METABOLOMICS AND
PROTEOMICS STUDIES OF BRAIN
TUMORS - A CHEMOMETRIC
BIOINFORMATICS APPROACH
Lina Mörén
Department of Chemistry
Umeå University, Sweden
2015
Copyright © Lina Mörén
ISBN: 978-91-7601-354-0
Front cover picture: Lina Mörén
Printed by: Lars Åberg, VMC-KBC, Umeå
Umeå, Sweden 2015
“If we knew what it was we were doing,
it would not be called research, would it?”
Albert Einstein
ABSTRACT
The WHO classification of brain tumors is based on histological features and the
aggressiveness of the tumor is classified from grade I to IV, where grade IV is the
most aggressive. Today, the correlation between prognosis and tumor grade is the
most important component in tumor classification. High grade gliomas,
glioblastomas, are associated with poor prognosis and a median survival of 14
months including all available treatments. Low grade meningiomas, usually
benign grade I tumors, are in most cases cured by surgical resection. However
despite their benign appearance grade I meningiomas can, without any
histopathological signs, in some cases develop bone invasive growth and become
lethal. Thus, it is necessary to improve conventional treatment modalities,
develop new treatment strategies and improve the knowledge regarding the basic
pathophysiology in the classification and treatment of brain tumors.
In this thesis, both proteomics and metabolomics have been applied in the search
for biomarkers or biomarker patterns in two different types of brain tumors,
gliomas and meningiomas. Proteomic studies were carried out mainly by surface
enhanced laser desorption ionization time of flight mass spectrometry (SELDITOF-MS). In one of the studies, isobaric tags for relative and absolute quantitation
(iTRAQ) labeling in combination with high-performance liquid chromatography
(HPLC) was used for protein detection and identification. For metabolomics, gaschromatography time-of-flight mass spectrometry (GC-TOF-MS) has been the
main platform used throughout this work for generation of robust global
metabolite profiles in tissue, blood and cell cultures. To deal with the complexity
of the generated data, and to be able to extract relevant biomarker patters or
latent biomarkers, for interpretation, prediction and prognosis, bioinformatic
strategies based on chemometrics were applied throughout the studies of the
thesis.
In summary, we detected differentiating protein profiles between invasive and
non-invasive meningiomas, in both fibrous and meningothelial tumors.
Furthermore, in a different study we discovered treatment induce protein pattern
changes in a rat glioma model treated with an angiogenesis inhibitor. We
identified a cluster of proteins linked to angiogenesis. One of those proteins,
HSP90, was found elevated in relation to treatment in tumors, following ELISA
validation. An interesting observation in a separate study was that it was possible
to detect metabolite pattern changes in the serum metabolome, as an effect of
treatment with radiotherapy, and that these pattern changes differed between
different patients, highlighting a possibility for monitoring individual treatment
response. In the fourth study of this work, we investigated tissue and serum from
glioma patients that revealed differences in the metabolome between
glioblastoma and oligodendroglioma, as well as between oligodendroglioma
grade II and grade III. In addition, we discovered metabolite patterns associated
to survival in both glioblastoma and oligodendroglioma. In our final work, we
identified metabolite pattern differences between cell lines from a subgroup of
glioblastomas lacking argininosuccinate synthetase (ASS1) expression, (ASS1
negative glioblastomas), making them auxotrophic for arginine, a metabolite
required for tumor growth and proliferation, as compared to glioblastomas with
normal ASS1 expression (ASS1 positive). From the identified metabolite pattern
differences we could verify the hypothesized alterations in the arginine
biosynthetic pathway. We also identified additional interesting metabolites that
may provide clues for future diagnostics and treatments. Finally, we were able to
verify the specific treatment effect of ASS1 negative cells by means of arginine
deprivation on a metabolic level.
Keywords: Glioblastoma, glioma, meningioma, metabolomics, proteomics, massspectrometry
CONTENTS
LIST OF PAPERS
Other papers by the author not appended in this thesis
POPULÄRVETENSKAPLIG SAMMANFATTNING PÅ SVENSKA
ABBREVATIONS AND NOTATIONS
Abbreviations
Notation
III
iv
V
VII
vii
viii
BACKGROUND
Some common genomic alterations underlying gliomagenesis
Common traits of tumor metabolism
Brain tumors
Standard treatment for glioma and meningioma
Glioma and meningioma biomarkers
Biological sample matrices
1
3
5
7
9
9
13
AIMS OF THIS WORK
16
METHODS
Generation of proteomics data
Generation of metabolomics data
Data processing
Chemometrics
Validation
17
17
19
21
23
25
RESULTS AND DISCUSSION OF THE PAPERS
Paper I
Paper II
Paper III
Paper IV
Paper V
28
28
30
32
34
37
CONCLUSION
40
REFERENCES
43
ACKNOWLEDGEMENTS
59
i
ii
LIST OF PAPERS
This thesis is based on the following papers, referred to in the text by the Roman
numerals in bold font:
I.
Wibom C, Mörén L, Aarhus M, Knappskog PM, Lund-Johansen M,
Antti H, Bergenheim AT. (2009) Proteomic profiles differ between
bone invasive and noninvasive benign meningiomas of fibrous and
meningothelial subtype. Journal Of Neuro-Oncology, 94, 321-331.
II.
Mörén L, Pernemalm M, Sandström M, Wibom C, Bergenheim AT,
Johansson M, Lehtiö J, Antti H. A two-step proteomics approach
reveals protein pattern changes in a rat glioma model in response to
tyrosine kinase inhibitor treatment (Manuscript)
III.
Mörén L, Wibom C, Bergström P, Johansson M, Antti H, Bergenheim
AT. Characterization of the serum metabolome following radiation
treatment in patients with high-grade gliomas. (Submitted
manuscript)
IV.
Mörén L, Bergenheim AT, Ghasimi S, Brännström T, Johansson M,
Antti H. (2015) Metabolomic screening of tumor tissue and serum in
glioma patients reveals diagnostic and prognostic information.
Metabolites, 5, 502-520.
V.
Mörén L, Perryman R, Langer J, O´Neill K, Syed N, Antti H.
Metabolomic analysis of glioblastoma multiforme reveal potential
diagnosis and treatment options for ASS1 negative tumours.
(Manuscript)
Paper I was reprinted with kind permission from Springer Science+Business
Media. For paper IV no permission was needed.
iii
Other papers by the author not appended in this
thesis
VI.
Wibom C, Surowiec I, Mörén L, Bergström P, Johansson M, Antti H,
Bergenheim AT. (2010) Metabolomic patterns in glioblastoma and
changes during radiotherapy – a clinical microdialysis study. Journal
of proteome research, 9, 2909-2919.
VII.
Sjöberg RL, Bergenheim T, Mörén L, Antti H, Lindgren C, Naredi S,
Lindvall P. (2015) Blood Metabolomic Predictors of 1-year Outcome
in Subarachnoid Hemorrhage. Neurocritical Care. 23, 225-232.
iv
POPULÄRVETENSKAPLIG
SAMMANFATTNING PÅ SVENSKA
Cancer är en växande folksjukdom. Enligt Socialstyrelsen så kommer var tredje
svensk att drabbas av cancer någon gång i livet. Bröst- och prostatacancer är de
vanligaste typerna av cancer och utgör ca 30 % av alla cancerfall hos kvinnor
respektive män. Maligna gliom, en typ av hjärntumör som denna avhandling till
största delen behandlar, utgör 2 % av alla cancerfall hos vuxna men på grund av
dess elakartade karaktär utgör det den fjärde vanligaste orsaken till dödsfall till
följd av cancer.
En hjärntumör uppstår när celler i hjärnan börjar dela sig okontrollerat. Detta
inträffar när cellens regleringsmekanismer för kontroll av celldelningen har blivit
skadade. Beroende på vilken typ av cell som skadats uppstår olika sorters tumörer
med varierande prognoser. Maligna gliom är en tumörtyp som uppstår från
gliaceller, hjärnans stödjeceller. Gliom är indelade i undergrupper och har olika
prognos beroende på vilken typ av gliaceller som de uppkommit från samt
beroende på hur aggressivt växtsätt de har. En av dessa undergrupper kallas
glioblastom vilken är en av de allvarligaste hjärntumörerna man kan drabbas av.
Detta beror på att tumörcellerna sprider sig snabbt från själva tumören till
omkringliggande vävnad vilket gör det i stort sett omöjligt att operera bort alla
tumörceller. Därför krävs kompletterande behandling med cellgifter och
strålning. Glioblastom är en aggressiv tumör där medianöverlevnaden är omkring
4 månader om patienten inte får adekvat behandling. Med all tillgänglig
behandling (operation, strålbehandling och cytostatika) är prognosen trots allt
dålig med en medianöverlevnad på endast 14-16 månader.
Många tumörformer är elakartade (benämns maligna) men vissa är godartade
(benämns benigna) till sin karaktär. Behandlingsmässigt är många av de benigna
botbara med radikal kirurgi. Med radikal kirurgi menas att man gör bedömningen
att ingen tumörvävnad finns kvar efter operation. En typ av benign tumör är
meningiom vilken vi även studerat i denna avhandling. Menigiom uppstår i
hinnorna som omsluter hjärnan och ger varierande symptom beroende på var i
skallkaviteten tumören växer. De flesta meningiom går att operera radikalt utan
några bestående men för patienten medan några tumörer sitter känsligt till vilket
gör radikal operation omöjlig. Dessa tumörrester kan ibland utveckla ett
aggressivt tillväxtbeteende och kan därför ibland även behandlas med
strålningsterapi. Ett problem är dock att man inte kan förutse vilka meningiom
som kommer att utveckla ett aggressiv tillväxtbeteende. Detta försvårar
planeringen av patientens behandling då man vill undvika att strålbehandla
patienter som inte är i behov av ytterligare behandling.
v
Vi har i denna avhandling mätt nivåer av proteiner och metaboliter (små
molekyler) i vävnadsprover, blodprover och cellinjer från både patienter och
försöksdjur med hjärntumörer. Detta för att om möjligt finna biomarkörer i form
av förändrade proteiner eller metaboliter, som ska kunna mätas för att ställa olika
diagnoser, göra en prognostisk bedömning eller bedöma tumörens svar på
behandling. I samtliga arbeten har vi har använt oss av masspektrometriska
analysmetoder vilka genererat stora mängder informationsrik data i form av
protein och metabolitmönster. För att analysera och hantera dessa stora
datamängder har vi använt oss av kemometriska eller multivariata
analysmetoder. Detta möjliggör en överblick av informationen samt ger möjlighet
att påvisa skillnader mellan olika grupper vi analyserat (t.ex. olika
gliomdiagnoser). I de två första arbetena analyserade vi proteinmönster. I det
första arbetet påvisade vi proteinmönster i tumörvävnaden som skilde invasivt
växande menigiom från icke-invasivt växande menigiom. I den andra studien, på
försöksdjur, fann vi proteiner i blod och i tumörvävnad som skilde råttor
behandlade med cytostatika från råttor som ej blivit behandlade med cytostatika.
I de följande tre arbetena analyserade vi metabolitmönster hos patienter med
gliom. Vid en undersökning av blodprover från patienter med glioblastom som
behandlats med strålningsterapi kunde vi, i det tredje arbetet, påvisa metabola
förändringar i blod till följd av strålbehandlingen. I det följande arbetet
analyserade vi både blod och vävnad från patienter med olika typer av gliom.
Utifrån metabolitmönster gick det att särskilja mellan olika diagnoser och
malignitetsgrad. Dessutom fann vi skillnader i metabolitmönster mellan patienter
som dog tidigt efter diagnos jämfört med de patienter som hade längre
överlevnad. Slutligen analyserade vi en undergrupp av gliom där patienterna
saknar ett enzym som leder till att de ej kan tillverka aminosyran arginin. Arginin
är en metabolit som krävas för att tumörcellen ska kunna växa. Vi jämförde
cellinjer som saknade enzymet, och därför inte kunde tillverka sitt eget aginin,
mot cellinjer med fungerande enzym. Vi fann signifikanta skillnader i
metabolitmönster mellan dessa två grupper. Därefter behandlade vi cellinjerna
med en behandlingsmetod som minskar mängden arginin utanför cellerna och på
så sätt ser till att cellerna inte får tillgång till och kan ta upp något arginin från
omgivningen. Således leder det till att cellerna som saknar enzymet inte får
tillgång till något arginin alls. Detta jämfördes mot celler som inte fått någon
behandling varpå vi kunde se metabola skillnader mellan behandlade och icke
behandlade celler.
Resultaten från samtliga arbeten visar att de metoder vi valt för analys av prover
är användbara för våra frågeställningar. Förhoppningsvis har vi genom detta
arbete kommit en liten bit närmare till att bättre förstå hjärntumörer vilket i
framtiden kan leda till att förbättrade behandlingsmetoder kan tas fram, att
prognostiska bedömningar kan göras baserat på individens egna biologiska
”fingeravtryck” samt att snabbare svar kan ges på om den behandling som valts
till en viss patient fungerar.
vi
ABBREVATIONS AND NOTATIONS
Abbreviations
1D
2D
BCAA
BCAT1
CNS
CV
CT
DA
DoE
ECM
EGF(R)
ELISA
EP
GC
GBM
HMCR
HIF
HSP90
IDH
IMAC
ITOT
iTRAQ
LC
LOH
MAPK
MGMT
mTOR
MRI
MS
MVA
m/z
OPLS
PCA
PIP2/3
PI3K
PLS
PTEN
RI
One Dimension
Two Dimension
Branched-Chain Amino Acid
Branched Chain Amino-Acid Transaminase 1
Central Nervous System
Cross Validation
Computer Tomography
Discriminant Analysis
Design of Experiments
Extracellular Matrix
Endothelial Growth Factor (Receptor)
Enzyme-Linked Immunosorbent Assay
Effect Projection
Gas Chromatography
Glioblastoma Multiforme
Hierarchical Multivariate Curve Resolution
Hypoxia-Inducible Factor
Heat Shock Protein 90
Isocitrate Dehydrogenase
Immobilized Metal Affinity Chromatography
Individual Treatment Over Time
Isobaric Tags For Relative And Absolute Quantitation
Liquid Chromatography
Loss of Heterozygosity
Mitogen-Activated Protein Kinase
Methylguanine-DNA Methyltransferas
Mammalian Target of Rapamycin
Magnetic Resonance Imaging
Mass Spectrometry
Multivariate Data Analysis
Mass to Charge Ratio
Orthogonal Projections to Latent Structures
Principal Component Analysis
Phosphatidylinositol Bisphosphate/Trisphosphate
Phosphatidylinositol-3 Kinase
Partial Least Squares
Phosphatase and Tensin Homolog
Retention Index
vii
ROI
RTK
SELDI
TCA
TERT
TOF
VEGF(R)
WHO
Region of Interest
Receptor Tyrosine Kinase
Surface-Enhanced Laser Desorption/Ionization
Tricarboxylic Acid
Telomerase Reverse Transcriptase
Time of Flight
Vascular Endothelial Growth Factor (Receptor)
World Health Organization
Notation
The following notation has been used throughout this thesis. Genes are denoted
by italic letters. Vectors are denoted by bold, lower case letters (e.g. t) and
matrices are denoted by bold capital letters (e.g. X). Vectors are assumed to be
column vectors unless indicated by transposition, (e.g. tT). A matrix inverse is
denoted as X-1 for a matrix X.
A
K
M
N
B
E
F
P
T
U
X
Y
c
p
t
u
w
Number of components in model
Number of columns in X
Number of columns in Y
Number of rows in X and Y
Matrix of regression coefficients for X, [K×M]
Residual matrix of predictor variables, [N×K]
Residual matrix of response variables, [N×M]
Matrix of loading vectors for X, [K×A]
Matrix of score vectors for X, [N×A]
Matrix of score vectors for Y, [N×A]
Matrix of predictor variables, [N×K]
Matrix of response variables, [N×M]
Weight vector for Y, [M×1]
Loading vector for X, [K×1]
Score vector for X, [N×1]
Score vector for Y, [N×1]
Weight or covariance loading vector for X, [K×1]
viii
BACKGROUND
Cancer is a widespread disease. Over the last two decades the incidence of tumors
in the Swedish population have increased by around two percent partly due to
higher exposure to risk factors, increased aging of the population, as well as better
methods for screening and diagnosis. According to the National Board of Health
and Welfare (Socialstyrelsen), one third of the Swedish population will be
diagnosed with cancer during their lifetime [1]. Therefore, understanding the
mechanism of disease is necessary in order to defeat cancer. Two types of brain
tumors, gliomas and meningiomas, have been studied and will be described in this
thesis. This section aims to provide a short background to why proteomic and
metabolomic changes and affected pathways are important to study in gliomas
and meningiomas.
Genetic alterations are probably the most studied molecular deviations in cancer.
Moreover, to support tumor growth and survival, the metabolism of tumor cells is
also different from normal cells. Metabolic changes can be linked to specific
genetic alterations in oncogenes and tumor suppressor genes, responsible for
regulation of cell cycle control, check points and cell signaling. The first discovered
alteration in metabolism in tumor cells was found by Otto Warburg. Warburg
discovered that cancerous cells, even in the presence of oxygen, tend to utilize
glycolysis instead of mitochondrial oxidative phosphorylation of glucose to supply
energy, thereby converting the majority of glucose to lactate [2]. This was coined
the Warburg effect. However, the Warburg effect is just one example of a
metabolic transformation in cancer cells. Tumor cells reprogram their metabolic
pathways to meet the needs during the process of tumor progression and show
increased metabolic self-sufficiency by taking up extracellular nutrients and
metabolizing them in pathways that supports growth and proliferation [3]. In the
year 2000, Hanahan and Weinberg published a review article stating that all
cancers share common hallmarks that leads the transformation of normal cells to
cancer cells [4]. In 2011 an updated version was published including altered
metabolism as a key hallmark [5]. The relatively new field of metabolomics is now
being used to analyze the cancer metabolome, with the aim to reveal further
details of tumor development that can be found connected to metabolic
perturbations.
The metabolome refers to the complete set of small-molecules found within a
biological sample or cell including both endogenous metabolites as well as
1
exogenous chemicals. Endogenous metabolites are naturally produced by the
organism and includes amino acids, organic acids, nucleic acids, fatty acids,
amines and sugars. Exogenous metabolites including food additives, drugs and
toxins are on the other hand not naturally produced by an organism. Metabolites
are involved in all aspects of cell growth, development and reproduction. They
also have stimulatory and inhibitory effects on enzymes and form the
intermediates and products of the metabolism. The metabolism defines all
chemical reactions involved in maintaining the living state of a cell. The cell
metabolism consists of an intricate network of metabolic pathways, the series of
steps a chemical can take to be transformed into another, that enable step-wise
synthesis and breakdown of molecules. These steps are catalyzed by enzymes
which also regulate pathways that respond to signals from other cells or to
changes in the environment of the cell. The field of metabolomics strives to
understand the relationship between metabolites present in a cell, tissue or organ
and the metabolic pathways affected by different biological mechanisms. This is
done by studying the differences between samples in relation to their metabolic
composition.
The field of proteomics strives to analyze the function and interactions of proteins
produced by the genes of a particular cell, tissue or organ. Proteins are the
molecules that put the cells genetic information into action. The human proteome,
the complete set of expressed proteins in a cell or an organism, at a specific time
and under specific conditions, consist of more than 200.000 protein species [6].
Proteins are long unbranched polymer chains that carry out most of the work in
the cells and are required for structure, function, and regulation of tissues and
organs in the body. Proteins are built from amino acids attached to each other
making up long polypeptide chains. There are 20 different types of naturally
occurring amino acids that have been found coded for to be translated into
proteins and each protein achieves a specific function according to its own
genetically specified sequence of amino acids and 3D shape. The protein sequence
of amino acids is specified from polynucleotides and proteins catalyzes many
chemical reactions including which new DNA molecules that are synthesized and
the genetic information in DNA contains specifications for thousands of proteins.
The total genetic information of a cell, the genome, is embodied in its complete
DNA sequence. The human genome is made up of 23 chromosome pairs and resent
research has estimated 190 000 human protein coding genes [7]. A gene is
encoded from segments of DNA which is transcribed and translated into single or
a set of protein variants and the cell adjust the rate of transcription and translation
of different genes independently according to temporal requirements. This
transfer of information from DNA to proteins is the central dogma of molecular
biology [8] and the metabolite expression at a given point in time is the result of
the proteins transcribed (figure 1).
2
Figure 1. Schematic presentation of the central dogma of molecular biology showing the hierarchical
levels from genotype to phenotype. Genomics, studies the function and structure of the genome.
Transcriptomics, studies the complete set of RNA transcripts that are produced by the genome.
Proteomics, large scale studies of protein structure and function including mass spectrometry based
screening of protein concentrations and modifications. Metabolomics, studies the molecular
mechanisms involving metabolites which are the end products of a cellular process and closest to the
phenotype.
Some common genomic alterations underlying
gliomagenesis
In normal cells, DNA density, mRNA synthesis and protein expression are highly
regulated processes that are kept under strict control. However, when cancer
arises, generally due to a series of genetic alterations, the control mechanisms of
the cells gets hijacked, resulting in uncontrolled proliferation and other functional
alterations constructing the hallmarks of cancer [5]. These occurrences, like
mutations, translocations, amplifications and deletions of genetic material, leads
to an accumulation of deviations in the regulatory pathways that affect various
processes and signaling pathways responsible for gliomagenesis [9]. Several of
these deviations affects the growth factor signaling pathways which in turn
regulate numerous cellular functions including differentiation, cell division, cell
survival and angiogenesis. Deregulation of these pathways by involvement of
oncogenes and/or tumor suppressor genes result in uncontrolled cell
proliferation and consequently cancer. Many fundamental oncogenic signaling
pathways congregate to adapt tumor cell metabolism in order to support growth
and survival.
3
For a normal cell to proliferate, activation of the mitogenic signaling pathways is
required. This is achieved through diffusible growth factor binding, cell-cell
adhesion and/or contact with extra cellular matrix (ECM) components. The
signals are then transduced intracellularly by transmembrane receptors that
usually activate the phosphatidylinositol-3 kinase/mitogen-activated protein
kinase (PI3K/MAPK) pathway. Tumor cells overcome this normal control of the
mitogenic pathway by acquiring genomic alterations that reduces the
dependence on exogenous growth stimulations that enables uncontrolled cell
division and survival through constitutive activation of these pathways. Gliomas
appear to do this mainly by constitutive activation of receptor tyrosine kinases
(RTKs). [10]
The epidermal growth factor receptor (EGFR) signaling pathway is one of the
most important pathways that regulate proliferation, differentiation, growth and
survival. EGFR gets activated through the binding of a ligand, for example
epidermal growth factor (EGF), to the extracellular domain which in turn recruits
PI3K to the cell membrane. PI3K then phosphorylates phosphatidylinositol
bisphosphate (PIP2) to phosphatidylinositol trisphosphate (PIP3) activating
downstream molecules like AKT and mammalian target rapamycin (mTOR)
resulting in cell proliferation and increased cell survival by blocking apoptosis
[11, 12].
Activated RTKs, by growth factor signaling, mutation or overexpression, activates
the MAPK pathway through the activation of RAS. When RAS gets activated it
activates the RAF kinase, which then regulates downstream signaling pathways
such as the MAPK pathway [13]. The MAPK pathways in turn regulate several
physiological responses, including apoptosis, cell proliferation, cell
differentiation, and tissue development. Hence, defects in the RAS/MAPK pathway
cause abnormal cell growth and proliferation, invasion and apoptosis.
Phosphatase and tensin homolog (PTEN), a tumor suppressor gene, inhibits
angiogenesis and is a negative regulator of the P13K/AKT pathway [14] thereby
inhibiting cell proliferation. The loss of PTEN leads to unregulated PI3K signaling,
AKT activation and up-regulation of mTOR signaling, which in turn increases
protein translation [15, 16]. Several growth factors including EGFR wield their
oncogenic effects through activation of the PI3K/AKT pathway [17] and
downstream of AKT is serine-threonine kinase, mTOR that controls cell growth by
regulating mRNA translation, metabolism, and autophagy [18]. mTOR signaling is
also involved in the hypoxic adaptation of tumors [19].
Another tumor suppressor gene, TP53, encodes a protein, TP53, involved in the
regulation of normal cell functions such as cell cycle control, DNA damage
response, cell death and differentiation. TP53 is also involved in the inhibition of
angiogenesis. If DNA damage occurs, TP53 plays a critical role for determination
between DNA repair or initiation of apoptosis and cell death. Hence, TP53 is a key
4
player in prevention of tumor development by halting the division of cells that
have acquired mutations or DNA damage. In the absence of TP53 the cell cycle
proceeds which lead to uncontrolled proliferation and tumor formation. [20]
Isocitrate dehydrogenase (IDH) mutations are common in low grade gliomas. The
IDH genes encode the enzyme IDH, which catalyzes the oxidative decarboxylation
of isocitrate into α-ketoglutarate, resulting in the production of NADPH in the
citric acid cycle. Both isozymes IDH1 and 2 are involved in several metabolic
processes including signal transduction, lipid synthesis, oxidative stress and
oxidative respiration [21]. It has been shown that forced expression of mutant
IDH1 in cultured cells reduces the formation of α-ketoglutarate and increases the
levels of hypoxia-inducible factor subunit (HIF)-1α which is a transcription factor
that enables tumor growth [22].
Tumor growth is dependent on angiogenesis. New vessels supplying the tumor
with oxygen and nutrients is essential for the tumors to grow in size and allows
cancer cells to invade adjacent tissue [5]. Hypoxia, often found in solid tumors, is
the major driving force behind vascular endothelial growth factor (VEGF)
activation. Fast growing tumors creates intracellular hypoxia and the
transcription factor, HIF, responds to the change in oxygen concentration [23, 24].
During normal conditions HIF is hydroxlated and acetylated and thus subjected to
von Hippel-Lindau (vHL)-mediated ubiquitin degradation. When the cells are in a
hypoxic state, HIF get accumulated and transported to the nucleus where it
induces expression of several target gene products and growth factors such as
VEGF [25]. This in turn results in endothelial cell proliferation, increased vascular
permeability, and cell migration. VEGF transcription can also be upregulated by
overexpression of oncogenes or the absence of tumor suppressor genes like TP53
or vHL.
Common traits of tumor metabolism
For energy supply, tumor cells rely mostly on glycolysis, the conversion of glucose
into pyruvate and further into lactate, even in the presence of adequate oxygen [2]
and tumor cells radically increases the rate of glycolysis to meet their high-energy
demand. The increased rate of glycolysis is achieved by increased glucose uptake
and expression of glucose transporters, GLUTs, at the cell membrane [26]. Glucose
together with glutamine [27] are the two major nutrients consumed by tumors
and they provide precursors for nucleic acid, proteins and lipids needed for
cellular growth. In glioblastoma (GBM) most of the acetyl-CoA originates from
glucose and essentially all of the oxaloacetate come from glutamine [28]. As
mentioned before, mutations in tumor suppressors and oncogenes have an effect
on the metabolic expression. For example, overexpression of Ras or Myc is enough
to drive glucose uptake in fibroblasts [29] and inhibition of AKT signaling in a GBM
cell line with genomic c-MYC amplification lead to decreased glycolysis [30]. It has
5
also been shown that over activation of the PI3K/AKT system facilitates an
increase in glucose and flux through the plasma membrane that may be
attributable in part in the activation of HIF1-α [31]. Also, stabilization of HIF1-α
increases the expression of glycolytic enzymes, glucose transporters and
inhibitory kinases for the pyruvate dehydrogenase complex, all of which aid to
increase lactate from glucose [32, 33]. Inhibiting β-oxidation is another effect of
PI3K activation, which could be one contributing factor to the glucose addiction of
tumor cells [34]. In contrast, TP53 reduces the intracellular glucose levels by
inhibiting the expression of glucose transporters and TP53 directly represses the
transcriptional activity of the GLUT1 and GLUT4 gene promoters [35].
As mentioned earlier, glutamine also serves as metabolic fuel for the tumor cells
[27] that after metabolizing to α-ketoglutarate can provide energy through
substrate level phosphorylation within the tricarboxylic acid (TCA) cycle [36].
Under hypoxic conditions, glutamine may be metabolized to citrate and fatty acids
through an IDH–dependent pathway [37, 38]. Glioma cells require glutamine
uptake for growth [39] since gliomas cannot synthesize their own glutamine [40,
41].
The levels of glutamate, a primary excitatory neurotransmitter, present in the
brain is synthesized de novo by astrocytes [42] and the blood brain barrier has
very low permeability to glutamate. Studies have shown that the extracellular
levels of glutamate are increased both in and around experimental glioma
implants in vivo and that glioma cells actively releases glutamate in vitro [43, 44].
Furthermore, it has been shown that glutamate secreting gliomas have a growth
advantage as compared to other tumors [45]. Gliomas can release glutamate in
sufficient quantities to cause neuronal cell death [46, 47]. By releasing glutamate
from glioma cells, gliomas have developed an ability to free space for their
expansion by destroying the surrounding cells [48]. Radiotherapy can also
increase the glutamate levels. Glutamate levels in the microenvironment will be
elevated due to oxidative tissue damage and necrosis caused by radiotherapy [45]
[49].
The metabolism of the branched chain amino acids (BCAA), leucine, isoleucine and
valine, plays an important role for the synthesis of nonessential amino acids [50].
The BCAA are initially catabolized by the transfer of a α-amino group to αketoglutarate through branched chain amino-acid transaminase 1 (BCAT1) or 2
(BCAT2) isoenzymes which in turn yields glutamate and the respective branchedchain α-ketoacids [51]. The main source of nitrogen to the glutamate synthesis
comes from the BCAAs [52] before they are further catabolized to acetyl-Co and
succinyl-CoA which are oxidized in the TCA cycle. The importance of BCAAs for
glioma growth have been presented in a publication by Tönjes et al, which showed
that, for wild type IDH glioma to sustain an aggressive growth phenotype
catabolism of BCAAs was required. In addition they also showed that branched
6
chain amino-acid transaminase 1 (BCAT1) was overexpressed in wild type IDH
but not expressed in the majority of gliomas with IDH mutations [53].
It has been demonstrated that intensified lipogenesis is also a feature of cancer
[54] and that tumor tissue lipid levels are higher compared to normal tissue in
malignant glioma [55]. This increased lipogenesis has been shown to directly
correlate with enhanced glucose and glutamine metabolism [38, 37, 56].
Enhanced lipid synthesis and uptake are partly regulated by the EGFR/PI3K/Akt
pathway through upregulation of sterol regulatory element-binding protein 1
(SREBP-1) activity [57, 58].
Brain tumors
Brain tumors are tumors confined within the skull cavity. They usually emerge
from the supportive cell population in brain tissue, neuroepithelial cells, the
meninges surrounding the brain or they can be secondary to tumors outside the
skull cavity. Brain tumors can be benign or malignant and arise in different parts
of the brain and the CNS. The two most common primary brain tumors are gliomas
that makes up around 50 percent of all brain tumors and meningiomas that makes
up around 20 percent. The cause of brain tumors is still unknown, nevertheless,
some factors like exposure to ionizing radiation or inherited conditions like LiFraumeni syndrome, neurofibromatosis and the Von Hippel-Lindau syndrome
may increase the risk of developing a brain tumor [59]. The symptoms vary
depending on the tumor size and location within the skull cavity and eventual
secondary brain edema. The most common symptoms are seizures, headache,
changes in personality and focal neurological deficits [60]. In case of suspected
brain tumor following neuroimaging, diagnosis is obtained after examination of
histopathological specimens by means of surgical resection or tumor biopsy.
Gliomas
Malignant gliomas constitute 2 % of all adult tumors but their aggressive nature
makes them the fourth largest cause of cancer death. Gliomas is a broad category
of brain and spinal cord tumors that arise from the supportive cells of the brain
called neuroepithelial or glial cells. The glia cells are divided into sub-groups were
the main types are astrocytes oligodendrocytes and ependymal cells. Glial cells, in
comparison to neurons, undergoes cell division and multiply to a large extent.
Therefore, if control of the regeneration process is lost a glial tumor is formed.
Gliomas are classified histologically and immunohistochemically based on their
cellular origin and the degree of malignancy which is graded from I-IV according
to the World Health Organization (WHO) grading system [61].
7
Astrocytoma
Astrocytes are the most abundant glial cell and astrocytomas are the most
common glioma tumor. Low grade astrocytomas, astrocytoma grade II, are
relatively uncommon. They are usually restricted and slow growing while high
grade astrocytomas, anaplastic astrocytoma or astrocytoma III, grows rapidly and
more invasively. Median survival time for low grade astrocytomas is 5-6 years,
while for high grade astrocytomas only 1.5 years [62].
Glioblastoma
Grade IV astrocytoma, also called glioblastoma (GBM), constitutes about 50 % of
all gliomas and is the most common intracranial malignancy [10]. The median
survival time for a patient with GBM is only 15 months after a multimodal
treatment approach including surgery, postoperative radiotherapy and
chemotherapy [63]. The poor prognosis of GBM is much due to characteristic
biological hallmarks of the GBM cells including invasive growth, rapid
proliferation, necrosis and intensive angiogenesis. The diffusely infiltrative
growth of GBM with invasion into the surrounding brain parenchyma [10] makes
radical surgical resection impossible in most cases.
The majority of GBMs are primary, meaning that they develop rapidly de novo
without indication of a less malignant precursor lesion. Secondary GBMs develops
gradually from low-grade or anaplastic astrocytomas. Primary and secondary
GBMs affect patients of different age. Primary GBMs mainly affects elderly patients
while secondary GBMs are more common in younger patients [64]. There are also
differences in the genetic pathway leading to tumorgenesis. Primary GBMs are
genetically characterized by loss of heterozygosity 10q, EGFR amplification, p16
deletion and PTEN mutations while in secondary GBMs TP53 mutations are more
common [65]. A more detailed description on the different genetic alterations can
be found in the following chapter below regarding glioma biomarkers.
Oligodendroglioma
Oligodendrogliomas are composed of cells that morphologically resembles
oligodendroglia. Oligodendrogliomas of WHO grade II are slow growing and welldifferentiated tumors that show tendencies to diffusely infiltrate the surrounding
brain. Anaplastic oligodendrogliomas WHO grade III show more malignant
histological features [61]. Survival time for the low grade oligodendrogliomas is
11.6-16.7 years [62, 66] while for anaplastic oligodendrogliomas the survival time
is 3.5-7.3 years [62, 67].
Meningioma
Meningioma is most often a benign, WHO grade I [68], brain tumor that arises
from arachnoid cap cells adjacent to the dura mater. They mostly appear within
the intracranial, orbital or spinal cavities [69]. Meningiomas can afflict patients of
all ages but occurs most often between the ages of 40-70 years and are more
8
common among women. Meningiomas can be divided in different histological
subtypes, usually fibrous, meningothelial or mixed [69]. Surgical resection is often
curative, however, some tumors cannot be entirely resected due to their
anatomical location and postoperative remaining tumor cells may develop an
aggressive and bone invasive growth behavior [70]. It is currently not possible to
foresee invasive tumor development which makes scheduling of treatment plans
for individual patients challenging.
Standard treatment for glioma and meningioma
The standard treatment plan for patients with low-grade astrocytomas and
oligodendrogliomas are very similar. If the patient is considered elderly, having a
slow growing tumor and do not experience any symptoms, the first treatment step
is to keep the tumor under observation. The tumor is regularly monitored with
MRI scans and if the tumor starts to grow, surgical resection is the first line of
treatment followed by radiotherapy if necessary. For grade III astrocytomas,
oligodendrogliomas and GBMs, the standard care is surgical resection of as much
of the tumor as possible followed by radiotherapy and sometimes chemotherapy.
Patients with anaplastic astrocytomas are usually offered postoperative
radiotherapy to 60 Gy in 2 Gy fractions. Chemotherapy using the alkylating agent
temozolomide is the most common choice for first line treatment at relapse. For
anaplastic oligodendroglioma surgery is followed by radiotherapy and adjuvant
chemotherapy using the alkylating agents lomustine and procarbazine in
combination with the antimitotic agent vincristine (PCV) for 6 cycles [71]. For
GBM, surgery is followed by postoperative radiochemotherapy, normally 60 Gy in
30 fractions over a six week period together with concomitant temozolomide 75
mg/m2. The radiochemotherapy is followed by 6 adjuvant chemotherapy courses
using 150-200 mg/m2/day during five days every 4 weeks [63].
For meningiomas, complete surgical resection of accessible tumor is the mainstay
of treatment [72]. For patients with critically located meningiomas where radical
surgery is not possible, radiotherapy is the treatment of choice [73, 74]. It has
been shown that the recurrence frequency in patients with benign meningiomas
decreases with adjuvant radiotherapy after subtotal resection [75, 76].
Nevertheless, due to the side effects it might be preferable in some patients to
delay the use of radiotherapy until signs of progressive disease [77].
Glioma and meningioma biomarkers
A biomarker is defined as “a biological molecule found in blood, other body fluids,
or tissues that is a sign of a normal or abnormal process, or of a condition or
disease” [78]. Much research today is focused on finding biomarkers for a variety
of diseases. A biomarker may be of importance when one wants to know if a
patient has a disease or not, what type of disease the patient has or how well a
9
specific treatment is working. Accordingly, finding biomarkers for gliomas is of
great interest. Histopathology is currently the golden standard for the grading and
classification of gliomas, however, histological classification of gliomas is not a
straightforward task, hence substantial variability occur between
histopathologists. Consequently, a beneficial diagnostic glioma biomarker should
be able to classify the tumors according to clinical outcome and thereby provide
sub-classifications within the specific histological subtype. The current most
useful prognostic factors for glioma patients are age, tumor type and grade, tumor
resectability and Karnofsky performance status [79, 80]. Therefore, an ideal
prognostic biomarker should be able to predict the overall survival beyond
already established parameters. GBMs are rapidly progressive making routine
evaluation of the current treatment response using MRI insufficient; as treatment
effects are not detectable by MRI until months after treatment initiation when it
is often too late to change strategy. A predictive biomarker reflecting therapeutic
effect of a given therapy should offer information on the response at an early stage
making it possible for the clinician to optimize the treatment for the individual
patient. Biomarkers could further help gain better understanding of the molecular
mechanisms involved in glioma formation and treatment resistance, which is
needed for identifying new treatment targets in order to overcome the poor
prognosis and strive for a better outcome for glioma patients.
Most research regarding glioma biomarkers, has so far been delimited to genetic
alterations. Currently, there is a number of genetic signatures that provide
additional information regarding the tumor grade and, to some extent, prediction
of treatment response and survival. The following genetic alterations are the most
common established markers for gliomas, some that are in clinical use today. A
summary of the alterations can be found in table 1.
Table 1. Biomarker candidate status for some common alterations in brain tumors.
Diagnosis Prognosis Prediction
1p/19q
√
√
√
EGFR
√
√
√
IDH
√
√
MGMT
√
√
MAPK
√
√
PTEN
√
√
PI3K/AKT
√
TP53
√
TERT
√
VEGF
√
NF2
√
Diagnosis, Prognosis and Prediction indicate if the genetic alteration provides information about
diagnosis, prognosis and/or prediction of treatment response. As shown, some genetic alterations are
biomarker candidates for all three categories while other genetic alterations are candidates for only
one of the categories.
10
The most common genetic alteration in oligodendrogliomas it the loss of
heterozygosity (LOH) of chromosome 1p and 19q [81] and is hence seen as a
diagnostic marker for oligodendrogliomas. Besides that 1p/19q deletion is a
positive predictive marker for procarbazine, lomustine and vincristine (PCV) [82]
or temozolomide chemotheraphy [83, 84]. It is not yet established which genes
are located at the sites, or how their loss are involved in the advancement of
oligodendroglioma growth, or by which mechanism they contribute to the more
promising therapeutic response [85]. Unfortunately, LOH of 1p/19q is uncommon
in primary GBMs.
Several growth factors and their receptors are upregulated in gliomas. EGFR is
mutated and amplified in approximately 50% of all primary GBMs [86] with
EGFRvIII as the most common aberration [87]. EGFRR promotes cell proliferation
and angiogenesis and has been shown to contribute to cell invasion as GBMs
harboring active EGRFVIII receptors display a more invasive phenotype than wild
type EGFR [15]. Identification of EGFR amplification is highly indicative of high
grade malignancy and can hence provide prognostic information. The detection of
EGFR amplification in low-grade gliomas strongly suggests that the tumors are
more malignant than indicated by the histopathological classification [88-90].
In the majority of low-grade gliomas, mutations of the IDH1/2 genes have been
identified. If IDH1 or IDH2 mutations are present in high grade gliomas they are
often secondary gliomas that have progressed from lower grade lesions [91, 9].
IDH1/2 mutations are associated with improvement in overall survival and glioma
patients harboring this mutation appear to have a prognostic advantage. In a
study by Yan et al, the overall survival of patients with IDH1/2 mutation was 31
months compared to 15 months in patients without IDH1/2 mutation [91]. Today,
these mutations are used as diagnostic markers for grade II and III gliomas and
secondary GBMs [21, 92].
O-6-Methylguanine-DNA-Methyltransferase (MGMT) encodes a repair protein
that removes alkylation at the O6 position of guanine which is a common site
altered by alkylating agents including temozolomide. Approximately 40% of
glioblastomas have the MGMT gene silenced by promotor hypermethylation and
thereby reduced MGMT expression levels. A study by Hegi et al. shows that
patients whose tumors had methylated MGMT promotor and where treated with
radiotherapy and temozolomide survived significantly longer than patients
whose tumors lacked MGMT promotor methylation. The study also showed that
MGMT promotor methylation did not significantly influence survival in patients
treated with radiotherapy alone. This indicates that MGMT hypermethylation is a
positive predictive marker for temozolomide [93, 94].
High grade astrocytomas display a defective RAS/MAPK pathway and
overexpression of RAS has been detected in several cases of GBMs [95-97]. It has
been shown that tumors with high levels of phosphorylated MAPK are associated
11
with more proliferative tumors and shorter survival time [98] The levels of MAPK
proteins appear to be both predictive and prognostic since phosphorylated MAPK
both impact overall survival and is associated with increased radiation resistance
[99].
LOH on chromosome 10q resulting in loss of PTEN has been associated with
reduced survival of GBM patients [100]. A frequent genetic alteration in GBMs is
the LOH of chromosome 10q [64] and PTEN which is a tumor suppressor gene
situated on chromosome 10q is deleted due to LOH in 50-70% of primary GBMs
[101]. PTEN is a protein/lipid phosphatase which main substrate is PIP3, the
product of PI3K. Since PI3K signaling is an vital mediator of cell growth and
proliferation its activation is associated with poor prognosis [15]. It is also
possible that PTEN could be a predictive marker of treatment response since upregulation of the PI3K/mTOR pathway predicts decreased survival, the use of
PI3K, AKT or mTOR inhibitors in treatment of gliomas could be a therapeutic
target.
A hallmark of low-grade gliomas is the inactivating mutation of the tumor
suppressor gene TP53 that encodes for the TP53 protein [87]. Since the TP53
mutation is overrepresented in low-grade gliomas [102], it provides a marker for
determination of tumor grade but also to identify secondary GBMs.
Telomerase reverse transcriptase (TERT) is found to be associated with highgrade disease [103]. TERT is undetectable or present in very low levels in normal
cells but frequently up-regulated in several tumor types [104, 105]. Mutations in
the TERT promotor results in lengthen telomeres through enhanced telomerase
activity and has been seen in both high-grade astrocytomas and low grade
oligodendrogliomas suggesting that telomere maintenance may be necessary for
the formation of brain cancers [106, 107].
GBM is one of the most vascularized tumors. GBM express a high level of VEGF and
VEFG signaling is a strong stimulator of angiogenesis [108, 109] . In gliomas,
expression levels of VEGF correlate with malignant progression and high vascular
density in solid tumors [110].
In meningiomas, mutations of the neurofibromatosis type 2 gene (NF2), located at
chromosome 22q12 are found in about 60% of sporadic meningiomas and are
suggested to be an early event in tumorgenesis [111]. The frequency of NF2 gene
mutations varies in different histological variants, up to 80% of patients with
fibroblastic meningiomas harbor NF2 mutations while the mutation can be
detected in only about 25% of meningothelial meningiomas [112].
The use of molecular biomarkers was recently demonstrated by Eckel-Passow et
al who showed that it was possible to group gliomas based on the use of TERT
promotor mutations, IDH1/2 mutations and 1p/19q codeletion [113]. They
12
observed that almost all stages of gliomas were matched into one of the five
subgroups defied by these markers: triple positive (mutations in TERT and IDH,
1p/19 codeletion), mutations in both TERT and IDH, mutations in IDH only,
mutations in TERT only and triple-negative. They found that specific histological
types were more frequent within certain groups. For example almost all triplepositive gliomas where oligodendrogliomas or mixed oligoastrocytomas. Tumors
with TERT mutations only and triple-negative tumors carried the highest risk for
patient mortality compared to patients with IDH mutations in any combination
with TERT or 1p/19q codeletion which supports that IDH1/2 mutations are
associated with improvement in overall survival.
So far, not many metabolites have been identified as glioma biomarkers.
However, glucose, as mentioned before is an interesting metabolite when it
comes to tumors. The brain uses glucose as the main fuel but glucose is also the
fuel for tumor cell glycolysis, which is the pathway that drives the growth of
most brain cancers. Elevated glucose is associated with poor prognosis [114,
115] and a study by Seyfried et al revealed that higher blood glucose levels were
directly related to faster tumor growth [115], stating that brain tumor growth
will be difficult to manage as long as circulating glucose levels remain elevated. A
few so called oncometabolites have also been identified. Oncometabolites can be
defined as metabolites that can promote tumorigenesis by altering the
epigenome [116]. The three most studied oncometabolites are fumarate,
succinate and 2-hydroxyglutarate. Loss-of-function mutations in genes encoding
tricarboxylic acid (TCA) cycle enzymes fumarate hydratase (FH) and succinate
dehydrogenase (SDH) lead to accumulation of fumarate and succinate,
respectively, whereas gain-of-function mutations in IDH1/2 cause increased
levels of 2-hydroxyglutarate [117, 118]. Since the IDH genes are associated with
gliomas, 2-hydroxyglutarate is a highly interesting metabolite in glioma
metabolism and it has been show that 2-hydroxyglutarate can be used as a
metabolic marker for detection of IDH mutations [119, 120].
Biological sample matrices
The papers in this thesis are based on biological samples from different sources.
In paper I and IV tumor tissue from patients are analyzed, in paper II tumor tissue
and serum from an animal glioma model is analyzed, in paper III and IV serum
from patients is analyzed and in paper VI glioma cell lines are analyzed. It is
important to choose the samples matrix based on the research question and
hypothesis defined. In our studies we have chosen sample matrices mainly based
on what was ethical to collect, what was possible to collect and what has been
collected.
13
Tumor tissue
Direct analysis of tumor tissue is a good approach for glioma-specific biomarker
discovery and much research have been carried out on tumor tissue and
biomarkers thereof [121-123]. Tumor tissue analysis can be useful when
searching for classification and progression markers and also when investigating
the molecular mechanisms involved in tumorgenesis. Tumor tissue samples can
be either resected tumor or a biopsy thereof. It is an invasive sampling method
and when analyzing human tissue it is obviously not possible to get normal brain
tissue as corresponding control samples. When working in animal models normal
tissue is feasible to use as control. Also, human samples from the extra cellular
compartment collected by the use of microdialysis may provide information from
both tumor and normal brain and thus allows for the use of patients as their own
control.
Serum/plasma
Human blood in terms of serum or plasma is a great source for discovery of
disease markers. The markers are easily accessible and may represent products
secreted from the involved tissues. Serum and plasma have been shown to be
valuable material for human glioma biomarker research [124-126]. Both plasma
and serum are derived from whole blood that has been submitted to different
biochemical processes after collection. Serum is obtained from blood that has
coagulated and to obtain plasma an anticoagulant is added before removal of
blood cells. Consequently, serum includes the proteins not used in blood clotting,
antibodies, antigens, hormones, electrolytes and any exogenous substances while
plasma contains dissolved proteins, clotting factors, hormones, electrolytes and
carbon dioxide. The accessibility and less invasive procedure of serum/plasma
sampling makes it possible to use in longitudinal studies, collecting several
samples over longer periods of time, and to get control samples from healthy
patients. One drawback by studying serum/plasma from glioma patients is the
blood-brain-barrier that may more or less influence the molecular exchange
between the extracellular tumor compartment and the blood [127].
Extracellular compartment
The extracellular compartment is a pool of molecules that is secreted by the cells
and providing structural and biochemical support to the surrounding cells and is
required for tissue homeostasis, morphogenesis and differentiation. The
extracellular compartment is tissue specific and heterogeneous, each tissue has
an extracellular compartment with a unique composition and topology that is
generated during tissue development and consists mainly of water, proteins and
polysaccharides. An advantage with studies where the extracellular compartment
is collected is that the patients can act as their own control and that several
samples can be collected during a period of time, enabling short-term longitudinal
studies. Sampling of the extracellular compartment is an invasive procedure.
However, the patients are fully mobile during the collection time.
14
Cell lines
Cell lines are frequently used in biomedical experiments and individual cancer cell
lines provide a snapshot of the tumor at the time the biopsy was taken. Analyzing
explant cultures from surgically removed GBM tumors was first established in
1933 to study GBM growth patterns [128]. Many years later, a set of glioma cell
lines were established resulting in a standard and cohesive method for studying
gliomas in vitro [129]. Several cell lines, derived from malignant gliomas, are now
commercially available and have become essential for research into the biological
mechanisms of GBMs [129] [130]. Using cell lines come with many advantages,
they are self-replicating yielding a large amount of biological material, they are
relatively homogenous, and they can be genetically manipulated for mechanistic
studies. Conversely, cell lines require their own specific growth factor cocktails
and their growth must be closely monitored to prevent overgrowth. They are also
likely to show both genotypic and phenotypic drifts during culturing and it is
possible for subpopulations to arise and cause phenotypic changes over time by
the selection of more rapidly growing clones within the population [131].
Animal models
Much of the research in human cancer and many drugs and treatments for human
diseases are developed in part with the guidance of animal models. This since
model organisms allow understanding of disease processes without the risk of
harming any humans. Utilization of animal models has the advantages of
providing highly controlled experimental conditions, homogeneity and
reproducibility. A good GBM animal model should also contain the key
histopathological and genetic features of the aggressive growth of GBM. There are
several rodent models of GBM tumors available. For example, a concurrent
retroviral expression of active Ras and Akt in a somatic gene-transfer model gives
rise to the formation of high-grade gliomas that are morphologically similar to
human GBM [132]. To obtain highly infiltrative GBM tumors an EGFR transgenic
mouse model can be used where the LOH of p16INK4α, p19ARF and PTEN work
together with the amplification of EGFR [133]. It is also possible to obtain highgrade gliomas which are histologically similar to human GBMs by the
simultaneous deletion of Tp53 and PTEN in the mouse central nervous system
[134]. However, these models have limitations since the tumor cells are not of
human origin. Xenograft models are models where human tumor cells are
transplanted, either subcutaneously or into the organ type in which the tumor
originated, into immunocompromised animals. Xenograft models of malignant
astrocytoma have been used to evaluate the function of several signaling
molecules or matrix proteins in glioma growth and invasion [135, 136]
15
AIMS OF THIS WORK
The overall aims of this work were to investigate if the proteome and metabolome
may harbor diagnostic and prognostic information as well as possible biomarkers
for treatment effects in meningioma and glioma. Our underlying hypothesis is that
molecular patterns of detected proteins or metabolites can contribute with more
predictive, specific and informative so called “latent biomarkers”.
To achieve this, the specific aims have been to:
I.
Evaluate the difference in protein profiles between invasive and noninvasive meningiomas.
II.
Investigate changes in protein profile in rat gliomas treated with an
anti-angiogenetic tyrosine kinase inhibitor.
III.
Evaluate radiation-induced changes in serum metabolome in glioma
patients.
IV.
Investigate the metabolome in tissue and serum from glioma patients
and relate it to diagnostic and prognostic information.
V.
Investigate the metabolic profiles of ASS1 positive and ASS1 negative
glioma cell lines and verify the treatment effects of arginine
deprivation therapy in ASS1 negative cell lines.
16
METHODS
Considering the high complexity of the proteome and metabolome there is today
no analytical method sufficient enough to cover all proteins or metabolites
present in a biological sample alone. Therefore, complementary analytical
methods have to be used to reach the best possible coverage. In this thesis,
different mass-spectrometry based methods have been used in combination with
bioinformatics approaches based on chemometrics to strive towards the aims of
the thesis. The methods are described and discussed below.
Generation of proteomics data
There are different approaches to proteomic studies. It is possible to use targeted
methods to analyze one or a few proteins in a sample to investigate or verify a
specific hypothesis or one could apply global screening approaches to analyze a
multitude of proteins in a sample in an unbiased manner. In our studies we have
analyzed the samples with two different global screening methods, Surface
Enhanced Laser Desorption Ionization Time of Flight Mass Spectrometry (SELDITOF-MS) and isobaric Tags for Relative and Absolute Quantitation (iTRAQ)
labeling in combination with high-performance liquid chromatography (HPLC).
Additionally we also used a targeted enzyme-linked immunosorbent assay
(ELISA) based method to analyze one specific protein (heat shock protein 90
(HSP90)) fort result verification on paper II.
SELDI-TOF-MS
To be able to globally analyze protein pattern differences in a high-throughput
fashion we utilized SELDI-TOF-MS as the initial screening method for analyzing
proteins in the papers regarding proteomics (papers I and II).
SELDI-TOF-MS is a chip based technology that combines chromatography and
mass spectrometry to achieve a protein expression profile. It allows for analysis
of a variety of complex biological materials where only a few µl of sample is
needed. Thanks to the high throughput of this technique the samples can be
analyzed using chips with different specific chromatographic surfaces. These can
be either chemically treated (anionic, cationic, hydrophobic, hydrophilic etc.),
and retain whole classes of proteins, or biochemically treated (antibody,
receptor etc.) to bind a specific target protein, thereby enabling the capture and
quantification of as much of the proteome as possible [137]. Before analyzing the
chip with the mass spectrometry reader, an energy absorbing molecule (EAM)
solution is added to the surface, which enhances the laser energy transfer and
analyte ionization. The reader is a laser desorption ionization instrument that
uses a pulsed UV nitrogen laser source. When the surface is hit by the laser the
proteins turn into gaseous ions. The ions then enter the TOF-MS region at the
17
end of which there is a detector that measures the mass to charge ratio (m/z) of
the retained proteins as well as their relative quantity [138, 139].
After the introduction of SELDI in 1993 by Hutchens and Yip [140, 137], as a
technique to analyze biological samples with high throughput using a simple
sample extraction method, SELDI became “the new black” in proteomics research.
Many different SELDI applications have been published since then [141-143].
However, a major disadvantage of SELDI-TOF-MS is the lack of direct
identification of the peaks discovered. The output, i.e. the mass to charge ratio and
the relative quantity of the retained proteins, from the SELDI technique is not
enough to be able to identify the individual proteins or modifications thereof,
which obstructs the biological interpretations of the results. Multiple
identification strategies have been adopted, including enrichment and
purification by chromatography and gel electrophoresis, followed by mass
spectrometric sequencing [144-146]. Although several biomarker candidates
have successfully been identified by this approach [147-149], the process is laborintensive, requires a rather large amount of sample and a relatively high
abundance of the protein of interest. Taken together, the issues surrounding the
actual identification of putative biomarkers are detracting the simple and highthroughput advantage with the SELDI-TOF-MS technique. This is probably why
the SELDI-TOF-MS technique is not used as extensively today as it was in the years
following its introduction.
iTRAQ labeling in combination with HPLC-MS/MS
In paper II, after verifying that changes in the proteome occurred as a result of
vandetanib treatment by SELDI-TOF_MS we made an in attempt to identify
proteins altered in relation the treatment. For that we analyzed the same samples
using another proteomic technique; isobaric Tags for Relative and Absolute
Quantitation (iTRAQ) labeling together with HPLC-MS/MS.
iTRAQ is an isobaric labeling method that uses stable isotope labeled molecules
that can be covalently bonded to the side chain amines and N-terminus of peptides
from protein digestions with tags of varying mass in order to determine the
amount of proteins from different sources in a single experiment [150]. The risk
for overlapping peaks in the MS scan is reduced by the isobaric mass of the
reagents which enables the peptides to appear as single peaks. When the labeled
peptides are analyzed by MS/MS, the mass balancing moiety is released as a
neutral fragment, freeing the isotope-encoded reporter ions which then provides
the relative quantitative information for the proteins. There are different iTRAQ
reagents available, 4-plex and 8-plex [151], which makes it possible to analyze a
set of four to eight samples in one MS run. The reporter ions from the attached
labels are used to relatively quantify the peptides and the proteins from which
they originate. A database search is applied to identify the labeled peptides, and
subsequently the corresponding proteins.
18
For comparative studies of small sample sets iTRAQ labeling has many advantages
since it is possible acquire both relative and absolute quantitation using internal
peptide standards [150]. Since all samples are combined in one run, the
instrument time for analyses can be reduced, and variations between different LCMS runs do not affect the results to any extent. A drawback when it comes to the
iTRAQ technology is that due to the enzymatic digestion of proteins prior to
labelling, the complexity of the sample increases. This results in the need for a
powerful, multi-dimensional fractionation method of peptides prior to the MS
identification.
ELISA
With iTRAQ labeling we could detect a specific protein, HSP90, which we wanted
to further investigate and validate. An enzyme-linked immunosorbent assay
(ELISA) technique was used to validate the concentrations of HSP90. ELISA is a
biochemical assay which is used to measure the concentration of an analyte in
solution. The ELISA technique uses antibodies and an enzyme-mediated color
change to detect the presence of either antigens (proteins, peptides, hormones,
etc.) or antibodies in a sample. There are different types of ELISA. In this work the
quantitative sandwich enzyme immunoassay technique was used to quantify a
specific protein, HSP90. In brief, the technique works as follows. Standards and
samples are pipetted into the wells of a microplate pre-coated with a monoclonal
antibody specific for HSP90. Any HSP90 present in the sample is bound by the
immobilized antibody. An enzyme-linked monoclonal antibody specific for HSP90
is added before unbound reagent is washed away. Finally, a substrate solution is
added causing color to develop in proportion to the amount of HSP90 bound. The
color development is stopped and the intensity of the color is measured in a UVvisible absorbance microplate reader.
Generation of metabolomics data
As for proteomics, there are different approaches in the design of metabolomic
studies and several different analytical techniques to choose from. Throughout
this work, a global screening approach using gas-chromatography time-of-flight
mass-spectrometry (GC-TOF-MS) has been the preferred choice. A targeted
approach for amino acids was utilized as a complement in paper V.
To be able to analyze the metabolites in a sample, the metabolites need to be
extracted from the sample. Depending on the extraction method different classes
of metabolites can be dissolved. The preferred extraction method may also differ
depending on the sample matrix. Serum and cell line samples have been extracted
using a polar extraction solution mix consisting of methanol (90%) and H2O
(10%). For the analysis of the more fatty tumor tissue, a less polar extraction
mixture consisting of chloroform, methanol and H2O (1:3:1) was selected.
19
GC-TOF-MS
GC-TOF-MS is a hyphenated technique that combines the separating power of gas
chromatography with the detection power of mass spectrometry. The gaschromatograph requires that the samples are derivatized prior to analysis to
make the polar compounds more volatile. The derivatization is performed in two
steps. Firstly, to stabilize the carbonyl moieties in the compounds, Omethylhydroxylamine
hydrochloride
is
used,
then
N-Methyl-Ntrimethylsilyltrifluoroacetamide
(MSTFA)
is
added
with
1%
trimethylchlorosilane (TMCS) as catalyst to convert the remaining functional
groups to trimethylsilyl (TMS)-derivatives.
After derivatization, the sample is injected into the gas-chromatograph (GC) were
the sample molecules are vaporized. GC uses a carrier gas (mobile phase) to
transport the sample components through the column containing the stationary
phase. The molecules interact with the stationary phase during their journey and
when they enter the detector the quantity of each molecular species is detected in
relative terms. Depending on the properties of the molecule, the degree of
interaction with the stationary phase differs, and the time it takes for the molecule
to travel through the column (retention time) will differ [152]. The difference in
retention time between different molecular species makes it possible to separate
the molecules in a sample. The mass spectrometer separates the ions in gas phase
according to their m/z [153]. There are different ionization methods. In our
studies electron impact (EI) has been used. In EI the analyte molecules are directly
ionized through a bombarding electron stream resulting in the removal of an
electron to form a radical cation species. This results in a mass spectrum which
together with information of the retention time (transformed to retention index)
can be used in data base searches to be able to identify the molecules [154]. In the
papers presented in this thesis the metabolites were identified by comparison of
retention indices and mass spectra with data in commercial, as well as in-house,
retention indexes and mass spectra libraries using NIST MS Search 2.0 (National
Institute of Standards and Technology, 2001).
GC-TOF-MS is a commonly used analytical platform for generating global
metabolic data [155-157]. The frequent used of GC-MS is probably due to its
relatively high sensitivity and reproducibility and that the identification process
by spectral databases is fairly straightforward. Nonetheless, GC-TOF-MS have one
important limitation in that it cannot detect non-volatile compounds and the
necessary derivatization complicates the sample preparation.
Two-dimensional gas-chromatography (2D GC) was also used for the gloval
screening in paper V. 2D GC uses two columns resulting in that the molecular
components are first separated on a polar primary column before entering a
nonpolar secondary column. Using two columns gives a higher degree of
chromatographic resolution and greater sensitivity. However, the preprocessing
20
of the data, explained in more detail below, becomes more complicated and the
time required for the analysis of one sample increases fivefold compared to 1D
GC.
Amino Acid Analysis by UPLC-LCTOFMS
To verify the amino acid pattern altered in response to ASS1 status seen with 1D
and 2D GC in paper V we decided to specifically evaluate the amino acid
concentrations in the samples.
A targeted approach utilizing the commercially available AccQ-Tag kit obtained
from Waters (Millford, MA) together with a UPLC-LCTOFMS instrument was used
to analyze amino acids. This method exploits pre-column derivatization of amino
acids with 6-aminoquinolyl-N-hydroxysuccinimidylcarbamate (AQC), which
converts both primary and secondary amino acids into stable fluorescent
derivatives that are amenable to UV-absorbance, fluorescence, electro-chemical,
and mass spectrometric detection [158].
Data processing
Pre-processing of SELDI-TOF-MS data
To enable an applicable analysis of the numerous mass spectra generated by
SELDI-TOF-MS, some pre-processing steps were necessary. Firstly, the baseline
was subtracted from the spectral data. Thereafter, to compensate for differing
energy absorbing molecule (EAM)-effects and possible variation in the amount of
retained sample between the different samples, each spectrum was normalized
with respect to its total ion current (TIC) by dividing each data point by the mean
of all data points in the region to be analyzed. Finally, to facilitate multivariate
comparison of the spectra, by ensuring that all spectra were represented by the
same variables, the normalized spectral data was binned into narrow mass
intervals, i.e. new spectral variables (bins) were then generated where each
variable represented the mean value of all original data points within the bin
interval. The bin sizes differed between different optimization mass-ranges,
depending on the frequency (i.e. data points per mass unit) of the original data
points. In the low mass region the frequency of data points was relatively high,
therefore the bin size was smaller than in the high mass region where the data
point frequency was relatively low [159].
Pre-processing of GC-TOF-MS data using HMCR
Analysis of multiple samples with GC-TOF-MS generates a three dimensional data
structure; the sample dimension, the spectral dimension and the chromatographic
dimension. Hierarchical Multivariate Curve Resolution (HMCR) is a multivariate
deconvolution method to translate these three dimensions into a two dimensional
data table where the cells consist of the integrated area under the resolved
21
chromatographic profile [160]. The HMCR procedure includes the following steps.
First, smoothing of each mass channel by a moving average, followed by alignment
which is done by finding maximum covariance between the samples total ion
current chromatograms. Thereafter, the data is divided into time windows and
each time windows is then put through multivariate curve resolution separately
which results in chromatographic profiles for each compound in each sample with
a corresponding spectral profile. The advantage with HMCR is the predictive
feature of the method [161], making it possible to deconvolute many samples in a
short time.
Targeted pre-processing
Another approach used to pre-process GC-TOF-MS data is an in-house MATLAB
script that processes the data based on an in-house spectral library of authentic
standard compounds. This will result in a data table containing only identified
compounds. It has the advantaged that all compounds in the data can be
biologically interpreted and that manual and time consuming identification steps,
comparing retention indices and mass spectra with commercial and in-house
retention indexes and mass spectra libraries, is not necessary to the same extent.
One drawback is the risk of losing compounds that could be of great importance
even though they have an unknown identity, since only metabolites present in the
spectra library will be found. However, when it comes to identification, regardless
of the method, manual verification of the identified compound is essential to be
confident of the identity of the metabolite.
Pre-processing of 2D GC-TOF-MS data
The second chromatographic dimension obtained from the 2D GC-TOF-MS
analysis, as compared to 1D GC-TOF-MS analysis, complicates the pre-processing
and at the time this work was performed no HMCR script was available for 2D GCMS. Consequently, other available software’s were used. For baseline correction,
peak detection and spectrum deconvolution, mass spectra library search for
identification and calculation of peak height/area the ChromaTOF software by
Leco (version 4.50.8.0) was used. For alignment, normalization with internal
standards and filtering, the data processing software Guineu (1.0.3 VTT; Espoo,
Finland) was utilized [162]. The library search was performed against publicly
available mass spectral libraries from US National Institute of Science and
Technology (NIST) and from the Max Planck Institute in Golm together with inhouse libraries established at Swedish Metabolomics Centre (SMC). After
processing in Guineu all peaks were manually investigated by using the average
spectra information, obtained from Guineu, in NIST MS Search 2.0 to search
against the same libraries as previously used to additionally confirm the putative
annotations of the metabolites and detect possible split peaks.
22
Pre-processing of amino acid analysis
Pre-processing of the amino acid analysis is a fairly straightforward process.
MATLAB 7.11.0 (R2014b) (Mathworks, Natick, MA) was used with an in-house
script for alignment and extraction of the integrated peak area for each amino
acid, resulting in a data table containing relative concentrations for each amino
acid in each sample.
Chemometrics
Chemometrics is an interdisciplinary science that contains tools from chemistry,
statistics and mathematics [163, 164]. Two of the cornerstones in chemometrics
are design of experiments (DOE) and multivariate analysis (MVA). To optimize the
experimental setup and ensure minimal covariance between factors of interest
(e.g. diagnosis or treatment response) and possible confounding factors (e.g. age,
gender, time in storage, etc.) DOE is employed in the initial stage of a study. When
the data has been acquired, MVA is used to analyze the obtained dataset to reveal
patterns built up by relationships between the measured variables (e.g. proteins
or metabolites).
Multivariate analysis of proteomic and metabolomic data
Multivariate analysis (MVA) methods are used to evaluate and interpret large
amounts of data. The MVA methods utilized in this thesis are primarily principal
component analysis (PCA) and orthogonal projections to latent structures (OPLS).
For simplicity all examples that follows are based on metabolic data but the same
principles applies for proteomic data.
Principal Component Analysis
Principal component analysis (PCA) is an essential tool in chemometrics [165]. It
is an unsupervised projection method used to visualize and interpret complex
multidimensional data. To be able to analyze the metabolic variation in the data,
the mass spectral information is stored in a matrix X, where each column
describes a measured metabolite and each row represents the relative
concentration of the metabolites in a sample. PCA compresses the variation of the
data by creating a window in the multidimensional space, generating orthogonal
principal components (PCs) containing the systematic variation. The number of
PCs can be calculated using different algorithms. The non-iterative partial least
squares (NIPALS) algorithm is the one used throughout this work [166]. The
systematic variation in the data, TPT, consist of the new PCs referred to as scores
(T) and the connection between the original data matrix (X) and the scores (T) is
explained by the loadings (P). An example of PCA scores and loadings is illustrated
in figure 2.
23
X = TP T + E
Where E represents the residuals, unexplained variation in data.
In this thesis PCA has mainly been used to get an overview of the data and detect
deviating samples prior to further more sophisticated analyses.
Figure 2. PCA A) scores plot (t[2] vs t[1]) and B) corresponding loadings plot (p[2] vs p[1]). The scores
show the first and the second principal component and how three different cell lines T98G (stars),
U118 (triangles) and LN229 (squares), behave in relation to each other based on the concentrations
of 75 metabolites (each symbol represents one cell line sample). Symbols close to each other in the
plot have similar properties based on the measured variables (metabolites). The loadings plot show
the corresponding loading vectors which show the distribution of the measured variables in relation
to each other. The scores and loadings can be interpreted together to better understand the differences
seen between the samples (cell lines). Variables close to each other in the loadings plot are describing
the same biological variation and the scores and loadings are directly comparable to each other. In this
specific case the interpretation reveals that alanine and phenylalanine are positively correlated to, i.e.
higher in concentration in, the T98G cell line while histidine and 2-oxoiscaproic acid are positively
correlated to the U118 cell line and citric acid and mannose are positively correlated to the LN229 cell
line.
Orthogonal Projections to Latent Structures
Orthogonal Projections to Latent Structures (OPLS) is a supervised projection
method which is an extension of the more commonly known partial least squares
(PLS) [167]. Supervised methods are useful when there is known information
about the samples in the data that can be utilized as response variables (Y) in the
calculation of multivariate regression models. OPLS describes the relations
between the response (Y) and the descriptive data matrix (X). The scores (T) are
calculated to describe the connection between X and Y. In contrast to PLS, OPLS
divides the systematic variation in X in two parts, one related to Y and the other
unrelated, orthogonal, to Y, which facilitates the interpretation of the data.
𝑿 = 𝑻𝒑 𝑷𝑻𝒑 + 𝑻𝐨 𝑷𝑻𝐨 + 𝑬
𝒀 = 𝑻𝒑 𝑪𝑻𝒑 + 𝑭
Tp PpT denotes the variation in X linearly related to Y, which hence can be used to
predict the response based on observed values. T0 P0T represents the variation
24
orthogonal to Y, and E and F describes the residuals, i.e. the non-systematic
variation in X.
OPLS – Discriminant Analysis
OPLS-Discriminant Analysis (OPLS-DA) [168] is a method to extract the variable
patterns in X that discriminates between pre-defined samples classes (e.g.
control-case). Y is in this case constructed as a binary dummy matrix holding
information only about the pre-defined sample class. If the sample belongs to a
pre-defined class it is given the value 1, if not, it is given the value 0. OPLS is then
used to correlate the experimental data (X) to the dummy matrix (Y) in order to
obtain a multivariate model for the separation between the classes. An example of
OPLS-DA scores and loadings is illustrated in figure 3.
Figure 3. OPLS-DA analysis of two sample classes, ASS1 negative cells (LN229; black) and ASS1
positive cells (U118 and T98G; grey). A) Scores plot and B) corresponding loadings plot. The scores
plot show the first predictive component t[1] and the first orthogonal component to[1], and how three
different cell lines T98G (stars), U118 (triangles) and LN229 (squares), behave in relation to each other
based on the concentrations of 75 metabolites (each symbol represents one cell line sample). Symbols
close to each other in the plot have similar properties based on the measured variables (metabolites).
The loadings plot show the corresponding loading vectors. pq[1], which is the first predictive X
loading p and Y loading q combined to one vector, and poso[1] which is the first orthogonal loading,
po, of the X-part and the projection of to onto Y (so), combined to one vector. Variables close to each
other in the loadings plot are describing the same biological variation and the scores and loadings are
directly comparable to each other. The scores plot show that the first predictive component (t1)
discriminates between ASS1 negative cell lines (black) and ASS1 positive cell lines (grey). The first
orthogonal component to[1] show the separation between the two ASS1 negative cell lines U118
(triangles) and T98G (stars). The loadings plot show that the metabolites separating the ASS1 negative
cell lines from the ASS1 positive cell lines are, among others, citric acid and mannose which are
positively correlated to, i.e. higher in concentration in, the ASS1 negative cell lines while alanine,
phenylalanine, histidine and 2-oxoiscaproic acid are positively correlated to ASS1 positive cell lines.
Validation
Validation of results is essential in all research. In our field of research an ideal
validation technique would be to analyze another set of samples with the same
analytical technique and receive the same results. Unfortunately, this is not always
25
possible since the accessibility of samples is limited. In our studies we have used
a couple of other available validation techniques instead.
Cross-Validation
When using OPLS there is always a risk of overestimating the predictive ability of
the models, i.e. incorporating random noise into the calculation of the scores. To
minimize this risk cross-validation (CV) is used [169]. In the cross-validation
process, parts of the data, normally 1/7, is initially held out when the model is
calculated. When samples are matched or if replicates are included they should
belong the same CV group, i.e., being held out the same time. Another approach is
to leave out one observation in each round, so called leave-one-out crossvalidation. The calculated regression model is then used to predict a response
value for the held out observations, which can then be compared to their true
response values. This is done until all observations have been held out and
predicted once and a prediction error term can be calculated, the Prediction Error
Sum of Squares (PRESS), which represents the sum of the square of the differences
between the observed and predicted value for each observation.
𝒏
𝑷𝑹𝑬𝑺𝑺 = ∑(𝒚𝒊 − 𝒚̂𝒊 )𝟐
𝒊=𝟏
𝑦 denotes the true value for each observation. 𝑦̂ denotes the predicted value and
𝑖 represents the number of observations and n is the number of observations. CVscore plots can then be created showing each observations predicted value. In all
paper, when score-plots are shown they are based on CV values.
Receiver operating characteristic curves
Another validation method is to use the area under a receiver operating
characteristic curve (ROC) curve as a measure of a variable’s predictive accuracy
by plotting the trade-off between sensitivity and specificity [170, 171]. For an
investigated variable, the true positive fraction is determined by calculating the
percentage of observations belonging to a specific class and displaying an
observed value for this variable greater than an arbitrary cut-off.
Correspondingly, the percentage of observations belonging to the other class, also
with an observed value greater than the arbitrary cut-off is calculated (false
positives). The cut-off is iteratively increased, the calculations repeated and the
results plotted against each other. This method can also be used to illustrate the
predictive accuracy of an OPLS model, by using the cross-validated (predicted)
score vector from a model as the variable for which the ROC curve is calculated.
The closer the curve follows the left-hand border and the top border of the ROC
space, the more accurate the test. In paper IV we used ROC curves to show the
reliability of the models.
26
p-values
Presenting p-values are a common way to display results and to validate the
significance of individual variables. p-values can be calculated by, for example,
using a student’s t-test or a Mann-Whitney U-test, depending on the data. To
validate the significance of an OPLS model a p-value be calculated using CVANOVA [172]. In all studies we calculate p-values for each individual variable to
validate the significance of the variables reported. In addition, we also calculated
a p-value based on CV-ANOVA to validate the models, molecular patterns, in all
studies.
Same samples, different methods
In paper V we analyzed the same samples with three different methods, 1D GCMS, 2D GC-MS and amino acid analysis. This approach, when analyzing different
samples with different methods should also be seen as a way of validating the
results. Having different analytical techniques showing the same results adds to
the reliability of the results presented.
Biological validation
The possibility to relate the detected proteins/metabolites of interest to a
biological context is essential when clinically validating our findings. This is not a
straightforward task since it is not always clear what is to be expected when it
comes to tumor biology and metabolite involvement. One problem is that many
metabolites are involved in many different cellular processes. To decide which of
the processes that are affected by a specific treatment is a complex task that may
be difficult, especially in the in vivo situation i.e. studying patient material. We
have used IPA® (Ingenuity Systems, Inc) as a pathway analysis tool to aid in the
biological validation.
27
RESULTS AND DISCUSSION OF THE
PAPERS
Paper I
Meningiomas of WHO grade I can, despite their benign appearance, display
invasive growth. Today it is not possible to predict which tumors that will develop
invasive growth, as there is no histopathologic difference between invasive and
non-invasive tumors. This makes the planning of an individual patient’s treatment
schedule difficult.
Objective
To extract protein expression patterns with the potential to discriminate between
invasive and non-invasive benign WHO grade I meningiomas.
Results
Expression patterns from the protein peaks obtained with SELDI-TOF-MS analysis
could discriminate between invasive and non-invasive tumors in both fibrous
meningiomas and meningothelial meningiomas. In fibrous tumors, 22 peaks
showed significantly different expression levels between invasive and noninvasive samples resulting in a complete separation between the groups (figure
4 A). It was also possible to partly separate invasive from non-invasive
meningothelial tumors based on six significantly different peaks (figure 4 B).
Figure 4. Cross-validated scores (tcv[1]) based on the final OPLS-DA models for A) fibrous tumors
and for B) meningothelial tumors. Non-invasive meningiomas are shown as black circles and invasive
meningiomas are shown as grey inverted triangles. Each symbol represents one tumor sample
described by 22 and six protein peaks in A and B respectively. There is a clear separation between
invasive and non-invasive fibrous meningiomas (p = 2.84*10-14) and a significant separation, however
with some overlap, between invasive and non-invasive meningothelial meningiomas (p = 1.6*10-9).
28
Comments and Discussion
SELDI-TOF-MS was shown to be a useful method to screen for protein pattern
differences between invasive and non-invasive meningiomas. We were able to
show that there is a clear and highly significant difference in protein pattern
between invasive and non-invasive fibrous tumors and a less but still significant
separation between meningothelial tumors (with some overlap in scores). We
found a panel of protein peaks, 22 in fibrous and six in meningothelial tumors,
with different m/z that were significantly altered between invasive and noninvasive tumors. Unfortunately, when using SELDI-TOF-MS the biological
interpretation is nonexistent since the applied technique does not allow
identification of the detected proteins or protein fragments based on the SELDI
data alone. However, the aim of this paper was to investigate whether protein
spectra differed between invasive and non-invasive benign meningiomas so it was
beyond the scope of this study to identify the proteins representing the peaks of
interest. In summary, this study shows that it is possible to distinguish invasive
from non-invasive grade I meningiomas based on protein pattern. The possibility
to be able to early identify the invasive tumors could, in the future, influence the
planning of treatment and be of aid when deciding on early or late postoperative
radiotherapy of meningiomas.
29
Paper II
The use of angiogenesis inhibitors is an established part of modern cancer
treatment regiments. As angiogenesis is associated with worse prognosis in GBM
[173] and VEGF is one of the regulators of angiogenesis, VEGF is a valid target for
treatment [110]. A commonly used angiogenesis inhibitor, bevacizumab, is an anti
VEGFR monoclonal antibody used as a second-line treatment of GBM. However,
the value of bevacizumab for glioma treatment and its precise mechanism of
action is controversial [174, 175]. Here we investigated another angiogenesis
inhibitor, vandetanib, which is a VEGFR/EGFR tyrosine kinase inhibitor, as a
model for targeted therapies and angiogenesis inhibition. Vandetanib has shown
promising effects on tumor growth in different subcutaneous tumor models
including intracranial rat glioma models [176, 177]. In this study, an animal
model, a BTC4 syngenic intracerebral rat glioma model [178, 179], was subjected
to the angiogenesis inhibitor. The BT4C model is a transplacental nitrosurea
induced rat tumor where the tumor cells originate from the same species and are
implanted in the same organ as the cells originated from. The tumor has
previously been characterized by infiltrative growth with a histopathological
picture resembling gliosarcoma.
Objective
To, in a first step, verify if proteomic patterns in brain tumor tissue and serum
could discriminate between rats treated with an angiogenesis inhibitor and
untreated animals using SELDI-TOF-MS. Furthermore, in a second step, to identify
differentiating proteins using iTRAQ labeling and HPLC.
Results
The SELDI-TOF-MS results showed that treated and untreated tumors had
different proteomic patterns (figure 5). With HPLC it was possible to identify
some proteins that were significantly different between treated and untreated
though it was not possible to identify the selected proteins/peptides of interest
found by the SELDI-TOF-MS analysis. In conclusion, we found 72 proteins that
were significantly altered by treatment. After pathway analysis using Ingenuity
Pathway analysis (IPA®) we could extract a cluster of 26 proteins that were
connected to HIF1-α signaling, a key signaling pathway in angiogenesis. Among
the 26 proteins HSP90 was one of the most interesting for which we carried out
an ELISA analysis to verify the identified alterations in the tumor samples. The
results showed that the HSP90 tissue levels were 4.4 times higher in tumor
samples from rats treated with the tyrosine kinase inhibitor as compared to
untreated tumors.
30
Figure 5. Cross-validated scores (tcv[1]) based on the final OPLS-DA model of SELDI-TOF-MS data of
rat tumors showing the difference between vandetanib treated (black diamonds) and untreated (grey
circles) samples (p =. 5*10-6).
Comments and Discussion
In this study we used SELDI-TOF-MS analysis as a screening method to verify our
hypothesis that treatment with an angiogenesis inhibitor would induce changes
to the serum proteome. Based on those findings we then proceeded to analyze
both serum and tissue samples with iTRAQ labeling in combination with HPLC to
obtain both quantified and identified proteins to enable a biological
interpretation. Correlating the identified proteins from the HPLC analysis with the
SELDI-TOF-MS analysis was not possible. Most of the interesting protein peaks
from the SELDI-TOF-MS analysis were below 10kDA while the majority of the
protein identified with iTRAQ were higher than 10kDA. The high amount of lowmolecular peaks from the SELDI-TOF-MS analysis also indicates that there might
be fragmentation of larger proteins in the sample which makes the identification
of what protein fragment they originate from difficult. Nevertheless, with the
iTRAQ technique in combination with HPLC, we were able to identify a number of
interesting proteins.
Following pathway analysis we found that several of the identified proteins
elevated in treated animals were directly connected to HIF1-α, a key protein in the
angiogenetic pathway. HIF1-α signaling responds to changes in oxygen
concentration and during hypoxia, HIF1-α increases the transcription of VEGF
which in turn trigger angiogenesis. HSP90-α, a chaperone protein required for
HIF1-α activation was elevated in treated samples compared to control indicating
more hypoxia in the treated samples compared to control. We further evaluated
HSP90 with an ELISA analysis and found that HSP90 tissue levels were 4.4 times
higher in samples treated with vandetanib than in the control samples. As the
interaction of HSP90 with HIF1-α occurs upstream from the VEGF signaling, it is
hard to draw any conclusion about the relationship between the treatment and
the elevation of HSP90 and HIF-α signaling. However, the possibility to find
angiogenesis-related proteins affected by treatment is an indication that this is a
useful approach to investigate treatment induced changes in gliomas.
31
Paper III
The standard treatment of high-grade gliomas is surgery followed by
radiotherapy and adjuvant temozolomide. If this first line of treatment is
unsuccessful, second line chemotherapy and re-irradiation can be considered in
specific cases. One of the major problems when treating high-grade gliomas is the
lack of immediate monitoring of the treatment response. Morphological changes
are in many cases detected too late to influence treatment modifications making
today’s treatment response evaluation with repeated CT or MRI insufficient.
Objective
To investigate if radiation of high-grade gliomas in patients induces any
metabolite changes in serum by means of GC-TOFMS. Additionally, to compare the
findings with the results from a previous study where radiation-induced
metabolite pattern changes was demonstrated in the extracellular compartment
of tumors and brain adjacent to tumor (BAT) in the same patients using
microdialysis.
Results
From the acquired GC-TOFMS data it was possible to detect metabolic changes in
serum during the first five days of radiotherapy for high grade gliomas. 84
metabolites, out of which 28 could be identified, differed significantly between
samples collected before treatment compared to samples collected during
treatment with radiotherapy. The final OPLS-DA model for the separation
between samples collected before and samples collected during treatment
consisted of 1 predictive and 1 orthogonal component and predicted 31.9% of the
response variation (Q2=0.319, p=0.006)). As shown in figure 6, it was possible to
follow the metabolic response over the specific time period and the majority of
the patients showed a different metabolic response pattern during treatment as
compared to before treatment. Some of the metabolites detected as significant in
serum were also found significant in the extracellular compartment of the same
patients in our previous study.
Figure 6. Cross-validated scores (tcv[1]) based on the final OPLS-DA model for the separation
between samples collected before and samples collected during treatment. The circles represent the
two time points before start of treatment and the triangles the time points during treatment. The labels
32
correspond to the sampling time point. 2* was calculated and used as a reference point in the ITOT
normalization of the treated samples.
Comments and Discussion
In this study we analyzed serum samples collected before and during five days of
radiotherapy with the aim to investigate potential treatment induced effects on
the serum metabolome. We found several metabolites that were altered due to
treatment in serum. In general there was a decrease in metabolite levels during
radiotherapy compared to the levels before treatment. All detected and identified
amino acids and fatty acids were among the metabolites that decreased in
concentration, while citric acid was among the metabolites that increased in
concentration. If this change in serum is dependent on what is happening in the
tumor during treatment, such as increased release or uptake of metabolites, or if
it is a direct effect on the blood compartment itself is a question that we can only
speculate about at this point. Comparing our findings in serum with the previous
study of the extracellular compartment from GBMs, we found some metabolites
that changed in the same direction during treatment and some metabolites that
where increased in serum and decreased in tumor extracellular compartment and
vice versa.
Considering individual patients, the majority showed a clear difference in
metabolite pattern due to treatment. Two patients, however, showed a more
moderate or no response (patient 6 and patient 11). The reason for this cannot be
made conclusive based on the acquired data in this limited sample set. These two
patients do not deviate in sex, age, performance status or survival time. Possibly
could their deviating pattern be associated with a different response to the given
treatment, something that is if great interest to investigate in a larger sample
material.
From our findings we could not draw any final conclusions on the biological
events that resulted in the metabolic changes. Also, it will be an issue for further
studies to investigate why some of the metabolites in the different sample
compartments responded differently to treatment. An interesting option could be
to analyze the CSF as an intermediate between serum and the extracellular
compartment which could have facilitated the interpretation of the flow of
metabolites.
33
Paper IV
The classification and grading of glioma is today based on histological features
which are not always easy to interpret. The most important feature of the
classification is the expected correlation between tumor grade and prognosis even
though large variations exist in the survival of patients concerning both GBMs and
low-grade gliomas.
Objective
To investigate the metabolic profiles in tumor tissue and serum from patients with
glial tumors of different histological type and grade, with the aim to retrieve
diagnostic and prognostic information.
Results
We found differentiating metabolite patterns between GBMs and
oligodendrogliomas (figure 7 A and B). The final OPLS-DA model on tumor
samples consisted of 1 predictive component predicting 34.1% of the response
variation (Q2 = 0.341; p = 2.46*10-8). The final OPLS-DA model for serum samples
consisted of 1 predictive component predicting 22.3% of the response variation
(Q2 = 0.223; p = 1.3*10-5). Furthermore, in oligodendrogliomas we could see a
pattern difference associated with tumor grade (grade II versus grade III) (figure
7 C and D). The final OPLS-DA model for tumor samples consisted of 1 predictive
component predicting 43.4% of the response variation (Q2 = 0.434; p = 0.01). The
final OPLS-DA model for serum samples consisted of 1 predictive component
predicting 58.9% of the response variation (Q2 = 0.589; p = 0.008). When
modelling survival it was found that patients with GBM that died shortly, within
four months, after diagnosis and patients that lived long, more than two years,
after diagnosis showed significantly different metabolite patterns in both tumor
tissue and serum (figure 8 A and B). The final OPLS-DA model for tumor samples
consisted of 1 predictive component predicting 42.7% of the response variation
(Q2 = 0.427; p = 0.006). The final OPLS-DA model for serum samples consisted of
1 predictive component predicting 47.8% of the response variation (Q2 = 0.478;
p = 0.004). Similar results were also obtained for patients with
oligodendrogliomas (figure 8 C and D) in tumor (Q2 = 0.767; p = 0.004) and in
serum (Q2 = 0.855; p = 0.01).
34
Figure 7. Upper panel. Cross-validated scores (tcv[1]) based on the final OPLS-DA models
discriminating between glioblastoma (black) and oligodendroglioma (grey) in A) tumor and B) serum.
Lower panel. Cross-validated scores (tcv[1]) discriminating between WHO grade II (diamonds) and
III (inverted triangles) in oligodendroglioma in C) tumor and D) serum.
Figure 8. Cross-validated scores (tcv[1]) from the final OPLS models illustrating short survival time
(boxes) compared to long survival time (triangles) in A) GBM tissue and B) serum and C)
oligodendroglioma tissue and D) serum.
35
Comments and Discussion:
In this study we carried out a global metabolomic screening of 104 gliomas. All
glioma tumors collected in Umeå during a four year time period were initially
analyzed resulting in a very diverse dataset. GBM is the most common type of
glioma therefore the majority of the tumors included were GBMs. The initial
dataset contained only four patients with low grade astrocytoma. It was not
meaningful to compare those to the large group of GBMs, subsequently the low
grade astrocytomas were excluded in the data analysis.
The metabolite profiles from patients with GBM and oligodendroglioma were
initially compared. Because of the major histopathological difference between the
two glioma subtypes, a clear difference in metabolic pattern may also have been
expected. The skewedness in sample size may, to some extent, explain the fairly
moderate separation seen between the groups. Additionally, the metabolic
profiles of grade II and III oligodendrogliomas were analyzed and we could
identify metabolites that were different in relation to tumor grade. From a clinical
point of view it would have been more interesting to be able to separate low grade
from high grade astrocytomas, however, the small number of low grade
astrocytomas made that comparison impossible.
Remarkably, we could show that there was an association between the metabolite
data and in survival time in patients with both GBM and oligodendrogliomas.
Initially, we tried to find patterns correlated to survival time based on a
continuous timeline. This appeared not be possible in this dataset. The patients
where then divided into four groups, one short time survival group, two
intermediary survival time groups and one longtime survival group. In the paper
we showed that there were metabolic differences between the long time survival
group and short time survival group. What is not shown is that it also was possible
to separate the two intermediary survival time groups. Both from each other and
from the two extremes. When comparing survival time it would have been
interesting to have had information about IDH1/2 mutations for these patients. As
mentioned earlier, it has been shown than IDH1/2 mutations are related to
survival. Nevertheless, we have identified 2-hydroxyglutaric acid in the tissue
samples. 2-hydroxyglutaric acid is converted from isocitrate by the IDH enzymes
and it has been shown that the concentrations of 2- hydroxyglutaric acid is
vanishingly low in normal tissue but extremely high in tumors with mutations in
the IDH1/2 genes [118]. In our data we see an increased level of 2hydroxyglutaric acid in oligodendrogliomas compared to GBMs indicating that it
could be possible that the oligodendrogliomas have IDH1/2 mutations. This is also
accordance with the earlier statement that IDH1/2 mutations are more common
in the lower grade gliomas compared to GBMs. IDH1/2 mutations and 2hydroxyglutaric acid would have been interesting to investigate further in followup studies as it has been shown that 2-hydroxyglutaric acid generated from IDH1
mutant gliomas can be monitored during surgery with ambient mass
spectrometry techniques [180] .
36
Paper V
Pegylated arginine deiminase ADI-PEG20 is an arginine deprivation therapy that
has been in trial for various cancers [181-183]. Arginine is a semi-essential amino
acid involved in several pathways including major cellular functions such as nitric
oxide production, creatine production and polyamine synthesis. Human cells can
synthesize arginine. However, endogenous production is insufficient when cells
are under stress or need to proliferate. Tumor cells have a high requirement for
arginine as arginine influences tumor cells growth and proliferation [184, 185].
Epigenetic studies in GBM cell lines by Syed et al [186] identified a subgroup of
GBMs that lacked argininosuccinate synthetase (ASS1) expression. ASS1 is an
enzyme that together with argininosuccinate lyase (ASL) catalyzes the synthesis
of arginine from citrulline via the urea cycle. This ASS1 negative group of tumors
are thus auxotrophic for arginine and sensitive to deprivation therapy by ADIPEG20 [187-189]. ADI-PEG20 works by catabolizing arginine to citrulline and
ammonia, depleting the extracellular arginine levels and subsequently depleting
the intracellular availability in ASS1 negative cells [190].
Objective
To evaluate differences in the metabolome between ASS1 positive and ASS1
negative glioma cell lines and to evaluate treatment effects of ASS1 negative cell
lines exposed to ADI-PEG20 therapy. Furthermore, to test our hypothesis that the
arginine biosynthetic pathway would be affected when comparing ASS1 negative
and ASS1 positive cell lines and subsequently that the ASS1 negative cells would
show lower levels of the metabolites involved in the arginine metabolism.
Results
In this study, as a complement to the global metabolomic screening methods, 1D
GC-TOF-MS and 2D GC-TOF-MS, we also performed a targeted analysis of 29
amino acids. This since amino acids dominated the arginine biosynthetic pathway.
All three techniques showed that the metabolites involved in arginine
biosynthesis were lowered in the ASS1 negative cell lines, which is in agreement
with our pre-defined hypothesis. In figure 9 the separation between ASS1
negative and ASS1 positive cell lines can be seen based on 1D GC-TOF-MS data of
the cell supernatant (p = 1.84*10-5) and cell pellet (p=0.03). This separation could
also be seen in the 2D-GC-TOF-MS data as well as in the data for the 29 amino
acids.
37
Figure 9. Cross-validated scores (tcv[1]) based on the final OPLS-DA models showing a complete
separation of ASS1 positive cell lines (T98G and U118; grey) and ASS1 negative cell lines (GAMG, LN229
and SNB19; black) for 1D GC-TOF-MS data in A) cell supernatant and B) cell pellet.
Furthermore, to evaluate the effect of arginine deprivation treatment on the
metabolic level we analyzed ASS1 negative and ASS1 positive cell lines together
with normal astrocytes (fetal cells) before and after treatment with. As
hypothesized, only ASS1 negative cells were affected by the treatment (p=0.001)
while the other cell types remained unaffected (figure 10).
Figure 10. PCA scores, second principal component (t[2]) for 1D GC-TOF-MS data of cell supernatant
from normal astrocytes (white), an ASS1 negative cell line (SNB19; black) and an ASS1 positive cell line
(U87; grey). Samples treated with ADI-PEG20 are denoted with a star (*). All cells were run in
triplicates. It can be clearly seen that the ASS1 negative samples are significantly affected by treatment
(p = 0.001) while normal astrocytes and ASS1 positive cells remains unaffected by treatment.
Comments and Discussion
Cell line samples were analyzed with three different methods, two global
screening methods (1D and 2D GC-TOFMS) and one targeted LC/MS method for
quantification of 29 amino acids. Subsequently, analysis of data from all three
methods showed a clear metabolic difference between ASS1 negative and ASS1
positive cell lines. As hypothesized, among the metabolites responsible for this
difference we identified that all metabolites involved in the arginine biosynthetic
pathway was elevated in ASS1 positive cell lines as compared to ASS1 negative cell
lines. Interestingly, we could also see an upregulation of many other metabolites
which can be supported by the fact that arginine is a substrate for other
metabolites including proline and creatine.
38
When analyzing the treatment induced effects of arginine deprivation therapy
using ADI-PEG20 we found that only ASS1 negative cell lines were affected by the
treatment when comparing to normal astrocytes and ASS1 positive cells. The lack
of effect on the arginine deprivation therapy in normal astrocytes and ASS1
positive cell lines should be expected because of their ability to synthesize their
own arginine, something ASS1 negative cell lines cannot.
This is a unique study presenting a metabolomics characterization of ASS1 status
in GBM cell lines. The results verify pre-defined hypotheses regarding ASS1
metabolism but also generate new hypotheses for future investigations; e.g. other
metabolic pathways for treatment of ASS1 positive tumors as well as potential
diagnostic markers for non-invasive diagnosis in-vivo, e.g. by magnetic resonance
spectroscopy (MRS). In addition the study shows proof of concept for an
interesting way of using metabolomics as a complement to other omics
techniques, in this case epigenetics, in order to create new possibilities for
detection or diagnosis of molecular tumor subgroups as well as to reveal new
metabolic targets for treatment of disease. To be able to carry out this type of
multi-omics studies, the bioinformatics strategy involving chemometric or
multivariate techniques is essential. In this way it is possible to detect markers of
metabolites, latent biomarkers that can be used as intermediate phenotypes in
combination with other omics descriptions to facilitate the molecular
subgrouping of tumors. It will be of high interest to continue to investigate the
possibilities of these types of multi-omics approaches where the systematic
variation in the metabolome is captured as multivariate phenotypic descriptors.
Hopefully this type of bioinformatic strategies can play a part in improving the
characterization, diagnosis and treatment of GBMs based on molecular evidence.
39
CONCLUSION
This thesis describes proteomic and metabolomic studies of brain malignancies
with the main objective to discover diagnostic and prognostic information in the
proteome and/or metabolome as well as to investigate the possibility to identify
predictive biomarkers in glioma and meningioma.
Initially, protein profiles from tumor tissue from meningioma patients were
investigated with SELDI-TOF-MS with the objective to find protein patterns that
differed between invasive and non-invasive tumors. We were able to show that
within specific subgroups of meningiomas (fibrous and meningothelial) there
were protein patterns that where altered between the groups. A finding that
demonstrates the possibility to find marker patterns that may indicate an invasive
growth of the tumors, a result which should be confirmed in prospective studies
In addition, proteomic profiles of tumor tissue and serum from a glioma rat model
where rats were treated with an angiogenesis inhibitor (vandetanib) were
investigated. SELDI-TOF-MS was used to evaluate the protein profiles from
samples collected from treated and untreated animals. As discussed before,
SELDI-TOF-MS does not allow for identification of interesting proteins. Therefore
the samples were, in a second step, analyzed with iTRAQ labeling together with
HPLC to allow for identification of altered proteins. The SELDI-TOF-MS data
showed proteomic patterns that differed between treated and non-treated
tumors, which made it worthwhile to continue with protein identification Using
iTRAQ labeling we found several proteins that were altered between the groups,
for which we could also assign identities. A cluster of these identified proteins
were connected to HIF1-α, a key protein in the angiogenetic pathway. To verify
the level of one of those proteins, HSP90, the samples were analyzed using an
ELISA which showed a significant increase of HSP90 in tumors treated with
vandetanib.
Serum samples from glioma patients treated with radiotherapy was submitted to
a global metabolomic screening approach using GC-TOF-MS. Several metabolites
were altered due to treatment indicating the possibility to follow treatmentinduced metabolic changes in blood. It was also evident that different patients had
different metabolic response to treatment. Follow-up studies confirming these
findings is off course necessary due to the low number of patients included in our
study. However, hopefully, this study may be small step forward on the road
towards personalized treatment.
GC-TOF-MS was also used to globally characterize gliomas of different
histopathological origin and grade. Serum and tumor tissue from the same
patients were analyzed. Results from both serum and tissue showed that there
40
were metabolic differences between GBM and oligodendroglioma, grade II and III
oligodendrogliomas and that the metabolome also may harbor information
associated with time of survival in patients with GBM as well as
oligodendrogliomas. These findings opens for the possibility to obtain diagnostic
and prognostic information not only from tumor tissue samples but also from
blood.
In the final study presented in this thesis, metabolite profiles of ASS1 positive and
ASS1 negative glioma cell lines were evaluated together with the effects of
arginine deprivation therapy in both cell lines. The cell line samples were
analyzed using both 1D and 2D GC-TOF-MS to be able to cover a larger proportion
of the metabolome than in our previous metabolomic studies. Since many of the
metabolites included in the arginine biosynthetic pathway, which was expected to
be different between the ASS1 positive and ASS1 negative glioma cell lines, are
amino acids, a targeted LC/MS approach for 29 amino acids was also applied. We
could show that ASS1 positive and ASS1 negative cell lines expressed different
metabolic phenotypes. Furthermore, as hypothesized, when exposed to arginine
deprivation therapy only cells lacking ASS1 expression (ASS1 negative), were
affected. This study demonstrates the possibility of using metabolomics as a
complement to other omics techniques and potentially opens up for the
opportunity to investigate other multi-omics approaches for detection of new
molecular tumor subgroups and reveal novel treatment options in GBM as well
other types of tumors.
Proteomics and metabolomics are up and coming fields in cancer research. As
compared to genes, proteins and their modifications, provide an insight to what is
happening in a cancer patient in real time. This enables an understanding of the
biological processes of cancer development as the proteins encoded by the genes
are the functional players that drive both normal and disease physiology.
Metabolites are the endpoint of the central dogma of molecular biology and
closest to the phenotype, thus reflecting the genetic composition as well as
external factors such as lifestyle and diet. By analyzing the metabolome it is
possible to get an insight into the biochemical events occurring downstream of
gene expression and protein activation.
Unbiased global metabolomic and proteomic screening methods are valid
approaches in the search for biomarkers as large scale screening studies open up
for the possibility to analyze and detect a variety of proteins and metabolites.
However, the interpretation of the proteins and metabolites in a biological context
becomes a more difficult task. In our proteomics studies, one issue was the
difficulties in identifying individual proteins from the protein spectra obtained.
Nevertheless, our studies clearly indicate that by analyzing protein spectra one
can obtain useful information on meningioma characteristics and treatment
effects in gliomas. Another issue, regarding metabolomics, was the difficulty to
understand the biological processes behind the metabolic differences/changes
41
demonstrated due to the fact that our methods did not cover the whole
metabolome. Although our attempts to understand the biological processes by
utilizing computerized pathway databases, the fact that many metabolites are
involved in many different metabolic processes makes the interpretation even
more complex. In the future, a recommendation in studies, such as those
presented in this thesis, is to rely more on hypothesis-based approaches targeting
specific pathways based on the knowledge obtained from our global screenings or
from other data sources, e.g. as in paper V.
The rapid development in the field of proteomics and metabolomics is likely to
provide new protein/metabolite biomarkers for glioma and other cancers.
However, the likelihood that single proteins/metabolites should revolutionize the
field, as more specific and predictive biomarkers, is not high. Instead the idea of
utilizing molecular patterns could be an interesting option for the future. Such, so
called, latent biomarkers built up by a number of co-varying proteins/metabolites
are in theory more likely to generate both more specific and predictively robust
entities, e.g. for diagnosis, treatment monitoring and prediction. In addition, these
latent descriptors of the systematic variation in the metabolome could also be
used in correlations with other alterations, e.g. in the genome, to allow for a more
detailed molecular characterization of heterogeneity within the same tumor
group, e.g. GBMs. In this thesis, we show some initial results of these ideas applied
to brain tumors. However, to extract these latent biomarkers in an optimal way
and perform multi-omics correlations based on them bioinformatics based on
chemometrics or similar methods taking variable co-variation into account must
be further developed and applied.
42
REFERENCES
1. Socialstyrelsen. Cancerincidens i Sverige 2013 – Nya diagnosticerade cancerfall år
2013 2013.
2. Warburg O. Origin of cancer cells. Science. 1956;123(3191):309-14.
doi:10.1126/science.123.3191.309.
3. Roslin M, Henriksson R, Bergstrom P, Ungerstedt U, Bergenheim AT. Baseline
levels of glucose metabolites, glutamate and glycerol in malignant glioma assessed by
stereotactic microdialysis. Journal of Neuro-Oncology. 2003;61(2):151-60.
doi:10.1023/a:1022106910017.
4. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell. 2000;100(1):57-70.
doi:10.1016/s0092-8674(00)81683-9.
5. Hanahan D, Weinberg RA. Hallmarks of Cancer: The Next Generation. Cell.
2011;144(5):646-74. doi:10.1016/j.cell.2011.02.013.
6. Nishimura T, Ogiwara A, Fujii K, Kawakami T, Kawamura T, Anyouji H et al.
Disease proteomics toward bedside reality. Journal of Gastroenterology. 2005;40:713. doi:10.1007/bf02990572.
7. Ezkurdia I, Juan D, Manuel Rodriguez J, Frankish A, Diekhans M, Harrow J et al.
Multiple evidence strands suggest that there may be as few as 19 000 human proteincoding genes. Human Molecular Genetics. 2014;23(22):5866-78.
doi:10.1093/hmg/ddu309.
8. Crick F. Central dogma of molecular biology. Nature. 1970;227(5258):561-&.
doi:10.1038/227561a0.
9. Parsons DW, Jones S, Zhang X, Lin JC-H, Leary RJ, Angenendt P et al. An
integrated genomic analysis of human glioblastoma Multiforme. Science.
2008;321(5897):1807-12. doi:10.1126/science.1164382.
10. Furnari FB, Fenton T, Bachoo RM, Mukasa A, Stommel JM, Stegh A et al.
Malignant astrocytic glioma: genetics, biology, and paths to treatment. Genes &
Development. 2007;21(21):2683-710. doi:10.1101/gad.1596707.
11. Carnero A, Blanco-Aparicio C, Renner O, Link W, Leal JFM. The
PTEN/PI3K/AKT signalling pathway in cancer, therapeutic implications. Current
Cancer Drug Targets. 2008;8(3):187-98. doi:10.2174/156800908784293659.
12. Engelman JA, Luo J, Cantley LC. The evolution of phosphatidylinositol 3-kinases
as regulators of growth and metabolism. Nature Reviews Genetics. 2006;7(8):606-19.
doi:10.1038/nrg1879.
13. Moodie SA, Willumsen BM, Weber MJ, Wolfman A. Complexes of ras.gtp with
raf-1 and mitogen-activated protein-kinase kinase. Science. 1993;260(5114):1658-61.
doi:10.1126/science.8503013.
43
14. Mellinghoff IK, Wang MY, Vivanco I, Haas-Kogan DA, Zhu SJ, Dia EQ et al.
Molecular determinants of the response of glioblastomas to EGFR kinase inhibitors.
New England Journal of Medicine. 2005;353(19):2012-24.
doi:10.1056/NEJMoa051918.
15. Fischer I, Aldape K. Molecular Tools: Biology, Prognosis, and Therapeutic
Triage. Neuroimaging Clinics of North America. 2010;20(3):273-+.
doi:10.1016/j.nic.2010.05.004.
16. Zoncu R, Efeyan A, Sabatini DM. mTOR: from growth signal integration to
cancer, diabetes and ageing. Nature Reviews Molecular Cell Biology. 2011;12(1):2135. doi:10.1038/nrm3025.
17. Kreisl TN, Lassman AB, Mischel PS, Rosen N, Scher HI, Teruya-Feldstein J et al.
A pilot study of everolimus and gefitinib in the treatment of recurrent glioblastoma
(GBM). Journal of Neuro-Oncology. 2009;92(1):99-105. doi:10.1007/s11060-0089741-z.
18. McBride SM, Perez DA, Polley M-Y, Vandenberg SR, Smith JS, Zheng S et al.
Activation of PI3K/mTOR pathway occurs in most adult low-grade gliomas and
predicts patient survival. Journal of Neuro-Oncology. 2010;97(1):33-40.
doi:10.1007/s11060-009-0004-4.
19. Van Meir EG, Hadjipanayis CG, Norden AD, Shu H-K, Wen PY, Olson JJ.
Exciting New Advances in Neuro-Oncology The Avenue to a Cure for Malignant
Glioma. Ca-a Cancer Journal for Clinicians. 2010;60(3):166-93.
doi:10.3322/caac.20069.
20. Levine AJ. p53, the cellular gatekeeper for growth and division. Cell.
1997;88(3):323-31. doi:10.1016/s0092-8674(00)81871-1.
21. Riemenschneider MJ, Jeuken JWM, Wesseling P, Reifenberger G. Molecular
diagnostics of gliomas: state of the art. Acta Neuropathologica. 2010;120(5):567-84.
doi:10.1007/s00401-010-0736-4.
22. Zhao S, Lin Y, Xu W, Jiang W, Zha Z, Wang P et al. Glioma-Derived Mutations
in IDH1 Dominantly Inhibit IDH1 Catalytic Activity and Induce HIF-1 alpha.
Science. 2009;324(5924):261-5. doi:10.1126/science.1170944.
23. Carmeliet P, Dor Y, Herbert JM, Fukumura D, Brusselmans K, Dewerchin M et
al. Role of HIF-1 alpha or in hypoxia-mediated apoptosis, cell proliferation and
tumour angiogenesis. Nature. 1998;394(6692):485-90. doi:10.1038/28867.
24. Vaupel P. The role of hypoxia-induced factors in tumor progression. Oncologist.
2004;9:10-7.
25. Forsythe JA, Jiang BH, Iyer NV, Agani F, Leung SW, Koos RD et al. Activation
of vascular endothelial growth factor gene transcription by hypoxia-inducible factor
1. Molecular and Cellular Biology. 1996;16(9):4604-13.
44
26. Macheda ML, Rogers S, Best JD. Molecular and cellular regulation of glucose
transporter (GLUT) proteins in cancer. Journal of Cellular Physiology.
2005;202(3):654-62. doi:10.1002/jcp.20166.
27. DeBerardinis RJ, Cheng T. Q's next: the diverse functions of glutamine in
metabolism, cell biology and cancer. Oncogene. 2010;29(3):313-24.
doi:10.1038/onc.2009.358.
28. DeBerardinis RJ, Mancuso A, Daikhin E, Nissim I, Yudkoff M, Wehrli S et al.
Beyond aerobic glycolysis: Transformed cells can engage in glutamine metabolism
that exceeds the requirement for protein and nucleotide synthesis. Proceedings of the
National Academy of Sciences of the United States of America. 2007;104(49):1934550. doi:10.1073/pnas.0709747104.
29. Flier JS, Mueckler MM, Usher P, Lodish HF. Elevated levels of glucose-transport
and transporter messenger-rna are induced by ras or src oncogenes. Science.
1987;235(4795):1492-5. doi:10.1126/science.3103217.
30. Wise DR, DeBerardinis RJ, Mancuso A, Sayed N, Zhang X-Y, Pfeiffer HK et al.
Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis
and leads to glutamine addiction. Proceedings of the National Academy of Sciences
of the United States of America. 2008;105(48):18782-7.
doi:10.1073/pnas.0810199105.
31. Pouyssegur J, Dayan F, Mazure NM. Hypoxia signalling in cancer and
approaches to enforce tumour regression. Nature. 2006;441(7092):437-43.
doi:10.1038/nature04871.
32. Semenza GL. Targeting HIF-1 for cancer therapy. Nature Reviews Cancer.
2003;3(10):721-32. doi:10.1038/nrc1187.
33. Kim JW, Tchernyshyov I, Semenza GL, Dang CV. HIF-1-mediated expression of
pyruvate dehydrogenase kinase: A metabolic switch required for cellular adaptation
to hypoxia. Cell Metabolism. 2006;3(3):177-85. doi:10.1016/j.cmet.2006.02.002.
34. DeBerardinis RJ, Lum JJ, Thompson CB. Phosphatidylinositol 3-kinasedependent modulation of carnitine palmitoyltransferase 1A expression regulates lipid
metabolism during hematopoietic cell growth. Journal of Biological Chemistry.
2006;281(49):37372-80. doi:10.1074/jbc.M608372200.
35. Schwartzenberg-Bar-Yoseph F, Armoni M, Karnieli E. The tumor suppressor p53
down-regulates glucose transporters GLUT1 and GLUT4 gene expression. Cancer
Research. 2004;64(7):2627-33. doi:10.1158/0008-5472.can-03-0846.
36. Weinberg JM, Venkatachalam MA, Roeser NF, Nissim I. Mitochondrial
dysfunction during hypoxia/reoxygenation and its correction by anaerobic
metabolism of citric acid cycle intermediates. Proceedings of the National Academy
of Sciences of the United States of America. 2000;97(6):2826-31.
doi:10.1073/pnas.97.6.2826.
45
37. Metallo CM, Gameiro PA, Bell EL, Mattaini KR, Yang J, Hiller K et al.
Reductive glutamine metabolism by IDH1 mediates lipogenesis under hypoxia.
Nature. 2012;481(7381):380-U166. doi:10.1038/nature10602.
38. Mullen AR, Wheaton WW, Jin ES, Chen P-H, Sullivan LB, Cheng T et al.
Reductive carboxylation supports growth in tumour cells with defective
mitochondria. Nature. 2012;481(7381):385-U171. doi:10.1038/nature10642.
39. Dranoff G, Elion GB, Friedman HS, Campbell GL, Bigner DD. Influence of
glutamine on the growth of human glioma and medulloblastoma in culture. Cancer
Research. 1985;45(9):4077-81.
40. Medina MA, Sanchezjimenez F, Marquez J, Quesada AR, Decastro in. Relevance
of glutamine-metabolism to tumor-cell growth. Molecular and Cellular Biochemistry.
1992;113(1):1-15.
41. Portais JC, Martin M, Canioni P, Merle M. Glutathione, but not glutamine, is
detected in c-13-nmr spectra of perchloric-acid extracts from c6 glioma-cells. Febs
Letters. 1993;327(3):301-6. doi:10.1016/0014-5793(93)81009-o.
42. Hertz L, Dringen R, Schousboe A, Robinson SR. Astrocytes: Glutamate
producers for neurons. Journal of Neuroscience Research. 1999;57(4):417-28.
doi:10.1002/(sici)1097-4547(19990815)57:4<417::aid-jnr1>3.0.co;2-n.
43. Ye ZC, Sontheimer H. Glioma cells release excitotoxic concentrations of
glutamate. Cancer Research. 1999;59(17):4383-91.
44. Ye ZC, Rothstein JD, Sontheimer H. Compromised glutamate transport in human
glioma cells: Reduction-mislocalization of sodium-dependent glutamate transporters
and enhanced activity of cystine-glutamate exchange. Journal of Neuroscience.
1999;19(24):10767-77.
45. Takano T, Lin JHC, Arcuino G, Gao Q, Yang J, Nedergaard M. Glutamate release
promotes growth of malignant gliomas. Nature Medicine. 2001;7(9):1010-5.
doi:10.1038/nm0901-1010.
46. Behrens PF, Langemann H, Strohschein R, Draeger J, Henning J. Extracellular
glutamate and other metabolites in and around RG2 rat glioma: an intracerebral
microdialysis study. Journal of Neuro-Oncology. 2000;47(1):11-22.
doi:10.1023/a:1006426917654.
47. Chung WJ, Lyons SA, Nelson GM, Hamza H, Gladson CL, Gillespie GY et al.
Inhibition of cystine uptake disrupts the growth of primary brain tumors. Journal of
Neuroscience. 2005;25(31):7101-10. doi:10.1523/jneurosci.5258-04.2005.
48. Sontheimer H. A role for glutamate in growth and invasion of primary brain
tumors. Journal of neurochemistry. 2008;105(2):287-95. doi:10.1111/j.14714159.2008.05301.x.
49. Dichiro G, Oldfield E, Wright DC, Demichele D, Katz DA, Patronas NJ et al.
Cerebral necrosis after radiotherapy and or intraarterial chemotherapy for brain-
46
tumors - pet and neuropathologic studies. American Journal of Roentgenology.
1988;150(1):189-97.
50. Lieth E, LaNoue KF, Berkich DA, Xu BY, Ratz M, Taylor C et al. Nitrogen
shuttling between neurons and glial cells during glutamate synthesis. Journal of
Neurochemistry. 2001;76(6):1712-23. doi:10.1046/j.1471-4159.2001.00156.x.
51. Ichihara A, Koyama E. Transaminase of branched chain amino acids .i. Branched
chain amino acids-alpha-ketoglutarate transaminase. Journal of Biochemistry.
1966;59(2):160-&.
52. Hutson SM, Sweatt AJ, LaNoue KF. Brached-chain amino acid metabolism:
Implications for establishing safe intakes'. Journal of Nutrition. 2005;135(6):1557S64S.
53. Toenjes M, Barbus S, Park YJ, Wang W, Schlotter M, Lindroth AM et al. BCAT1
promotes cell proliferation through amino acid catabolism in gliomas carrying wildtype IDH1. Nature Medicine. 2013;19(7):901-+. doi:10.1038/nm.3217.
54. Radlwimmer B, Toenjes M, Barbus S, Park YJ, Wang W, Weibrecht I et al.
BCAT1 promotes cell proliferation through amino acid catabolism in gliomas
carrying wild-type IDH1. European Journal of Cancer. 2014;50:S83-S.
55. Gopal K, Grossi E, Paoletti P, Usardi M. Lipid composition of human intracranial
tumors: a biochemical study. Acta neurochirurgica. 1963;11:333-47.
doi:10.1007/bf01402012.
56. Wise DR, Ward PS, Shay JES, Cross JR, Gruber JJ, Sachdeva UM et al. Hypoxia
promotes isocitrate dehydrogenase-dependent carboxylation of alpha-ketoglutarate to
citrate to support cell growth and viability. Proceedings of the National Academy of
Sciences of the United States of America. 2011;108(49):19611-6.
doi:10.1073/pnas.1117773108.
57. Guo D, Prins RM, Dang J, Kuga D, Iwanami A, Soto H et al. EGFR Signaling
Through an Akt-SREBP-1-Dependent, Rapamycin-Resistant Pathway Sensitizes
Glioblastomas to Antilipogenic Therapy. Science Signaling. 2009;2(101).
doi:10.1126/scisignal.2000446.
58. Guo D, Reinitz F, Youssef M, Hong C, Nathanson D, Akhavan D et al. An LXR
Agonist Promotes Glioblastoma Cell Death through Inhibition of an
EGFR/AKT/SREBP-1/LDLR-Dependent Pathway. Cancer Discovery.
2011;1(5):442-56. doi:10.1158/2159-8290.cd-11-0102.
59. Sanai N, Chang S, Berger MS. Low-grade gliomas in adults. Journal of
Neurosurgery. 2011;115(5):948-65. doi:10.3171/2011.7.jns10238.
60. Chandana SR, Movva S, Arora M, Singh T. Primary brain tumors in adults.
American Family Physician. 2008;77(10):1423-30.
47
61. Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A et al. The
2007 WHO classification of tumours of the central nervous system. Acta
neuropathologica. 2007;114(2):97-109. doi:10.1007/s00401-007-0243-4.
62. Ohgaki H, Kleihues P. Population-based studies on incidence, survival rates, and
genetic alterations in astrocytic and oligodendroglial gliomas. Journal of
Neuropathology and Experimental Neurology. 2005;64(6):479-89.
63. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJB et al.
Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. New
England Journal of Medicine. 2005;352(10):987-96.
64. Ohgaki H, Dessen P, Jourde B, Horstmann S, Nishikawa T, Di Patre PL et al.
Genetic pathways to glioblastoma: A population-based study. Cancer Research.
2004;64(19):6892-9. doi:10.1158/0008-5472.can-04-1337.
65. Ohgaki H, Kleihues P. Genetic pathways to primary and secondary glioblastoma.
American Journal of Pathology. 2007;170(5):1445-53.
doi:10.2353/ajpath.2007.070011.
66. Olson JD, Riedel E, DeAngelis LM. Long-term outcome of low-grade
oligodendroglioma and mixed glioma. Neurology. 2000;54(7):1442-8.
67. Puduvalli VK, Hashmi M, McAllister LD, Levin VA, Hess KR, Prados M et al.
Anaplastic oligodendrogliomas: Prognostic factors for tumor recurrence and survival.
Oncology. 2003;65(3):259-66. doi:10.1159/000074479.
68. Lee JH, Sade B, Choi E, Golubic M, Prayson R. Meningothelioma as the
predominant histological subtype of midline skull base and spinal meningioma.
Journal of Neurosurgery. 2006;105(1):60-4. doi:10.3171/jns.2006.105.1.60.
69. Paul Kleinhues WKC. Pathology & Genetics, Tumours of the Nervous System.
1997.
70. Bikmaz K, Mrak R, Al-Mefty O. Management of bone-invasive, hyperostotic
sphenoid wing meningiomas. Journal of Neurosurgery. 2007;107(5):905-12.
doi:10.3171/jns-07/11/0905.
71. van den Bent MJ, Brandes AA, Taphoorn MJB, Kros JM, Kouwenhoven MCM,
Delattre J-Y et al. Adjuvant Procarbazine, Lomustine, and Vincristine Chemotherapy
in Newly Diagnosed Anaplastic Oligodendroglioma: Long-Term Follow-Up of
EORTC Brain Tumor Group Study 26951. Journal of Clinical Oncology.
2013;31(3):344-50. doi:10.1200/jco.2012.43.2229.
72. Simpson D. The recurrence of intracranial meningiomas after surgical treatment.
Journal of Neurology Neurosurgery and Psychiatry. 1957;20(1):22-39.
doi:10.1136/jnnp.20.1.22.
73. Kondziolka D, Lunsford LD, Flickinger JC. The application of stereotactic
radiosurgery to disorders of the brain. Neurosurgery. 2008;62(2):707-19.
doi:10.1227/01.neu.0000296953.45771.66.
48
74. Mendenhall WM, Morris CG, Amdur RJ, Foote KD, Friedman WA. Radiotherapy
alone or after subtotal resection for benign skull base meningiomas. Cancer.
2003;98(7):1473-82. doi:10.1002/cncr.11645.
75. Barbaro NM, Gutin PH, Wilson CB, Sheline GE, Boldrey EB, Wara WM.
Radiation-therapy in the treatment of partially resected meningiomas. Neurosurgery.
1987;20(4):525-8.
76. Milker-Zabel S, Zabel-du Bois A, Huber P, Schlegel W, Debus J. Fractionated
stereotactic radiation therapy in the management of benign cavernous sinus
meningiomas - Long-term experience and review of the literature. Strahlentherapie
Und Onkologie. 2006;182(11):635-40. doi:10.1007/s00066-006-1548-2.
77. Miralbell R, Linggood RM, Delamonte S, Convery K, Munzenrider JE,
Mirimanoff RO. The role of radiotherapy in the treatment of subtotally resected
benign meningiomas. Journal of Neuro-Oncology. 1992;13(2):157-64.
78. NIH. National Cancer Institute Dictionary of Cancer Terms. National Institutes of
Health.
79. Hormigo A, Gu B, Karimi S, Riedel E, Panageas KS, Edgar MA et al. YKL-40
and matrix metalloproteinase-9 as potential serum biomarkers for patients with highgrade gliomas. Clinical Cancer Research. 2006;12(19):5698-704. doi:10.1158/10780432.ccr-06-0181.
80. Becker KP, Yu J. Status Quo-Standard-of-Care Medical and Radiation Therapy
for Glioblastoma. Cancer Journal. 2012;18(1):12-9.
doi:10.1097/PPO.0b013e318244d7eb.
81. Gladson CL, Prayson RA, Liu WM. The Pathobiology of Glioma Tumors. Annual
Review of Pathology-Mechanisms of Disease. 2010;5:33-50. doi:10.1146/annurevpathol-121808-102109.
82. Cairncross JG, Ueki K, Zlatescu MC, Lisle DK, Finkelstein DM, Hammond RR et
al. Specific genetic predictors of chemotherapeutic response and survival in patients
with anaplastic oligodendrogliomas. Journal of the National Cancer Institute.
1998;90(19):1473-9.
83. Kouwenhoven MCM, Kros JM, French PJ, Biemond-ter Stege EM, Graveland
WJ, Taphoorn MJB et al. 1p/19q loss within oligodendroglioma is predictive for
response to first line temozolomide but not to salvage treatment. European Journal of
Cancer. 2006;42(15):2499-503. doi:10.1016/j.ejca.2006.05.021.
84. Brandes AA, Tosoni A, Cavallo G, Reni M, Franceschi E, Bonaldi L et al.
Correlations between O6-methylguanine DNA methyltransferase promoter
methylation status, 1p and 19q deletions, and response to temozolomide in anaplastic
and recurrent oligodendroglioma: A prospective GICNO study. Journal of Clinical
Oncology. 2006;24(29):4746-53. doi:10.1200/jco.2006.06.3891.
49
85. Maher EA, Furnari FB, Bachoo RM, Rowitch DH, Louis DM, Cavenee WK et al.
Malignant glioma: genetics and biology of a grave matter. Genes & Development.
2001;15(11):1311-33. doi:10.1101/gad.891601.
86. Wong AJ, Ruppert JM, Bigner SH, Grzeschik CH, Humphrey PA, Bigner DS et
al. Structural alterations of the epidermal growth-factor receptor gene in human
gliomas. Proceedings of the National Academy of Sciences of the United States of
America. 1992;89(7):2965-9. doi:10.1073/pnas.89.7.2965.
87. Noda S-e, El-Jawahri A, Patel D, Lautenschlaeger T, Siedow M, Chakravarti A.
Molecular Advances of Brain Tumors in Radiation Oncology. Seminars in Radiation
Oncology. 2009;19(3):171-8. doi:10.1016/j.semradonc.2009.02.005.
88. Jeuken JWM, Sprenger SHE, Boerman RH, von Deimling A, Teepen H, van
Overbeeke JJ et al. Subtyping of oligo-astrocytic tumours by comparative genomic
hybridization. Journal of Pathology. 2001;194(1):81-7. doi:10.1002/path.837.
89. Jeuken JWM, Sprenger SHE, Wesseling P, Macville MVE, von Deimling A,
Teepen H et al. Identification of subgroups of high-grade oligodendroglial tumors by
comparative genomic hybridization. Journal of Neuropathology and Experimental
Neurology. 1999;58(6):606-12. doi:10.1097/00005072-199906000-00005.
90. Smith JS, Tachibana I, Passe SM, Huntley BK, Borell TJ, Iturria N et al. PTEN
mutation, EGFR amplification, and outcome in patients with anaplastic astrocytoma
and glioblastoma multiforme. Journal of the National Cancer Institute.
2001;93(16):1246-56. doi:10.1093/jnci/93.16.1246.
91. Yan H, Parsons DW, Jin G, McLendon R, Rasheed BA, Yuan W et al. IDH1 and
IDH2 mutations in gliomas. The New England journal of medicine. 2009;360(8):76573. doi:10.1056/NEJMoa0808710.
92. Jansen M, Yip S, Louis DN. Molecular pathology in adult gliomas: diagnostic,
prognostic, and predictive markers. Lancet Neurology. 2010;9(7):717-26.
doi:10.1016/s1474-4422(10)70105-8.
93. Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N, Weller M et al.
MGMT gene silencing and benefit from temozolomide in glioblastoma. The New
England journal of medicine. 2005;352(10):997-1003. doi:10.1056/NEJMoa043331.
94. Hegi ME, Liu L, Herman JG, Stupp R, Wick W, Weller M et al. Correlation of O6-Methylguanine methyltransferase (MGMT) promoter methylation with clinical
outcomes in glioblastoma and clinical strategies to modulate MGMT activity. Journal
of Clinical Oncology. 2008;26(25):4189-99. doi:10.1200/jco.2007.11.5964.
95. Tsurushima H, Tsuboi K, Yoshii Y, Ohno T, Meguro K, Nose T. Expression of
N-ras gene in gliomas. Neurologia medico-chirurgica. 1996;36(10):704-8.
doi:10.2176/nmc.36.704.
96. Blumenstock M, Prosenc N, Patt S, Pfanne K, Drum F, Cervosnavarro J. In
contrast to egfr gene overexpression, h-ras gene-expression decreases in human
gliomas. Anticancer Research. 1991;11(3):1353-7.
50
97. Arvanitis D, Malliri A, Antoniou D, Linardopoulos S, Field JK, Spandidos DA.
Ras p21 expression in brain tumors: elevated expression in malignant astrocytomas
and glioblastomas multiforme. In vivo (Athens, Greece). 1991;5(4):317-21.
98. Mawrin C, Diete S, Treuheit T, Kropf S, Vorwerk CK, Boltze C et al. Prognostic
relevance of MAPK expression in glioblastoma multiforme. International Journal of
Oncology. 2003;23(3):641-8.
99. Pelloski CE, Lin E, Zhang L, Yung WKA, Colman H, Liu J-L et al. Prognostic
associations of activated mitogen-activated protein kinase and Akt pathways in
glioblastoma. Clinical Cancer Research. 2006;12(13):3935-41. doi:10.1158/10780432.ccr-05-2202.
100. Schmidt MC, Antweiler S, Urban N, Mueller W, Kuklik A, Meyer-Puttlitz B et
al. Impact of genotype and morphology on the prognosis of glioblastoma. Journal of
Neuropathology and Experimental Neurology. 2002;61(4):321-8.
101. Simpson L, Parsons R. PTEN: Life as a tumor suppressor. Experimental Cell
Research. 2001;264(1):29-41. doi:10.1006/excr.2000.5130.
102. Okamoto Y, Di Patre PL, Burkhard C, Horstmann S, Jourde B, Fahey M et al.
Population-based study on incidence, survival rates, and genetic alterations of lowgrade diffuse astrocytomas and oligodendrogliomas. Acta Neuropathologica.
2004;108(1):49-56. doi:10.1007/s00401-004-0861-z.
103. Di Stefano AL, Enciso-Mora V, Marie Y, Desestret V, Labussire M, Boisselier
B et al. Association between glioma susceptibility loci and tumour pathology defines
specific molecular etiologies. Neuro-Oncology. 2013;15(5):542-7.
doi:10.1093/neuonc/nos284.
104. DeMasters BKK, Markham N, Lillehei KO, Shroyer KR. Differential telomerase
expression in human primary intracranial tumors. American Journal of Clinical
Pathology. 1997;107(5):548-54.
105. Hiraga S, Ohnishi T, Izumoto S, Miyahara E, Kanemura Y, Matsumura H et al.
Telomerase activity and alterations in telomere length in human brain tumors. Cancer
Research. 1998;58(10):2117-25.
106. Killela PJ, Reitman ZJ, Jiao Y, Bettegowda C, Agrawal N, Diaz LA, Jr. et al.
TERT promoter mutations occur frequently in gliomas and a subset of tumors derived
from cells with low rates of self-renewal. Proceedings of the National Academy of
Sciences of the United States of America. 2013;110(15):6021-6.
doi:10.1073/pnas.1303607110.
107. Vinagre J, Pinto V, Celestino R, Reis M, Populo H, Boaventura P et al.
Telomerase promoter mutations in cancer: an emerging molecular biomarker?
Virchows Archiv. 2014;465(2):119-33. doi:10.1007/s00428-014-1608-4.
108. Salmaggi A, Eoli M, Frigerio S, Silvani A, Gelati M, Corsini E et al.
Intracavitary VEGF, bFGF, IL-8, IL-12 levels in primary and recurrent malignant
51
glioma. Journal of Neuro-Oncology. 2003;62(3):297-303.
doi:10.1023/a:1023367223575.
109. Plate KH, Breier G, Weich HA, Risau W. Vascular endothelial growth-factor is
a potential tumor angiogenesis factor in human gliomas invivo. Nature.
1992;359(6398):845-8. doi:10.1038/359845a0.
110. Samoto K, Ikezaki K, Ono M, Shono T, Kohno K, Kuwano M et al. Expression
of vascular endothelial growth-factor and its possible relation with neovascularization
in human brain-tumors. Cancer Research. 1995;55(5):1189-93.
111. Ruttledge MH, Sarrazin J, Rangaratnam S, Phelan CM, Twist E, Merel P et al.
Evidence for the complete inactivation of the NF2 gene in the majority of sporadic
meningiomas. Nature Genetics. 1994;6(2):180-4. doi:10.1038/ng0294-180.
112. Lee JH, Sundaram V, Stein DJ, Kinney SE, Stacey DW, Golubic M. Reduced
expression of schwannomin/merlin in human sporadic meningiomas. Neurosurgery.
1997;40(3):578-87. doi:10.1097/00006123-199703000-00031.
113. Eckel-Passow JE, Lachance DH, Molinaro AM, Walsh KM, Decker PA, Sicotte
H et al. Glioma Groups Based on 1p/19q, IDH, and TERT Promoter Mutations in
Tumors. New England Journal of Medicine. 2015;372(26):2499-508.
doi:10.1056/NEJMoa1407279.
114. Derr RL, Ye X, Islas MU, Desideri S, Saudek CD, Grossman SA. Association
Between Hyperglycemia and Survival in Patients With Newly Diagnosed
Glioblastoma. Journal of Clinical Oncology. 2009;27(7):1082-6.
doi:10.1200/jco.2008.19.1098.
115. Seyfried TN, Sanderson TM, El-Abbadi MM, McGowan R, Mukherjee P. Role
of glucose and ketone bodies in the metabolic control of experimental brain cancer.
British Journal of Cancer. 2003;89(7):1375-82. doi:10.1038/sj.bjc.6601269.
116. Nowicki S, Gottlieb E. Oncometabolites: tailoring our genes. Febs Journal.
2015;282(15):2796-805. doi:10.1111/febs.13295.
117. Pollard PJ, Briere JJ, Alam NA, Barwell J, Barclay E, Wortham NC et al.
Accumulation of Krebs cycle intermediates and over-expression of HIF1 alpha in
tumours which result from germline FH and SDH mutations. Human Molecular
Genetics. 2005;14(15):2231-9. doi:10.1093/hmg/ddi227.
118. Dang L, White DW, Gross S, Bennett BD, Bittinger MA, Driggers EM et al.
Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature.
2009;462(7274):739-U52. doi:10.1038/nature08617.
119. Kalinina J, Carroll A, Wang L, Yu Q, Mancheno DE, Wu S et al. Detection of
"oncometabolite" 2-hydroxyglutarate by magnetic resonance analysis as a biomarker
of IDH1/2 mutations in glioma. Journal of Molecular Medicine-Jmm.
2012;90(10):1161-71. doi:10.1007/s00109-012-0888-x.
52
120. Esmaeili M, Vettukattil R, Bathen TF. 2-Hydroxyglutarate as a Magnetic
Resonance Biomarker for Glioma Subtyping. Translational Oncology. 2013;6(2):928. doi:10.1593/tlo.12424.
121. Iwadate Y, Sakaida T, Hiwasa T, Nagai Y, Ishikura H, Takiguchi M et al.
Molecular classification and survival prediction in human gliomas based on proteome
analysis. Cancer Research. 2004;64(7):2496-501. doi:10.1158/0008-5472.can-031254.
122. Wibom C, Sandstrom M, Henriksson R, Antti H, Johansson M, Bergenheim AT.
Vandetanib alters the protein pattern in malignant glioma and normal brain in the
BT4C rat glioma model. International Journal of Oncology. 2010;37(4):879-90.
doi:10.3892/ijo_00000739.
123. Schwartz SA, Weil RJ, Johnson MD, Toms SA, Caprioli RM. Protein profiling
in brain tumors using mass spectrometry: Feasibility of a new technique for the
analysis of protein expression. Clinical Cancer Research. 2004;10(3):981-7.
doi:10.1158/1078-0432.ccr-0927-3.
124. Sreekanthreddy P, Srinivasan H, Kumar DM, Nijaguna MB, Sridevi S, Vrinda M
et al. Identification of Potential Serum Biomarkers of Glioblastoma: Serum
Osteopontin Levels Correlate with Poor Prognosis. Cancer Epidemiology Biomarkers
& Prevention. 2010;19(6):1409-22. doi:10.1158/1055-9965.epi-09-1077.
125. Zhang H, Wu G, Tu H, Huang F. Discovery of serum biomarkers in astrocytoma
by SELDI-TOF MS and proteinchip technology. Journal of Neuro-Oncology.
2007;84(3):315-23. doi:10.1007/s11060-007-9376-5.
126. Petrik V, Saadoun S, Loosemore A, Hobbs J, Opstad KS, Sheldon J et al. Serum
alpha(2)-HS glycoprotein predicts survival in patients with glioblastoma. Clinical
Chemistry. 2008;54(4):713-22. doi:10.1373/clinchem.2007.096792.
127. Abbott NJ, Ronnback L, Hansson E. Astrocyte-endothelial interactions at the
blood-brain barrier. Nature Reviews Neuroscience. 2006;7(1):41-53.
doi:10.1038/nrn1824.
128. Bland DSRaOW. A study of gliomas by the method of tissue culture. The
Journal of Pathology and Bacteriology. 1933:273-83.
129. Ponten J, Macintyr.Eh. Long term culture of normal and neoplastic human glia.
Acta Pathologica Et Microbiologica Scandinavica. 1968;74(4):465-&.
130. Giard DJ, Aaronson SA, Todaro GJ, Arnstein P, Kersey JH, Dosik H et al. Invitro cultivation of human tumors - establishment of cell lines derived from a series of
solid tumors. Journal of the National Cancer Institute. 1973;51(5):1417-23.
131. Osborne CK, Hobbs K, Trent JM. Biological differences among mcf-7 humanbreast cancer cell-lines from different laboratories. Breast Cancer Research and
Treatment. 1987;9(2):111-21. doi:10.1007/bf01807363.
53
132. Holland EC, Celestino J, Dai CK, Schaefer L, Sawaya RE, Fuller GN. Combined
activation of Ras and Akt in neural progenitors induces glioblastoma formation in
mice. Nature Genetics. 2000;25(1):55-7. doi:10.1038/75596.
133. Zhu H, Acquaviva J, Ramachandran P, Boskovitz A, Woolfenden S, Pfannl R et
al. Oncogenic EGFR signaling cooperates with loss of tumor suppressor gene
functions in gliomagenesis. Proceedings of the National Academy of Sciences of the
United States of America. 2009;106(8):2712-6. doi:10.1073/pnas.0813314106.
134. Zheng H, Ying H, Yan H, Kimmelman AC, Hiller DJ, Chen A-J et al. p53 and
Pten control neural and glioma stem/progenitor cell renewal and differentiation.
Nature. 2008;455(7216):1129-U13. doi:10.1038/nature07443.
135. Ding Q, Grammer JR, Nelson MA, Guan JL, Stewart JE, Gladson CL. P27(Kip)
and cyclin D1 are necessary for focal adhesion kinase regulation of cell cycle
progression in glioblastoma cells propagated in vitro and in vivo in the scid mouse
brain. Journal of Biological Chemistry. 2005;280(8):6802-15.
doi:10.1074/jbc.M409180200.
136. Golembieski WA, Ge S, Nelson K, Mikkelsen T, Rempel SA. Increased SPARC
expression promotes U87 glioblastoma invasion in vitro. International Journal of
Developmental Neuroscience. 1999;17(5-6):463-72. doi:10.1016/s07365748(99)00009-x.
137. Weinberger SR, Dalmasso EA, Fung ET. Current achievements using
ProteinChip (R) array technology. Current Opinion in Chemical Biology.
2002;6(1):86-91. doi:10.1016/s1367-5931(01)00282-4.
138. Issaq HJ, Veenstra TD, Conrads TP, Felschow D. The SELDI-TOF MS
approach to proteomics: Protein profiling and biomarker identification. Biochemical
and Biophysical Research Communications. 2002;292(3):587-92.
139. Merchant M, Weinberger SR. Recent advancements in surface-enhanced laser
desorption/ionization-time of flight-mass spectrometry. Electrophoresis.
2000;21(6):1164-77.
140. Hutchens TW, Yip TT. New desorption strategies for the mass-spectrometric
analysis of macromolecules. Rapid Communications in Mass Spectrometry.
1993;7(7):576-80. doi:10.1002/rcm.1290070703.
141. Wei J, Buriak JM, Siuzdak G. Desorption-ionization mass spectrometry on
porous silicon. Nature. 1999;399(6733):243-6.
142. Bonam-rani A, Ternier J, Ratel D, Benabid A-L, Issartel J-P, Brambilla E et al.
Direct-tissue SELDI-TOF mass spectrometry analysis: A new application for clinical
proteomics. Clinical Chemistry. 2006;52(11):2103-6.
143. Schaub S, Wilkins J, Weiler T, Sangster K, Rush D, Nickerson P. Urine protein
profiling with surface-enhanced laser-desorption/ionization time-of-flight mass
spectrometry. Kidney International. 2004;65(1):323-32. doi:10.1111/j.15231755.2004.00352.x.
54
144. Fertin M, Burdese J, Beseme O, Amouyel P, Bauters C, Pinet F. Strategy for
purification and mass spectrometry identification of SELDI peaks corresponding to
low-abundance plasma and serum proteins. Journal of Proteomics. 2011;74(4):42030. doi:10.1016/j.jprot.2010.12.005.
145. Li JN, Orlandi R, White CN, Rosenzweig J, Zhao J, Seregni E et al. Independent
validation of candidate breast cancer serum biomarkers identified by mass
spectrometry. Clinical Chemistry. 2005;51(12):2229-35.
doi:10.1373/clinchem.2005.052878.
146. Malik G, Ward MD, Gupta SK, Trosset MW, Grizzle WE, Adam BL et al.
Serum levels of an isoform of apolipoprotein A-II as a potential marker for prostate
cancer. Clinical Cancer Research. 2005;11(3):1073-85.
147. Carrette O, Demalte I, Scherl A, Yalkinoglu O, Corthals G, Burkhard P et al. A
panel of cerebrospinal fluid potential biomarkers for the diagnosis of Alzheimer's
disease. Proteomics. 2003;3(8):1486-94. doi:10.1002/pmic.200300470.
148. Ward DG, Suggett N, Cheng Y, Wei W, Johnson H, Billingham LJ et al.
Identification of serum biomarkers for colon cancer by proteomic analysis. British
Journal of Cancer. 2006;94(12):1898-905. doi:10.1038/sj.bjc.6603188.
149. Rosty C, Christa L, Kuzdzal S, Baldwin WM, Zahurak ML, Carnot F et al.
Identification of hepatocarcinoma-intestine-pancreas/pancreatitis-associated protein I
as a biomarker for pancreatic ductal adenocarcinoma by protein biochip technology.
Cancer Research. 2002;62(6):1868-75.
150. Ross PL, Huang YLN, Marchese JN, Williamson B, Parker K, Hattan S et al.
Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive
isobaric tagging reagents. Molecular & Cellular Proteomics. 2004;3(12):1154-69.
doi:10.1074/mcp.M400129-MCP200.
151. Choe L, D'Ascenzo M, Relkin NR, Pappin D, Ross P, Williamson B et al. 8-Plex
quantitation of changes in cerebrospinal fluid protein expression in subjects
undergoing intravenous immunoglobulin treatment for Alzheimer's disease.
Proteomics. 2007;7(20):3651-60. doi:10.1002/pmic.200700316.
152. Pasikanti KK, Ho PC, Chan ECY. Gas chromatography/mass spectrometry in
metabolic profiling of biological fluids. Journal of Chromatography B-Analytical
Technologies in the Biomedical and Life Sciences. 2008;871(2):202-11.
doi:10.1016/j.jchromb.2008.04.033.
153. El-Aneed A, Cohen A, Banoub J. Mass Spectrometry, Review of the Basics:
Electrospray, MALDI, and Commonly Used Mass Analyzers. Applied Spectroscopy
Reviews. 2009;44(3):210-30. doi:10.1080/05704920902717872.
154. Schauer N, Steinhauser D, Strelkov S, Schomburg D, Allison G, Moritz T et al.
GC-MS libraries for the rapid identification of metabolites in complex biological
samples. Febs Letters. 2005;579(6):1332-7. doi:10.1016/j.febslet.2005.01.029.
55
155. Fiehn O, Kopka J, Dormann P, Altmann T, Trethewey RN, Willmitzer L.
Metabolite profiling for plant functional genomics. Nature Biotechnology.
2000;18(11):1157-61. doi:10.1038/81137.
156. Jonsson P, Gullberg J, Nordstrom A, Kusano M, Kowalczyk M, Sjostrom M et
al. A strategy for identifying differences in large series of metabolomic samples
analyzed by GC/MS. Analytical Chemistry. 2004;76(6):1738-45.
doi:10.1021/ac0352427.
157. Welthagen W, Shellie RA, Spranger J, Ristow M, Zimmermann R, Fiehn O.
Comprehensive two-dimensional gas chromatography-time-of-flight mass
spectrometry (GC x GC-TOF) for high resolution metabolomics: biomarker discovery
on spleen tissue extracts of obese NZO compared to lean C57BL/6 mice.
Metabolomics. 2005;1(1):65-73. doi:10.1007/s11306-005-1108-2.
158. Salazar C, Armenta JM, Shulaev V. An UPLC-ESI-MS/MS Assay Using 6Aminoquinolyl-N-Hydroxysuccinimidyl Carbamate Derivatization for Targeted
Amino Acid Analysis: Application to Screening of Arabidopsis thaliana Mutants.
Metabolites. 2012;2(3):398-428. doi:10.3390/metabo2030398.
159. Wibom C, Pettersson F, Sjostrom M, Henriksson R, Johansson M, Bergenheim
AT. Protein expression in experimental malignant glioma varies over time and is
altered by radiotherapy treatment. British Journal of Cancer. 2006;94(12):1853-63.
doi:10.1038/sj.bjc.6603190.
160. Jonsson P, Johansson AI, Gullberg J, Trygg J, A J, Grung B et al. Highthroughput data analysis for detecting and identifying differences between samples in
GC/MS-based metabolomic analyses. Analytical Chemistry. 2005;77(17):5635-42.
doi:10.1021/ac050601e.
161. Jonsson P, Johansson ES, Wuolikainen A, Lindberg J, Schuppe-Koistinen I,
Kusano M et al. Predictive metabolite profiling applying hierarchical multivariate
curve resolution to GC-MS datas - A potential tool for multi-parametric diagnosis.
Journal of Proteome Research. 2006;5(6):1407-14. doi:10.1021/pr0600071.
162. Castillo S, Mattila I, Miettinen J, Oresic M, Hyotylainen T. Data Analysis Tool
for Comprehensive Two-Dimensional Gas Chromatography/Time-of-Flight Mass
Spectrometry. Analytical Chemistry. 2011;83(8):3058-67. doi:10.1021/ac103308x.
163. Wold S. Chemometrics; What do we mean with it, and what do we want from it?
Chemometrics and Intelligent Laboratory Systems. 1995;30(1):109-15.
doi:10.1016/0169-7439(95)00042-9.
164. Trygg J, Holmes E, Lundstedt T. Chemometrics in metabonomics. Journal of
Proteome Research. 2007;6(2):469-79. doi:10.1021/pr060594q.
165. Wold S, Esbensen K, Geladi P. Principal component analysis. Chemometrics
and Intelligent Laboratory Systems. 1987;2(1-3):37-52.
166. H W. Nonlinear estimation by iterative least squares procedures. New York
Ed.Wiley1966.
56
167. Trygg J, Wold S. Orthogonal projections to latent structures (O-PLS). Journal of
Chemometrics. 2002;16(3):119-28. doi:10.1002/cem.695.
168. Bylesjo M, Rantalainen M, Cloarec O, Nicholson JK, Holmes E, Trygg J. OPLS
discriminant analysis: combining the strengths of PLS-DA and SIMCA classification.
Journal of Chemometrics. 2006;20(8-10):341-51. doi:10.1002/cem.1006.
169. Wold S. Cross-validatory estimation of number of components in factor and
principal components modelS. Technometrics. 1978;20(4):397-405.
doi:10.2307/1267639.
170. Lasko TA, Bhagwat JG, Zou KH, Ohno-Machado L. The use of receiver
operating characteristic curves in biomedical informatics. Journal of Biomedical
Informatics. 2005;38(5):404-15. doi:10.1016/j.jbi.2005.02.008.
171. Li C-R, Liao C-T, Liu J-P. A non-inferiority test for diagnostic accuracy based
on the paired partial areas under ROC curves. Statistics in Medicine.
2008;27(10):1762-76. doi:10.1002/sim.3121.
172. Eriksson L, Trygg J, Wold S. CV-ANOVA for significance testing of PLS and
OPLS (R) models. Journal of Chemometrics. 2008;22(11-12):594-600.
doi:10.1002/cem.1187.
173. Leon SP, Folkerth RD, Black PM. Microvessel density is a prognostic indicator
for patients with astroglial brain tumors. Cancer. 1996;77(2):362-72.
174. Tabatabai G, Felsberg J, Sabel M, Hofer S, Westphal M, Weller M et al.
Bevacizumab failure in glioblastomas. Journal of Clinical Oncology. 2012;30(15).
175. Claes A, Wesseling P, Jeuken J, Maass C, Heerschap A, Leenders WPJ.
Antiangiogenic compounds interfere with chemotherapy of brain tumors due to vessel
normalization. Molecular Cancer Therapeutics. 2008;7(1):71-8. doi:10.1158/15357163.mct-07-0552.
176. Sandstrom M, Johansson M, Andersson U, Bergh A, Bergenheim AT,
Henriksson R. The tyrosine kinase inhibitor ZD6474 inhibits tumour growth in an
intracerebral rat glioma model. British Journal of Cancer. 2004;91(6):1174-80.
doi:10.1038/sj.bjc.6602108.
177. Wedge SR, Ogilvie DJ, Dukes M, Kendrew J, Chester R, Jackson JA et al.
ZD6474 inhibits vascular endothelial growth factor signaling, angiogenesis, and
tumor growth following oral administration. Cancer Research. 2002;62(16):4645-55.
178. Bergenheim AT, Elfverson J, Gunnarsson PO, Edman K, Hartman M,
Henriksson R. Cytotoxic effect and uptake of estramustine in a rat glioma modeL.
International Journal of Oncology. 1994;5(2):293-9.
179. Johansson M, Bergenheim AT, D'Argy R, Edman K, Gunnarsson PO, Widmark
A et al. Distribution of estramustine in the BT4C rat glioma model. Cancer
Chemotherapy and Pharmacology. 1998;41(4):317-25. doi:10.1007/s002800050745.
57
180. Santagata S, Eberlin LS, Norton I, Calligaris D, Feldman DR, Ide JL et al.
Intraoperative mass spectrometry mapping of an onco-metabolite to guide brain
tumor surgery. Proceedings of the National Academy of Sciences of the United States
of America. 2014;111(30):11121-6. doi:10.1073/pnas.1404724111.
181. Ascierto PA, Scala S, Castello G, Daponte A, Simeone E, Ottaiano A et al.
Pegylated arginine deiminase treatment of patients with metastatic melanoma: Results
from phase I and II studies. Journal of Clinical Oncology. 2005;23(30):7660-8.
doi:10.1200/jco.2005.02.0933.
182. Izzo F, Marra P, Beneduce G, Castello G, Vallone P, De Rosa V et al. Pegylated
arginine deiminase treatment of patients with unresectable hepatocellular carcinoma:
Results from phase I/II studies. Journal of Clinical Oncology. 2004;22(10):1815-22.
doi:10.1200/jco.2004.11.120.
183. Tomlinson BK, Thomson JA, Bomalaski JS, Diaz M, Akande T, Mahaffey N et
al. Phase I trial of arginine deprivation therapy with ADI-PEG 20 plus docetaxel in
patients with advanced malignant solid tumors. Clinical Cancer Research 2015.
184. Caso G, McNurlan MA, McMillan ND, Eremin O, Garlick PJ. Tumour cell
growth in culture: dependence on arginine. Clinical Science. 2004;107(4):371-9.
doi:10.1042/cs20040096.
185. Wheatley DN, Campbell E. Arginine deprivation, growth inhibition and tumour
cell death: 3. Deficient utilisation of citrulline by malignant cells. British Journal of
Cancer. 2003;89(3):573-6. doi:10.1038/sj.bjc.6601134.
186. Syed N, Langer J, Janczar K, Singh P, Lo Nigro C, Lattanzio L et al. Epigenetic
status of argininosuccinate synthetase and argininosuccinate lyase modulates
autophagy and cell death in glioblastoma. Cell Death & Disease. 2013;4.
doi:10.1038/cddis.2012.197.
187. Kim RH, Coates JM, Bowles TL, McNerney GP, Sutcliffe J, Jung JU et al.
Arginine Deiminase as a Novel Therapy for Prostate Cancer Induces Autophagy and
Caspase-Independent Apoptosis. Cancer Research. 2009;69(2):700-8.
doi:10.1158/0008-5472.can-08-3157.
188. Ensor CM, Holtsberg FW, Bomalaski JS, Clark MA. Pegylated arginine
deiminase (ADI-SS PEG(20,000) (mw)) inhibits human melanomas and
hepatocellular carcinomas in vitro and in vivo. Cancer Research. 2002;62(19):544350.
189. Gong H, Zolzer F, von Recklinghausen G, Havers W, Schweigerer L. Arginine
deiminase inhibits proliferation of human leukemia cells more potently than
asparaginase by inducing cell cycle arrest and apoptosis. Leukemia. 2000;14(5):8269. doi:10.1038/sj.leu.2401763.
190. Shen LJ, Beloussow K, Shen WC. Modulation of arginine metabolic pathways
as the potential anti-tumor mechanism of recombinant arginine deiminase. Cancer
Letters. 2006;231(1):30-5. doi:10.1016/j.canlet.2005.01.007.
58
ACKNOWLEDGEMENTS
Jag skulle vilja tacka alla som på något vis bidragit till att detta
avhandlingsarbete har gått att genomföra.Tack!
Först så skulle jag vilja tacka min huvudhandledare Henrik Antti. För att du
trott på mig, alltid uppmuntrat mig till till att våga testa nya saker inom
forskningen samt för att du väglett mig i allt från studiedesign till konsten att
piffa till torra texter. Utan ditt inflytande hade detta blivit en avhandling helt
utan bilder.
Tack även till mina bihandledare Tommy Bergenheim och Mikael Johansson
för att ni gett mig möjligheten att jobba med detta spännande ämne och för att ni
bidragit med era expertiser. Ni har lärt mig mycket.
Underbaraste, bästaste Knatti, utan dig hade det inte blivit någon avhandling.
Tack för du inspirerat, peppat och kommit med idéer och förslag när jag irrat
bort mig bland silylartefakter och röriga signalvägar. Tack för alla rolig upptåg
på labb, alla kvällar med stickning och Gilmore Girls, alla tusen diskussioner om
allt mellan himmel och jord samt allt röj i mosh pits till diverse obskyra band.
Alla nuvarande och forna gruppmedlemmar: Elin Näsström, Elin Chorell, Elin
Thysell, Anna Wuolikainen, Pär Jonsson, Benny Björkblom, Tommy Öman,
Izabella Surowiec och Carl Wibom. Elin Näsström-för att du stått ut att dela
kontor med mig i en herrans massa år och för allt ditt tålamod med mina tusen
frågor om allt från engelsk grammatik till när-var-och-hur det var med diverse
möten och andra tider som jag borde passa. Carl Wibom-för att du introducerade
mig till hela forskningsområdet, lärde mig om SELDI-analyser samt hur man
strukturerar upp sitt mapp-system så att man faktiskt hittar saker.
Alla övriga medarbetar på kemin, speciellt Sofie Knutsson, Tommy Löfstedt,
Mattias Hedenström, Hans Stenlund, Rasmus Madsen och Johan Trygg. Stort tack
till all administrativ och teknisk personal som hjälpt till med allt från datorstrul
till blanketter som jag inte alls förstår mig på. Ett extra tack till Maria Jansson,
Rosita Nilsson och Lars-Göran Adolfsson.
Alla på SMC, där för övrigt allt labb-arbete är gjort. Ett extra stort tack till IngaBritt Carlsson som alltid gjort att det känts roligt att gå utpå labb. Tack även till
Maria Ahnlund och Annika Johansson som hjälpte mig med aminosyra analysen.
Janne Letiö och Maria Pernemalm för att jag fick vara hos er på KI samt för att ni
visade mig hur man kör iTRAQ analys.
59
Nelofer Syed and Julia Langer for a great collaboration. Thank you for showing
me a good time in London.
Konstantinos Kouremenos for showing me how to run LC and GCGC, thank you.
Tack till alla medarbetare på sjukhuset. Speciellt Charlotta Nordström som först
lärde upp mig på labbet. Soma Ghasimi, Kristin Nyman och Mikael Kimdal för all
hjälp med att plocka fram prover. Maria Sandström och Beatrice Melin för att ni
delat med er av er expertis gällande både kliniska och genetiska aspekter när det
kommer till cancer och hjärntumörer. Annika Nordstrand, för att du blev min
vän när jag var alldeles ny och vilsen på onkologen.
Tack alla kära vänner och familj. Björn-för att du tycker att jag är sjukt häftig
som håller på med cancerforskning. Det är bra för mitt ego. Lottinen-för alla
trevliga häng med mycket kaffe och för att du tog dig tid att korrekturläsa min
svenska sammanfattning.
Viktor, min älskade, du är rätt praktisk att ha ibland. Alva, mitt tokiga lilla
bustroll och lillebror/syster som frodas i magen. Tack för att ni finns och
påminner om vad som är viktigast här i livet.
60