Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Mapping transcription factor networks linked to glioblastoma multiform: identifying target genes of the oncogenic transcription factor ZFX in glioblastoma multiforme
(USC Thesis Other)
Mapping transcription factor networks linked to glioblastoma multiform: identifying target genes of the oncogenic transcription factor ZFX in glioblastoma multiforme
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Mapping transcription factor networks linked to glioblastoma multiform:
Identifying target genes of the oncogenic
transcription factor ZFX in glioblastoma multiforme
by
Zexun Wu
A Thesis Presented to the
FACULTY OF THE USC KECK SCHOOL OF MEDICINE UNIVERSITY OF SOUTHERN
CALIFORNIA
In Partial Fulfilment of
the Requirements for the Degree
MASTER OF SCIENCE
(BIOCHEMISTRY AND MOLECULAR MEDICINE)
December 2021
Copyright 2021 Zexun Wu
ii
Acknowledgements
I would like to thank my mentor Dr. Suhn Rhie. She is a great mentor who is willing to
provide patient instructions on her students. Without her help, I cannot switch my career into
bioinformatics smoothly. She is not only a superior scientist who contributes the most passion
into research but also a thoughtful friend who is willing to listen and encourage her students. I
cannot express my thankfulness anymore and I always feel fortunate, proud, and grateful
entering the Rhie lab.
I would like to thank Dr. Peggy Farnham. She has given me lots of advices on my project
and academic presentation. I really appreciate her experienced leadership to not only give the
academic direction on research but also harmonize every lab member during the pandemic
COVID-19 year.
I would like to thank all my lab members: Shannon Schreiner and Stephanie Ni who
performed the ChIP and siRNA experiments in U87, Daniel Mullen who helps me run
TENET2.0 bioinformatics tool, Beoung Hun Lee with whom together I designed the
bioinformatics pipelines, Lauren Han who helps download and integrate TCGA datasets and
perform the survival analysis. And all the previous lab members who contributed to perform the
siRNA and ZFX ChIP experiments in other cell lines.
I would also like to thank my thesis committee members, Dr. Yali Dou and Dr. Dan
Weisenberger who give many advices on my project.
iii
Table of Contents
Acknowledgements ......................................................................................................................... ii
List of Tables ...................................................................................................................................v
List of Figures ................................................................................................................................ vi
List of Abbreviations ................................................................................................................... viii
Abstract ............................................................................................................................................x
CHAPTER 1 INTRODUCTION .....................................................................................................1
1.1 Glioblastoma multiforme .................................................................................................................... 1
1.2 Regulatory elements ............................................................................................................................ 1
1.3 Histone modification and DNA methylation ...................................................................................... 2
1.4 Transcription factors ........................................................................................................................... 4
1.5 Transcription factors dysregulated in GBM ....................................................................................... 4
1.6 ZFX family and its regulation in GBM ............................................................................................... 5
CHAPTER 2 MATERIALS AND METHODS ..............................................................................7
2.1 Expression level of ZFX across TCGA cancer types .......................................................................... 7
2.2 siRNA knockdown and RNA isolation ............................................................................................... 7
2.3 RNA-seq ............................................................................................................................................. 7
2.4 ChIP-seq .............................................................................................................................................. 9
2.5 Motif analysis on GBM-specific ZFX peaks and enhancers ............................................................. 11
2.6 The candidate TFs interacting with ZFX and differentially expressed TFs in TCGA-GBM dataset 11
2.7 TENET2.0 ......................................................................................................................................... 12
CHAPTER 3 Identifying genes regulated by ZFX in GBM ..........................................................14
3.1 The expression of ZFX across different cancer types using TCGA data .......................................... 14
3.2 ZFX binding sites in GBM ................................................................................................................ 15
3.3 Differentially expressed genes upon knockdown of ZFX ................................................................. 17
iv
3.4 Genes directly regulated by ZFX ...................................................................................................... 18
CHAPTER 4 Genes that are specifically regulated by ZFX in GBM ............................................20
4.1 Differential expressed genes upon knockdown of ZFX in multiple cell lines .................................. 20
4.2 Identification of the common and cell-type specific genes regulated by ZFX ................................. 25
4.3 Signaling pathway analysis of genes regulated by ZFX ................................................................... 27
4.4 Identification of cell-cycle genes ...................................................................................................... 28
4.5 Identification of housekeeping genes and cell-type specific genes .................................................. 29
CHAPTER 5 Transcription factors that work with ZFX in GBM ..............................................32
5.1 ZFX binding sites specifically found in GBM .................................................................................. 32
5.2 Motif analysis of GBM-specific ZFX binding sites .......................................................................... 34
5.3 Transcription factors that potentially interact with ZFX ................................................................... 35
CHAPTER 6 Key transcription factors linked to GBM ................................................................38
6.1 Transcription factors expressed in GBM using TCGA ..................................................................... 38
6.2 Identification and motif analysis on GBM-specific enhancers ......................................................... 40
6.3 Key transcription factors liked to GBM identified by TENET2.0 .................................................... 43
CHAPTER 7 DISCUSSION ..........................................................................................................49
7.1 ZFX binding pattern in GBM ............................................................................................................ 49
7.2 Genes that are specifically regulated by ZFX in GBM ..................................................................... 50
7.3 Transcription factors that potentially interact with ZFX in GBM ..................................................... 51
7.4 Candidate oncogenic transcription factors identified by TENET2.0 in GBM .................................. 52
REFERENCES ..............................................................................................................................55
APPENDIX ....................................................................................................................................63
v
List of Tables
Table 1 GSM ID of GBM H3K27ac ChIP-seq data used in this study ........................................ 63
Table 2 GSM ID of normal brain H3K27ac ChIP-seq data used in this study ............................. 64
Table 3 A list of genes regulated by ZFX involved in cell-cycle found common across cell types
....................................................................................................................................................... 64
Table 4 A list of GBM-specific genes regulated by ZFX involved in cell-cycle ......................... 65
Table 5 A list of TFs with more than 200 linked GBM-specific enhancer probes identified by
TENET2.0 ..................................................................................................................................... 67
Table 6 TCGA cancer types abbreviations ................................................................................... 69
vi
List of Figures
Figure 1.1 Transcription factors and gene activation in GBM ....................................................... 5
Figure 2.1 RNA-seq pipeline workflow. ......................................................................................... 8
Figure 2.2 ChIP-seq analysis workflow .......................................................................................... 9
Figure 3.1 The violin plots of gene expression level of ZFX in tumor vs normal samples across
23 cancer types using TCGA RNA-seq data. ............................................................................... 15
Figure 3.2 Characterization of ZFX binding sites in GBM. ......................................................... 17
Figure 3.3 The differential gene expression analysis of ZFX knockdown experiments .............. 18
Figure 3.4 Venn diagram of the percentage of the genes bound by ZFX in downregulated genes
and upregulated genes. .................................................................................................................. 19
Figure 4.1 The volcano plots of siZFX vs siControl in U87, C42B, MCF7 and HEK293T as well
as siBoth (siZFX & siZNF711) vs siControl in MCF7 and HEK293T. ....................................... 23
Figure 4.2 The upset plot of downregulated and upregulated genes upon knockdown of
siZFX/siBoth in four cell types. .................................................................................................... 24
Figure 4.3 The pie chart of the percentage of DEG bound by ZFX in U87, C42B, MCF7 and
HEK293T. ..................................................................................................................................... 25
Figure 4.4 The upset plot of downregulated and upregulated genes upon knockdown of
siZFX/siBoth and bound by ZFX in four cell types. .................................................................... 27
Figure 4.5 The IPA analysis results of U87, C42B, MCF7 and HEK293T. ................................. 28
Figure 4.6 The Venn diagrams of down regulated genes and cell-cycle genes bound by ZFX. .. 29
Figure 4.7 The overlap analysis of non-DEGs among two-cell-type comparison. ....................... 31
vii
Figure 4.8 The number of cell-type specific overexpressed genes. .............................................. 31
Figure 5.1 The total number of ZFX peaks found in HEK293T, U87, HCT116, MCF7, C42B and
PrEC. ............................................................................................................................................. 33
Figure 5.2 The overlap analysis of ZFX peaks among PrEC, C42B, MCF7, HCT116, U87 and
HEK293T. ..................................................................................................................................... 34
Figure 5.3 The top TF motifs enriched in GBM-specific ZFX promoter peaks ......................... 35
Figure 5.4 The scatter plot of the TF that potentially interact with ZFX. ..................................... 36
Figure 6.1 The heatmap of Top50 downregulated transcription factors in GBM ranked by fold
change. .......................................................................................................................................... 39
Figure 6.2 The heatmap of Top50 upregulated TFs in GBM ranked by fold change. .................. 40
Figure 6.3 TF-only volcano plots GBM and normal samples generated using TCGA. ............... 40
Figure 6.4 The workflow to identify GBM-specific enhancers. ................................................... 42
Figure 6.5 The top 10 transcription factors with -log (P value) reported by Homer motif analysis
on GBM-specific enhancers .......................................................................................................... 42
Figure 6.6 The workflow of TENET2.0. ...................................................................................... 44
Figure 6.7 The amount of different types of DNA methylation enhancers probes identified for
GBM using TENET2.0 ................................................................................................................. 45
Figure 6.8 The UCSC genome browser snapshot of the example unmethylated, hypermethylated
and hypomethylated enhancer probes ........................................................................................... 45
Figure 6.9 The frequency of TFs with different number of linked probes ................................... 46
Figure 6.10 The top 10 TF with the most numerous linked enhancer probes .............................. 47
Figure 6.11 Circos plot of HES6 with all the linked enhancer probes. ........................................ 47
Figure 6.12 The heatmap of top 10 TFs identified by TENET2.0 in TCGA-GBM samples. ...... 48
viii
List of Abbreviations
GBM: Glioblastoma Multiforme
TSS: Transcription Start Site
TF: Transcription Factor
RNA-seq: RNA Sequencing
ChIP-seq: Chromatin Immunoprecipitation Sequencing
ATAC-seq: Assay for Transposase-Accessible Chromatin with sequencing
siRNA: Small Interfering RNA
CGI: CpG Islands
TGF-β1: Transforming Growth Factor-β1
GABP: GA-binding protein
EMT: Epithelial to Mesenchymal Transition
TERT: Telomerase Reverse Transcriptase
GSC: Glioblastoma Stem Cells
CTCF: CCCTC-binding Factor
NGS: Next Generation Sequencing
PBC: PCR Bottleneck Coefficient
NSC: Normalized Strand Cross-correlation coefficient
RSC: Relative Strand Cross-correlation coefficient
TENET2.0: Tracing Enhancer Networks using Epigenetic Traits 2.0
GEO: Gene Expression Omnibus
NCBI: National Center for Biotechnology Information
ix
TCGA: The Cancer Genome Atlas
DEG: Differentially Expressed Genes
NSCLC: Non-Small Cell Lung Cancer
IPA: Ingenuity Pathway Analysis
AP1: Activator Protein-1
x
Abstract
Glioblastoma multiforme (GBM) is the most common intrinsic brain tumor in adults and is
universally fatal. My strategy to study molecular mechanisms leading to GBM is to investigate
transcription factors (TFs) that bind to GBM-specific regulatory elements. Among TFs, ZFX has
been shown as a key driver TF for GBM. Several studies have reported the significant role of
ZFX in carcinogenesis, however, the molecular mechanisms of TF ZFX that drive GBM still
remain unknown. Therefore, my project aims to characterize the function of ZFX in GBM by
identifying its binding sites and target genes through ChIP-seq and siRNA followed by RNA-seq
analysis. Besides ZFX, I also integrated other publicly available multi-omic datasets and
performed analysis using in-house developed bioinformatic tool, TENET2.0 to identify other key
TFs linked to GBM. This study provides new insights into understanding of TF networks driving
GBM.
1
CHAPTER 1 INTRODUCTION
1.1 Glioblastoma multiforme
Glioblastoma multiforme (GBM) is the most common intrinsic brain tumors in adults and
universally fatal. The incidence rate of GBM is 2.96 cases per 100,000 population every year in
the United States (Yin et al., 2019). The median survival of GBM patients remains 12-15
months, and the 5 years survival rate of GBM is less than 5% (Yin et al., 2019). Based on
genomic abnormalities, GBM can be classified into four molecular subtypes: classical,
mesenchymal, neural, and proneural subtypes (Yin et al., 2019). Classical molecular subtype
tumors have chromosome 7 amplification paired with chromosome 10 loss (Yin et al., 2019).
The majority of mesenchymal subtype tumors have lower NF1 expression levels (Yin et al.,
2019). Neural subtype is typified by the expression of neuron markers such as NEFL, GABRA1,
SYT1 and SLC12A5 (Yin et al., 2019). Proneural subtype has two major features including the
amplification and mutations of PDGFRA and point mutations in IDH1 (Verhaak et al., 2010).
However, the molecular mechanisms underlying GBM subtypes still remain unclear, and thus the
available therapies to recovery GBM patients are limited. Therefore, it is in urgent need to
identify the promising therapeutic targets and biomarkers of GBM. Epigenetics is the study of
how cells control gene expression beyond DNA sequence. Studying epigenetics can reveal
molecular mechanisms that drive carcinogenesis by providing additional insights into gene
regulation and further facilitate the development of new therapeutic treatments.
1.2 Regulatory elements
2
My strategy to study epigenetics in GBM is to investigate cis-regulatory elements, including
enhancers, promoters, and insulators, which control the expression of genes. Promoter regions
are DNA sequences which are located at upstream, 5’ end of the transcription start site (TSS) of
a gene, and they are defined by RNA polymerase binding. Promoters define the direction of
transcription and indicate which DNA strand will be transcribed. Enhancer regions are located
distal from TSS of their target gene. They can be either upstream or downstream from TSS of
their target gene and stimulate the transcription by binding of transcription factors (TFs). Since
DNA can fold in the nucleus, enhancers may physically interact with promoters located near
TSSs in close proximity although they are located thousands of base pairs away on a linear
chromosome. Additionally, enhancers can locate at either forward or reversed sequence
orientations but to increase gene transcription (Levo et al., 2014). Insulators, which are marked
by CTCF (CCCTC-binding factor), are another regulatory elements that may block the
interaction between adjacent chromatin domains through two different mechanisms (Gaszner et
al., 2006). One is to develop domains that split enhancers and promoters to inhibit the interaction
while another way is to form a barrier against the spread of heterochromatin (Gaszner et al.,
2006).
1.3 Histone modification and DNA methylation
A principle component of chromatin that plays a key role in the regulation of transcription
is histone modification. DNA is wrapped around histones forming the nucleosome core, which
includes two H2A-H2B dimers and a H3-H4 tetramer. Most of histone modifications are
chemical residues added to amino acids near N terminus of the histone. Histone modification
often indicates chromatin states of genomic regions. For example, the enrichment of histone
3
modification H3K4me3 has been shown as a hallmark of promoter (Sharifi-Zarchi et al., 2017).
The histone modification H3K27ac has been shown to mark active enhancers while poised
enhancers contain H3K4me1 alone (Creyghton et al., 2010). Chromatin immunoprecipitation
coupled with sequencing (ChIP-seq) is a commonly used method to identify genomic regions
bound by the DNA-interacted proteins, and thus the histone mark ChIP-seq can be used to
identify promoter and enhancer regions.
DNA methylation is an epigenetic mark representing the covalent transfer of a methyl group
to the C-5 position of the cytosine ring of DNA by DNA methyltransferases. In mammals, most
of the CpG dinucleotides are methylated but the CpG dinucleotides within regulatory elements
tend to be excluded from methylation. CpG islands (CGI) are pivotal genomic regulatory
elements involved in the transcriptional initiation of genes (Bell et al., 2017). CGI located at
active promoters have high CpG density and low DNA methylation level, which allow more
accessible and TFs to bind to initiate transcription (Bell et al., 2017). Not only for promoter
regions, DNA methylation at distal regulatory elements such as enhancers can also regulate gene
expression. High DNA methylation levels at CGI promoters are reported to result in epigenetic
silencing in various tumor samples (Chang et al., 2013). A study of DNA methylation has shown
that enhancer regions are aberrantly methylated in different human cancer cell lines (Heyn et al.,
2016). Furthermore, a cancer epigenome, marked by a global DNA hypomethylation, may cause
abnormal binding of TF and thus result in gene dysregulation (Sharma et al., 2010). Therefore,
the DNA hypomethylation and DNA hypermethylation at CpG sites are commonly considered
the epigenetic cancer hallmarks (Ehrlich, 2002; Sharma et al., 2010). However, which regulatory
elements are differentially methylated in GBM and what TFs bind to those regions is not yet
characterized.
4
1.4 Transcription factors
TFs are DNA-binding proteins involved in the process of transcribing DNA to RNA. TFs
bind to regulatory elements including promoters and enhancers throughout human genome.
Some TFs bind to promoters located near the TSS and help generate the transcription-initiation
complex while other TFs bind to distal regulatory elements such as enhancers. The expression
level and activity of TFs are altered in numerous cancer types (Bushweller, 2019). For example,
Zhang el al. (Zhang et al., 2013) found that the knockdown of KLF8, a downstream TF of
transforming growth factor-β1 (TGF-β1), blocked TGF-β1-promoted cell migration, invasion
and motility in gastric cancer cells. Noura el al. (Morita et al., 2017) found the RUNX1-p53-
CBFB autoregulatory loop which can drive acute myeloid leukemia. Multiple studies have
shown RUNX2 as an essential factor in breast cancer and prostate cancer through driving
epithelial to mesenchymal transition (EMT).
1.5 Transcription factors dysregulated in GBM
There are many TFs that have been reported to be involved in GBM. For example, Bell el
al. (Bell et al., 2015) found that the highly recurrent telomerase reverse transcriptase (TERT)
promoter mutations could abnormally recruit multimeric GA-binding protein (GABP) in GBM,
causing the activation of TERT, which is a gene reactivated in the majority of human cancers
while stay silenced in healthy somatic cells. Wang el al.(Wang et al., 2015) found that FoxM1
plays a critical role in GBM via the interaction with other TFs involved in cell proliferation,
epithelial to mesenchymal transition (EMT), and invasion. Cheng et al.(Cheng et al., 2019)
identified 68 differentially expressed TFs in TCGA GBM data, and LHX2, MEOX2, SNAI2 and
5
ZNF22 are identified from the univariate regression differential gene analysis between tumor and
normal. Moreover, Zhang et al.(Zhang et al., 2018) reported the gain-of-function mutation of
P53 in GBM patients and showed that the de-regulated P53 pathway may function on a set of
genes other than those regulated by wile type P53. However, it is not clear which TFs bind to
regulatory elements to drive GBM carcinogenesis.
Figure 1.1 Transcription factors and gene activation in GBM Shown is the illustration of how transcription factors in GBM can
activate the genes that are silenced in normal.
1.6 ZFX family and its regulation in GBM
Among TFs, there have been studies showing that ZFX is a key driver TF for GBM.
Overexpression of ZFX in GBM associated with poor patient survival (Yan et al., 2016). Zhu et
al. (Zhu Z, 2013) found that the knockdown of ZFX using shRNA in U87 GBM cells can inhibit
the formation of tumor compared with scrambled shRNA-infected cells. Moreover, they showed
that knockdown of ZFX can affect the proliferation and survival of U87 and U373 glioma cells.
Fang et al. (Fang et al., 2014) discovered that ZFX is preferentially expressed in glioblastoma
stem cells (GSCs), and silencing ZFX by shRNA can inhibit GSC self-renewal and tumor
growth. Furthermore, ZFX has been also reported as a critical factor in other cancer types such as
prostate cancer (Zhu et al., 2017), pancreatic cancer (Song et al., 2018) and breast cancer
Distal regulatory elements
(enhancers, insulators)
Promoters
Transcript
Gene
Inactive Inactive
X
No Transcript
Active
Active
Transcript
Distal regulatory elements
(enhancers, insulators)
Promoters
Transcript
Gene
Inactive Inactive
X
No Transcript
Active
Active
Transcript
Promoters Enhancers
Distal regulatory elements
(enhancers, insulators)
Promoters
Transcript
Gene
Inactive Inactive
X
No Transcript
Active
Active
Transcript
Distal regulatory elements
(enhancers, insulators)
Promoters
Transcript
Gene
Inactive Inactive
X
No Transcript
Active
Active
Transcript
Normal
GBM
TF
TF
TF
//
//
TF
Active Active
6
(Pourkeramati et al., 2019). Above studies indicate a potentially significant role of ZFX in
carcinogenesis. However, the role of ZFX in GBM related to gene regulation is not clear.
Under ZFX family, ZFY and ZNF711 are other TFs which have similar structure as ZFX.
ZFY has 96% overall similarity with ZFX and 99% similarity in zinc finger domains while
ZNF711 has 67% overall similarity with ZFX and 87% similarity in zinc finger domain (Ni et al.,
2020). ZFX and ZFY has 13 zinc finger domains at C-terminal while ZNF711 only has 11 ZFs
with disrupted ZF3 and ZF7 domains (Ni et al., 2020). It has been reported that ZNF711 and ZFY
bind to the similar genomic regions as ZFX in prostate cancer cells (22Rv1) (Ni et al., 2020).
Another study shows that ZNF711 has the similar binding pattern with ZFX and binds with a
subset of active promoters under ZFX-bound active promoters, leading to the consideration that
ZNF711 might play a subordinate role to ZFX in breast cancer cells (C42B) (Rhie et al., 2018).
Because of the structure similarity, ZNF711 and ZFY need to be taken into account when the
function of ZFX is characterized. Herein, I aimed to determine the role of ZFX in GBM and
characterize the key TFs networks linked to GBM.
7
CHAPTER 2 MATERIALS AND METHODS
2.1 Expression level of ZFX across TCGA cancer types
In order to measure the expression level of ZFX across TCGA cancer types, RNA-seq
HTseq count files of all 33 TCGA cancer types are downloaded from Genomic Data Commons
Data Portal (GDC) (https://portal.gdc.cancer.gov) and merged into genes count table. Among 33
TCGA cancer types, 23 of them (Seen in Table 6) have the normal samples that can be used to
compare with the tumors. Student t-test is performed between normal and tumor and the p value
is adjusted by Benjamini and Hochberg (BH) procedure.
2.2 siRNA knockdown and RNA isolation
The siRNA experiment is performed to knockdown the expression of ZFX in U87 GBM
cells. Cells were transfected in three replicates with 100nM of siRNA targeting ZFX (Dharmacon
Cat# L006572000005) or control (Dharmacon Cat# D0018101005) using Lipofectamine 3000
(ThermoFisher Cat#L3000015) according to the instructions from manufacturer. Cells are
incubated for 24 hours then lysed in TRI Reagent (Zymo Cat#R2050-1-200) and RNA was
isolated. Total RNA was converted to cDNA using iScript (Bio-Rad Cat#1708841BUN). RT-
qPCR was carried out using SYBR Green (Bio-Rad Cat#1725275) on a BioRad CFX96 machine
(Bio-Rad, Cat#1855196)..
2.3 RNA-seq
8
Figure 2.1 RNA-seq pipeline workflow. Software programs used to analyze RNA-seq are listed in boxes. File formats used for
each software is shown in parentheses. The function of each software is shown at the bottom of the boxes.
RNA-seq analysis is performed using U87 ZFX knockdown samples as well as HEK293T,
C42B, and MCF7 siZFX, siZNF711, and siBoth knockdown samples (Rhie et al., 2018). The
RNA-seq library sequenced files (.fastq) were trimmed by Trimgalore
(https://github.com/FelixKrueger/TrimGalore) to filter out the low quality reads as well as the
adapter sequences. And then, the trimmed library files were aligned onto the human reference
genome version38 (hg38) using the annotation GENCODE version35 (Frankish et al., 2018) by
STAR (Dobin et al., 2012) to get the alignment files (.bam), which contain the mapping
information for each read. To quantify the reads counts for each gene and output a gene counts
table for library file, HTseq (Anders et al., 2015) was used with the bam files as input. DEseq2
(Love et al., 2014), an R package which applies a linear model to calculate P value for each gene
between experimental and control group is used to identify differentially expressed genes
(DEGs). The P values calculated by DEseq2 were adjusted into false discovery rate (FDR) to
correct multiple comparison (Jafari et al., 2019). The DE genes are selected based on false
discovery rate (FDR) and fold change (FDR < 0.05, |Fold Change| > 1.5).
After differential expression analysis, Ingenuity Pathway Analysis (IPA) (QIAGEN lnc.,
https://www.qiagenbio- informatics.com/products/ingenuity-pathway-analysis) is performed to
determine the most relevant pathway of the DEGs with the corresponding FDR. The top 5
functional categories are identified as most enriched biological process pathways. TopGO
(https://bioconductor.org/packages/release/bioc/html/topGO.html), which is an R package to
perform Gene Ontology (GO) analysis, is used to get a list of all the genes that are involved in
9
any relevant GO terms containing the key words including “cell cycle” or “cellular proliferation”
using the setting nodeSize=1.
In order to identify the housekeeping genes and the cell-type specific genes, DE analysis is
performed among U87, C42B, MCF7 and HEK293T cell lines using siControl data. For each DE
analysis, the non-DEGs are identified using adjusted p value ³ 0.05 or |Fold Change| £ 1.5. The
overexpressed genes for two cell types are genes that are upregulated in the cell type that is being
compared and the genes that are downregulated in the cell type that is being compared to, in the
two cell type comparison. As a result, 6 comparisons will generate 6 sets of non-DEGs and 12
sets of overexpressed genes, which means that each cell type has 3 sets of overexpressed genes,
compared to the other three cell types. I defined the housekeeping genes as genes that are
commonly found across 6 sets of non-DEGs, and the cell-type specific genes as genes that are
commonly overexpressed when the individual cell type was compared with the other three cell
types.
2.4 ChIP-seq
Figure 2.2 ChIP-seq analysis workflow The ChIP-seq data is processed through ENCODE ChIP-seq pipeline using bwa as the
aligner and MACS2 method to call the peaks.
ChIP was performed as described in previous studies (Rhie et al., 2018). ChIP experiments
were performed in U87 GBM cell line using ZFX (CST Cat#5419S), CTCF (Active Motif
Cat#3418), H3K4me3 (CST Cat#9751S), and H3K27ac (Active Motif Cat#39133) antibodies.
The ChIP experiment used 30ul antibody to target ZFX in 400ug chromatin, 5ul antibody to
10
target CTCF in 20ug chromatin, 10ul antibody to target H3K4me3 in 10ug chromatin and 4ul
antibody to target H3K27ac in 10ug chromatin.
The ENCODE ChIP-seq pipeline was used to analyze above U87 ChIP-seq data as well as
the ChIP-seq data generated from multiple cell lines including PrEC (prostate epithelial),
HCT116 (colon), HEK293T (kidney), C42B (osteotropic prostate), and MCF7 (breast) in
previous studies (Rhie et al., 2018). PrEC ZFX ChIP-seq only has one replicate while the other
cell lines were performed in two replicates. The ChIP-seq library sequenced files were first
aligned by bwa onto the human reference genome (hg38) (Schneider et al., 2017). To determine
the quality of library, library complexity is computed to output quality metrics. The quality
metrics including PCR Bottleneck Coefficient (PBC), Normalized Strand Cross-correlation
coefficient (NSC), Relative Strand Cross-correlation coefficient (RSC) were computed. PBC is a
measure of library complexity that calculates PCR bias. The very low values of PBC indicate a
technical problem of PCR bias. NSC is a ratio of the maximal cross-correlation value (which
occurs at strand shift equal to fragment length) divided by the background cross-correlation
(minimum cross-correlation value over all possible strand shifts). Higher values of NSC indicate
more enrichment. RSC is the ratio of the ratio of the fragment-length cross-correlation value
minus the background cross-correlation value, divided by the phantom-peak cross-correlation
value minus the background cross-correlation value.
MACS2 (Zhang et al., 2008) is used to call peaks, which indicates areas with the
enrichment of aligned reads in the genome. The input data is required as background to
distinguish and false positive and true peaks. To identify binding sites of TFs, which tend to bind
to short sequence regions, the narrowPeak method is used. After the peak files of all the samples
are generated, they are used to do the following down-stream analysis. After peak calling,
11
Irreproducibility Discovery Rate (IDR) (https://github.com/nboley/idr) is performed to identify
the reproducible ZFX peaks which have the high consistency between two biological samples.
In order to visualize the ChIP-seq tracks in UCSC genome browser, the bwa-aligned bam
files are converted into bedGraph files using the makeTagDirectory and makeUCSCfile
functions in Homer (Heinz et al., 2010). The bedGraph files generate by Homer are then
uploaded onto UCSC genome browser to visualize the ChIP-seq tracks.
2.5 Motif analysis on GBM-specific ZFX peaks and enhancers
To determine the TFs motifs enriched in GBM-specific ZFX peaks, an overlap analysis is
performed using the IDR-identified ZFX peaks of U87, C42B, MCF7, HCT116, PrEC and
HEK293T. The identified U87 GBM-specific ZFX peaks are used for motif analysis using
findMotifsGenome.pl function from Homer (Heinz et al., 2010) to determine the enriched motifs
and calculate the P values with the default settings. The top 10 most enriched motifs within
GBM-specific ZFX peaks were identified (Figure 5.3).
2.6 The candidate TFs interacting with ZFX and differentially
expressed TFs in TCGA-GBM dataset
TCGA-GBM (https://portal.gdc.cancer.gov) database includes 174 samples of RNA-seq
data, consisting of 169 tumor samples and 5 normal samples. To identify the candidate TFs
interacting with ZFX, Pearson correlation analysis was performed using the RNA-seq FPKM UQ
data of 169 tumor samples. The FPKM-UQ data were log transformed before the correlation
analysis. The potential interacting TFs are selected using the Pearson correlation factor r > 0.5.
12
To identify the differentially expressed TFs in TCGA-GBM dataset, 174 RNA-seq HTseq
(Anders et al., 2015) count files are downloaded and integrated into a gene counts table. The
differential gene expression analysis is performed via DEseq2 using FDR < 0.05 and |Fold
Change| > 1.5 within a list of 1639 transcription factors (Lambert et al., 2018). The top 50 most
differentially expressed TFs ranked by their fold change are identified and used to generate
Figure 6.1 and 6.2.
2.7 TENET2.0
Tracing Enhancer Networks using Epigenetic Traits 2.0 (TENET2.0) (Mullen et al., 2020),
a bioinformatic tool designed in-house is used to identify differentially activated enhancers and
their associated transcription factors between GBM tumor and normal tissue samples. The
studies from Encyclopedia of DNA Elements (ENCODE) and the Roadmap Epigenome
Mapping Consortium (REMC) have shown that active enhancers are marked by H3K27ac and
open chromatin regions (Abascal et al., 2020; Kundaje et al., 2015). However, it remains unclear
which enhancers interact with which TFs. Although the interaction between enhancers and TFs
can be determined using the ChIP-seq data of TFs, the ChIP-seq data of all the TFs are not
available because generating the ChIP-seq data for most of TFs is infeasible due to lack of
corresponding antibodies. ChIP-seq is more laborious to perform in tissues than DNA
methylation profiling and RNA-seq because it requires a large amount of cells.
TENET2.0 integrates information from ChIP-seq and open chromatin assays combined
with numerous epigenome and transcriptome datasets to determine different types of enhancers
in normal and tumor tissues, and it uses the DNA methylation levels of probes as an indicator of
enhancer activity. Moreover, by liking enhancer activities with TF gene expression levels, TFs
13
that potentially bind to enhancers can be identified by TENET 2.0. Therefore, TENET2.0 is an
advantaged tool to identify the TFs interacting with active enhancers in tissues because it only
uses DNA methylation and RNA-seq data, which are easier to generate. I performed TENET2.0
to identify the key TFs in GBM with the settings including methcutoff = 0.7, hypocutoff = 0.6,
unmethcutoff = 0.3, hypercutoff = 0.4, minTumor = 5. To perform TENET2.0 on GBM, I
downloaded DNA methylation, RNA-seq, and ATAC-seq data from TCGA-GBM
(https://portal.gdc.cancer.gov). I downloaded H3K27ac ChIP-seq data of 46 primary GBM
samples (Mack et al., 2019) from the Gene Expression Omnibus (GEO); GSE119755 is a GEO
dataset ID, and GSM ID of data used in this study is listed in Table1. H3K27ac ChIP-seq and
ATAC-seq data were used to identify the enhancer regions in GBM while DNA methylation and
RNA-seq data were utilized to measure the activities of the enhancer probes and the TFs linked
to the enhancer probes.
14
CHAPTER 3 Identifying genes regulated by ZFX in GBM
3.1 The expression of ZFX across different cancer types using
TCGA data
ZFX is not only studied in relation with GBM. Actually, it has been implicated in various
human cancers including prostate cancer, breast cancer, colorectal cancer gastric cancer and
renal carcinoma. To measure the expression level of ZFX in other cancer types, I used the gene
expression data of ZFX from 33 TCGA cancer types (See Table 6). However, among 33 cancer
types, 23 of them have the adjacent normal samples and thus I perform student t-test using these
23 cancer types in tumors compared to normal samples (Figure 3.1). The result shows that ZFX
is overexpressed (FDR < 0.05) in Liver Hepatocellular Carcinoma (LIHC) and
Cholangiocarcinoma (CHOL) but underexpressed (FDR < 0.05) in Colon Adenocarcinoma
(COAD), Kidney Renal Papillary Cell Carcinoma (KIRP), Prostate Adenocarcinoma (PRAD),
Thyroid Carcinoma (THCA) and Uterine Corpus Endometrial Carcinoma (UCEC), compared to
corresponding normal samples.
ZFX is not statistically significantly differentially expressed between normal and all GBM
samples, but a subset of tumors has higher expression than normal. Furthermore, the patient
survival analysis shows that high expression of ZFX is associated with poor survival of GBM
neural subtype patients (data not shown), suggesting that ZFX may act as an oncogene in a subset
of GBM samples.
15
Figure 3.1 The violin plots of gene expression level of ZFX in tumor vs normal samples across 23 cancer types using TCGA
RNA-seq data. The Y axis represents the expression level of ZFX. The X axis represents different cancer types (See Table 6 for
the meaning of abbreviations). The statical significance is measured using t-test between normal (shown in green) and tumor
(shown in dark red) samples; * p value <0.05, ** p value < 0.01, *** p value < 0.001, **** p value < 0.0001
3.2 ZFX binding sites in GBM
In order to characterize the binding pattern of ZFX in GBM, ChIP-seq targeting ZFX was
performed in GBM cell line U87. U87, also known as U-87 Malignant Glioblastoma, is an
epithelial cell line isolated in 1996 from a female patient with stage 3 glioblastoma. U87 has
been discovered to have a large number of chromosomal abnormalities. A study about the
genomic features of U87 found a large number of insertions, deletions and translocations within
the protein coding sequences and most of the mutated genes were involved in cellular adhesion
(Clark et al., 2010). Over 500+ genes were found to be homozygously mutated and the
homozygous mutation in PTEN was robustly reported (Clark et al., 2010). The transcriptome
16
analysis also shows that U87 has the mesenchymal property since many of the mesenchymal
marker genes are overexpressed in U87 including COL1A1, COL1A2, TGFB1 and DAB2
(Breznik et al., 2017).
To annotate the genomic location of ZFX binding sites in U87, ChIP-seq experiments
targeting CTCF, histone modifications H3K4me3 and H3K27ac were also performed in the same
cell line. CTCF is a TF with zinc finger domain, and it is considered as the signal of insulators,
DNA sequences that control the interaction of promoters and enhancers (Kubo et al., 2021).
H3K4me3 and H3K27ac are histone modifications that are reported to mark promoter and
enhancer regions, respectively. The ChIP-seq data are analyzed through standard ENCODE3
ChIP-seq bioinformatic pipeline using hg38 reference genome and IDR analysis was performed
to identify the reproducible peaks for each target; 9640 ZFX peaks, 17413 H3K4me3 peaks,
30092 H3K27ac peaks, and 29627 CTCF peaks were identified.
ZFX ChIP-seq peaks are then overlapped with H3K4me3, H3K27ac and CTCF peaks,
representing promoter, enhancer and insulator regions, respectively. Since H3K27ac and CTCF
signals are often enriched at promoter regions, enhancer and insulator regions are defined as
regions with unique H3K27ac and CTCF signals that are distal from promoters, respectively.
When the zoom-out genomic context of H3K4me3 and ZFX ChIP-seq data in U87 GBM
cell line are visualized using the UCSC genome browser (Kent et al., 2002), it was seen that most
of the promoters marked with H3K4me3 signals are bound by ZFX (Figure 3.1A). When the
percentiles of ZFX ChIP-seq peaks that belong to different regulatory elements are calculated
(Figure 3.1B), most of ZFX peaks in the U87 GBM cell line are located at promoters (H3K4me3,
94.5%). A limited number of ZFX ChIP-seq peaks binds to enhancer (H3K27ac, 4.9%), insulator
(CTCF, 0.21%) and others (0.35%) regions that are distal from promoters. This result indicates
17
that ZFX binds to the majority of promoters of genes expressed in GBM. Furthermore, among
the U87 ZFX promoter peaks, 95.5% of them are at CpG island promoter regions, suggesting that
ZFX binds to the CpG island promoters.
Figure 3.2 Characterization of ZFX binding sites in GBM. A) The UCSC genome browser snapshot of chr5q21.3-q34 shows that
most of promoters marked with H3K4me3 are bound by ZFX. B) The percentile of different regulatory elements of ZFX ChIP-
seq peaks in U87 GBM cell line is shown.
3.3 Differentially expressed genes upon knockdown of ZFX
To further identify the genes regulated by ZFX in U87, siRNA experiment was performed to
knockdown ZFX in U87 followed by RNA-seq. RNA-seq is a transcriptome profiling approach
which can examine the expression level of RNA sequences in a sample using next generation
sequencing (NGS). RNA-seq indicates which genes are turned on and what their expression
levels are. As a negative control, siControl samples were prepared by transfecting scrambled
siRNA. The RNA-seq data was analyzed through bioinformatic pipeline (TrimGalore – STAR –
HTseq – DEseq2) (Figure 2.1) to identify the differentially expressed genes.
When siZFX (n=3) and siControl (n=3) samples are compared,1106 genes are found to be
downregulated and 679 genes are found to be upregulated in ZFX knockdown cells (FDR < 0.05,
|Fold Change| > 1.5). Among the top ranked genes, many of them have been reported to drive
cancer. For example, ZMYND11 is a transcriptional corepressor and candidate tumor suppressor
with several histone binding domains (Wen et al., 2014). ZDHHC20 has been shown to function
as an oncogenic protein and the overexpression of ZDHHC20 has been measured in ovarian,
18
breast, kidney, prostate and colon cancer (Draper et al., 2010). NAE1 is an encoding gene that
activate Neddylation, which is a post-translational modification that is overexpressed in multiple
human cancers (Zhou et al., 2019).
Figure 3.3 The differential gene expression analysis of ZFX knockdown experiments A) The volcano plot shows all the
differentially expressed genes labeled in red. B) The green dots in volcano plot represent the DEGs bound by ZFX. Blue labels
indicate top 20 genes ranked by -log2(adjusted P value) in both down and up regulated genes.
3.4 Genes directly regulated by ZFX
Genes that are responsive to changes in the level of ZFX include direct target genes and
indirect target genes of ZFX. One method to identify direct ZFX target genes is to identify DEGs
that are bound by ZFX. Therefore, the genomic coordinates of promoters of all DEGs are used to
overlap with ZFX ChIP-seq peaks in U87. From this analysis, 1,106 genes are found to be
downregulated upon knockdown of ZFX, and 843 of them are bound by ZFX. 679 genes are
upregulated, but only 200 of them are bound by ZFX (Figure 3.2A, green dots). More percentage
(76.22%) of downregulated regulated genes are bound by ZFX than upregulated ones (29.93%),
suggesting that ZFX acts as a transcriptional activator in GBM U87 cell line (Figure 3.3).
19
Figure 3.4 Venn diagram of the percentage of the genes bound by ZFX in downregulated genes and upregulated genes. The blue
part represents the genes bound by ZFX and the orange part represents the genes not bound by ZFX
20
CHAPTER 4 Genes that are specifically regulated by ZFX in
GBM
4.1 Differential expressed genes upon knockdown of ZFX in multiple
cell lines
ZFX has been found to be critical in different human cancers such as prostate cancer, breast
cancer, colorectal cancer, glioma, gastric cancer, and gallbladder adenocarcinoma in addition to
GBM. To learn the mechanism of ZFX in GBM transcription, I integrated GBM U87 siZFX
RNA-seq data (see Chapter 3) with RNA-seq data (Figure4.1), which were generated after
knocking down of ZFX in different cell types that include C42B (prostate cancer), MCF7 (breast
cancer) and HEK293T (embryonic kidney) cells (Rhie et al., 2018). When RNA-seq data was
processed using the same pipeline (Figure 2.1) in U87, 1106 genes are found to be
downregulated and 679 genes are upregulated in siZFX group compared to siControl group. In
C42B, 1355 genes are downregulated and 699 genes are upregulated in siZFX compared to
siControl group. In HEK293T, 351 genes are downregulated and 28 genes are upregulated. In
MCF7, 280 genes are downregulated and 112 genes are upregulated. siZFX has slighter effects
on the global changes of transcriptome in MCF7 and HEK293T than U87 and C42B.
The possible reason that not many genes are differentially expressed upon knockdown of
ZFX in HEK293T and MCF7 cells could be due to the gene expression difference of ZNF711,
one of the C2H2 zinc finger proteins belong to the ZFX family, which has the similar overall
structure especially at zinc finger domains and thus is considered having potentially similar
21
function as ZFX (see 1.6). The expression of ZNF711 in U87 and C42B is too low to be
considered but high in HEK293T and MCF7 cells. Therefore, the siZNF711 along with siZFX
(siBoth) experiments are performed in MCF7 and HEK293T and processed by the same RNA-
seq pipeline. I compared DE analysis results of siZFX vs siControl in U87 and C42B cells with
siZFX and siZNF711 (siBoth) vs siControl data in HEK293T and MCF7 cells (Figure 4.1). In
MCF7 siBoth, 691 gene are downregulated and 819 genes are upregulated. In HEK293T siBoth,
506 genes are downregulated and 99 genes are upregulated. More genes are downregulated in
HEK293T, MCF7, and C42B than upregulated genes upon knockdown while more genes are
upregulated in MCF7.
22
23
Figure 4.1 The volcano plots of siZFX vs siControl in U87, C42B, MCF7 and HEK293T as well as siBoth (siZFX & siZNF711)
vs siControl in MCF7 and HEK293T. The green dots represent the DEG bound by ZFX and the red dot represent the DEG not
bound by ZFX. The grey dots represent the genes that fail to pass the cutoff.
To determine whether the effect brought by siZFX is cell-type (GBM) specific or common,
the overlap analysis is done using the all the downregulated and upregulated genes in four cell
types (Figure 4.2). The overlap analysis shows that 52.8% downregulated genes and 79.82%
upregulated genes are cell-type specific while 19.08% downregulated genes and only 1.48%
upregulated genes are found among three cell types or more, suggesting that the transcriptome
change brought by siZFX is more cell-type specific.
24
Figure 4.2 The upset plot of downregulated and upregulated genes upon knockdown of siZFX/siBoth in four cell types. The X-
axis represents the intersection of DE genes (top: downregulated, bottom: upregulated) found among four cell types. The Y-axis
represents the number of overlapped DE genes.
To identify direct target genes of ZFX in multiple cell lines, ZFX ChIP-seq data generated in
C42B, HEK293T, and MCF7 cells were obtained from previous studies (Rhie et al., 2018)
and
processed through ENCODE ChIP-seq pipeline. A higher percentage of downregulated genes is
bound by ZFX than upregulated genes across four cell types, suggesting that ZFX is acting as a
transcriptional activator, which is not a unique property in U87 (Figure 4.3).
25
Figure 4.3 The pie chart of the percentage of DEG bound by ZFX in U87, C42B, MCF7 and HEK293T. The percentage of DEG
bound by ZFX are calculated downregulated and upregulated groups for U87, C42B, MCF7 and HEK293T.
4.2 Identification of the common and cell-type specific genes
regulated by ZFX
Considering ZFX as a transcriptional activator, ZFX may induce cancer through the
regulation on genes that are downregulated by siZFX and bound by ZFX. To determine whether
genes regulated by ZFX is shared or cell-type specific, I performed an overlap analysis using all
of the down regulated genes bound by ZFX of four cell types. The commonly found genes may
provide the insight into the general function of ZFX while cell-type specific genes can show the
distinct function of ZFX in different cell types.
26
The overlap analysis shows that the genes regulated by ZFX are more cell-type specific,
indicating that ZFX may have distinct functions in each cell type (Figure 4.4). To identify
commonly regulated genes, I selected genes detected in three cell types or above. Those genes
might characterize the general function of ZFX among various cell types. Among the top
downregulated genes upon knockdown of ZFX and bound by ZFX, LRRC41 is commonly found
across four cell types while ANKRD52 and CPNE3 are found in 3 cell types. ANKRD52 is found
in MCF7, HEK293T, and C42B cell lines, and CPNE3 is found in HEK293T, C42B, and U87
cell lines. LRRC41 is a substrate recognition component of Elongin BC-CUL2/5-SOCS-box
protein E3 ubiquitin ligase complex, which modifies the ubiquitination and proteasomal
degradation of proteins (Schenková et al., 2012). ANKRD52, ankyrin repeat domain 52, is
reported to interacts with protein phosphatase 6 to regulate cell mobility (Lee et al., 2021).
Moreover, a previous study showed that the knockdown of ANKRD52 significantly promotes
lung cancer cell mobility, suggesting that ANKRD52 acts as a tumor suppressor in lung cancer
(Lee et al., 2021). CPNE3 is a member of a Ca
2+
-dependent phospholipid-binding protein family
and is reported to be involved in metastasis of non-small cell lung cancer (Lin et al., 2018).
27
Figure 4.4 The upset plot of downregulated and upregulated genes upon knockdown of siZFX/siBoth and bound by ZFX in four
cell types. The X-axis represents the intersection of DE genes and bound by ZFX (top: downregulated, bottom: upregulated)
found among four cell types. The Y-axis represents the number of overlapped DE genes bound by ZFX.
4.3 Signaling pathway analysis of genes regulated by ZFX
To further characterize the genes that are activated by ZFX, I performed ingenuity
pathway analysis (IPA) using all the downregulated genes bound by ZFX with the corresponding
FDR calculated by DE analysis. The pathway analysis results show that the genes regulated by
28
ZFX are involved in cell cycle or other similar categories like cell death and survival, suggesting
that target genes of ZFX are critical in controlling the cellular growth and proliferation (Figure
4.5).
Figure 4.5 The IPA analysis results of U87, C42B, MCF7 and HEK293T. Shown is the top5 biological functional categories,
which potential ZFX target genes are enriched in each cell type. These categories are selected after ranking them by P values.
4.4 Identification of cell-cycle genes
In order to identify all the cell-cycle genes that are downregulated and bound by ZFX in
each cell type, TopGO analysis (Described in 2.2) is performed. Figure 4.6 shows the overlap
analysis of downregulated genes bound by ZFX and downregulated cell-cycle genes bound by
ZFX among different cell types. 23 genes of 157 genes found in 3 cell types or above are
involved in cell cycle process, and 8 of 47 commonly found genes in all four cell types are
involved in cell cycle process. For GBM, 51.70% of U87 ZFX-bound downregulated genes are
cell-type specific, and 22.2% of U87-specific downregulated genes are involved in cell cycle,
suggesting that ZFX regulates distinct sets of genes in different cell types. The common and
GBM-specific genes and the corresponding cell-cycle involvement are listed in Table3 and
Table4.
29
Among the commonly found cell-cycle genes in four cell types, TWSG1, twisted
gastrulation BMP signaling modulator, is known as a direct antagonist of BMP signaling
pathway and thus considered as a tumor suppressor (Yuan et al., 2018). ARNT2, a member of the
basic helix-loop-helix/PER-ARNT-SIM (bHLH/PAS) TF family, has been shown lowly
expressed in gastric cancer tissues and the overexpression of ARNT2 will cause the decreased
cellular proliferation (Jia et al., 2019). Furthermore, ZMYND11, which is one of the top regulated
ZFX-bound genes reported in U87, C42B, and HEK293T, is also identified as a cell-cycle gene.
ZMYND11 is known as a tumor suppressor through the interaction with H3K36me3 on the
genome and its expression level is correlated with tumor cell growth (Wen et al., 2014).
Therefore, ZMYND11 might be a promising candidate gene regulated by ZFX to affect cell-cycle
process.
Figure 4.6 The Venn diagrams of down regulated genes and cell-cycle genes bound by ZFX. The overlap analysis is performed
among all the downregulated ZFX-bound genes (left) and downregulated ZFX-bound cell-cycle genes (right) to identify the
common and cell-type specific genes regulated by ZFX.
4.5 Identification of housekeeping genes and cell-type specific genes
30
To determine how many genes regulated by ZFX are housekeeping or cell-type specific,
two-cell-type comparison DE analysis was performed using siControl RNA-seq data of 4 cell
lines (U87, HEK293T, C42B, and MCF7). For each analysis, non-differentially expressed genes
(non-DEGs) are collected to identify the housekeeping genes and differentially-expressed genes
are collected to identify the cell type specific genes (see Methods 2.3). Figure 4.7 shows the
overlap analysis using non-DEGs from six comparisons (U87 vs HEK293T, U87 vs C42B, U87
vs MCF7, C42B vs MCF7, C42B vs HEK293T, MCF7 vs HEK293T). 6498 genes are not
differentially expressed genes across cell types and these genes are considered as housekeeping
genes. Figure 4.8 shows the cell-type specific genes reported to be overexpressed in one cell type
compared to the other 3 cell types. For example, 3625 genes are U87-specific.
In U87, among all the 11495 genes bound by ZFX in U87, 1277 (11.11%) genes are
housekeeping genes and 1711 (14.88%) are U87-specific genes. 19.7% of housekeeping genes
are bound by ZFX at their promoters while 47.2% of U87-specific genes are bound by ZFX at
their promoters. Moreover, when I investigated the genes regulated by ZFX integrating ChIP-seq
and RNA-seq data, among all the 854 downregulated genes bound by ZFX, 45 (5.27%) genes are
housekeeping genes and 176 (20.61%) genes are U87-specific genes. Among 442 U87-specific
downregulated genes bound by ZFX, 21 (4.75%) genes are housekeeping genes while 119
(26.92%) genes are U87-specific genes. Among 98 U87-specific downregulated genes bound by
ZFX and involved in cell-cycle, 6 (6.12%) genes are housekeeping genes and 20 (20.41%) genes
are U87-specific genes. Taken together, these results suggest that more cell-type specific genes
are regulated by ZFX than housekeeping genes.
31
Figure 4.7 The overlap analysis of non-DEGs among two-cell-type comparison. The non-differentially expressed genes (non-
DEGs) identified in U87 vs HEK293T, U87 vs C42B, U87 vs MCF7, C42B vs MCF7, C42B vs HEK293T, and MCF7 vs
HEK293T are overlapped with each other to identify the housekeeping genes.
Figure 4.8 The number of cell-type specific overexpressed genes. The cell-type specific genes are identified in DE analysis. Each
cell type was compared with the other three cell types and overexpressed genes that were commonly found in each comparison
were considered as cell-type specific genes. For example, 3625 genes are overexpressed in U87 compared to the other 3 cell
types.
32
CHAPTER 5 Transcription factors that work with ZFX in
GBM
5.1 ZFX binding sites specifically found in GBM
To understand the genome-wide ZFX binding patterns among different cancer cell types,
ZFX ChIP-seq analysis was performed using U87 GBM, K562 blood, HCT116 colon, HepG2
liver, C42B prostate, MCF7 breast cancer, HEK293T embryonic kidney cell lines, and PrEC
normal primary prostate epithelial cells. The sequencing data were processed through ENCODE
ChIP-seq pipeline and then IDR analysis was performed to identify the reproducible peaks,
which were then used to do the downstream data analysis. Because a majority of ZFX peaks are
located at promoter regions in U87, ZFX peaks are classified into two categories, promoter peaks
and non-promoter peaks. I found 9931 reproducible peaks in HEK293T, 9242 peaks in U87,
8653 peaks in HCT116, 8870 peaks in MCF7 and 8145 peaks in C42B and 6616 peaks in PrEC
(Figure 5.1). The number of ZFX binding peaks in PrEC was considerably smaller than other cell
types, suggesting that increased ZFX binding pattern is related to carcinogenesis.
The overlap analysis (figure 5.2) shows that a largest number of ZFX promoter peaks
(n=4079) are found common across PrEC, C42B, MCF7, HCT116, U87, and HEK293T. GBM-
specific ZFX promoters peaks (n = 626) are also detected. However, the non-promoter ZFX
peaks are more cell-type specific. I found 433 GBM-specific ZFX non-promoter peaks. 49.1% of
U87 ZFX promoter peaks are commonly shared with other cell types while only 7.6% U87 ZFX
non-promoter peaks are shared with other cell types, suggesting that ZFX peaks are more
33
commonly shared at promoter regions and ZFX peaks at non-promoter regions are more cell-type
specific.
Figure 5.1 The total number of ZFX peaks found in HEK293T, U87, HCT116, MCF7, C42B and PrEC. The numbers of ZFX
peaks classified into the promoter peaks (left) and the non-promoter peaks (right) are shown.
34
Figure 5.2 The overlap analysis of ZFX peaks among PrEC, C42B, MCF7, HCT116, U87 and HEK293T. The ZFX peaks are
classified into promoter peaks (upper) and non-promoter peaks (bottom) and the number of peaks found in six cell-types are
plotted after performing overlap analysis.
5.2 Motif analysis of GBM-specific ZFX binding sites
Each TF binds to specific sequences called motifs. Previous studies identified that ZFX
binds to AGGCCTAG motif (Chen et al., 2008). Using U87 ZFX ChIP-seq data, I performed
motif search and found that AGGCCTRG motif is highly enriched in these sites (31.42%, p = 1e-
115). TF motif analysis can be performed with the genomic regions as the input to identify the
relevant TFs based on known binding motifs of TFs. Therefore, TF motif analysis may identify
the potential TFs that interact with ZFX. Because ZFX binds to promoters throughout the genome
in different cancer cell lines, other TFs that bind to ZFX-bound promoter regions might be an
attractive set to better understand cell-type specific gene regulation of ZFX. Therefore, I searched
motifs enriched in the 626 GBM-specific ZFX promoter peaks (Figure 5.2) using Homer (Heinz
et al., 2010) against the default background sequences. Figure 5.3 shows the top TF motifs
35
enriched in GBM-specific ZFX promoter peaks. I found high enrichment of TGASTCCA, the
motif from Fos family member in GBM-specific ZFX promoter peaks.
Figure 5.3 The top TF motifs enriched in GBM-specific ZFX promoter peaks The unique TF families are shown with the top TF
under the family and the corresponding motifs.
Fosl2 belongs to the Activator Protein-1 (AP1) TF family, which includes different Fos and
Jun genes that have TGASTCCA motif. Fosl2 has been proven as a key factor in TGF-b pathway
and it can facilitate TGF-b-induced migration via the interaction with Smad3 in non-small cell
lung cancer and thus is considered as the potential therapeutic target (Wang et al., 2014). Besides
bZIP TF motif, AP2, and Homeobox motifs are identified as enriched TF motifs for GBM-
specific ZFX promoter peaks. The Rbpj1 with the motif TTTCCCASG, known as a central
transcriptional regulator of NOTCH signaling activity, localizes to active promoters and
enhancers and has been identified as a target therapy in GBM (Xie et al., 2016).
5.3 Transcription factors that potentially interact with ZFX
Previous studies have shown that the expression level of genes are correlated with the
protein interactions (Langfelder et al., 2008; Mullen et al., 2020; Rhie et al., 2016). Therefore, I
used TCGA-GBM RNA-seq (n=169 tumors) data to perform correlation analysis to identify TFs
that potentially interact with ZFX. When I calculated correlation coefficient between ZFX
36
expression level and expression levels of each TF from the list of 1639 TFs, which are defined
from Lambert et al. (Lambert et al., 2018), three TFs, SP3, REST and ZNF217 were identified
using the cutoff of correlation coefficient, r > 0.5 (Figure 5.4).
Figure 5.4 The scatter plot of the TF that potentially interact with ZFX. Pearson correlation analysis identifies three TFs, SP3,
REST and ZNF217, using the cut off r > 0.5.
SP3 is the member of the Sp/Krüppel-like family and it is highly expressed in cancer cell
lines. The knockdown experiment (siRNA) of SP3 shows that it regulates genes involved in cell
proliferation survival and migration (Hedrick et al., 2016). The acetylation-modified SP3 has
been reported as a transcriptional activator in MCF7 (Ammanamanchi et al., 2003). REST is a
member of the Kruppel-type zinc finger TF family and it is known as a transcriptional repressor
which inhibits the neuronal genes in non-neuronal tissues (Huang et al., 2012). A study has
shown that targeting REST for proteasomal degradation might inhibit the growth of GSCs which
are derived from GBM primary tumors (Huang et al., 2012). ZNF217 is an oncogenic TF that
plays critical role in different cancers while its genomic loci 20q13 is a frequently amplified
chromosomal region in human tumors (Cohen et al., 2015). Therefore, ZNF217 is considered as
a promising candidate target for anti-cancer therapies (Cohen et al., 2015).
Although three TFs are detected through Pearson correlation analysis, the reported TFs are
not convincing candidates because their correlation coefficients are barely passed the cutoff (r >
37
0.5). Moreover, these TFs are neither found to be downregulated in RNA-seq of U87 siZFX data
nor identitifed to be enriched from the motif analysis on GBM-specific ZFX ChIP-seq peaks.
38
CHAPTER 6 Key transcription factors linked to GBM
6.1 Transcription factors expressed in GBM using TCGA
To further investigate TF networks linked to GBM, I performed DE analysis using TCGA
GBM RNA-seq data, which has 5 normal samples and 169 tumor samples in total. Next, I used
1,639 TF list (Lambert et al., 2018) to identify differentially expressed TFs between GBM and
normal samples. Among all the 1639 TFs, 1635 of them have the actual gene count value. After
DE analysis, 260 TFs are downregulated while 392 TF are upregulated. Among downregulated
TFs in GBM compared to normal, some of TFs are previously reported to be involved in brain
development (Figure 6.1). For example, MTY1L is a gene expressed in human brain and it is
reported to give an important instruction for brain development (Melhuish et al., 2018). It has
also been shown to inhibit the proliferation of GBM cells (Melhuish et al., 2018). Another
identified TF, NEUROD2 is reported to be involved in neurogenic differentiation in non-
neuronal cells in Xenopus embryos and play a role in the determination and maintenance of
neuronal cell fates (Agrawal et al., 2018). This gene has been reported as a tumor suppressor in
GBM (Agrawal et al., 2018).
On the other hand, upregulated TFs in GBM compared to normal samples included many
HOX genes (Figure 6.2). HOX genes specify the characteristics of “position”, ensuring that the
correct structures are formed at the correct places in the body. A study has shown that the
knockdown of HOX genes caused the suppression of the invasion ability of GBM U-118 and U-
138 cells (Guo et al., 2016).
39
The most statistically significantly upregulated TF was MYBL2 (adjusted P value is
6.39*10
-46
). Besides MYBL2, FOXM1 and CENPA, which are reported to induce cancer are also
significantly upregulated (Liao et al., 2018; Mullen et al., 2020; Musa et al., 2017). Moreover,
E2F TFs such as E2F2, E2F7 and E2F8 are upregulated. Previous studies report that these
factors may not drive cancer because they are activated by cell growth (de Bruin et al., 2003;
Pusapati et al., 2010).
Figure 6.1 The heatmap of Top50 downregulated transcription factors in GBM ranked by fold change. Shown is the heat map of
top 50 TFs downregulated in GBM. Each row represents TFs and each column represents TCGA GBM samples. The upper color
bar indicates whether sample is normal (blue) or tumor (red).
40
Figure 6.2 The heatmap of Top50 upregulated TFs in GBM ranked by fold change. Shown is the heat map of top 50 TFs
upregulated in GBM. Each row represents TFs and each column represents TCGA GBM samples. The upper color bar indicates
whether sample is normal (blue) or tumor (red).
Figure 6.3 TF-only volcano plots GBM and normal samples generated using TCGA. The top 20 TFs with highest values of -
log2(adjusted P-value) are labeled. The red dots represent the differentially expressed TFs while the grey spots represent TFs
which are not differentially expressed; The cutoff FDR < 0.05 and |Fold Change| > 1.5 are used.
6.2 Identification and motif analysis on GBM-specific enhancers
It is reported that enhancer activity is tightly linked to cell-type specificity (Abascal et
al., 2020). Therefore, I hypothesize that TFs driving carcinogenesis bind to GBM-specific
41
enhancers. In order to identify GBM-specific enhancers, multiple next generation sequencing
datasets are integrated. First, I downloaded the H3K27ac ChIP-seq data generated in 46 primary
GBM samples from GEO (GSE119755) and processed using ENCODE3 ChIP-seq pipeline to
call peaks and merged them to identify all H3K27ac signals detected in GBM (n=594427). To
identify enhancer regions, I selected H3K27ac ChIP-seq peaks located at 2KB upstream or
downstream window of TSSs of genes using GENCODE V28 annotation (Frankish et al., 2018)
provided by UCSC table browser(n=76564). Because the previous study has shown that the
nucleosome-depleted-regions (NDR) in enhancers are bound with TFs (Grossman et al., 2018), I
overlapped the GBM-specific enhancers with GBM ATAC-seq peaks file downloaded from
TCGA dataset to more accurately identify the nucleosome-depleted-regions in the enhancers
(n=21840). Furthermore, because these enhancers may include ones active in normal samples, I
downloaded H3K27ac ChIP-seq data generated in six sub-anatomical regions of normal brain
(The GSM IDs of used datasets are listed in Table 2) from RoadMap (Kundaje et al., 2015)
(http://www.roadmapepigenomics.org) and processed using ENCODE3 ChIP-seq pipeline to call
peaks and merged them to identify the H3K27ac signals in normal brain (n=289996). Next, with
the GBM and normal brain enhancer files, I performed overlap analysis and identified the GBM-
specific enhancers (n=3115). As a result, I found 3,115 NDR regions in GBM-specific
enhancers, and I used those regions to perform TF motif analysis to identify TFs that bind to
those enhancers. Figure 6.4 describes the workflow to identify GBM-specific enhancers.
42
Figure 6.4 The workflow to identify GBM-specific enhancers. The H3K27ac ChIP-seq data from GEO and RoadMap, ATAC-seq
from TCGA, TSS from GENCODE V28 are used to identify GBM-specific enhancers.
Top TF motifs enriched in GBM-specific enhancers are TGASTCA, which is from AP-1 TF
family (Figure 6.5). The reported TFs under bZIP family are similar with the motif analysis of
GBM-specific ZFX promoter peaks. Many of reported genes under AP-1 family are found both
and the motif ATGASTCA is enriched at GBM-specific enhancers as well as GBM-specific ZFX
promoter peaks.
Figure 6.5 The top 10 transcription factors with -log (P value) reported by Homer motif analysis on GBM-specific enhancers The
top10 motifs enriched in GBM-specific enhancers are shown. The corresponding motifs of unique TF families are selected.
43
6.3 Key transcription factors liked to GBM identified by TENET2.0
To further identify TFs linked to GBM and map enhancer-TF networks, I used TENET 2.0,
in-house developed bioinformatic tools which correlate TF gene expression levels with DNA
methylation levels of cell-type specific enhancers (Mullen et al., 2020). I performed TENET2.0
using TCGA-GBM RNA-seq data (n=60483) as well as DNA methylation data (n=485577),
TCGA-GBM ATAC-seq data (n=77164), H3K27ac ChIP-seq data (n=594427) generated in 46
primary GBM samples, downloaded from GEO (GSE119755), and H3K27ac ChIP-seq data
(n=289996) generated in six sub-anatomical regions of brain, downloaded from RoadMap (GSM
ID listed in Table1 and Table2). As the first step of TENET 2.0, DNA methylation probes on
enhancer regions were identified by overlapping them with H3K27ac ChIP-seq data and open
chromatin regions. Secondly, the identified enhancer probes are classified as hypomethylated,
hypermethylated, methylated, and unmethylated probes based on their DNA methylation level
between GBM vs normal using TCGA DNA methylation data. Next, different types of enhancer
probes are linked with the expression of genes in TCGA RNA-seq data to identify key TFs with
the largest number of linked enhancer probes (Figure 6.6).
44
Figure 6.6 The workflow of TENET2.0. GBM ATAC-seq from TCGA, H3K27ac ChIP-seq data of 46 primary tumor samples
from GEO and six sub-anatomical regions of normal brain sample from Roadmap, TCGA-GBM DNA methylation and RNA-seq
data are used to perform TENET2.0 analysis to identify key transcription factors activated in GBM.
I found 55644 probes located in enhancers. Among those, 20368 methylated probes, 16817
hypomethylated probes, 9759 hypermethylated probes, and 8700 unmethylated probes are
identified (Figure 6.7). Figure 6.8 shows the example unmethylated, hypomethylated, and
hypermethylated enhancer probes identified by TENET2.0. The unmethylated probe shows an
active enhancer region in both normal and GBM. The hypermethylated probe shows an active
enhancer region found in only normal sample while the hypomethylated enhancer probe displays
marks in only GBM samples. The hypomethylated enhancer probes that are only activated in
GBM samples are my interest, ones located in GBM-specific enhancers. I used these 16817
identified hypomethylated enhancer probes for downstream analysis to identify key TFs linked to
GBM using TENET2.0.
45
Figure 6.7 The amount of different types of DNA methylation enhancers probes identified for GBM using TENET2.0 Methylated
(meth) enhancer probes are methylated in both GBM and normal samples, hypomethylated (Hypometh) enhancer probes are
methylated only in normal samples, hypermethylated (Hypermeth) probes are methylated only in GBM samples, unmethylated
(Unmeth) probes are not methylated in both GBM and normal samples.
Figure 6.8 The UCSC genome browser snapshot of the example unmethylated, hypermethylated and hypomethylated enhancer
probes H3K27ac ChIP-seq signal in hippocampus (normal) and U87 (tumor) cells are shown near example unmethylated
enhancer probe cg10786279 in chr9 q34.3, hypermethylated enhancer probe cg19548763 in chr19 p11.21, and hypomethylated
enhancer probe cg00034336 in chr2 p25.1.
When DNA methylation levels of hypomethylated (GBM-specific) enhancer probes are
correlated with expression levels of TFs, most of TFs are linked to a limited number of DNA
methylation enhancer probes while only a few TFs have the large number of linked enhancer
probes. These TFs are potential TFs that bind to numerous GBM-specific enhancers, driving
carcinogenesis and thus can be the key TFs in GBM. Figure 6.9 shows top 10 TFs that have the
most numerous linked enhancer probes. The other TFs with more than 200 linked enhancer
probes can be checked in Table5. For example, HES6 is linked to 775 enhancer probes
Unmethylated
enhancer probe
Hypermethylated
enhancer probe
Hypomethylated
enhancer probe
HM450 probe
Normal
H3K27ac
Tumor
H3K27ac
46
throughout genome (Figure 6.11). Among the top10 identified TFs, 8 TFs are highly expressed
across GBM samples compared to normal samples according to TCGA-GBM DE analysis results
(Figure 6.12) while expression of 2 TFs (NKX3-1 and ZNF730) are higher in a subset of GBM
samples. Some top TFs including HES6 (Carvalho et al., 2015), NKX3-1 (Gurel et al., 2010) and
SOX11 (Tsang et al., 2020) have been previously reported to drive carcinogenesis. Taken
together, the identification of these key TFs may provide new insights for developing the
promising biomarkers and therapeutic targets of GBM.
Figure 6.9 The frequency of TFs with different number of linked probes Number of hypomethylated (GBM-specific) enhancer
probes linked to each TF is calculated using TENET 2.0, and the frequency of TFs is generated.
47
Figure 6.10 The top 10 TF with the most numerous linked enhancer probes
Top 10 TFs with the most numerous linked hypomethylated (GBM-specific) enhancer probes identified by TENET 2.0 is shown.
Figure 6.11 Circos plot of HES6 with all the linked enhancer probes. HES6, identified by TENET2.0, has 775 linked enhancer
probes in GBM.
48
Figure 6.12 The heatmap of top 10 TFs identified by TENET2.0 in TCGA-GBM samples. The expression level of top 10 TFs
across 174 TCGA-GBM samples are shown.
49
CHAPTER 7 DISCUSSION
7.1 ZFX binding pattern in GBM
To characterize the ZFX binding pattern in GBM, ChIP-seq analysis is performed in GBM
U87 cell line targeting ZFX as well as other histone modifications including H3K4me3,
H3K27ac and CTCF. These different histone marks are used to identify the regulatory elements
in the genome. H3K4me3, H3K27ac and CTCF are considered as the chromatin hallmarks for
promoters, enhancers and insulators, respectively. The percentile of ZFX peaks in regulatory
elements shows that ZFX binds to promoters mostly. This result is consistent with our previous
study about ZFX binding pattern in different cancer cell types (Rhie et al., 2018). Analysis of
genomic distribution of ZFX ChIP-seq peaks showed that 80% of the ZFX-binding sites are
located in promoter region, among which the most are in CpG island promoters (Rhie et al.,
2018). The ZFX binding site in promoter regions are largely shared among four cell types while
the distal site bound by ZFX are not always the same in different cell types (Rhie et al., 2018).
To further identify the unique binding sites of ZFX in GBM, ChIP-seq targeting ZFX is also
performed in HEK293T embryonic kidney, HCT116 colon cancer, C42B prostate cancer, MCF-7
breast cancer cell lines and the normal prostate PrEC cell line. ZFX ChIP-seq peaks in human
normal prostate epithelial cells (PrEC) were considerable smaller than other cancer cell types
(U87, HCT116, C42B, MCF-7), suggesting that ZFX binding is associated with carcinogenesis.
From the overlap analysis of ZFX ChIP-seq data, I identified 626 GBM-specific ZFX peaks at
promoter regions. Compared with other cell types, it is more than MCF7 (n = 387), PrEC (n =
257), C42B (n = 250), and HCT116(n =24) specific ZFX peaks but less than HEK293T (n = 752)
50
specific peaks. HEK293T has shared ZFX peaks more common with ZFX peaks in U87 since
over 80.3% of U87 ZFX peaks are shared with HEK293T. Noticeably, these GBM-specific peaks
have significantly (p=2.2e-16) lower signal value (mean signalValue = 6.97) than the other ZFX
peaks (mean signalValue = 13.33). The GBM-specific peaks are also narrower than the other
peaks (1191bp < 1405bp, p = 9.14e-15).
7.2 Genes that are specifically regulated by ZFX in GBM
The RNA-seq analysis of siZFX vs siControl in U87 GBM cell line revealed thousands of
DEGs. Combined with ZFX ChIP-seq data, I found that a higher percentage of downregulated
genes are bound by ZFX than upregulated genes in GBM. This result suggests the role of ZFX as
a transcriptional activator in the genome. However, this activator function of ZFX is not a unique
property in GBM because same pattern are found in other cancer cell lines including C42B,
MCF7 and HEK293T (Rhie et al., 2018), suggesting the general function of ZFX to activate
genes. The overlap analysis on downregulated bound by ZFX suggests that the sets of genes
regulated by ZFX are more cell-type specific. The housekeeping and U87-specific gene analyses
suggest that more U87-specific genes are regulated by ZFX than housekeeping genes.
Based on the consideration that ZFX regulates transcriptome by activating genes, the
downregulated genes bound by ZFX are put into IPA analysis to identify signaling pathways, in
which genes regulated by ZFX are involved. The results show that the genes activated by ZFX
are involved in cell cycle or other relevant categories such as cell death and survival or cellular
growth and proliferation, indicating the molecular mechanism through which ZFX may promote
GBM. The overlap analysis on cell-cycle genes regulated by ZFX still shows a cell-type specific
trend.
51
In U87 siZFX data, among the 854 downregulated genes bound by ZFX, 47 genes are bound
by the 626 GBM-specific ZFX peaks and 10 of them (DLC1, GLI2, CHST11, WWTR1, SPRY2,
GDNF, CEND1, CSPG4, EML1, HAS2) are involved in cell-cycle, reported by topGO analysis.
Among these 10 genes, GLI2 (Tanigawa et al., 2021), WWTR1 (also known as TAZ) (Yuan et
al., 2015), SPRY2 (Park et al., 2018), GDNF (Chen et al., 2018), CEND1 (Mukherjee, 2020),
CSPG4 (Pellegatta et al., 2018), HAS2 (Lim et al., 2017) have been shown to drive GBM
carcinogenesis.
When I studied genes regulated by ZFX across cell types, 23 of 157 commonly found genes
in three cell types or more are involved in cell-cycle process. Among these 23 genes, ZMYND11
is a promising candidate because it is detected as a top downregulated gene bound by ZFX with
extremely low P value in U87, C42B, HEK293T (not a DEG in MCF7). Previous study has
shown ZMYND11 as a tumor suppressor of GBM (Yang et al., 2017). However, more work is
needed to characterize the biological process on how ZFX gets involved in cell cycle and cellular
proliferation.
7.3 Transcription factors that potentially interact with ZFX in GBM
In order to identify the potential TFs interacting with ZFX in GBM, correlation analysis
using TCGA-GBM RNA-seq data and the motif analysis on GBM-specific ZFX ChIP-seq peaks
are performed.
The correlation study found that three TFs (SP3, REST and ZNF217) passed the cutoff
0.5 of r, correlation coefficient. Although previous studies have proven the relevance of these
TFs in carcinogenesis, whether they are biologically interacting with ZFX is not clear. First
reason is that Pearson correlation factor of these TFs are not high but just pass the 0.5 cutoff.
52
Secondly, the standard deviation of the expression level of ZFX in TCGA-GBM tumor samples
is quite lower than the other DEGs. Among all the 1639 TFs, the standard deviation of ZFX
ranked at 1400 (sd = 0.4459). Lastly, none of these three reported TFs are found in either DEGs
of RNA-seq U87 siZFX data or motif analysis report. Therefore, the three TFs reported by
correlation analysis might not be promising candidates that are interacting with ZFX.
From the motif analysis, AP-1 TF family motif TGASTCCA is found to be enriched in
both GBM-specific ZFX promoter peaks and GBM-specific enhancers. Most of these TFs like
Fos, Fra1, Fra2 or JunB belong to AP-1 family, which is known as the first TF family critical in
cellular growth and proliferation (Garces de Los Fayos Alonso et al., 2018). This is interesting
because two independent analysis, using totally different source of data (in-house ZFX ChIP-seq
data vs publicly available H3K27ac ChIP-seq data) to target two regulatory elements in genome,
give the similar result. AP-1 family TFs are not highly enriched in MCF7, HEK293T and C42B
ZFX promoter peaks. Two studies have shown that AP-1 family TFs get involved in GBM
carcinogenesis through the interaction with Interleukin 13 receptor alpha 2 (IL-13Rα2), which is
overexpressed in over 80% GBM samples (Bhardwaj et al., 2018; Wu et al., 2010). Another
study has shown the correlation between Neurofibromatosis type1 genes (NF1) signaling and
mesenchymal GBM sub cell type through FOSL1, a member of AP-1 family (Marques et al.,
2019). Furthermore, a study shows that the migration and invasion of glioblastoma is induced via
NF-kB/AP-1- mediated IL-8 regulation (Ahn et al., 2016). Therefore, AP-1 family may be
critical in GBM for its interaction with other genes to induce glioblastoma growth and migration.
7.4 Candidate oncogenic transcription factors identified by TENET2.0 in GBM
53
Among the top 10 key TFs linked to GBM using TENET2.0, 8 of them (HES6, ASCL1,
SOX2, ZNF730, SOX11, VAX2, SALL3, FOXD1) are detected as upregulated in TCGA-GBM
DE analysis, indicating that these TFs are highly expressed across GBM samples unlike NKX3-1
and ZNF730. Interestingly, ASCL1 and SOX2 are reported to be enriched in GBM-specific
enhancer motif analysis (Figure 6.5). ASCL1 is under bHLH TF family with the motif
CAGCTG. A previous study has shown that ASCL1 is the gene classifier for the proneural GBM
subtype and it can repress the molecular features of mesenchymal GBM subtype and thus is a
critical biomarker for GBM (Narayanan et al., 2019). A previous ASCL-1 ChIP-seq study has
shown that ASCL-1 binds with enhancer regions by opening closed chromatin to activate
neurogenic genes (Park et al., 2017). SOX2 is also identified as a therapeutic target in GBM for
its overexpression in glioblastoma, and the increased level of SOX2 is associated with poor
patient outcome (Garros-Regulez et al., 2016). The activity of SOX2 has been correlated with the
undifferentiated state of cancer stem cells in brain tissue (Garros-Regulez et al., 2016). A ChIP-
seq study of SOX2 has also shown that most of SOX2 binding signals in neural stem cells are in
intronic and intergenic regions labeled with H3K27ac marks, suggesting that SOX2 also binds
with enhancer regions (Bertolini et al., 2019). Therefore, these two TFs identified by TENET2.0
are potential key TFs driving GBM, regulating enhancer networks.
Besides the above two TFs, other key TFs reported by TENET2.0 have been also found in
GBM carcinogenesis. HES6 was linked to the most numerous GBM-specific enhancers. HES6 is
a HES family member, and it is known to inhibit Notch signaling pathway which has been
identified in the pathogenesis of various cancer types (Carvalho et al., 2015) and recognized as
an important regulator of glioma proliferation (Haapa-Paananen et al., 2012). SOX11 is a
member of the SOXC groups and it is expressed during embryogenesis period while lost its
54
expression in adult differentiated tissues (Tsang et al., 2020). It is shown that SOX11 is involved
in brain differentiation and prevents tumorigenesis of glioma by the inhibition of the oncogene
plagl1 (Hide et al., 2009). A study shows that SOX11 controls chromatin accessibility by
affecting lineage-specific enhancers (Decaesteker et al., 2020). In summary, this analysis
pinpointed key TFs linked to GBM involved in enhancer networks, and it provides new insight
into better understanding of GBM carcinogenesis.
55
REFERENCES
Abascal, F., Acosta, R., Addleman, N. J., Adrian, J., Afzal, V., Aken, B., . . . The, E. P. C.
(2020). Perspectives on ENCODE. Nature, 583(7818), 693-698. doi:10.1038/s41586-020-2449-8
Agrawal, R., Garg, A., Benny Malgulwar, P., Sharma, V., Sarkar, C., & Kulshreshtha, R. (2018).
p53 and miR-210 regulated NeuroD2, a neuronal basic helix-loop-helix transcription factor, is
downregulated in glioblastoma patients and functions as a tumor suppressor under hypoxic
microenvironment. Int J Cancer, 142(9), 1817-1828. doi:10.1002/ijc.31209
Ahn, S.-H., Park, H., Ahn, Y.-H., Kim, S., Cho, M.-S., Kang, J. L., & Choi, Y.-H. (2016).
Necrotic cells influence migration and invasion of glioblastoma via NF-κB/AP-1-mediated IL-8
regulation. Scientific Reports, 6(1), 24552. doi:10.1038/srep24552
Ammanamanchi, S., Freeman, J. W., & Brattain, M. G. (2003). Acetylated sp3 is a
transcriptional activator. J Biol Chem, 278(37), 35775-35780. doi:10.1074/jbc.M305961200
Anders, S., Pyl, P. T., & Huber, W. (2015). HTSeq--a Python framework to work with high-
throughput sequencing data. Bioinformatics, 31(2), 166-169. doi:10.1093/bioinformatics/btu638
Bell, J. S. K., & Vertino, P. M. (2017). Orphan CpG islands define a novel class of highly active
enhancers. Epigenetics, 12(6), 449-464. doi:10.1080/15592294.2017.1297910
Bell, R. J., Rube, H. T., Kreig, A., Mancini, A., Fouse, S. D., Nagarajan, R. P., . . . Costello, J. F.
(2015). Cancer. The transcription factor GABP selectively binds and activates the mutant TERT
promoter in cancer. Science, 348(6238), 1036-1039. doi:10.1126/science.aab0015
Bertolini, J. A., Favaro, R., Zhu, Y., Pagin, M., Ngan, C. Y., Wong, C. H., . . . Wei, C. L. (2019).
Mapping the Global Chromatin Connectivity Network for Sox2 Function in Neural Stem Cell
Maintenance. Cell Stem Cell, 24(3), 462-476.e466. doi:10.1016/j.stem.2019.02.004
Bhardwaj, R., Suzuki, A., Leland, P., Joshi, B. H., & Puri, R. K. (2018). Identification of a novel
role of IL-13Rα2 in human Glioblastoma multiforme: interleukin-13 mediates signal
transduction through AP-1 pathway. Journal of Translational Medicine, 16(1), 369.
doi:10.1186/s12967-018-1746-6
Breznik, B., Motaln, H., Vittori, M., Rotter, A., & Lah Turnšek, T. (2017). Mesenchymal stem
cells differentially affect the invasion of distinct glioblastoma cell lines. Oncotarget, 8(15),
25482-25499. doi:10.18632/oncotarget.16041
Bushweller, J. H. (2019). Targeting transcription factors in cancer — from undruggable to
reality. Nature Reviews Cancer, 19(11), 611-624. doi:10.1038/s41568-019-0196-7
56
Carvalho, F. L., Marchionni, L., Gupta, A., Kummangal, B. A., Schaeffer, E. M., Ross, A. E., &
Berman, D. M. (2015). HES6 promotes prostate cancer aggressiveness independently of Notch
signalling. J Cell Mol Med, 19(7), 1624-1636. doi:10.1111/jcmm.12537
Chang, K., Creighton, C. J., Davis, C., Donehower, L., Drummond, J., Wheeler, D., . . .
Biospecimen Core Resource, C. (2013). The Cancer Genome Atlas Pan-Cancer analysis project.
Nature Genetics, 45(10), 1113-1120. doi:10.1038/ng.2764
Chen, M., Ba, H., Lu, C., Dai, J., & Sun, J. (2018). Glial Cell Line-Derived Neurotrophic Factor
(GDNF) Promotes Angiogenesis through the Demethylation of the Fibromodulin (FMOD)
Promoter in Glioblastoma. Med Sci Monit, 24, 6137-6143. doi:10.12659/msm.911669
Chen, X., Xu, H., Yuan, P., Fang, F., Huss, M., Vega, V. B., . . . Ng, H. H. (2008). Integration of
external signaling pathways with the core transcriptional network in embryonic stem cells. Cell,
133(6), 1106-1117. doi:10.1016/j.cell.2008.04.043
Cheng, Q., Huang, C., Cao, H., Lin, J., Gong, X., Li, J., . . . Huang, J. (2019). A Novel
Prognostic Signature of Transcription Factors for the Prediction in Patients With GBM. 10(906).
doi:10.3389/fgene.2019.00906
Clark, M. J., Homer, N., O'Connor, B. D., Chen, Z., Eskin, A., Lee, H., . . . Nelson, S. F. (2010).
U87MG decoded: the genomic sequence of a cytogenetically aberrant human cancer cell line.
PLoS Genet, 6(1), e1000832. doi:10.1371/journal.pgen.1000832
Cohen, P. A., Donini, C. F., Nguyen, N. T., Lincet, H., & Vendrell, J. A. (2015). The dark side of
ZNF217, a key regulator of tumorigenesis with powerful biomarker value. Oncotarget, 6(39),
41566-41581. doi:10.18632/oncotarget.5893
Creyghton, M. P., Cheng, A. W., Welstead, G. G., Kooistra, T., Carey, B. W., Steine, E. J., . . .
Jaenisch, R. (2010). Histone H3K27ac separates active from poised enhancers and predicts
developmental state. Proceedings of the National Academy of Sciences, 107(50), 21931.
doi:10.1073/pnas.1016071107
de Bruin, A., Maiti, B., Jakoi, L., Timmers, C., Buerki, R., & Leone, G. (2003). Identification
and characterization of E2F7, a novel mammalian E2F family member capable of blocking
cellular proliferation. J Biol Chem, 278(43), 42041-42049. doi:10.1074/jbc.M308105200
Decaesteker, B., Louwagie, A., Loontiens, S., De Vloed, F., Roels, J., Vanhauwaert, S., . . .
Speleman, F. (2020). SOX11 is a lineage-dependency factor and master epigenetic regulator in
neuroblastoma. bioRxiv, 2020.2008.2021.261131. doi:10.1101/2020.08.21.261131
Dobin, A., Davis, C. A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., . . . Gingeras, T. R.
(2012). STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 29(1), 15-21.
doi:10.1093/bioinformatics/bts635 %J Bioinformatics
Draper, J. M., & Smith, C. D. (2010). DHHC20: a human palmitoyl acyltransferase that causes
cellular transformation. Mol Membr Biol, 27(2-3), 123-136. doi:10.3109/09687681003616854
57
Ehrlich, M. (2002). DNA methylation in cancer: too much, but also too little. Oncogene, 21(35),
5400-5413. doi:10.1038/sj.onc.1205651
Fang, X., Huang, Z., Zhou, W., Wu, Q., Sloan, A. E., Ouyang, G., . . . Bao, S. (2014). The zinc
finger transcription factor ZFX is required for maintaining the tumorigenic potential of
glioblastoma stem cells. Stem Cells, 32(8), 2033-2047. doi:10.1002/stem.1730
Frankish, A., Diekhans, M., Ferreira, A.-M., Johnson, R., Jungreis, I., Loveland, J., . . . Flicek, P.
(2018). GENCODE reference annotation for the human and mouse genomes. Nucleic Acids
Research, 47(D1), D766-D773. doi:10.1093/nar/gky955
Garces de Los Fayos Alonso, I., Liang, H. C., Turner, S. D., Lagger, S., Merkel, O., & Kenner,
L. (2018). The Role of Activator Protein-1 (AP-1) Family Members in CD30-Positive
Lymphomas. Cancers (Basel), 10(4). doi:10.3390/cancers10040093
Garros-Regulez, L., Garcia, I., Carrasco-Garcia, E., Lantero, A., Aldaz, P., Moreno-Cugnon,
L., . . . Matheu, A. (2016). Targeting SOX2 as a Therapeutic Strategy in Glioblastoma. Front
Oncol, 6, 222. doi:10.3389/fonc.2016.00222
Gaszner, M., & Felsenfeld, G. (2006). Insulators: exploiting transcriptional and epigenetic
mechanisms. Nat Rev Genet, 7(9), 703-713. doi:10.1038/nrg1925
Grossman, S. R., Engreitz, J., Ray, J. P., Nguyen, T. H., Hacohen, N., & Lander, E. S. (2018).
Positional specificity of different transcription factor classes within enhancers. Proceedings of
the National Academy of Sciences, 115(30), E7222. doi:10.1073/pnas.1804663115
Guo, Y. B., Shao, Y. M., Chen, J., Xu, S. B., Zhang, X. D., Wang, M. R., & Liu, H. Y. (2016).
Effect of overexpression of HOX genes on its invasive tendency in cerebral glioma. Oncol Lett,
11(1), 75-80. doi:10.3892/ol.2015.3893
Gurel, B., Ali, T. Z., Montgomery, E. A., Begum, S., Hicks, J., Goggins, M., . . . De Marzo, A.
M. (2010). NKX3.1 as a marker of prostatic origin in metastatic tumors. Am J Surg Pathol,
34(8), 1097-1105. doi:10.1097/PAS.0b013e3181e6cbf3
Haapa-Paananen, S., Kiviluoto, S., Waltari, M., Puputti, M., Mpindi, J. P., Kohonen, P., . . .
Kallioniemi, O. (2012). HES6 gene is selectively overexpressed in glioma and represents an
important transcriptional regulator of glioma proliferation. Oncogene, 31(10), 1299-1310.
doi:10.1038/onc.2011.316
Hedrick, E., Cheng, Y., Jin, U. H., Kim, K., & Safe, S. (2016). Specificity protein (Sp)
transcription factors Sp1, Sp3 and Sp4 are non-oncogene addiction genes in cancer cells.
Oncotarget, 7(16), 22245-22256. doi:10.18632/oncotarget.7925
Heinz, S., Benner, C., Spann, N., Bertolino, E., Lin, Y. C., Laslo, P., . . . Glass, C. K. (2010).
Simple combinations of lineage-determining transcription factors prime cis-regulatory elements
required for macrophage and B cell identities. Mol Cell, 38(4), 576-589.
doi:10.1016/j.molcel.2010.05.004
58
Heyn, H., Vidal, E., Ferreira, H. J., Vizoso, M., Sayols, S., Gomez, A., . . . Esteller, M. (2016).
Epigenomic analysis detects aberrant super-enhancer DNA methylation in human cancer.
Genome Biol, 17, 11. doi:10.1186/s13059-016-0879-2
Hide, T., Takezaki, T., Nakatani, Y., Nakamura, H., Kuratsu, J., & Kondo, T. (2009). Sox11
prevents tumorigenesis of glioma-initiating cells by inducing neuronal differentiation. Cancer
Res, 69(20), 7953-7959. doi:10.1158/0008-5472.Can-09-2006
Huang, Z., & Bao, S. (2012). Ubiquitination and deubiquitination of REST and its roles in
cancers. FEBS Lett, 586(11), 1602-1605. doi:10.1016/j.febslet.2012.04.052
Jafari, M., & Ansari-Pour, N. (2019). Why, When and How to Adjust Your P Values? Cell J,
20(4), 604-607. doi:10.22074/cellj.2019.5992
Jia, Y., Hao, S., Jin, G., Li, H., Ma, X., Zheng, Y., . . . Wang, Y. (2019). Overexpression of
ARNT2 is associated with decreased cell proliferation and better prognosis in gastric cancer. Mol
Cell Biochem, 450(1-2), 97-103. doi:10.1007/s11010-018-3376-y
Kent, W. J., Sugnet, C. W., Furey, T. S., Roskin, K. M., Pringle, T. H., Zahler, A. M., &
Haussler, D. (2002). The human genome browser at UCSC. Genome Res, 12(6), 996-1006.
doi:10.1101/gr.229102
Kubo, N., Ishii, H., Xiong, X., Bianco, S., Meitinger, F., Hu, R., . . . Ren, B. (2021). Promoter-
proximal CTCF binding promotes distal enhancer-dependent gene activation. Nature Structural
& Molecular Biology, 28(2), 152-161. doi:10.1038/s41594-020-00539-5
Kundaje, A., Meuleman, W., Ernst, J., Bilenky, M., Yen, A., Heravi-Moussavi, A., . . . Principal,
i. (2015). Integrative analysis of 111 reference human epigenomes. Nature, 518(7539), 317-330.
doi:10.1038/nature14248
Lambert, S. A., Jolma, A., Campitelli, L. F., Das, P. K., Yin, Y., Albu, M., . . . Weirauch, M. T.
(2018). The Human Transcription Factors. Cell, 172(4), 650-665. doi:10.1016/j.cell.2018.01.029
Langfelder, P., & Horvath, S. (2008). WGCNA: an R package for weighted correlation network
analysis. BMC Bioinformatics, 9(1), 559. doi:10.1186/1471-2105-9-559
Lee, T. F., Liu, Y. P., Lin, Y. F., Hsu, C. F., Lin, H., Chang, W. C., . . . Wu, C. W. (2021). TAZ
negatively regulates the novel tumor suppressor ANKRD52 and promotes PAK1
dephosphorylation in lung adenocarcinomas. Biochim Biophys Acta Mol Cell Res, 1868(2),
118891. doi:10.1016/j.bbamcr.2020.118891
Levo, M., & Segal, E. (2014). In pursuit of design principles of regulatory sequences. Nature
Reviews Genetics, 15(7), 453-468. doi:10.1038/nrg3684
Liao, G.-B., Li, X.-Z., Zeng, S., Liu, C., Yang, S.-M., Yang, L., . . . Bai, J.-Y. (2018). Regulation
of the master regulator FOXM1 in cancer. Cell Communication and Signaling, 16(1), 57.
doi:10.1186/s12964-018-0266-6
59
Lim, E. J., Suh, Y., Yoo, K. C., Lee, J. H., Kim, I. G., Kim, M. J., . . . Lee, S. J. (2017). Tumor-
associated mesenchymal stem-like cells provide extracellular signaling cue for invasiveness of
glioblastoma cells. Oncotarget, 8(1), 1438-1448. doi:10.18632/oncotarget.13638
Lin, H., Zhang, X., Liao, L., Yu, T., Li, J., Pan, H., . . . Yao, M. (2018). CPNE3 promotes
migration and invasion in non-small cell lung cancer by interacting with RACK1 via FAK
signaling activation. J Cancer, 9(22), 4215-4222. doi:10.7150/jca.25872
Love, M. I., Huber, W., & Anders, S. (2014). Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2. Genome Biol, 15(12), 550. doi:10.1186/s13059-014-
0550-8
Mack, S. C., Singh, I., Wang, X., Hirsch, R., Wu, Q., Villagomez, R., . . . Rich, J. N. (2019).
Chromatin landscapes reveal developmentally encoded transcriptional states that define human
glioblastoma. J Exp Med, 216(5), 1071-1090. doi:10.1084/jem.20190196
Marques, C., Unterkircher, T., Kroon, P., Izzo, A., Gargiulo, G., Kling, E., . . . Squatrito, M.
(2019). NF1 regulates mesenchymal glioblastoma plasticity and aggressiveness through the AP-1
transcription factor FOSL1. bioRxiv, 834531. doi:10.1101/834531
Melhuish, T. A., Kowalczyk, I., Manukyan, A., Zhang, Y., Shah, A., Abounader, R., & Wotton,
D. (2018). Myt1 and Myt1l transcription factors limit proliferation in GBM cells by repressing
YAP1 expression. Biochim Biophys Acta Gene Regul Mech, 1861(11), 983-995.
doi:10.1016/j.bbagrm.2018.10.005
Morita, K., Noura, M., Tokushige, C., Maeda, S., Kiyose, H., Kashiwazaki, G., . . . Kamikubo,
Y. (2017). Autonomous feedback loop of RUNX1-p53-CBFB in acute myeloid leukemia cells.
Sci Rep, 7(1), 16604. doi:10.1038/s41598-017-16799-z
Mukherjee, S. (2020). Quiescent stem cell marker genes in glioma gene networks are sufficient
to distinguish between normal and glioblastoma (GBM) samples. Scientific Reports, 10(1),
10937. doi:10.1038/s41598-020-67753-5
Mullen, D. J., Yan, C., Kang, D. S., Zhou, B., Borok, Z., Marconett, C. N., . . . Rhie, S. K.
(2020). TENET 2.0: Identification of key transcriptional regulators and enhancers in lung
adenocarcinoma. PLOS Genetics, 16(9), e1009023. doi:10.1371/journal.pgen.1009023
Musa, J., Aynaud, M.-M., Mirabeau, O., Delattre, O., & Grünewald, T. G. P. (2017). MYBL2
(B-Myb): a central regulator of cell proliferation, cell survival and differentiation involved in
tumorigenesis. Cell Death & Disease, 8(6), e2895-e2895. doi:10.1038/cddis.2017.244
Narayanan, A., Gagliardi, F., Gallotti, A. L., Mazzoleni, S., Cominelli, M., Fagnocchi, L., . . .
Galli, R. (2019). The proneural gene ASCL1 governs the transcriptional subgroup affiliation in
glioblastoma stem cells by directly repressing the mesenchymal gene NDRG1. Cell Death &
Differentiation, 26(9), 1813-1831. doi:10.1038/s41418-018-0248-7
60
Ni, W., Perez, A. A., Schreiner, S., Nicolet, C. M., & Farnham, Peggy J. (2020). Characterization
of the ZFX family of transcription factors that bind downstream of the start site of CpG island
promoters. Nucleic Acids Research, 48(11), 5986-6000. doi:10.1093/nar/gkaa384
Park, J.-W., Wollmann, G., Urbiola, C., Fogli, B., Florio, T., Geley, S., & Klimaschewski, L.
(2018). Sprouty2 enhances the tumorigenic potential of glioblastoma cells. Neuro-Oncology,
20(8), 1044-1054. doi:10.1093/neuonc/noy028
Park, N. I., Guilhamon, P., Desai, K., McAdam, R. F., Langille, E., O'Connor, M., . . . Dirks, P.
B. (2017). ASCL1 Reorganizes Chromatin to Direct Neuronal Fate and Suppress Tumorigenicity
of Glioblastoma Stem Cells. Cell Stem Cell, 21(2), 209-224.e207.
doi:10.1016/j.stem.2017.06.004
Pellegatta, S., Savoldo, B., Di Ianni, N., Corbetta, C., Chen, Y., Patané, M., . . . Dotti, G. (2018).
Constitutive and TNFα-inducible expression of chondroitin sulfate proteoglycan 4 in
glioblastoma and neurospheres: Implications for CAR-T cell therapy. Science Translational
Medicine, 10(430), eaao2731. doi:10.1126/scitranslmed.aao2731
Pourkeramati, F., Asadi, M. H., Shakeri, S., & Farsinejad, A. (2019). Differential Expression
Profile of ZFX Variants Discriminates Breast Cancer Subtypes. Iran Biomed J, 23(1), 47-56.
doi:10.29252/.23.1.47
Pusapati, R. V., Weaks, R. L., Rounbehler, R. J., McArthur, M. J., & Johnson, D. G. (2010).
E2F2 suppresses Myc-induced proliferation and tumorigenesis. Mol Carcinog, 49(2), 152-156.
doi:10.1002/mc.20584
Rhie, S. K., Guo, Y., Tak, Y. G., Yao, L., Shen, H., Coetzee, G. A., . . . Farnham, P. J. (2016).
Identification of activated enhancers and linked transcription factors in breast, prostate, and
kidney tumors by tracing enhancer networks using epigenetic traits. Epigenetics & Chromatin,
9(1), 50. doi:10.1186/s13072-016-0102-4
Rhie, S. K., Yao, L., Luo, Z., Witt, H., Schreiner, S., Guo, Y., . . . Farnham, P. J. (2018). ZFX
acts as a transcriptional activator in multiple types of human tumors by binding downstream of
transcription start sites at the majority of CpG island promoters. Genome Res, 28(3), 310-320.
doi:10.1101/gr.228809.117
Schenková, K., Lutz, J., Kopp, M., Ramos, S., & Rivero, F. (2012). MUF1/leucine-rich repeat
containing 41 (LRRC41), a substrate of RhoBTB-dependent cullin 3 ubiquitin ligase complexes,
is a predominantly nuclear dimeric protein. J Mol Biol, 422(5), 659-673.
doi:10.1016/j.jmb.2012.06.016
Schneider, V. A., Graves-Lindsay, T., Howe, K., Bouk, N., Chen, H.-C., Kitts, P. A., . . . Church,
D. M. (2017). Evaluation of GRCh38 and de novo haploid genome assemblies demonstrates the
enduring quality of the reference assembly. Genome Research, 27(5), 849-864.
doi:10.1101/gr.213611.116
Sharifi-Zarchi, A., Gerovska, D., Adachi, K., Totonchi, M., Pezeshk, H., Taft, R. J., . . . Araúzo-
Bravo, M. J. (2017). DNA methylation regulates discrimination of enhancers from promoters
61
through a H3K4me1-H3K4me3 seesaw mechanism. BMC Genomics, 18(1), 964.
doi:10.1186/s12864-017-4353-7
Sharma, S., Kelly, T. K., & Jones, P. A. (2010). Epigenetics in cancer. Carcinogenesis, 31(1),
27-36. doi:10.1093/carcin/bgp220
Song, X., Zhu, M., Zhang, F., Zhang, F., Zhang, Y., Hu, Y., . . . Liu, Y. (2018). ZFX Promotes
Proliferation and Metastasis of Pancreatic Cancer Cells via the MAPK Pathway. Cellular
Physiology and Biochemistry, 48(1), 274-284. doi:10.1159/000491727
Tanigawa, S., Fujita, M., Moyama, C., Ando, S., Ii, H., Kojima, Y., . . . Nakata, S. (2021).
Inhibition of Gli2 suppresses tumorigenicity in glioblastoma stem cells derived from a de novo
murine brain cancer model. Cancer Gene Ther. doi:10.1038/s41417-020-00282-5
Tsang, S. M., Oliemuller, E., & Howard, B. A. (2020). Regulatory roles for SOX11 in
development, stem cells and cancer. Seminars in Cancer Biology, 67, 3-11.
doi:https://doi.org/10.1016/j.semcancer.2020.06.015
Verhaak, R. G., Hoadley, K. A., Purdom, E., Wang, V., Qi, Y., Wilkerson, M. D., . . . Hayes, D.
N. (2010). Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma
characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell, 17(1), 98-110.
doi:10.1016/j.ccr.2009.12.020
Wang, J., Sun, D., Wang, Y., Ren, F., Pang, S., Wang, D., & Xu, S. (2014). FOSL2 Positively
Regulates TGF-β1 Signalling in Non-Small Cell Lung Cancer. PLOS ONE, 9(11), e112150.
doi:10.1371/journal.pone.0112150
Wang, Z., Zhang, S., Siu, T. L., & Huang, S. (2015). Glioblastoma multiforme formation and
EMT: role of FoxM1 transcription factor. Curr Pharm Des, 21(10), 1268-1271.
doi:10.2174/1381612821666141211115949
Wen, H., Li, Y., Xi, Y., Jiang, S., Stratton, S., Peng, D., . . . Shi, X. (2014). ZMYND11 links
histone H3.3K36me3 to transcription elongation and tumour suppression. Nature, 508(7495),
263-268. doi:10.1038/nature13045
Wu, A., Ericson, K., Chao, W., & Low, W. C. (2010). NFAT and AP1 are essential for the
expression of a glioblastoma multiforme related IL-13Ra2 transcript. Cell Oncol, 32(5-6), 313-
329. doi:10.3233/clo-2010-0524
Xie, Q., Wu, Q., Kim, L., Miller, T. E., Liau, B. B., Mack, S. C., . . . Rich, J. N. (2016). RBPJ
maintains brain tumor–initiating cells through CDK9-mediated transcriptional elongation. The
Journal of Clinical Investigation, 126(7), 2757-2772. doi:10.1172/JCI86114
Yan, X., Shan, Z., Yan, L., Zhu, Q., Liu, L., Xu, B., . . . Gao, Y. (2016). High expression of
Zinc-finger protein X-linked promotes tumor growth and predicts a poor outcome for stage II/III
colorectal cancer patients. Oncotarget, 7(15), 19680-19692. doi:10.18632/oncotarget.7547
62
Yang, J. P., Yang, J. K., Li, C., Cui, Z. Q., Liu, H. J., Sun, X. F., . . . Jiao, B. H. (2017).
Downregulation of ZMYND11 induced by miR-196a-5p promotes the progression and growth of
GBM. Biochem Biophys Res Commun, 494(3-4), 674-680. doi:10.1016/j.bbrc.2017.10.098
Yin, W., Tang, G., Zhou, Q., Cao, Y., Li, H., Fu, X., . . . Jiang, X. (2019). Expression Profile
Analysis Identifies a Novel Five-Gene Signature to Improve Prognosis Prediction of
Glioblastoma. Front Genet, 10, 419. doi:10.3389/fgene.2019.00419
Yuan, J., Xiao, G., Peng, G., Liu, D., Wang, Z., Liao, Y., . . . Yuan, X. (2015). MiRNA-125a-5p
inhibits glioblastoma cell proliferation and promotes cell differentiation by targeting TAZ.
Biochemical and Biophysical Research Communications, 457(2), 171-176.
doi:https://doi.org/10.1016/j.bbrc.2014.12.078
Yuan, J., Zeng, J., Shuai, C., & Liu, Y. (2018). TWSG1 Is a Novel Tumor Suppressor in Gastric
Cancer. DNA Cell Biol, 37(6), 574-583. doi:10.1089/dna.2018.4188
Zhang, H., Liu, L., Wang, Y., Zhao, G., Xie, R., Liu, C., . . . Fan, D. (2013). KLF8 involves in
TGF-beta-induced EMT and promotes invasion and migration in gastric cancer cells. J Cancer
Res Clin Oncol, 139(6), 1033-1042. doi:10.1007/s00432-012-1363-3
Zhang, Y., Dube, C., Gibert, M., Jr., Cruickshanks, N., Wang, B., Coughlan, M., . . . Abounader,
R. (2018). The p53 Pathway in Glioblastoma. Cancers (Basel), 10(9).
doi:10.3390/cancers10090297
Zhang, Y., Liu, T., Meyer, C. A., Eeckhoute, J., Johnson, D. S., Bernstein, B. E., . . . Liu, X. S.
(2008). Model-based analysis of ChIP-Seq (MACS). Genome Biol, 9(9), R137. doi:10.1186/gb-
2008-9-9-r137
Zhou, L., Jiang, Y., Luo, Q., Li, L., & Jia, L. (2019). Neddylation: a novel modulator of the
tumor microenvironment. Molecular Cancer, 18(1), 77. doi:10.1186/s12943-019-0979-1
Zhu, W., Hu, J., Hu, M., & Jiang, H. J. T. C. R. (2017). Downregulation of ZFX is associated
with inhibition of prostate cancer progression by baicalein. 2017, 6(5), 960-967. Retrieved from
http://tcr.amegroups.com/article/view/16170
Zhu Z, L. K., Xu D, Liu Y, Tang H, Xie Q, Xie L, Liu J, Wang H, Gong Y, Hu Z, Zheng J.
(2013). ZFX regulates glioma cell proliferation and survival in vitro and in vivo. J Neurooncol.
63
APPENDIX
Table 1 GSM ID of GBM H3K27ac ChIP-seq data used in this study
GSM_ID Sample_ID
GSM3382183 GBM_H3K27ac_1
GSM3382185 GBM_H3K27ac_17
GSM3382187 GBM_H3K27ac_21
GSM3382189 GBM_H3K27ac_28
GSM3382191 GBM_H3K27ac_30
GSM3382193 GBM_H3K27ac_31
GSM3382195 GBM_H3K27ac_32
GSM3382197 GBM_H3K27ac_33
GSM3382199 GBM_H3K27ac_4
GSM3382201 GBM_H3K27ac_45
GSM3382203 GBM_H3K27ac_46
GSM3382205 GBM_H3K27ac_47
GSM3382207 GBM_H3K27ac_48
GSM3382209 GBM_H3K27ac_49
GSM3382211 GBM_H3K27ac_5
GSM3382213 GBM_H3K27ac_50
GSM3382215 GBM_H3K27ac_51
GSM3382217 GBM_H3K27ac_52
GSM3382219 GBM_H3K27ac_53
GSM3382221 GBM_H3K27ac_54
GSM3382223 GBM_H3K27ac_55
GSM3382225 GBM_H3K27ac_57
GSM3382227 GBM_H3K27ac_58
GSM3382229 GBM_H3K27ac_59
GSM3382231 GBM_H3K27ac_60
GSM3382233 GBM_H3K27ac_61
GSM3382235 GBM_H3K27ac_62
GSM3382237 GBM_H3K27ac_63
GSM3382239 GBM_H3K27ac_64
GSM3382241 GBM_H3K27ac_66
GSM3382243 GBM_H3K27ac_67
GSM3382245 GBM_H3K27ac_68
64
GSM3382247 GBM_H3K27ac_69
GSM3382249 GBM_H3K27ac_70
GSM3382251 GBM_H3K27ac_71
GSM3382253 GBM_H3K27ac_73
GSM3382255 GBM_H3K27ac_74
GSM3382257 GBM_H3K27ac_75
GSM3382259 GBM_H3K27ac_76
GSM3382261 GBM_H3K27ac_78
GSM3382263 GBM_H3K27ac_79
GSM3382265 GBM_H3K27ac_80
GSM3382267 GBM_H3K27ac_81
GSM3382269 GBM_H3K27ac_82
GSM3382271 GBM_H3K27ac_83
GSM3382273 GBM_H3K27ac_84
Table 2 GSM ID of normal brain H3K27ac ChIP-seq data used in this study
GSM_ID Sample_ID
GSM772832 Brain_Anterior_Caudate.H3K27ac.149
GSM772995 Brain_Inferior_Temporal_Lobe.H3K27ac.149
GSM773011 Brain_Cingulate_Gyrus.H3K27ac.149
GSM773015 Brain_Mid_Frontal_Lobe.H3K27ac.149
GSM773016 Brain_Angular_Gyrus.H3K27ac.149
GSM773020 Brain_Hippocampus_Middle.H3K27ac.149
GSM916035 Brain_Hippocampus_Middle.H3K27ac.150
Table 3 A list of genes regulated by ZFX involved in cell-cycle found common across cell types
Gene_Name U87* C42B* HEK293T* MCF7* Cell cycle & proliferation*
ENSG00000128791.12 TWSG1 1 1 1 1 1
ENSG00000262919.8 CCNQ 1 1 1 1 1
ENSG00000149679.11 CABLES2 1 1 1 1 1
ENSG00000172379.21 ARNT2 1 1 1 1 1
ENSG00000128829.12 EIF2AK4 1 1 1 1 1
ENSG00000082146.13 STRADB 1 1 1 1 1
ENSG00000099910.17 KLHL22 1 1 1 1 1
ENSG00000066629.18 EML1 1 1 1 1 1
ENSG00000137103.20 TMEM8B 1 1 1 1
ENSG00000164114.19 MAP9 1 1 1 1
ENSG00000145604.16 SKP2 1 1 1 1
ENSG00000173757.10 STAT5B 1 1 1 1
65
ENSG00000109103.12 UNC119 1 1 1 1
ENSG00000185340.15 GAS2L1 1 1 1 1
ENSG00000096872.16 IFT74 1 1 1 1
ENSG00000253368.4 TRNP1 1 1 1 1
ENSG00000015171.20 ZMYND11 1 1 1 1
ENSG00000142208.16 AKT1 1 1 1 1
ENSG00000148400.12 NOTCH1 1 1 1 1
ENSG00000055332.18 EIF2AK2 1 1 1 1
ENSG00000116117.18 PARD3B 1 1 1 1
ENSG00000087448.11 KLHL42 1 1 1 1
ENSG00000276141.4 WHAMMP3 1 1 1 1
* genes belong to each category are marked with “1” in the cells
Table 4 A list of GBM-specific genes regulated by ZFX involved in cell-cycle
Gene_Name U87* C42B* HEK293T* MCF7* Cell cycle & proliferation*
ENSG00000005007.13 UPF1 1 1
ENSG00000007372.23 PAX6 1 1
ENSG00000007968.7 E2F2 1 1
ENSG00000018408.15 WWTR1 1 1
ENSG00000019485.13 PRDM11 1 1
ENSG00000054598.9 FOXC1 1 1
ENSG00000058453.17 CROCC 1 1
ENSG00000074047.22 GLI2 1 1
ENSG00000080824.19 HSP90AA1 1 1
ENSG00000082458.12 DLG3 1 1
ENSG00000084207.18 GSTP1 1 1
ENSG00000085999.13 RAD54L 1 1
ENSG00000088808.18 PPP1R13B 1 1
ENSG00000090447.12 TFAP4 1 1
ENSG00000100014.20 SPECC1L 1 1
ENSG00000100697.15 DICER1 1 1
ENSG00000101057.16 MYBL2 1 1
ENSG00000103653.16 CSK 1 1
ENSG00000104290.11 FZD3 1 1
ENSG00000104312.8 RIPK2 1 1
ENSG00000105173.14 CCNE1 1 1
ENSG00000106348.18 IMPDH1 1 1
ENSG00000107443.16 CCNJ 1 1
ENSG00000110172.12 CHORDC1 1 1
66
ENSG00000112658.8 SRF 1 1
ENSG00000112701.18 SENP6 1 1
ENSG00000113522.14 RAD50 1 1
ENSG00000113648.16 MACROH2A1 1 1
ENSG00000115758.13 ODC1 1 1
ENSG00000116285.13 ERRFI1 1 1
ENSG00000116661.11 FBXO2 1 1
ENSG00000116717.13 GADD45A 1 1
ENSG00000119772.17 DNMT3A 1 1
ENSG00000120738.8 EGR1 1 1
ENSG00000121068.14 TBX2 1 1
ENSG00000121274.13 TENT4B 1 1
ENSG00000123358.20 NR4A1 1 1
ENSG00000124006.15 OBSL1 1 1
ENSG00000125266.8 EFNB2 1 1
ENSG00000128714.6 HOXD13 1 1
ENSG00000128944.13 KNSTRN 1 1
ENSG00000128989.11 ARPP19 1 1
ENSG00000132275.11 RRP8 1 1
ENSG00000132612.16 VPS4A 1 1
ENSG00000133119.13 RFC3 1 1
ENSG00000134057.15 CCNB1 1 1
ENSG00000136104.21 RNASEH2B 1 1
ENSG00000136158.12 SPRY2 1 1
ENSG00000136630.13 HLX 1 1
ENSG00000136982.6 DSCC1 1 1
ENSG00000138162.19 TACC2 1 1
ENSG00000138346.15 DNA2 1 1
ENSG00000141646.14 SMAD4 1 1
ENSG00000143862.8 ARL8A 1 1
ENSG00000143878.10 RHOB 1 1
ENSG00000144381.17 HSPD1 1 1
ENSG00000149474.13 KAT14 1 1
ENSG00000152518.8 ZFP36L2 1 1
ENSG00000152669.9 CCNO 1 1
ENSG00000153904.21 DDAH1 1 1
ENSG00000154734.15 ADAMTS1 1 1
ENSG00000156273.16 BACH1 1 1
ENSG00000156345.18 CDK20 1 1
67
ENSG00000156374.16 PCGF6 1 1
ENSG00000159086.15 PAXBP1 1 1
ENSG00000159267.15 HLCS 1 1
ENSG00000159899.15 NPR2 1 1
ENSG00000163785.13 RYK 1 1
ENSG00000163932.15 PRKCD 1 1
ENSG00000164045.12 CDC25A 1 1
ENSG00000164087.8 POC1A 1 1
ENSG00000164162.14 ANAPC10 1 1
ENSG00000164741.15 DLC1 1 1
ENSG00000168621.15 GDNF 1 1
ENSG00000169594.13 BNC1 1 1
ENSG00000170873.19 MTSS1 1 1
ENSG00000170961.7 HAS2 1 1
ENSG00000171310.11 CHST11 1 1
ENSG00000171612.7 SLC25A33 1 1
ENSG00000173546.7 CSPG4 1 1
ENSG00000174775.17 HRAS 1 1
ENSG00000175197.12 DDIT3 1 1
ENSG00000176396.11 EID2 1 1
ENSG00000177602.5 HASPIN 1 1
ENSG00000179598.6 PLD6 1 1
ENSG00000181555.21 SETD2 1 1
ENSG00000181773.7 GPR3 1 1
ENSG00000183527.12 PSMG1 1 1
ENSG00000184524.6 CEND1 1 1
ENSG00000185697.16 MYBL1 1 1
ENSG00000185721.13 DRG1 1 1
ENSG00000186350.12 RXRA 1 1
ENSG00000186575.19 NF2 1 1
ENSG00000189079.17 ARID2 1 1
ENSG00000197461.13 PDGFA 1 1
ENSG00000198171.13 DDRGK1 1 1
ENSG00000198793.13 MTOR 1 1
ENSG00000204991.11 SPIRE2 1 1
* genes belong to each category are marked with “1” in the cells
Table 5 A list of TFs with more than 200 linked GBM-specific enhancer probes identified by TENET2.0
Gene name ENSG ID Frequency
68
HES6 ENSG00000144485 775
ASCL1 ENSG00000139352 766
ZFP69 ENSG00000187815 744
NKX3-1 ENSG00000167034 653
SOX2 ENSG00000181449 559
ZNF730 ENSG00000183850 552
SOX11 ENSG00000176887 547
VAX2 ENSG00000116035 525
SALL3 ENSG00000256463 481
FOXD1 ENSG00000251493 466
PCGF2 ENSG00000277258 466
PAX1 ENSG00000125813 451
ZNF677 ENSG00000197928 451
GSX1 ENSG00000169840 427
CBX2 ENSG00000173894 421
KLF6 ENSG00000067082 421
BARX1 ENSG00000131668 412
VENTX ENSG00000151650 393
HOXC6 ENSG00000197757 391
MYBL2 ENSG00000101057 389
RUNX1 ENSG00000159216 387
IRX5 ENSG00000176842 358
ZNF610 ENSG00000167554 340
ONECUT2 ENSG00000119547 337
FOXM1 ENSG00000111206 329
MYB ENSG00000118513 328
E2F2 ENSG00000007968 316
INSM1 ENSG00000173404 313
FEV ENSG00000163497 312
ZNF681 ENSG00000196172 311
SOX12 ENSG00000177732 304
KLF17 ENSG00000171872 291
ZNF300 ENSG00000145908 285
BATF ENSG00000156127 283
SNAI1 ENSG00000124216 282
ELK3 ENSG00000111145 279
SIX1 ENSG00000126778 276
ZFP37 ENSG00000136866 267
RUNX2 ENSG00000124813 260
69
GCM1 ENSG00000137270 255
RUNX3 ENSG00000020633 244
VDR ENSG00000111424 244
GLMP ENSG00000198715 230
CEBPB ENSG00000172216 224
FOSL2 ENSG00000075426 220
ZNF286A ENSG00000187607 212
CENPA ENSG00000115163 211
E2F7 ENSG00000165891 209
SOX6 ENSG00000110693 208
ZNF229 ENSG00000278318 204
ZNF600 ENSG00000189190 202
CEBPD ENSG00000221869 200
Table 6 TCGA cancer types abbreviations
Study Abbreviation Study Name
BLCA Bladder Urothelial Carcinoma
BRCA Breast invasive carcinoma
CESC Cervical squamous cell carcinoma and endocervical adenocarcinoma
CHOL Cholangiocarcinoma
COAD Colon adenocarcinoma
ESCA Esophageal carcinoma
GBM Glioblastoma multiforme
HNSC Head and Neck squamous cell carcinoma
KICH Kidney Chromophobe
KIRC Kidney renal clear cell carcinoma
KIRP Kidney renal papillary cell carcinoma
LIHC Liver hepatocellular carcinoma
LUAD Lung adenocarcinoma
LUSC Lung squamous cell carcinoma
PAAD Pancreatic adenocarcinoma
PCPG Pheochromocytoma and Paraganglioma
PRAD Prostate adenocarcinoma
READ Rectum adenocarcinoma
SARC Sarcoma
STAD Stomach adenocarcinoma
THYM Thymoma
70
THCA Thyroid carcinoma
UCEC Uterine Corpus Endometrial Carcinoma
Abstract (if available)
Abstract
Glioblastoma multiforme (GBM) is the most common intrinsic brain tumor in adults and is universally fatal. My strategy to study molecular mechanisms leading to GBM is to investigate transcription factors (TFs) that bind to GBM-specific regulatory elements. Among TFs, ZFX has been shown as a key driver TF for GBM. Several studies have reported the significant role of ZFX in carcinogenesis, however, the molecular mechanisms of TF ZFX that drive GBM still remain unknown. Therefore, my project aims to characterize the function of ZFX in GBM by identifying its binding sites and target genes through ChIP-seq and siRNA followed by RNA-seq analysis. Besides ZFX, I also integrated other publicly available multi-omic datasets and performed analysis using in-house developed bioinformatic tool, TENET2.0 to identify other key TFs linked to GBM. This study provides new insights into understanding of TF networks driving GBM.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Do the ZFX and ZFY transcription factors have redundant or unique functions?
PDF
Identification of target genes and protein partners of ZNF711 in glioblastoma cells
PDF
Characterizing ZFX-mediated gene regulation to reveal possible candidates for clinical intervention
PDF
Evaluation of preservatives in blood collection tubes for cell-free RNA transcriptional profiles in human plasma
PDF
Do ZFX and ZNF711 regulate the same genes in HEK293T cells?
PDF
Evaluating the therapeutic potential of targeting CBX8 in MLLr leukemia
PDF
Interleukin-11: a study of its effects in glioblastoma multiforme
PDF
Determining the epigenetic contribution of basal cell identity in cystic fibrosis
PDF
Evaluating the impact of long non-coding RNAs on tumor mutational burden in cancer
PDF
Analysis of human brain endothelial cells versus human tumor associated brain endothelial cell tight junction and adherens junction expression differences in glioblastoma multiforme
PDF
Integrated genomic & epigenomic analyses of glioblastoma multiforme: Methods development and application
PDF
Molecular targets for treatment of glioblastoma multiforme
PDF
Targeting glioma cancer stem cells for the treatment of glioblastoma multiforme
PDF
Application of tracing enhancer networks using epigenetic traits (TENET) to identify epigenetic deregulation in cancer
PDF
Functional study of Snord118 in microglila
PDF
Characterization of the ZFX family of transcription factors that bind downstream of the start site of CpG island promoters
PDF
MSI2 level alters histone transcription rate in HepG2 cells
PDF
Examination of the effect of oleandrin on head and neck cancer cells
PDF
A novel screening approach to identify regulators of polycomb-dependent gene silencing
PDF
Mbnl1 and Mbnl2 play roles in brain structural integrity
Asset Metadata
Creator
Wu, Zexun
(author)
Core Title
Mapping transcription factor networks linked to glioblastoma multiform: identifying target genes of the oncogenic transcription factor ZFX in glioblastoma multiforme
School
Keck School of Medicine
Degree
Master of Science
Degree Program
Biochemistry and Molecular Medicine
Degree Conferral Date
2021-12
Publication Date
11/16/2021
Defense Date
06/02/2021
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
bioinformatics,epigenomics,glioblastoma multiforme,OAI-PMH Harvest,transcription factor,ZFX
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Rhie, Suhn Kyong (
committee chair
), Dou, Yali (
committee member
), Weisenberger, Daniel (
committee member
)
Creator Email
632650149@qq.com,zexunwu@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC17138502
Unique identifier
UC17138502
Legacy Identifier
etd-WuZexun-10230
Document Type
Thesis
Format
application/pdf (imt)
Rights
Wu, Zexun
Type
texts
Source
20211117-wayne-usctheses-batch-897-nissen
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
bioinformatics
epigenomics
glioblastoma multiforme
transcription factor
ZFX