F 1988 - SA SNM T1 Nx M0
8 L Epi M3 12x10 75 F 1978 1989 1989 T3 Nx M0
9 L Spi M3 9x6 62 F 1982 1992 1992 T1 Nx M0
10 N Mix M3 7x8x9 70 F 1986 - 2004 T2 Nx M0
11 N Spi D3 14 59 F 1989 - SA SNM RNA T2 Nx M0
12 N Spi D3 9x9x6 52 F 1988 - SA SNM RNA T1 Nx M0
13 N Spi D3 10x10
x7
51 F 1987 - SA SNM RNA T2 Nx M0
14 E Epi M3 10x8 66 F 1981 1983 1983 T2 Nx M0
15 E Mix M3 15x3 39 M 1983 1985 1985 SEM RNA T2 Nx M0
16 E Mix M3 10x10 85 F 1989 1991 1991 SEM RNA T3 Nx M0
17 E Spi M3 12 35 M 1989 1992 1992 SEM RNA T3 Nx M0
18 L Mix M3 15x15
x6
61 M 1982 1993 1994 T2 Nx M0
19 L Mix D3 11x9x
6
57 M 1976 1991 1991 T2 Nx M0
20 L Spi M3 10x6 36 F 1979 1988 1988 T2 Nx M0
41
Supplementary table 2: DMRcate analysis identifying 36 DMRs (Stouffer<0.05) between donor 2, 15, 16 and
17 (“Subset Early metastasis”) vs samples 7, 11, 12 and 13 (“Subset No metastasis”). Coordinate: Chromosome
number, start and end. No.CpGs: Number of CpGs. beta FC: Mean beta fold change; negative value:
differentially less methylated in “Subset Early metastasis” vs “Subset No metastasis” and positive value:
differentially more methylated in “Subset Early metastasis” vs” Subset No metastasis”. Gene(s): Genes
associated with overlapping promoters.
Coordinates No.CpGs betaFC Gene (s)
chr11:68621650-68621969 2 -0.58
chr17:40935998-40936820 8 -0.55 WNK4
chr18:5890293-5892245 9 -0.53 TMEM200C
chr7:92237896-92238364 6 -0.43
chr15:41952429-41952827 7 -0.34 MGA
chr7:129912487-129913310 6 -0.31
chr2:223176683-223177742 10 -0.28
chr10:2978022-2978687 6 -0.27 RP11
chr10:1205222-1205942 10 -0.16 LINC00200
chr12:6756088-6757257 7 0.24 ACRBP
chr6:32123034-32123651 5 0.25 PPT2. PPT2-EGFL8.
chr19:17877419-17877846 6 0.32
chr10:42971011-42971732 5 0.32 LINC00839
chr15:78556178-78557584 13 0.32 DNAJA4, RP11
chr2:220041926-220042451 5 0.34 FAM134A, CNPPD1.
chr4:111397134-111397581 7 0.35 ENPEP
chr19:36246395-36246816 5 0.36 HSPB6, LIN37
chr1:8013974-8014650 6 0.38 PARK7
chr8:145728138-145729106 14 0.39 GPT. PPP1R16A. CTD
chr4:79861272-79861398 3 0.39 PAQR3
chr2:20551058-20551234 2 0.41 PUM2
chr16:3225044-3225401 3 0.41
chr12:12867669-12867753 3 0.41 CDKN1B
chr11:68611260-68611806 5 0.41 CPT1A
chr2:177418561-177418905 4 0.43
chr14:24779959-24780691 8 0.43 CIDEB, LTB4R
chr4:140216130-140216770 4 0.44 NDUFC1
chr8:2075209-2075820 4 0.45 MYOM2
chr8:145638881-145639181 3 0.45 SLC39A4
chr13:30077315-30077489 3 0.46
chr19:47288039-47288263 5 0.49 SLC1A
chr10:106088702-106089003 4 0.51 ITPRIP
chr20:52199520-52199778 5 0.52 ZNF217
chr17:47091038-47091521 3 0.52 IGF2BP1, RP11
chr17:80008917-80009015 2 0.58 GPS1, RFNG
chr14:73712902-73712967 2 0.66 RNU6, RP4
42
Supplementary table 3: List of the ten most up- and downregulated genes, from the microarray one-way
ANOVA (≥ 1.5 fold up- or down- regulated, p<0.05) in uveal melanoma sample 2, 15, 16 and 17 (“Subset
Early metastasis”) vs sample 7, 11, 12 and 13 (“Subset No metastasis”).
Gene Fold Change Gene Fold Change
ADGRG1 6.56 ZNF667-AS1 -2.38 HTR2B 6.12 KCNK2 -2.74 ADAM23 5.14 MIR548V -2.86 LINC01531 5.07 DTWD1 -3.02 CAPN3 4.79 HPGD -3.18 WARS 4.73 PRRT3-AS1 -3.54 MAP2 4.69 SEMA3C -4.45 LINC00152 4.63 CNTN3 -4.57 ANXA2 4.50 ROBO1 -4.74 PTPRM 4.37 PALMD -4.94
43
Supplementary table 4. Changes in DNA methylation (unadjusted p<0.05) correlated to fold change (FC);
relative changes in gene expression in uveal melanoma between “Subset Early metastasis” vs “Subset No
metastasis”. P-value for gene expression FC. The left side of table shows changes in gene expression (FC≥1.5,
p<0.05) and the right side of the table show changes in methylation (betaFC≥0.2, p<0.05). betaFC: beta fold
change; negative value: differentially less methylated in “Subset Early metastasis” vs “Subset No metastasis”
and positive value: differentially more methylated in “Subset Early metastasis” vs” Subset No
metastasis”.Location of the methylation site given by gene region, relation to island, Chr = chromosome and
CpG site. 22 genes downregulated/48 DMPs hypermethylated and 110 genes upregulated/174 DMPs
hypomethylated.
Gene FC p-value
Gene region Relation to Island
betaFC p-value Chr CpG site
PALMD -4.94 0.004 1stExon;5'UTR OpenSea 0.46 0.015 chr1 cg24603803
PALMD -4.94 0.004 TSS200 OpenSea 0.42 0.018 chr1 cg27563423
PALMD -4.94 0.004 Body OpenSea 0.49 0.024 chr1 cg02080641
ROBO1 -4.74 0.008 Body OpenSea 0.32 0.010 chr3 cg06807029
ROBO1 -4.74 0.008 Body N_shelf 0.33 0.018 chr3 cg15442678
ROBO1 -4.74 0.008 Body OpenSea 0.22 0.024 chr3 cg07107976
SEMA3C -4.45 0.033 Body OpenSea 0.41 0.010 chr7 cg01556677
SEMA3C -4.45 0.033 Body OpenSea 0.50 0.013 chr7 cg09587880
SEMA3C -4.45 0.033 5'UTR N_shore 0.26 0.029 chr7 cg14796406
KCNK2 -2.74 0.015 Body OpenSea 0.36 0.003 chr1 cg04923840
RNF13 -2.03 0.040 TSS1500 N_shore 0.34 1,49E-05 chr3 cg15108553
RNF13 -2.03 0.040 TSS1500 N-shore 0.34 0.0004 chr3 cg04118462
RNF13 -2.03 0.040 TSS200 Island 0.25 0.002 chr3 cg10802371
ZNF217 -1.87 0.035 TSS200 S_shore 0.65 2,85E-06 chr20 cg22164891
ZNF217 -1.87 0.035 TSS200 S_shore 0.60 4,81E-06 chr20 cg09228833
ZNF217 -1.87 0.035 TSS200 S_shore 0.60 2,15E-05 chr20 cg20979153
ZNF217 -1.87 0.035 5'UTR;1stExon S_shore 0.56 0.0001 chr20 cg09029902
MIR641 -1.81 0.006 TSS1500;5'UTR N_shore 0.41 0.001 chr19 cg07815521
MIR641 -1.81 0.006 TSS1500;5'UTR N_shore 0.34 0.001 chr19 cg06055845
MIR641 -1.81 0.006 TSS1500;5'UTR N_shore 0.24 0.005 chr19 cg26620021
MIR641 -1.81 0.006 TSS1500;5'UTR N_shore 0.39 0.027 chr19 cg09380135
MMP24 -1.74 0.0004 TSS1500 N_shore 0.35 0.0001 chr20 cg15270813
MMP24 -1.74 0.0004 TSS1500 N_shore 0.30 0.0008 chr20 cg12483876
CPS1 -1.73 0.004 Body OpenSea 0.24 0.030 chr2 cg21967368
HYAL1 -1.70 0.004 TSS1500;5'UTR OpenSea 0.42 3,63E-05 chr3 cg12930727
HYAL1 -1.70 0.004 TSS200;5'UTR OpenSea 0.23 0.004 chr3 cg14943722
SYNGAP1 -1.69 0.026 Body Island 0.28 0.006 chr6 cg18466911
HOXA6 -1.66 0.0129 TSS200 Island 0.31 0.014 chr17 cg14044640
HOXA6 -1.66 0.0129 1stExon Island 0.44 0.010 chr17 cg23129930
FBXO17 -1.65 0.0009 5'UTR N_shore 0.49 0.001 chr19 cg08820801
HSPB7 -1.60 0.028 1stExon;5'UTR OpenSea 0.32 0.049 chr1 cg16110455
HSPB7 -1.60 0.028 TSS200 OpenSea 0,34 0.002 chr1 cg13320181
KCNG4 -1.60 0.001 Body N_shore 0.26 0.001 chr16 cg01992487
LEKR1 -1.60 0.005 5'UTR S_shore 0.49 0.043 chr3 cg02354125
LEKR1 -1.60 0.005 TSS1500 N_shore 0.25 0.023 chr3 cg08832018
LEKR1 -1.60 0.005 TSS1500 N_shore 0.21 0.044 chr3 cg01201279
44
CTDSP1 -1.57 0.035 Body;TSS200 S_shore 0.24 0.026 chr2 cg14045814
COL5A3 -1.56 0.009 Body Island 0.26 0.036 chr19 cg20478934
COL5A3 -1.56 0.009 Body N_shore 0.33 0.04 chr19 cg17177699
ZNF876P -1.56 0.047 TSS200 N_shore 0.30 0.030 chr4 cg23063647
PAX6 -1.54 0.004 Body S_shore 0.27 0.001 chr11 cg04598774
PAX6 -1.54 0.004 5'UTR S_shore 0.31 0.014 chr11 cg09217215
PELI2 -1.53 0.029 TS1500 N_shore 0.57 6.77E-05 chr14 cg06744740
PELI2 -1.53 0.029 Body OpenSea 0.26 0.007 chr14 ch.14.689341R
PELI2 -1.53 0.029 Body S_shore 0.26 0.027 chr14 cg01208566
PELI2 -1.53 0.029 Body S_shore 0.32 0.036 chr14 cg11121623
PELI2 -1.53 0.029 Body OpenSea 0.23 0.037 chr14 cg18148021
PRR18 -1.53 0.042 TSS1500 Island 0.25 0.034 chr6 cg01243879
ZNF609 1.51 0.018 Body Open Sea -0.35 0.021 chr15 cg26916780
BAZ1B 1.51 0.011 Body Open Sea -0.30 0.006 chr7 cg12446543
NFIC 1.51 0,037 Body N_Shore -0.35 0.020 chr19 cg24848615
AKAP10 1.51 0,041 Body Open Sea -0.30 0.034 chr17 cg04132472
HDAC4 1.52 0.002 Body Open Sea -0.48 0.003 chr2 cg17410431
HDAC4 1.52 0.002 Body Open Sea -0.47 0.008 chr2 cg10071550
HDAC4 1.52 0.002 3'UTR N_Shore -0.42 0.0002 chr2 cg15964153
USP47 1.52 0.011 Body Open Sea -0.24 0.003 chr11 cg02269797
CAPN2 1.52 0.032 Body Island -0.42 0.009 chr1 cg06756211
CAPN2 1.52 0.032 Body Island -0.39 0.041 chr1 cg19598416
PIK3R5 1.52 0.021 Body Open Sea -0.23 0.035 chr17 cg05244974
PIK3R5 1.52 0.021 5'UTR;TSS200 Open Sea -0.21 0.011 chr17 cg24251850
LYN 1.53 0.004 5'UTR Open Sea -0.21 0.008 chr8 cg05973028
TAPBP 1.54 0.033 Body N_Shore -0.26 0.033 chr6 cg01253676
PYGO1 1.56 0.028 TSS1500 N_Shore -0.20 0.001 chr15 cg13878116
RELL1 1.57 0.012 3'UTR Open Sea -0.23 0.050 chr4 cg19029127
CCND2 1.58 0.002 Body Open Sea -0.43 0.0002 chr12 cg14834893
CCND2 1.58 0.002 Body Open Sea -0.34 0.010 chr12 cg17558623
SREBF1 1.58 0.032 Body Open Sea -0.26 0.040 chr17 cg09796270
TFDP1 1.58 0.018 Body S_Shore -0.34 0.003 chr13 cg21258259
TFDP1 1.58 0.018 Body N_Shore -0.23 0.001 chr13 cg00590320
LIMD1 1.59 0,011 TSS200 S_Shore -0.28 0.015 chr3 cg18779283
LIMD1 1.59 0,011 1stExon S_Shore -0.25 0.019 chr3 cg04037228
LIMD1 1.59 0,011 Body Open Sea -0.24 0.024 chr3 cg25437886
LIMD1 1.59 0,011 TSS1500 S_Shore -0.23 0.012 chr3 cg08534342
LIMD1 1.59 0,011 TSS200 S_Shore -0.22 0.029 chr3 cg03534662
ZEB1 1.59 0.007 TSS1500 N_Shore -0.40 0.006 chr10 cg03719128
ZEB1 1.59 0.007 TSS1500 N_Shore -0.32 0.028 chr10 cg00520933
ADARB2 1.60 0.005 Body N_shore -0,43 0.005 chr10 cg16646662
ADARB2 1.60 0.005 Body Open Sea -0.36 0.005 chr10 cg06422309
ADARB2 1.60 0.005 Body N_shore -0.31 0.038 chr10 cg17285208
ADARB2 1.60 0.005 Body Open Sea -0.27 0.0005 chr10 cg20423602
ADARB2 1.60 0.005 Body S_Shore -0.21 0.011 chr10 cg12438430
ADARB2 1.60 0.005 Body Island -0.20 0.011 chr10 cg01561194
45
HLA-DMA 1.60 0.043 TSS1500 Open Sea -0.23 0.01 chr6 cg17940902
HERC2 1.60 0.006 Body N_Shore -0.23 0.04 chr15 cg10839322
RAB4A 1.60 0.007 TSS1500 Island -0.22 0.009 chr1 cg09850632
CD44 1.61 0.011 Body Open Sea -0.28 0.016 chr11 cg15239179
EPSTI1 1.62 0.016 TSS200 Island -0.51 0.019 chr13 cg18882819
EPSTI1 1.62 0.016 TSS200 Island -0.38 0.027 chr13 cg22125968
EPSTI1 1.62 0.016 TSS200 S_Shore -0.34 0.020 chr13 cg03478249
WWC2 1.62 0.016 Body Open Sea -0.31 0.006 chr4 cg12998503
NCOR2 1.62 0.0017 Body N_Shore -0.41 0.003 chr12 cg03507593
NCOR2 1.62 0.0017 Body N_Shore -0.40 0.004 chr12 cg21626573
NCOR2 1.62 0.0017 Body N_Shore -0.35 0.012 chr12 cg05596926
NCOR2 1.62 0.0017 Body N_Self -0.33 9,03E-05 chr12 cg09267427
NCOR2 1.62 0.0017 Body Open Sea -0.22 0.017 chr12 cg12157156
NCOR2 1.62 0.0017 Body Open Sea -0.20 0.015 chr12 cg25754673
IGF1R 1.63 0.013 Body Open Sea -0.28 0.014 chr15 cg26272088
STAU2 1.63 0.025 Body Open Sea -0.44 0.011 chr8 cg21585977
IGF2R 1.64 0.030 Body Open Sea -0.42 0.006 chr6 cg02092589
EXOSC8 1.64 0.022 Body S_Shelf -0.24 0.010 chr13 cg27249554
SERINC1 1.64 0.005 Body Open Sea -0.37 0.002 chr6 cg12480176
HDLBP 1.65 0.01 Body Open Sea -0.25 0.015 chr2 cg16881309
HDLBP 1.65 0.01 5'UTR N_Shore -0.23 0.013 chr2 cg09564509
BMPR1B 1.65 0.050 5'UTR S_Shore -0.30 0.019 chr4 cg27391693
UBE2H 1.67 0.031 Body Open Sea -0.33 0.004 chr17 cg00993830
CPNE3 1.68 0.002 3'UTR Open Sea -0.34 0.003 chr8 cg02712949
PANK2 1.68 0.005 5'UTR N_Shore -0.20 0.019 chr22 cg14810501
YAP1 1.68 0.005 Body Open Sea -0.29 0.012 chr11 cg15999356
NUMA1 1.68 0.012 Body Open Sea -0.23 0.001 chr11 cg05032348
GPX1 1.60 0.031 3'UTR N_Shore -0.29 0.039 chr3 cg18642234
DYNC1H1 1.68 0.038 3'UTR Open Sea -0.49 0.011 chr14 cg21186263
DYNC1H1 1.68 0.038 Body Open Sea -0.26 0.012 chr14 cg20471297
DYNC1H1 1.68 0.038 Body Open Sea -0.22 0.008 chr14 cg02879081
PRMT2 1.69 0.009 Body Island -0.27 0.022 chr17 cg21461082
GFPT1 1.70 0.002 Body Open Sea -0.24 0.01 chr2 cg15899800
SLC25A13 1.70 0.002 Body Open Sea -0.25 0.040 chr7 cg20502039
RUNX1 1.70 0.036 Body Island -0.52 0.005 chr21 cg11498607
RUNX1 1.70 0.036 Body Island -0.48 0.029 chr21 cg05000748
RUNX1 1.70 0.036 5'UTR;1stExon;Body N_shore -0.27 0.024 chr21 cg01725383
PITPNA 1.71 0.006 Body Island -0.25 0.007 chr17 cg03804148
SH3PXD2A 1.70 0.017 Body N_shore -0.27 0.014 chr10 cg04688330
SH3PXD2A 1.70 0.017 Body Open Sea -0.26 0.014 chr10 cg12636499
MDM4 1.72 0.040 Body Open Sea -0.25 0.027 chr1 cg17158762
PGM1 1.73 0.007 TSS200 N_Shore -0.42 0.022 chr1 cg03373115
B4GALT7 1.73 0.021 Body N_Shore -0.28 0.003 chr5 cg13095737
CMIP 1.74 0.005 Body Open Sea -0.37 0.01 chr16 cg01799671
SNX9 1.74 0.030 Body Open Sea -0.21 0.004 chr6 cg05163325
TRAF3IP2 1.75 0.014 Body;5'UTR Open Sea -0.21 0.038 chr6 cg24634333
TANC1 1.76 0.030 Body Open Sea -0.28 0.001 chr2 cg23966795
46
DYNC1I2 1.76 0.014 TSS1500 N_Shore -0.40 0.01 chr2 cg11046380
B4GALNT3 1.76 0,014 Body Island -0.42 0.036 chr12 cg26388816
B4GALNT3 1.76 0,014 Body Open Sea -0.23 0.032 chr12 cg23491661
B4GALNT3 1.76 0,014 Body Open Sea -0.21 0.017 chr12 cg08965685
MFN2 1.78 0.004 5'UTR S_Shore -0.28 0.025 chr1 cg13216073
PMP22 1.78 0.003 Body Open Sea -0.37 0.0016 chr17 cg07710335
COG2 1.78 0.013 TSS1500 N_Shore -0.38 0.05 chr1 cg13234110
PALLD 1.79 0.053 5'UTR; Body Island -0.42 0.011 chr4 cg13573928
EXT1 1.80 0.033 Body Open Sea -0.31 0.040 chr8 cg14485744
EXT1 1.80 0.033 Body Open Sea -0.29 0.049 chr8 cg20547777
EXT1 1.80 0.033 Body Open Sea -0.28 0.0005 chr8 cg11064524
VDAC1 1.80 0.001 Body Open Sea -0.28 0.004 chr5 cg23256802
EWSR1 1.80 0.002 Body Open Sea -0.30 0.006 chr22 cg24351767
ARPC1B 1.80 0.003 TSS200 Island -0.20 0.011 chr7 cg08798295
TOB1 1.81 0.025 1stExon N_Shore -0.25 0.004 chr17 cg14494812
TPP2 1.81 0.042 3'UTR Open Sea -0.26 9,88E-05 chr13 cg13519549
B3GALT4 1.82 0.008 1st
Exon Island -0.63 0.011 chr6 cg03108070
B3GALT4 1.82 0.008 1st
Exon Island -0.40 0.043 chr6 cg06753439
B3GALT4 1.82 0.008 1st
Exon Island -0.26 0.008 chr6 cg17416730
ERLIN2 1.82 0.006 3'UTR Open Sea -0.24 0.012 chr8 cg26393977
FNBP1 1.83 0.001 Body N_Shore -0.23 0.005 chr9 cg06901890
TRAK2 1.83 0.012 TSS200;5'UTR;1stExon Island -0.22 0.015 chr2 cg14417676
PI4KA 1.86 0.003 Body Open Sea -0.48 0.04 chr22 cg02953144
CAMKK2 1.89 0.010 1stExon;5'UTR;TSS1500 S_Shore -0.20 0.009 chr12 cg00500936
SGCD 1.89 0.003 Body Open Sea -0.35 0.016 chr5 cg26439139
GATAD2B 1.92 0.037 5'UTR Open Sea -0.24 0.026 chr1 cg04137323
TMEM200C 1.94 0.0006 1stExon Island -0.73 3,44E-06 chr18 cg16086373
TMEM200C 1.94 0.0006 1stExon Island -0.71 0.002 chr18 cg21447871
TMEM200C 1.94 0.0006 1stExon Island -0.66 0.001 chr18 cg09366312
TMEM200C 1.94 0.0006 TSS200 S_Shore -0.61 0.003 chr18 cg27093273
TMEM200C 1.94 0.0006 1stExon S_Shore -0.52 0.0003 chr18 cg00058163
TMEM200C 1.94 0.0006 1stExon Island -0.50 0.0002 chr18 cg18139195
TMEM200C 1.94 0.0006 TSS200 S_Shore -0.49 1,29E-05 chr18 cg26438105
TMEM200C 1.94 0.0006 TSS200 S_Shore -0.27 0.006 chr18 cg17586988
TMEM200C 1.94 0.0006 1stExon Island -0.26 0.046 chr18 cg12899381
GNAS 1.94 0,002 5'UTR;1stExon Island -0.27 0.0002 chr20 cg16737409
MET 1.96 0.012 Body Open Sea -0.43 0.015 chr7 cg05997059
MET 1.96 0.012 Body Open Sea -0.28 0.0002 chr7 cg18285813
CNNM2 1.97 0.011 Body Open Sea -0.45 0.002 chr10 cg03493300
RGS10 1.98 0.033 Body Island -0.51 1,03E-05 chr10 cg04041960
LRP1 1.98 0.014 Body Island -0.33 0.004 chr12 cg12146864
LRP1 1.98 0.014 Body Open Sea -0.30 0.027 chr12 cg16766632
LRP1 1.98 0.014 Body Island -0.25 0.034 chr12 cg01276169
CSNK2A1 2.00 0.035 Body Open Sea -0.20 0.019 chr20 cg07789225
EXOC1 2.02 0.016 TSS1500 N_Shore -0.23 0.003 chr4 cg26329992
CUX1 2.02 0.003 Body Open Sea -0.45 0.0001 chr7 cg02169185
CUX1 2.02 0.003 Body Open Sea -0.50 0.0001 chr7 cg17148755
47
CUX1 2.02 0.003 Body Open Sea -0.25 0.004 chr7 cg18498598
PDGFRA 2.05 0.019 TSS200 S_Shore -0.29 0.040 chr4 cg02170478
RBPMS 2.05 0.030 Body Open Sea -0.35 0.004 chr8 cg24705939
RBPMS 2.05 0.030 Body Open Sea -0.40 0.001 chr8 cg00997969
FGFR1 2.05 0.003 5'UTR N_Shore -0.63 0.0008 chr8 cg00400221
FGFR1 2.05 0.003 5'UTR N_Shore -0.60 0.0007 chr8 cg13123964
ADAM12 2.07 0.028 Body N_shore -0.46 3,07E-05 chr10 cg16018302
ADAM12 2.07 0.028 Body Open Sea -0.32 0.0002 chr10 cg08993079
ADAM12 2.07 0.028 Body Open Sea -0.38 0.024 chr10 cg17287034
RASAL2 2.09 0.027 Body Open Sea -0.21 0.004 chr1 cg15462736
ATXN7L1 2.10 0.046 Body Open Sea -0.31 0.029 chr7 cg11932891
ATXN7L1 2.10 0.046 Body Open Sea -0.22 0.013 chr7 cg22661239
UTRN 2.10 0.005 Body Open Sea -0.55 0.001 chr6 cg12121162
RNF213 2.11 0.005 Body Open Sea -0.22 0.010 chr17 cg18784565
ADARB1 2.13 0.021 5'UTR;Body S_Shore -0,21 0.005 chr21 cg05516004
HMGB1 2.13 0.019 TSS1500 Open Sea -0.34 0.001 chr13 cg05818394
PAM 2.21 0.003 Body Open Sea -0.42 1,65E-05 chr5 cg22911687
PAM 2.21 0.003 TSS200 N_Shore -0.39 0.002 chr5 cg15999165
PAM 2.21 0.003 TSS1500 N_Shore -0.40 0.019 chr5 cg23021168
PAM 2.21 0.003 TSS1500 N_Shore -0.25 0.018 chr5 cg20131596
UBE2M 2.23 0.002 TSS200 Island -0.23 0.043 chr19 cg23186294
SLCO2B1 2.23 0.013 TSS1500 Open Sea -0.26 0.001 chr11 cg16244299
GAB1 2.29 0,029 Body Open Sea -0.31 0.004 chr4 cg12710519
GAB1 2.29 0,029 Body N_shore -0.27 0.009 chr6 cg05966641
CHST11 2.36 0.001 Body S_Shore -0.20 0.004 chr12 cg23855505
CHRNA10 2.39 0.049 TSS1500 Open Sea -0.44 0.0008 chr11 cg07484827
CHRNA10 2.39 0.049 TSS1500 Open Sea -0.32 0.023 chr11 cg26745143
HLA-DRA 2.51 0.011 Body Open Sea -0.23 0.02 chr6 cg23732629
CD74 2.55 0.0004 TSS200 Open Sea -0.32 0.005 chr4 cg01601628
CD74 2.55 0.0004 TSS200 Open Sea -0.25 0.050 chr4 cg24548564
CD74 2.55 0.0004 1stExon;5'UTR;TSS200 Open Sea -0.27 0.038 chr4 cg26129545
STK32A 2.76 0.016 Body Open Sea -0.25 0.012 chr5 cg23346625
ELFN1 2.83 0.003 5'UTR Open Sea -0.40 0.005 chr7 cg09160231
ELFN1 2.83 0.003 5'UTR N_Shore -0.25 0.008 chr7 cg17324095
ELFN1 2.83 0.003 5'UTR S_Shore -0.23 0.007 chr7 cg22104371
ZDHHC7 2.83 0.007 Body Open Sea -0.32 0.001 chr16 cg16671652
B2M 2.99 0.003 TSS1500 N_shore -0.25 0.009 chr15 cg18555073
IGFBP7 3.02 0.030 Body Open Sea -0.52 0.006 chr4 cg14824921
CSNK1A1 3,22 0.003 Body Open Sea -0.26 0.04 chr5 cg21229718
PTP4A3 3.65 0.003 Body N_Shore -0.22 0.017 chr8 cg02059849
SPTBN1 3.80 0.015 Body S_Shore -0.41 0.002 chr2 cg10929758
SGK1 3.81 0.006 TSS1500 S_shore -0.46 0.001 chr6 cg24937675
SGK1 3.81 0.006 TSS1500 S_shore -0.41 0.014 chr6 cg12871835
ANXA2 4.50 0.001 Body Open Sea -0.53 0.0002 chr15 cg03957109
ANXA2 4.50 0.001 Body Open Sea -0.37 0.0002 chr15 cg22581200
CAPN3 4.79 0.003 Body Open Sea -0.23 0.035 chr15 cg18425651
48
Appendix A: 348 hypo-/ hypermethylated probes in “Subset Early metastasis” vs “ Subset No metastasis”
subgroup. Name; Name of CpG probe. Chr; Chromosome. Log FC; log2(mean M-value of contrast group 1
samples) - log2(mean M-value of contrast group 2 samples. Negative value: differentially less methylated in
“Subset Early metastasis” vs “Subset No metastasis”. Positive value differentially more methylated in “Subset
Early metastasis” vs “Subset No metastasis”. Islands name; Genomic location of CpG island. Relation to CpG
Island. UCSC RefGene Accession; UCSC reference gene accession number. UCSC reference gene group;
genomic locations in relation to genes.
III
