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
/
Roles of circadian clock genes in cancer
(USC Thesis Other)
Roles of circadian clock genes in cancer
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
ROLES OF CIRCADIAN CLOCK GENES IN CANCER
by
YUANZHONG PAN
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
December 2024
ii
Acknowledgements
As this thesis comes into being, I have a strong feeling that I made this far because I had
more help than diQiculties. Therefore, I have more gratefulness than my references.
First and foremost, I would like to thank my mother. You carried me through all the jokes life
put on us, and the only reason I can always luxuriously be myself is because of the little
niche you created for me. It’s not a big one, but it’s so strong that I never needed to look
back while moving forward. This PhD is as much yours as it is mine.
To Dr. Steve Kay, I’m extremely lucky to have you as my advisor. And I’m all the time grateful
for your generous support to pursue the science that is 100% of my true interest without
other concerns, and your wise guidance on the science, on my career, especially when very
often my critical thinking was blindfolded by my excitements. Your professionalism,
passion, integrity, and care are living models for me to grow as not only a scientist but also
a person.
To Dr. Evanthia Roussos Torres, my scientific journey took real momentum when you took
me as a mentee. In the beginning, every time we met, I felt overwhelmed by your
intelligence power, and I’m still not sure I can catch up with you on this day. At the same
time, you are extremely supportive of me and all your other mentees for what we want to
pursue. The most I admire on a human being is the combination of high achieving and
kindness. I’m lucky to have both of my mentors like that.
To my friends who are chosen family to me, Eilidh and Rin, for making me feel loved and
having a family in this country.
iii
To the MCB coQee club, Falk (the wise spirit), Janielle (our literal “sugar” mummy), Oscar
(endless jokes), Lynne, Emily, and Shahd. You guys make it always fun to go in the lab even
when work itself goes on its rainy days.
To c-STEM members, Rudra, Kevin, Aneesh, Holly, Rachel, it has been an incredibly
rewarding experience to lead a good team. I’m extremely proud of what we have achieved.
Thank you to my qualifying and defense committee member, Drs. Mumenthaler, Finley,
Shen, and Newton for your mentorship and guidance, which are indispensable in making
me a scientist today and a better one in the future.
And lastly, thanks to all my other family and friends for supporting me through this journey.
iv
Table of Contents
ACKNOWLEDGEMENTS ..................................................................................................................... II
LIST OF TABLES .............................................................................................................................. VIII
LIST OF FIGURES .............................................................................................................................. IX
ABBREVIATIONS............................................................................................................................... XI
ABSTRACT...................................................................................................................................... XIV
CHAPTER 1. INTRODUCTION...............................................................................................................1
1.1 CANCER BIOLOGY ........................................................................................................................ 1
1.1.1 Current Scheme of Cancer Therapy .......................................................................................1
1.1.3 Glioblastoma .......................................................................................................................4
1.1.4 Tumor Microenvironment (TME).............................................................................................5
1.2 CIRCADIAN RHYTHM ..................................................................................................................... 5
1.2.1 Molecular Mechanism of the Circadian Rhythm in Mammals ......................................................5
1.2.2 Role of the Circadian Rhythm Genes in Cancer...........................................................................6
1.3 SMALL MOLECULES FOR TARGETING THE CIRCADIAN RHYTHM GENE NETWORK ......................................... 12
1.3.1 Targeting BMAL1 and CLOCK ...................................................................................................12
1.3.2 REV-ERB agonists ...................................................................................................................13
1.3.3 CRY stabilizer .........................................................................................................................14
1.3.4 CK2 inhibitor...........................................................................................................................14
1.3.5 Other modulators of the core circadian genes ..........................................................................15
1.4 REGULATION OF EUKARYOTIC TRANSCRIPTION ................................................................................... 16
1.4.1 Mechanism of gene transcription in eukaryotes ........................................................................16
1.4.2 Targeting the Transcription Mechanisms in Cancer ...................................................................16
1.5 OBJECTIVES AND AIMS ................................................................................................................. 17
CHAPTER 2. CLINICAL SIGNATURES OF CIRCADIAN GENES IN BREAST CANCER.................................18
v
2.1 INTRODUCTION, OBJECTIVE, AND RATIONALE............................................................................................ 18
2.2DATA AND METHODS .......................................................................................................................... 19
2.2.1 The TCGA breast carcinoma (BRCA) patient cohort...................................................................19
2.2.2 The METABRIC breast cancer cohort........................................................................................20
2.2.3 Statistical Methods .................................................................................................................20
2.3 RESULTS AND DISCUSSION................................................................................................................... 20
2.3.1 Expression features of the circadian clock genes in breast cancer .............................................20
2.3.2 A diagnostic signature of circadian genes for breast cancer.......................................................22
2.3.4 Triple negative breast cancer has the highest inferred activities of BMAL1 and CLOCK ................25
2.4CONCLUSION ................................................................................................................................... 26
CHAPTER 3. TARGETING CIRCADIAN GENES IN TRIPLE NEGATIVE BREAST CANCER ............................27
3.1 INTRODUCTION, OBJECTIVE, AND RATIONALE............................................................................................ 27
3.2MATERIALS AND METHODS................................................................................................................... 28
3.2.1 Cell culture ............................................................................................................................28
3.2.2 shRNA lentivirus production ....................................................................................................28
3.2.4 Quantitative RT-PCR and RNA sequencing................................................................................30
3.2.5 Small molecule screening and synergy analysis .......................................................................30
3.2.6 RNA-seq data analysis ............................................................................................................31
3.2.8 Cancer Cell Line Encyclopedia (CCLE) data analysis ................................................................31
3.2.9 Luciferase reporter assay ........................................................................................................32
3.2.10 STARR-seq Library Preparation and Cloning............................................................................32
3.2.11 STARR-seq Screening ............................................................................................................35
3.2.12 STARR-seq data analysis .......................................................................................................38
3.2.13 Other statistical methods ......................................................................................................38
3.3 RESULTS AND DISCUSSION................................................................................................................... 39
vi
3.3.1Knockdown of BMAL1 and CLOCK disrupts the proliferation of metastatic mesenchymal stem-like
TNBC cells......................................................................................................................................39
3.3.2 CRY2 stabilizer SHP1705 represses the negative arm genes in the core circadian network ..........42
3.3.3 CRY stabilizer and proteasome inhibitors synergize to repress BMAL1 and CLOCK activity and
proliferation in mMSL TNBC cells .....................................................................................................46
3.3.4 Combination of SHP1705 and MG132 inhibits the circadian transcription program.....................48
3.3.5 SHP1705 and MG132 repress gene transcription through a cis-regulatory mechanism................51
3.3.5 The combination of SHP1705 and MG132 suppresses a circadian transcription program by
inhibiting selected types of CREs .....................................................................................................57
3.4DISCUSSIONS AND CONCLUSION .......................................................................................................... 65
CHAPTER 4. EFFECTS OF CK2 INHIBITORS ON THE MICROENVIRONMENT OF GLIOBLASTOMA.............70
4.1 INTRODUCTION, OBJECTIVE, AND RATIONALE............................................................................................ 70
4.2MATERIALS AND METHODS................................................................................................................... 70
4.2.1 Cell Culture............................................................................................................................70
4.2.2 Small molecule information ....................................................................................................71
4.2.3 shRNA gene knockdown..........................................................................................................71
4.2.5 Chemotaxis assay...................................................................................................................71
4.2.4 Neural sphere invasion assay ..................................................................................................72
4.3 RESULTS.......................................................................................................................................... 72
4.3.1Microfluidic chemotaxis assay shows reduced microglia recruitment by GSC cells after knockdown
of CLOCK........................................................................................................................................72
4.3.2 The e]ect of small molecules on the proliferation and TME-related gene expression in
glioblastoma cells...........................................................................................................................74
4.3.3 CK2 inhibitors repress the invasion of GSC cells into the matrix .................................................75
4.4DISCUSSION AND CONCLUSION ............................................................................................................ 76
CHAPTER 5. E-BOX-BINDING TRANSCRIPTION FACTORS IN CANCER..................................................77
vii
5.1 INTRODUCTION ................................................................................................................................. 77
5.2 IMPORTANT EBTFS IN CANCER .............................................................................................................. 79
5.2.1 MYC family proteins ................................................................................................................79
5.2.2 Hypoxia-Inducible Factors.......................................................................................................84
5.2.3 BMAL1 and CLOCK circadian clock proteins.............................................................................86
5.2.4 EMT Transcription Factors .......................................................................................................88
5.2.5 Other E-box Binding TFs reported in cancer..............................................................................90
5.3 REGULATORY FEATURES OF E-BOX-CONTAINING REGULATORY ELEMENTS.......................................................... 90
5.4MUTUAL REGULATION OF DIFFERENT EBTFS IN CANCER............................................................................... 94
5.4.1 MYC-HIF interactions..............................................................................................................96
5.4.2 MYC-BMAL1/CLOCK interactions ............................................................................................97
5.4.3 HIF-BMAL1 interactions ..........................................................................................................98
5.4.4 EMT TFs interactions ...............................................................................................................99
5.4.5 Interactions within the same family........................................................................................100
5.5 PERSPECTIVES OF EBTFS IN CANCER .....................................................................................................102
5.5.1 Tumor initiation.....................................................................................................................102
5.5.2 Metabolism ..........................................................................................................................103
5.5.3 Immune evasion and inflammation........................................................................................104
5.5.4 Angiogenesis and other tumor microenvironments .................................................................105
5.5.5 Cancer Stem Cells................................................................................................................106
5.6 TARGETING EBTFS IN CANCER .............................................................................................................108
5.7 OUTLOOK .......................................................................................................................................109
CHAPTER 6 CONCLUSIONS AND FUTURE WORK.............................................................................. 111
REFERENCES ................................................................................................................................. 113
viii
List of Tables
Table 2-1. Details of a logistic regression model ………………………………………………………………….. 22
Table 2-2. Multivariate Cox regression model summary ……………………………………………………….. 25
Table 3-1. TNBC cell line panel screened for BMAL1 and CLOCK function ………………………………. 39
Table 3-2. GSEA analysis on hallmarks pathways after 8 hours of SHP1705 treatment …………………… 45
ix
List of Figures
Figure 1-1. Small molecules that target the core circadian gene network …………………………………15
Figure 2-1. Correlation of circadian genes in TCGA and METABRIC cohorts …………………………….. 21
Figure 2-2. DiOerential expression of circadian genes in tumor versus normal tissues ………………. 22
Figure 2-3. Overall survival analysis on single circadian genes ………………………………………………. 23
Figure 2-4. Survival analysis of the logistic tumor model on both TCGA and METABRIC cohort …... 24
Figure 2-5. Validation of the Cox regression model on TCGA and METABRIC cohort ………………….. 25
Figure 3-1. EOect of BMAL1 and CLOCK knockdown on the proliferation of TNBC cells …………….. 40
Figure 3-2. Quantified cell proliferation after BMAL1 and CLOCK knockdown ……………………………… 41
Figure 3-3. GSEA results after knockdown of BMAL1 and CLOCK …………………………………………… 42
Figure 3-4. IC50 values of SHP1705, SR29065, and GO289 in mMSL TNBC cells ……………………… 43
Figure 3-5. Expression of BMAL1 and CLOCK target genes after drug treatment ………………………. 44
Figure 3-6. Volcano plot of DE genes after SHP1705 treatment ……………………………………………… 45
Figure 3-7. Enrichment analysis after SHP1705 treatment ………………………………………………………… 45
Figure 3-8. Role of proteasome in BMAL1 and CLOCK transcriptional activity ………………………….. 47
Figure 3-9. Synergy analysis of SHP1705 and proteasome inhibitors ……………………………………… 47
Figure 3-10. Validation of RNA-seq results ………………………………………………………………………….. 48
Figure 3-11. Comparison of DE genes after single and dual drug treatment ……………………………… 49
Figure 3-12. GSEA results on combination-treated cells ……………………………………………………….. 50
Figure 3-13. TF enrichment analysis on DE genes ………………………………………………………………… 51
Figure 3-14. Intersection of drug-induced DE genes and cycling genes ……………………………………. 51
Figure 3-15. Refining a list of motifs that represent the binding sites of the enriched TFs ……………. 53
Figure 3-16. Motif cooccurrence and expression correlation of EBTF-cofactors ………………………… 54
Figure 3-17. Promoter sequence and their activity of CDC20 and RUNX2 ………………………………… 55
x
Figure 3-18. Promoter sequence and their activity of HSPA5 and HMOX1 ………………………………… 55
Figure 3-19. Validation of promoter responsivity to drug treatment …………………………………………. 56
Figure 3-20. STARR-seq experiment design ………………………………………………………………………… 57
Figure 3-21. STARR-seq library characterization ………………………………………………………………….. 58
Figure 3-22. Landscape of all TFBS in all STARR peaks ………………………………………………………….. 59
Figure 3-23. Normalized motif counts after treatments ………………………………………………………… 60
Figure 3-24. Co-occurrence analysis of TF-binding motifs ……………………………………………………. 61
Figure 3-25. PCA analysis on the motif counting matrix ………………………………………………………… 63
Figure 3-26. Summarized CRE types that are potentially repressed ………………………………………… 63
Figure 3-27. Contributors to the repression of CREs …………………………………………………………….. 65
Figure 4-1. TME-related gene expression of HMC3 after treated by conditioned media ………………. 73
Figure 4-2. Chemotaxis assay showing eOect of CLOCK KD on microglia ………………………………… 74
Figure 4-3. EOect of small molecules on the expression of TME-related genes …………………………. 75
Figure 4-4. EOect of GO289 on invasion of tumoroids into the matrix ……………………………………… 76
Figure 5-1. Structural overview of EBTFs ……………………………………………………………………………. 83
Figure 5-2. Functional features of E-box elements in transcriptional regulation ……………………….. 93
Figure 5-3. Physiological functions of EBTFs and their mutual regulation ………………………………… 95
Figure 5-4. EBTFs Contributing to the hallmarks of cancer …………………………………………………… 107
xi
Abbreviations
ADC Antibody drug conjugate
ALL Acute lymphoid leukemia
AML Acute myeloid leukemia
AR Androgen receptor
B/C BMAL1/CLOCK
bHLH Basic helix-loop-helix
BL Basal-like
BMAL1 Brain and Muscle ARNT-like 1
BME Basement membrane extract
BRCA Breast carcinoma
CAR-T Chimera antigen receptor-T cell
CCG Clock-controlled genes
CCLE Cancer cell line encyclopedia
CFZ Carfilzomib
ChIP-seq chromatin immunoprecipitation sequencing
CK Casein kinase
CLOCK Circadian locomotor output cycles kaput
CRC Colorectal cancer
CRE Cis-regulatory element
CRY cryptochrome
CSC Cancer stem cell
CYCLOPS cyclic ordering by periodic structure
DE DiOerentially expressed
DMSO dimethylsulfoxide
EBTF E-box-binding transcription factor
EMT Epithelial mesenchymal transition
ER Estrogen receptor
FDA Food and drug administration
xii
FISH fluorescent in situ hybridization
GBM Glioblastoma multiforme
GO Gene ontology
GSC Glioblastoma stem cell
GSEA Gene set enrichment analysis
HCC Hepatocellular carcinoma
HER2 Human epidermal growth factor receptor 2
HIF Hypoxia inducible factor
HNF4A hepatic nuclear receptor 4a
HR Hormone receptor
HSF Heat shock factor
IC50 half-inhibitory concentration
KD Knockdown
KEGG Kyoto encyclopedia of gene and genomes
LAR Luminal androgen receptor
MAX Mic-associated protein
METABRIC The Molecular Taxonomy of Breast Cancer International Consortium
mMSL Metastatic mesenchymal stem-like
MOI multiplicity of infection
MYC myelocytomatosis oncogene
NES Normalized enrichment score
NM Non-malignant
NR Nuclear receptor
NSC Neural stem cell
NSCLC non-small cell lung cancer
PAS PER-ARNT-SIM
PCA Principle component analysis
PCR Polymerase chain reaction
PER Period
PHD prolyl-4-hydroxylase
xiii
PIC Pre-initiation complex
QC quality control
REV-ERB Nuclear Receptor Subfamily 1 Group D Member
ROR RAR-related orphan receptor
RT reverse-transcription
SCN Suprachiamatic nucleus
SE Super enhancer
shRNA Short hairpin RNA
SNP Single nucleotide polymorphism
STARR-seq Self-transcribing active regulatory region sequencing
TAD transactivation domains
TCGA The cancer genome atlas
TF Transcription factor
TFBS Transcription factor binding site
TME Tumor microenvironment
TNBC Triple negative breast cancer
USF Upstream stimulatory factor
VEGF Vascular endothelial growth factor
VHL Von Hippel Lindau
xiv
Abstract
Circadian genes form a network that regulates the 24-hour rhythmic patterns of cellular
and organismal activities. Recent studies have uncovered various connections between
circadian genes and cancer outcomes, cancer cell progression, and the modulation of the
tumor microenvironment, paving the way for further research into these genes to advance
cancer biology and patient care. However, these studies also highlight the complex and
context-dependent roles of circadian genes, necessitating detailed follow-up studies to
fully understand and leverage their potential in cancer treatment.
To address these challenges, this thesis explores the roles of circadian genes from three
perspectives: their potential as biomarkers in cancer patients, their role and targetability as
therapeutic targets, and their capacity to modulate the tumor microenvironment. Firstly,
we identified expression patterns of circadian genes in breast cancer that could serve as
diagnostic and prognostic biomarkers. Secondly, we demonstrated that the core circadian
genes BMAL1 and CLOCK are promising targets in triple-negative breast cancer and
proposed a strategy to target their transcriptional programs through a cis-regulatory
mechanism. Thirdly, we developed bioengineered models of the glioblastoma tumor
microenvironment to test the eQects of drugs targeting the circadian gene CK2. These
studies enhance our understanding of circadian genes in cancer and suggest potential
strategies for incorporating them into cancer treatment.
1
Chapter 1. Introduction
1.1 Cancer Biology
Since its first description in the 1600s BCE, cancer remains one of the deadliest diseases in
human and the leading cause of death worldwide. It is a disease of uncontrolled
proliferation of cells that are transformed from normal tissues. As they progress, cancer
cells form tumor tissues and disrupt the functions of normal body parts, and ultimately
lead to the demise of the organism. Cancer can aQect a wide range of species in animals,
from fish to birds and to mammals.1 It can also arise from almost every tissue and cell
types. Thus, the development of cancer lies in the early branches of the evolutionary tree of
multicellular organisms and is an essential scientific question in biology.
The past decades witnessed great progress in understanding the biology of cancer and
treating it as a disease. These findings also revealed more complexity of cancer as a
disease, such as its immunity, its microenvironment, and its evolution. The growing
knowledge builds the foundation of future study of the biology and treatment of cancer.
1.1.1 Current Scheme of Cancer Therapy
Although most cancer types remain incurable, the development of new therapy has greatly
improved the expected lifetime of cancer patients. These progresses benefited from the
growing understanding of various parts of the biology of cancer. Generally, current
treatments for cancer include multiple types of therapy, such as surgery, chemotherapy,
radiotherapy, hormonal and targeted therapy, immune therapy.
2
Surgical removal of the tumor mass, often combined with other forms of therapy, is the
mainstay treatment for cancer when patient condition permits.2,3 Residual disease after
surgery is the main reason of cancer relapse.4 Fluorescent tools that can help better
visualize tumor boundary and minimize residual tumor tissue is an active area of research
to improve surgical therapy of cancer.5 Adjuvant and neo-adjuvant therapies for primary
surgical therapy are also eQective ways to improve surgery outcome.6,7
Chemotherapy uses cytotoxic agents to inhibit mitosis or induce DNA damage in
proliferating cells and thus trigger apoptosis. Chemotherapy agents are not highly selective
towards cancer cells, and thus cause relatively high side eQects. Despite its toxicity,
chemotherapy is still widely used to treat cancers lacking better options. It is also a
common form of adjuvant or neo-adjuvant therapy for primary surgery.8 Chemotherapy
fails when cancer cells develop resistance to the agents.9
Targeted therapy and hormonal therapy inhibit specific signaling molecules or pathways
that drives tumor cell proliferation, such as targeting human epidermal growth factor
receptor 2 (HER2) in multiple HER2-positive tumor types.10,11 Recent progress in targeted
therapy is built on the study and identification of “driver genes” in tumor patients. Finding
each specific patient’s tumor also promoted the development of precise medicine, the
goal of which is to provide optimal treatment according to each patient’s disease
features.
12 When tumor has actionable drivers, targeted therapy exhibits high eQicacy and
lower toxicity. However, resistance would eventually develop and lead to failure of the
therapy.
3
Immune therapy has brought significant progress in the treatment of cancer.13 It
antagonizes the pathways (e.g. immune check point) that tumor cells utilize to repress the
body’s natural reaction to recognize and kill cancer cells. Immune therapy can also be used
in combination with other therapies. However, only a portion of patients respond to
immune therapy and resistance will eventually develop.
Other new treatments include cell therapy, especially chimera antigen receptor T cell (CART cell) therapy, gene therapy, and oncolytic virotherapy.
14–16 The limitations of current
treatments for cancer requires new paradigms of the biology and treatment of cancer to
bring cancer therapy to the next level.
1.1.2 Breast Cancer
Breast cancer is one of the most common cancer types.
17 Because of its high prevalence,
breast cancer is among the most researched cancer types and has become a paradigm for
cancer biology. Clinically, breast cancers are treated according to their hormone receptor
(HR) and HER2 expression status, which are major drivers of the progression of
corresponding cancer types.18 HR-positive breast cancers benefit from hormonal therapies
including tamoxifen and aromatase inhibitors, with a 5-year survival rate of more than
95%.18 CDK4/6 inhibitors are new therapeutic leverages for HR-positive HER2 nonamplified
breast cancer and greatly improved the prognosis of late-stage tumors of this type. 19
HER2-positive breast cancer is characterized by amplified level of HER2 defined by
immunohistochemistry or RNA fluorescent in situ hybridization (RNA FISH). This type of
breast cancer can be targeted by antagonizing HER2 using antibodies such as trastuzumab
and related drugs. Recent years, antibody-drug conjugates (ADCs) based on trastuzumab,
4
such as trastuzumab emtansine (T-DM1) and trastuzumab deruxtecan (T-DXd) have
achieved significant improvement on the outcomes of diQerent HER2-positive cancers. 20
Breast cancers lacking both HR types (estrogen and progesterone) and HER2 are termed
triple-negative breast cancer (TNBC). TNBC is the most aggressive subtype of breast
cancer, for which no recurrent drivers or therapeutic targets have been recognized
currently, making it the subtype with the poorest prognosis.18 TNBC is also a highly
heterogenous disease. Based on its gene expression profile, TNBC can be classified into
diQerent subtypes. A commonly used way of subtyping classifies TNBC cells as basal-like,
mesenchymal, mesenchymal stem-like, and claudin-low.21 Based on genetic data of large
patient population metabolic pathways, more subtyping strategies have been proposed,
but these methods have not yet provided implications in the clinical treatment of TNBC,
which still largely rely on surgery and chemotherapy.22,23 Therefore, new biological and
therapeutical paradigms are urgently needed for the care of TNBC.
1.1.3 Glioblastoma
Glioblastoma is a highly malignant form of brain tumor. It is a relatively rare disease, but its
biology is poorly understood. Like TNBC, glioblastoma is highly heterogenous and exhibit
stem cell-like features.24 Current treatment for glioblastoma largely relies on surgery and
chemotherapy by temozolomide.
25,26 Based on single-cell RNA sequencing of patientderived tumors, a recent study classified glioblastoma into four subtypes, providing new
insights of biology of glioblastoma, but the clinical translation of such knowledge requires
more detailed study.27
5
1.1.4 Tumor Microenvironment (TME)
TME is the complex local environment that tumors reside and grow. Its component
depends on the tissue type where the tumor grows. In general, TME comprises the
extracellular matrices, blood vessels, immune cells, cancer-associated fibroblast cells,
and other tissue-specific cell types.
28,29 TME plays essential role in the development and
metastasis of tumor, thus becomes an intriguing target for therapy as well.30,31
1.2 Circadian Rhythm
Circadian rhythms comprise the observed 24-hour cycle of physiological activities of living
organisms as a result of adapting to the light-dark cycle that comes from the rotation of
earth. They are observed across almost all species, including cyanobacteria, plants,
worms, flies, and all vertebrates.
In mammals, the master pacemaker of the circadian rhythm is the suprachiasmatic
nucleus (SCN) in the brain. It synchronizes to the blue light signal from the retina and
controls the rhythms of physiological activities, such as sleep, blood pressure, body
temperature and food intake. In this context light is called a zeitgeber for being an input of
the biological time of the body. Another important zeitgeber is food intake. Circadian
rhythms also exhibit on the cellular level. For example, cells tend to repair DNA at night.
Circadian rhythms are essential for maintaining health.32
1.2.1 Molecular Mechanism of the Circadian Rhythm in Mammals
In mammals, circadian rhythms are thought to be formed by inter-connected negative
feedback loops of gene transcription and translation.33 Molecularly, two transcription
6
factors Brain and Muscle ARNT-like 1 (BMAL1, also known as ARNTL) and Circadian
Locomotor Output Cycles Kaput (CLOCK) form a heterodimer and activate the
transcription of a large set of genes, termed clock-controlled genes (CCGs), including
repressors of themselves such as PERs, CRYs, and REV-ERBs. These repressors are
translated and relocated back to the nucleus to repress the transcriptional activity (CRYs
and PERs) or expression (REV-ERBs) of BMAL1 and CLOCK, thus forming the negative arm
of the feedback loop. The alternating active and repressed cycles of BMAL1 and CLOCK
activity form the master oscillator of circadian rhythms. 33
1.2.2 Role of the Circadian Rhythm Genes in Cancer
A growing body of evidence reveals that the circadian clock is tightly linked to cancer on
multiple levels. Early evidence came from epidemiological studies showing that the
disruption of circadian rhythms by factors such as shift work was associated with
increased risk of breast cancer, later supported by additional epidemiological data
confirming the association between shift work and cancer risk.34,35 This evidence motivated
the international agency for research on cancer to list “shift work leading to a disruption in
circadian rhythm” as a probable human carcinogen.
36 Although this connection requires
further confirmation, it inspired researchers to investigate the direct correlation between
the clock and cancer using prospective physiological and genetic experiments in animal
models. For example, chronic jet lag mimicked by shifted lighting or SCN ablation in mice
can induce spontaneous hepatocellular carcinoma (HCC).37,38 In mice that harbor mutated
p53 or Ras genes, alteration or knock-out of core clock genes such as Bmal1 and Per can
accelerate the initiation and progression of multiple cancers.39,40 Epidemiological and
7
animal model studies are reviewed in detail by Pariollaud et al.41 More recently, a pancancer analysis of the cancer genome atlas showed that alteration of clock genes at
transcriptional and genetic levels is pervasive in cancer, and transcriptional dysregulation
is strongly associated with survival. Furthermore, the transcription of clock genes and
many drug target genes are closely associated with each other.42 This work underscored the
clinical significance of clock genes in cancer treatment.
Although a broader association between a disrupted clock and tumorigenesis is well
acknowledged, the detailed molecular relationship remains elusive. First, it is clear that
many cancer cells harbor disrupted clock machinery where the circadian rhythm of the
cells is dampened or completely diminished, but no data thus far have argued clearly
whether the disruption of the clock is the cause or eQect of malignant cell transformation.
Furthermore, in certain types of cancer such as glioblastoma (GBM), the circadian rhythm
in tumor cells remains intact and is in fact required for the maintenance of the disease.
Second, the isoform-specificity, activity, and function of the clock components highly
depend on their tissue background, which implies that diQerent roles may be played by the
clock in cancers originating from diQerent organs. This explains how certain clock genes
can act as tumor suppressors in some tumor types, but oncogenes in others. Third,
because of the essential role of the circadian clock in regulating cellular physiology, the
downstream processes regulated by the clock machinery in the cells are also broad,
weaving complex regulating networks and making it diQicult to build clear and definite
relationships. The interplay between clock components and the generally recognized
hallmarks of cancer (summarized in Ref 43) have been recently reviewed by Sulli et al.44
8
More detailed studies are being carried out to scrutinize the roles of clock molecules in
cancer.
The core clock in cancer
BMAL1 and CLOCK are the central modulators of circadian output in cells and their
pathological functions in cancer are the most thoroughly studied so far. BMAL1:: CLOCK
was initially suggested to be tumor suppressive. Low expression of BMAL1 is reported to be
associated with tumor progression and poor diagnosis in melanoma, pancreatic
adenocarcinoma, and breast cancer.45–47 Consistently, overexpression of BMAL1 in HCC,
osteosarcoma, ovarian and hematologic cancer cells reduced tumor cell growth, and
BMAL1 knockout in hamsters results in HCC.37,48–51 In addition, it is becoming clear that
BMAL1:: CLOCK works diQerently across organs by interacting with tissue-specific factors.
The molecular mechanism of this tissue-specific control has been exemplified by the liver,
where the hepatic nuclear receptor 4a (HNF4A) was shown to repress the BMAL1::CLOCK
transcriptional activity, adding another layer of modulation to the circadian network52.
HNF4A and BMAL1 seem to collaboratively regulate the proliferation of HCC cell lines,
pointing out the functional role of tissue-specific circadian clocks in carcinogenesis48. In
acute myeloid leukemia (AML) and GBM, BMAL1 and CLOCK are essential for the survival
of the cancer stem cell (CSC) population. In AML, knockdown of BMAL1 and CLOCK led to
stem cell diQerentiation and disease regression in a murine model, while keeping physical
hematopoiesis intact
53. This selective role of BMAL1::CLOCK was also observed in GBM,
where more details were revealed54. Glioblastoma stem cells (GSCs), normal neural stem
cells (NSCs), and non-malignant (NM) neuronal cultures excised from epilepsy patients all
9
display rhythmic expression of BMAL1 but disruption of BMAL1 by shRNA only specifically
blocked cell proliferation of GSCs, showing that GSCs have a specific and exquisite
reliance on BMAL1 activity for proliferation. GSCs’, but not NSCs’, self-renewal was also
impaired upon BMAL1 or CLOCK disruption. This observation was explained by chromatin
immunoprecipitation sequencing (ChIP-seq) analysis, which showed that in GSCs, BMAL1
had largely skewed binding activities compared to NSCs, including newly gained binding
sites and motifs that are enriched in genes involved with glucose and lipid metabolism.
Functional assays confirmed that BMAL1 and CLOCK indeed maintained metabolic
homeostasis in GSCs. These data highlight epigenetic regulation and metabolism as
primary processes regulated by the clock proteins in cancer. Disruption of BMAL1 or
CLOCK or BMAL1::CLOCK transcriptional activity by shRNA or small molecules drastically
reduced the tumor burden and improved survival in the mouse model of GBM, indicating
that BMAL1::CLOCK-fueled tumor progression is a viable target for anti-cancer therapy.
54 In
an independent study gain-of-function screen of epigenetic regulators, CLOCK was shown
to enhance GSC self- renewal. CLOCK::BMAL1 induced the recruitment of immunesuppressive microglia to the GBM microenvironment and triggered pro-tumor immune
activity, hence helping tumor maintenance and progression.55 These studies underscore
the role of the clock machinery in cancer and suggest there is great potential for clinical
benefits as a result of directly targeting the clock to treat cancer.
Repressor of the core clock
The negative regulator arms of the clock are also clearly implicated in cancer etiology.
Levels of the PER family genes were found repressed in clinical samples from many cancer
10
types, including PER1 in glioma, stomach, non-small cell lung cancer (NSCLC), breast and
prostate cancers, PER2 in lymphoma, leukemia, lung, stomach, and breast cancers, as
well as PER3 in colorectal cancer (CRC)56. Interestingly, an increase in PER3 expression
was observed in patients with AML and acute lymphoid leukemia (ALL) in remission, but
not in those who relapsed after treatment, suggesting that upregulation of PER3 is
associated with better clinical outcome and may have a therapeutic potential in acute
leukemia.57 Reduced PER1 and PER2 expression were associated with shorter survival in
glioma and gastric cancer. Accordingly, overexpression of PER1 in cell lines can suppress
tumor growth and induce apoptosis. Based on reporter assay in cultured cell lines, PER1
was shown to bind to the androgen receptor (AR) and inhibited AR-mediated gene
activation thus repressing prostate cancer cell growth.58 PER2 suppresses estrogen
receptor-α (ERα) transcriptional activation and is required for proteasomal-mediated
degradation of ERα in the MCF-7 breast cancer cell line. Furthermore, the addition of 17-b
estradiol (E2) induces PER2 expression, and PER2 overexpression results in inhibition of
growth and colony formation and induced apoptosis, implying the clock’s potential role as
part of a feedback mechanism in ER + breast cancer.59 In addition, PER2 may act as a
negative regulator of epithelial-mesenchymal transition (EMT) in normal mammalian
epithelial cells by interacting with OCT1.
60
The nuclear receptor REV-ERBs are relatively less studied in cancer. High expression of
REV-ERBβ was reported to be associated with poor overall survival in HCC.
61 In a cell line
study, REV-ERBα was shown to inhibit the proliferation of gastric cancer cells by regulating
glucose metabolism and the pentose phosphate pathway.62 Remarkably, pharmacological
11
activation of REV-ERBs was reported to induce apoptosis by regulating autophagy and de
novo lipogenesis and to be lethal to multiple types of tumor cells, including brain cancer,
CRC, breast cancer, leukemia, and melanoma. 63
The CRY family genes also play apparently complex roles in many cancer types, perhaps
explained by the diQerential role and tissue expression of the two isoforms, CRY1, and
CRY2. Single nucleotide polymorphisms (SNPs) in CRY1 were associated with a higher risk
of breast cancer. Increased CRY1 in CRC cells correlated with tumor advancement and
poor prognosis.64 Recently, it was found that CRY1 expression is regulated by androgen
hormones in that androgen stimulation via the addition of dihydrotestosterone leads to AR
binding to the CRY1 locus in prostate cancer cells. CRY1 was found to regulate DNA repair,
homologous recombination, and G2/M transition in the cell cycle, thereby promoting
prostate cancer cell survival and metastasis.65 By contrast, high expression levels of CRY2
correlated with better survival in breast cancer, and, consistently, a low level of CRY2 was
associated with shorter survival in HCC.66 Expression of CRY2 was also found to be lowered
in papillary and follicular thyroid carcinoma.67 These data imply a tumor- suppressive role
of CRY2. On the other hand, ablation of both Cry1/2 seems to suppress tumor
development and improve prognosis in mice with mutated p53, implying that CRY2 can
also be tumorigenic.68 Knocking out both Cry genes in hamsters resulted in the
development of HCC.
37 The role of CRY2 in tumor suppression may involve its direct
regulation of the oncogene MYC which is discussed in the following section.
The expression of core clock genes and negative regulators of the clock has been also
found to be dysregulated in esophageal and cervical cancer cell lines compared to their
12
normal counterparts, with low levels of CLOCK, CRY1, and RORA expression in cancer
cells due in part to promoter hypermethylation. Notably, overexpression of CLOCK and
PER2 attenuated cell proliferation while activation of RORα and REV-ERBα via small
molecule agonists (SR9011 and SR1078, respectively) led to increasing in apoptosis in
cervical and esophageal cancer cells, with a lesser eQect on non-cancerous control cells.69
1.3 Small Molecules for Targeting the Circadian Rhythm Gene Network
Tue to the extensive influence of circadian rhythm on various physiological processes, it
has been of interest to develop small molecules that can target the circadian genes as
potential agents for the treatment of metabolic disease, sleep disorder, and cancer. A
summary of small molecules used in this study is demonstrated in Figure 1-1.
1.3.1 Targeting BMAL1 and CLOCK
Designing small molecules that can specifically modulate the activity of BMAL1 and
CLOCK is challenging. These proteins function primarily through protein-protein
interactions and DNA binding, which are typically more diQicult to target with drugs
compared to enzymes or receptors with well-defined active sites.
One leverage for designing small molecules for BMAL1 and CLOCK is the PAS domain in
their structure. The PAS-B domain in the CLOCK protein contains a small “pocket”-like
structure that can potentially hold a small molecule.70 Targeting the PAS domain to interfere
the dimerization of bHLH-PAS TFs has been successful for HIF2A and leads to the approval
of Belzutivan by the FDA to treat tumors associated with von-Hippel Lindau syndrome.71
13
Currently there is one reported small molecule that binds to CLOCK and can interfere its
dimerization with BMAL1 and lengthen the amplitude of circadian rhythm in cells. Further
characterizations of this compounds are needed to validate its target and test its potential
use in disease treatment.
1.3.2 REV-ERB agonists
REV-ERBs are a type of TF called nuclear receptors (NR), which usually has hydrophobic
ligands that can travel through the plasmic membrane and bind to their target. Upon ligand
binding, NRs go through conformational change and activate or repress target genes. This
property makes them more amenable targets than other TFs.
GSK4112, also known as SR6452, was the first synthetic REV-ERB agonist identified,
targeting both REV-ERB isoforms and mimicking the action of heme, the natural ligand for
REV-ERBs. It could reset circadian rhythms and regulating metabolic pathways but had low
systemic exposure in vivo.72 Its derivatives, SR9009 and SR9011, exhibit improved potency,
eQicacy, and pharmacokinetic properties compared to GSK4112 and have been extensively
studied in various disease models.73 They also do not show apparent toxicity in animal
models.73
In later studies, REV-ERB-independent eQects has been reported for SR9009 in mouse
models.74 To further improve the pharmacokinetics and target specificity of SR9009, more
derivatives were developed and are being tested in animal models.75
14
1.3.3 CRY stabilizer
The first CRY stabilizer KL001 was identified in a cell-based phenotypic screening, then was
found to directly interact with CRY1/2 at the flavin adenine dinucleotide binding pocket and
inhibit their ubiquitination mediated by FBXL3 and subsequent degradation.76
Multiple derivatives of KL001 are developed to improve its pharmacological properties.
77
SHP656 is a derivative with improved bioavailability and has been investigated in diQerent
cancer models. It shows selective eQect in reducing tumor growth of glioblastoma cells
cultured in vitro and prolong survival in mouse glioblastoma model. It is also reported to
have synergistic eQect with anti-VEGF therapies in colorectal cancer mouse models.
Isotype selectivity of CRY stabilizers has also been investigated.
78,79 The selectivity for CRY
stabilizers for diQerent CRY isoforms is useful for studying the similarities and diQerences
between the CRY isoforms, their functional roles, and how the regulation of the CRYs plays
a role in a variety of diseases, including cancer.
1.3.4 CK2 inhibitor
CK2 is a clock-related protein kinase that plays important roles in cell growth and
apoptosis and tumorigenesis. Currently the most advanced drug targeting CK2 is CX-4945,
also known as Silmitasertib. CX-4945 inhibits CK2 by competitively binding to its ATPbinding site, which prevents the phosphorylation of its substrates.80 It is currently enrolled
in multiple clinical trials testing safety and eQicacy in diQerent types of cancers. However,
CX-4945 has been shown to bind to a few other targets such as Cdc2-like kinases, PIMs
and DMPKs.
81 A newer CK2 inhibitor, GO289, was identified as a period lengthening
compound through a chemical screen using U2OS mBmal1-dLuc reporter cells then went
15
through target identification.
82 In vitro kinase assays showed that other kinases are
minimally aQected by GO289. It inhibits the phosphorylation of PER2. Anti-cancer eQects
have been shown in a few cancer cell lines in culture.82 More recently, a rationally designed
drug compound was tested for CK2 and a new inhibitor, SGC-CK2-1, was proposed to be a
selective CK2 inhibitor.83 These new CK2 inhibitors will play important role in investigating
and manipulating CK2 in diseases including cancer.
1.3.5 Other modulators of the core circadian genes
Except for the small molecules that are described in the previous sections and used in this
study, others that can modulate the proteins in diQerent ways or other core circadian
components are also under active investigation, such as REV-ERB antagonists, ROR
agonists, CRY inhibitors, and CK1 inhibitors. A summary including selective small molecule
modulators of core circadian genes can be found in Ref 43.
Figure 1-1. Diagram of the small molecules that target the core circadian gene network to repress
the activity of BMAL1 and CLOCK. CRY stabilizers prevent CRY from being ubiquitinated and
16
degraded by the proteasome, thus prolong the repression of BMAL1 and CLOCK dimer activity. CK2
inhibitors achieve this through preventing the phosphorylation of PER proteins and eventually their
degradation. REV-ERB agonists elevates REV-ERB proteins’ repressive function on the transcription
of BMAL1 gene.
1.4 Regulation of Eukaryotic Transcription
1.4.1 Mechanism of gene transcription in eukaryotes
The transcription of DNA to RNA is a highly regulated process that involves multiple steps
and regulators. Protein-coding genes in eukaryotes are transcribed by RNA polymerase II.
Generally, transcription of nascent mRNA starts from assembly of the pre-initiation
complex (PIC) near the promoter of genes and recruitment of RNAPII onto the transcription
initiation site, then the RNAPII will go through release from the PIC, pausing, elongation,
and termination.84 These steps are regulated by the phosphorylation of the tail of RNAPII by
a series of cyclin-dependent kinases, transcription factors (TFs) that binds to DNA, and cisregulatory elements such as promoters and enhances that are bond to by TFs, and other
epigenetic regulators.
84
1.4.2 Targeting the Transcription Mechanisms in Cancer
Because gene transcription is a fundamental process for all cells, it used to be thought of
as not targetable in tumor while sparing normal cells. However, recent progress has shown
that tumor cells rely on the transcription mechanism diQerently from normal cells.85 This
inspired multiple strategies to selectively target cancer cells, such as small molecules that
repress the activity of cancer-driving super enhancers.86
17
1.5 Objectives and Aims
As discussed above, better understanding of the biology of cancer is needed to advance
the treatment and prolong the life span and quality of cancer patients. Circadian clock
genes are emerging as essential regulators of cancer progression, and their biology in
cancer remains to be investigated. This thesis aims to study the role of circadian genes in
cancer as potential biomarkers and therapeutic targets. Towards this objective, we propose
the following aims:
1.5.1 Aim 1. Investigating the clinical relevance of circadian clock genes in breast cancer
and their potential as diagnostic and therapeutic biomarker.
1.5.2 Aim 2. Targeting circadian transcriptional programs in triple negative breast cancer
through a cis-regulatory mechanism.
1.5.3 Aim 3. Modulating tumor microenvironment of glioblastoma through small molecules
targeting the circadian clock genes.
18
Chapter 2. Clinical Signatures of Circadian Genes in Breast Cancer
2.1 Introduction, Objective, and Rationale
Disrupted circadian rhythm among shift workers has been one of the first reported
connections between circadian rhythm and cancer. However, genetic connections
between circadian rhythm and cancer have been largely underexplored. This lack of study
limited the potential of circadian genes as biomarkers in cancer patients. One challenge of
such study is that traditional methods focus on building the association between
mutations of potential risk genes and patient outcome. However, core circadian genes are
rarely found mutated in cancer patients, and the functional implications of the mutations
are unclear. Another challenge to study circadian genes in clinical samples is that usually
only a single time point of data is available for each patient, and the sampling time is
mostly unmarked. This prevents researchers from defining rhythmic phenotypes in clinical
samples and establishing corresponding usage in the clinic.
Recent years, benefitting from large genomic datasets of cancer patients such as The
Cancer Genome Atlas and others, new studies have been reported to address these
challenges. In a systematic pan-cancer study of circadian genes, it is confirmed that
mutations of circadian genes are rare across diQerent cancer types.42 However, their
expression levels are mostly changed in tumor tissues compared to normal tissues. These
changes are diQerent in diQerent cancer types, suggesting that cancer-specific studies and
signatures are needed for future investigation. This study revealed that using gene
19
expression level for future genetic study would have more potential than using mutation
data. But in turn, it emphasized the importance of the second challenge.
To extract rhythmicity properties from unmarked data, Wu et al. proposed CYCLOPS, an
auto-regressor neural network that is trained on time-marked patient tissues to predict the
“phase” of unmarked samples.87 CYCLOPS revealed that in the liver, health tissues exhibit
a reconstructed rhythmicity, but the hepatocellular carcinoma tissues lost their
rhythmicity. Later the authors successfully applied CYCLOPS to cancer genomic datasets
and found that in breast cancer the rhythm is also largely disrupted.88 The loss of
rhythmicity showed by these results provided a rationale that for studying circadian genes
in cancer types where rhythmicity of gene expression is lost, we can study their expression
signature on the population level without factoring in time of sampling.
The objective of this chapter is to study the expression features of circadian genes in breast
cancer patient samples and test if they have the potential to be biomarkers of diagnosis
and prognosis in the clinic.
2.2 Data and Methods
2.2.1 The TCGA breast carcinoma (BRCA) patient cohort
TCGA BRCA cohort data was downloaded from UCSC Xena
(https://xenabrowser.net/datapages/?cohort=TCGA%20Breast%20Cancer%20(BRCA)&re
moveHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443). The cohort contains
1096 patients with gene expression data, of which patients have data from paired nontumor tissue. Gene expression level is assayed by next generation sequencing of mRNA
using the Illumina 2000 platform. Gene annotation and normalization was done through
20
the UCSC Xena protocol, and the gene-level transcription estimates are shown as log2(x+1)
transformed RSEM normalized count.
2.2.2 The METABRIC breast cancer cohort
The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) database
is a Canada-UK Project which contains targeted sequencing data of 1,980 primary breast
cancer samples. Clinical and genomic data was downloaded from cBioPortal
(https://www.cbioportal.org/study/summary?id=brca_metabric). The mRNA level was
assayed using Illumina HT12 V3 microarray platform.
2.2.3 Statistical Methods
DiQerential expression was tested using Wilcoxon rank sum test. Survival time of patients
in diQerent cohort was tested using log-rank test using the survminer package in R.
Logistic regression was performed on normalized estimates of gene counts using the glm
function in R. Multivariate Cox regression was performed on normalized estimates of gene
counts using the coxph function in R.
2.3 Results and Discussion
2.3.1 Expression features of the circadian clock genes in breast cancer
To confirm that the rhythmicity of circadian genes is lost on the population level, we tested
the correlation among the positive and negative arm genes in both datasets. In the normal
breast tissues from the TCGA BRCA cohort, BMAL1 exhibits negative correlation with CRY2,
PER3, REV-ERB1and REV-ERB2. This negative correlation is lost in tumor tissues from both
21
TCGA and METABRIC cohort. This result is consistent with those reported by Li et al.88 and
confirmed that rhythmicity is lost in the patient population of the two datasets.
Figure 2-1. Pair-wise Pearson correlations between genes in the core circadian rhythm network in
the TCGA and METABRIC cohort. Normal and tumor tissues in the TCGA dataset are analyzed
separately.
We then tested the change of circadian genes in cancer tissues compared to normal
tissues. By contrasting all normal and all tumor tissues expression data in the TCGA cohort,
we found that most diQerences in gene expression were in the negative arm genes (6/7, Fig.
1-3). They were downregulated in tumor tissues, whereas the positive arm genes showed
either small or insignificant diQerences. We then visualized tumor versus normal tissues
within the same patient group and observed clear trends of downregulation in tumors were
observed for PER1, PER2, and CRY2 (Figure 1a). This result implies that there is a global
pattern of change in the expression of core clock genes in breast cancer.
22
Figure 2-2. DiOerential expression analysis of core circadian genes in tumor samples versus normal
samples. (A). Fold change of core circadian genes contrasting all tumor samples versus all normal
samples in the TCGA cohort. P-value was calculated using Wilcoxon rank-sum test. (B). Parallel
comparison of core circadian gene level in paired tumor and normal tissues in the TCGA cohort.
2.3.2 A diagnostic signature of circadian genes for breast cancer
The diQerential expression result implies that there is a global pattern of change in the
expression of core clock genes in breast cancer. To identify a “tumor signature” of clock
genes, we ran logistic regression on the expression data from the cohort with paired tumor
and normal tissue (n=114) and then tested this model on the remaining tumor-only
samples (n=982, Table 2-1). The model resulted in a 6% training error and an 11% testing
error, and the coeQicients of significant factors (p < 0.01) generally agree with the
diQerential expression gene trends.
Table 2-1. Details of a logistic regression model that is trained on the expression of core circadian
genes to predict the tumor status of a tissue.
23
2.3.3 Prognostic signatures of circadian genes
We then sought to test if clock genes have prognostic implications for breast cancer
patients. For single circadian genes, when we stratified cohorts by their median expression
level, only CRY1 showed a significant diQerence in overall survival across all samples
(Figure 2-3, left). Within subtypes, only REV-ERBα and PER3 in TNBC are significant (Figure
2-3, middle and right).
Figure 2-3. Survival analysis results of stratifying patient cohorts by median expression level of core
clock genes across all subtypes or by subtype.
To test if the “tumor signature of clock genes” from the logistic model also have prognostic
implications, we calculated a “tumor clock score” by summing up the clock gene
expression levels weighted by the logistic model coeQicients, then tested it on both TCGA
24
and METABRIC datasets. Interestingly, using the median score to stratify cohorts, a clear
diQerence in survival between the two cohort is not observed. In the METABRIC cohort it
gives a statistically significant result in median overall survival, but the predicted outcome
tends to overlap towards the end of the period (Figure 2-4).
Figure 2-4. Survival analysis of the logistic tumor model on both TCGA and METABRIC cohort.
Patients are stratified by median score. P-values were calculated by log-rank test.
To find a prognostic signature of clock genes, we ran multivariate Cox regression on the
TCGA dataset. The Cox model successfully generated a risk signature (Table 2-2 and Figure
2-5, left), prompting us to test it in the METABRIC dataset as a validation. We summed
clock gene expression levels weighted by coeQicients from the Cox model and found that
the higher-score group indeed had significantly shorter overall survival (Figure 2-5, right).
Table 2-2. Multivariate Cox regression model summary
25
Figure 2-5. (Left). Validation of the Cox regression model on TCGA cohort. (Right) Results of the Cox
model tested on the METABRIC cohort data. P-values were calculated by log-rank test.
2.3.4 Triple negative breast cancer has the highest inferred activities of BMAL1 and CLOCK
BMAL1 and CLOCK are reported to be required in highly malignant tumor types with
unknown drivers such as glioblastoma and acute myeloid leukemia. These results imply a
higher likelihood that BMAL1 and CLOCK are involved in TNBC.53,54 We constructed a clock
scoring system that sums the z-scores of the expression of positive arm genes (BMAL1 and
CLOCK) and subtracts the z-score of negative arm genes (CRY1/2, PER1/2/3, and REVERBα/β). This score indicates the overall balance between positive- and negative-arm
26
genes in the core circadian circuitry. Among diQerent subtypes in the TCGA samples, TNBC
has the highest clock score (Figure 1e), supporting our hypothesis that BMAL1 and CLOCK
play a larger role in TNBC compared to other breast cancer types.
2.4 Conclusion
In this chapter we analyzed the expression signatures of circadian clock genes in breast
cancer. The results showed that these genes exhibit distinct expression features in cancer
tissues versus normal tissues. The characteristics of the expression are predictive of
whether a tissue is tumor, and provides information on the overall survival time for breast
cancer patients. These results imply important biological functions of these genes in breast
cancer and prompted us to further investigate their biology in the following chapter.
27
Chapter 3. Targeting Circadian Genes in Triple Negative Breast Cancer
3.1 Introduction, Objective, and Rationale
The core circadian genes have been reported to regulate tumor cell activities. However,
unlike traditional definition of oncogenes or tumor-suppressor genes, the role of circadian
genes in cancer can be dichotomous, which means that they can be oncogenic or tumorsuppressing in diQerent types of tumors. This requires careful definition of potential patient
groups for diQerent types and subtypes of cancers.
In this chapter, we investigate the role of the two master positive arm genes, BMAL1 and
CLOCK, in triple negative breast cancer. As the results from the previous chapter revealed,
TNBC has the highest potential activity of BMAL1 and CLOCK. In addition, in HR-positive
and HER2-positive breast cancers, the master drivers of the diseases are relatively clear,
suggesting that BMAL1 and CLOCK have a smaller chance of being the drivers. Previous
genetic studies also suggested that in ER-positive breast cancer cell lines, knocking down
BMAL1 and CLOCK do not aQect the proliferation of the cells. Therefore, we decided to
focus on investigating their role in TNBC.
TNBC is a highly heterogeneous disease. Currently a few molecular subtypes have been
defined for TNBC cells by their gene expression signature.21 The objective of this chapter is
to first define which subtypes of TNBC are sensitive to BMAL1 and CLOCK disruption. Then
we tested small molecules that modulates the circadian gene proteins to suppress the
activity of BMAL1 and CLOCK.
28
3.2 Materials and Methods
3.2.1 Cell culture
All cell lines were obtained from ATCC. MDA-MB157 (HTB-24), MDA-MB-231 (HTB-26),
MDA-MB-436 (HTB-130), MDA-MB-453 (HTB-131), HCC70 (CRL-2315), and HCC 1143 (CRL2321) were cultured in RPMI 1640 (Invitrogen Cat. 72400120) supplemented with 10% FBS.
BT549 (HTB-122) was cultured in RPMI 1640 containing 10% FBS and 0.023 U/mL insulin.
Hs-578T was cultured in DMEM with 10% FBS and 0.01 mg/mL insulin. All cells are cultured
in a 37 °C incubator with 5% CO2.
3.2.2 shRNA lentivirus production
Lentivirus vector clones expressing shBMAL1 and shCLOCK were obtained from SigmaAldrich (Sigma). shBMAL1: TRCN0000019097 and TRCN0000019096 (NM_001178.3-
1536s1c1, NM_001178.3-689s1c1), shCLOCK: TRCN0000018976 and TRCN0000018978
(NM_004898.2-1053s1c1, NM_004898.2-1494s1c1).
Lentivirus was produced by transfecting the shRNA vector with packaging (psPAX2,
Addgene #12260) and envelop (pMD2.G, Addgene #12259) vectors (both are gifts from
Didier Trono), into HEK293T cells using Lipofectamine 3000 following manufacturer’s
protocol. Briefly, HEK 293T cells were seeded in 6-well plate on day 0 to achieve a >90%
confluency on the morning of day 1 in 2 mL packaging media (OptiMEM with 5% FBS and
Sodium Pyruvate). In the morning of day 1, make Lipofectamine mix (125 µL OptiMEM + 7
µL Lipo 3000 for each well, vortex briefly) and DNA mix (125 µL OptiMEM + P3000 enhancer,
then add plasmid DNA) respectively, then mix the two by pipetting to make transfection and
29
incubate at room temperature for 15 minutes. Discard 1 mL of packaging media from each
well of HEK293T, then add 250 µL of transfection mix to each well. Six hours after
transfection, change the transfection reaction media into fresh packaging media. Twentyfour hours after transfection, the media containing lentivirus was harvested and
multiplicity of infection (MOI) was determined. Lentivirus were stored at -80 ºC until
transduction.
3.2.3 shRNA gene knockdown screening
For proliferation screening after shRNA KD, cells were plated in 6-well plates to reach a 30-
50% confluency by the time of transduction. Lentiviruses were diluted in culture media
containing 5µg/mL Polybrene (Sigma cat. TR-1003-G) to the same MOI and added to cells.
48 hours after transduction, KD eQiciency and puromycin selection were performed. To
validate KD, RNA was extracted and reverse-transcribed into cDNA to run quantitatiPCR as
described in the following section. To select for KD cells, media was exchanged into fresh
culture media containing 2 µg/mL puromycin. 72 hours after puromycin addition, cells
were stained with crystal violet. The screening was repeated three times.
For quantifying the proliferation of mMSL cell lines after KD, lentiviral transduction was
performed. After forty-eight hours, cells were treated with 2µg/mL puromycin for another
48 hours, then plated into black clear-bottom 96-well plates (Falcon cat. 353219) 48 hours
after transduction in a density of 10,000 cells per well. At indicated time points (day 0, 2,
and 4), 50µL CellTiter-Glo (Promega cat. G7572) was added to each well, incubated for 15
minutes at room temperature, and read in a TECAN luminescence reader. Proliferation was
normalized to day 0.
30
3.2.4 Quantitative RT-PCR and RNA sequencing
Total RNA was extracted with the NEB Monarch RNA miniprep kit (NEB cat. T2010S)
following manufacturer’s manual with on-column DNA digestion. For shRNA KD
experiments, total RNA was extracted 48 hours after transduction. For small moleculestreated cells, total RNA was extracted 8 or 24 hours after adding small molecules.
For quantitative PCR, 500 ng total RNA from each sample was used to perform reverse
transcription using SuperScript IV VILO Master Mix (Invitrogen cat. 11756050) following
manufacturer’s manual. Real-time PCR was done using PowerUp SYBR Green Master Mix
(Applied Biosystems cat. A25742) in a thermal cycler (Bio-Rad). Cq value was defined using
regression, normalized to PPIA89, and transformed to 2ΔΔCq.
For RNA sequencing, mRNA was selected through poly-A tail selection, library preparation
and sequencing were done with Azenta on Illumina HiSeq platforms.
3.2.5 Small molecule screening and synergy analysis
Cells were seeded in clear-bottom black 96-well plates (10,000 cells per well). Twenty-four
hours after seeding, fresh media was added in each well and small molecules of indicated
concentration were added via two-fold serial dilution. CellTiter-Glo (Promega cat. G7572)
was added to each well 72 hours after the addition of small molecules. The plate was
incubated at room temperature for 15 minutes and luminescence was determined in a
TECAN plate reader. All small molecule experiments were repeated at least three times.
Synergistic eQects were quantified using the SyngergyFinder software90.
31
3.2.6 RNA-seq data analysis
Pair-end reads in FASTQ format were processed with trimmomatic91 or trimgalore92 to
trim oQ adaptor sequences, only paired reads were kept, and sequencing quality was
checked with fastqc93 and multiqc94. QC-passed data were aligned to human genome
hg38 with hisat295 and gene counts were summarized using featureCounts96.
DiQerential gene expression was done with DEseq297. GSEA was performed using the
GSEA software98 or with fgsea99. GO and TF enrichment was performed using
gProfiler2100 and enrichr101. Volcano plots of DE genes are plotted using the R
package EnhancedVolcano102.
3.2.7 MDA-MB-231 H3K27ac ChIP-seq analysis
Raw sequencing reads were downloaded from GEO (GSE85158)103. Following trimming and
quality control using trimmomatic and fastqc respectively, reads were aligned to
human genome hg38 using bowtie2104, and broad peaks were called using MACS2105. Motifs
in the peaks were counted using the countPWM function from the Biostrings106 package
in R with matching threshold set to 87%. Co-occurrence test was performed using Fisher’s
exact test in R, and the p values were corrected using the Bonferroni method.
3.2.8 Cancer Cell Line Encyclopedia (CCLE) data analysis
Gene expression data of CCLE lines were downloaded from the Broad Institute DepMap
portal (https://depmap.org/portal/download/all/).107 Correlation analyses were
implemented and plotted using the corrplot package in R.
32
3.2.9 Luciferase reporter assay
Promoters of genes of interest were cloned from the genomic DNA of MDA-MB-231 cells
and cloned into an EcoRI-linearized luciferase reporter vector using In-Fusion cloning
(Takara Bio cat. 638948). Lentivirus was packaged using Lipofectamine 3000 following
manufacturer’s instructions. The reporter cell lines were established by lentiviral
transduction followed by 10 µg/mL blasticidin (Gibco cat. A1113902) selection for 96
hours. For single time point luciferase assay, cells were cultured for 8 hours after small
molecules were added and BrightGlo (Promega cat. G2650) luciferase assay substrate was
added to each well and luminescence was read in a TECAN plate reader. For prolonged
luciferase activity recording, small molecules were added in circadian luciferase assay
media, which was added in cells, and luminescence was recorded every hour for the
indicated time length in a TECAN plate reader. Triplicates were performed for each
condition and mean values were plotted.
3.2.10 STARR-seq Library Preparation and Cloning
STARR-seq library cloning was done as described in the detailed protocol by with minor
modifications.108 hSTARR-seq_ORI vector (Addgene #99296) was obtained from Addgene.
Genomic Library Insert Generation. Genomic DNA was extracted from cultured MDA-MB231 cells using NEB Monarch Genomic DNA Purification Kit (NEB cat. T3010). The genomic
DNA (5 µg gDNA in 80 µL water, in 1.7 mL sonification tube) was then sonicated in a
Diagenode® BIORUPTOR 300 ultrasonic processor (15 s sonication followed by 15 s pause
in 4°C water bath, 3 cycles on high strength), 20 reactions were performed. Fragmented
gDNA was run in 1% agarose gel in 20 wells at 140V for 30 minutes. Band of size range 500-
33
750 bp was cut and purified using Macherey-Nagel Gel extraction kit (MN cat. 740609)
following manufacturer’s protocol. The elution fractions were pooled and cleaned with
QIAquick PCR purification kit (QIAGEN cat. 28104), then eluted in 50µL EB buQer. The
resulting library was then ligated with Illumina sequencing adaptor using NEBNext Ultra II
DNA Library Prep Kit for Illumina (E7645S). End preparation reaction:
Reaction Program
End Prep Enzyme Mix 3 µL 20°C 30 minutes
End Prep Reaction BuQer 7 µL 65 °C 30 minutes
Fragmented DNA (500 ng) 50 µL 4 °C Hold
Total vol. 60 µL Lid 75 °C
Ligation reaction:
Reaction Program
End Prep Reaction Mixture 60 µL 20°C 15 minutes
Ligation Master Mix 30 µL 4 °C Hold
Ligation Enhancer 1 µL Lid OQ
Adaptor for Illumina 2.5 µL
Total vol. 93.5 µL
Add 3µL of USERTM Enzyme to the ligation mixture, mix well and incubate at 37 °C for 15
minutes. The reaction was purified twice using Agencourt AMPure XP beads (Beckman
Coulter Cat. A63881) following protocols provided by Muerdter and Boryn el al.
Amplification of adaptor-ligated DNA library. The library insert was amplified by PCR using
KAPA HiFi Hotstart ReadyMix (Roche cat. KK2601) in 30 reactions.
34
Reaction Program
Adaptor-ligated DNA library 1 µL 98°C 45 secs
Library cloning primer_fwd (10µM) 2.5 µL 98 °C 15 secs
Library cloning primer_rev (10µM) 2.5 µL 65 °C 30 secs
KAPA HiFi Mix 2X 25 µL 72 °C 45 secs
H2O To 50 µL Go to Step 2 10 cycles
Total vol. 50 µL 72 °C 60 secs
Every 10 PCR reactions were pooled and cleaned with Agencourt AMPure XP beads and
QIAquick PCR purification kit following the aforementioned protocol.
Restriction digest and purification of STARR-seq vector. 25 µg of STARR-seq screening
vector was used for restriction digest in a total of 500 µL reaction (25 µL AgeI-HF, 25 µL SalIHF, 50 µL CutSmart BuQer (10X), and water up to 500 µL). The reaction was incubated at
37 °C for 2 hours, heat activated at 65 °C for 20 mins, and then run on 1% agarose gel at 140
V for 30 minutes in 10 wells. Th de 3000 bp bands were cut out and extracted using MN Gel
and PCR purification kit, then cleaned with QIAquick PCR purification kit and QIAGEN
MinElute PCR purification kit (QIAGEN cat. 28006).
In-fusion HD reaction. Four reactions were pooled in one tube and a total of 20 reactions
were performed.
One Reaction Program
Linearized Plasmid 125 ng 50°C 15 minutes
PCR amplified library insert 100 ng 4 °C Hold
In-Fusion HD Enzyme Premix 2 µL
Water To 10 µL
35
Total vol. 10 µL
DNA precipitation was performed for each pooled In-Fusion reaction. The volume of InFusion reaction was adjusted to 250 µL using EB buQer, 25 µL 3M NaAc pH5.2 was added
and vortexed, then 750 µL ice-cold (-20 °C) 100% ethanol was added, followed by vortex.
The mixture was stored at -20 °C for 16 hours. Then DNA precipitation was spin down at full
centrifuge speed at 4°C and washed 3 times with 750 µL ice-cold 75% ethanol. After the
last wash, the pellets were first dried at room temperature and resuspended in 12.5 µL EB.
Library expansion. Cloning reactions are pooled before transformation and distributed to
pre-cooled 1.5 mL DNA LoBind tubes (2.5 µL in each tube of total 20 tubes). 20 µL
electrocompetent cells (MegaX DH10BTM T1R) to each tube and mixed well with DNA. The
mixture was pipetted into pre-cooled cuvettes and electroporated at 2kV, 25 µF, 200 ohms
with BioRad GenePulser. Pre-warmed recovery media (1mL per cuvette) was added
immediately after electroporation, and cells were recovered at 37 °C for 1 hour. All
transformation reactions were pooled after recovery and distributed evenly to 12 L prewarmed LB media with 100 µg/mL Ampicillin. The cells were cultured overnight while
shaking at 37 °C till the OD600 is between 2-2.6. Bacteria cells were spun down and
pooled, DNA library was extracted using QIAGEN Plasmid Giga Kit (QIAGEN cat. 12991)
following manufacturer’s protocol.
3.2.11 STARR-seq Screening
MDA-MB-231 cells were cultured in 15 cm tissue culture plates and reached 90%
confluency on the day of transformation (5 plates, total 1 x 108 cells for each treatment
group). Before transformation, drugs corresponding to each group (0.1% DMSO, 10µM
36
SHP1705, 100 nM MG132, and combination) were diluted in fresh media without FBS or
antibiotics and exchanged the culture media. Transformation was done with Lipofectamine
3000 following manufacturer’s protocol. For each 15 cm plate, 2.5 µg STARR library DNA
was diluted into 1250 µL OptiMEM, then 50 µL P3000 enhancer was added to make the
DNA mix. Then 1250 µL OptiMEM containing 65 µL well mixed Lipofectamine 3000 was add
to the DNA mix and mixed by pipetting. The transformation reaction was incubated at room
temperature for 15 minutes and added dropwise into the cells containing drugs. Eight
hours after transformation, cells were harvested using TrypLE treatment, and total RNA
was isolated with QIAGEN RNeasy Maxi Kit (QIAGEN cat. 75162) following manufacturer’s
protocol.
mRNA was isolated from total RNA using Oligo-dT magnetic beads (Dynabeads Oligo(dT)25,
Invitrogen cat. 61005) following manufacturer’s protocol. Briefly, the beads were washed
first with Binding BuQer and resuspended to the volume of the RNA solution to be purified.
Then equal volume of RNA was added and incubated on a rolling shaker for 10 minutes at
room temperature. The samples were then put on magnet for 2 minutes at room
temperature, and the supernatants were discarded. The beads were washed with Washing
BuQer B, then RNA was eluted in 10 mM Tris-HCl on an 80 °C heating block for 3 minutes
shaken at 750 rpm. Selected RNA was cleaned using NEB Monarch RNA miniprep kit with
on-column DNA digestion.
Reverse transcription (RT) was done with SuperScript III (Invitrogen cat. 18080093). For
each treatment group, 5 µg PolyA+ RNA was used to perform RT in 2 PCR tubes (2.5 µg each
tube). The reactions are set up per the following table:
37
RNA Mix RT Mix
polyA+ RNA 2.5 µg 10X RT buQer 10 µL
GSP (2 µM) 5 µL 25 mM MgCl2 20 µL
dNTP (10 mM) 5 µL 0.1 M DTT 10 µL
Water To 50 µL RNase OUT 5 µL
SuperScript III RT 5 µL
Total vol. 50 µL Total vol. 50 µL
RNA mixes were prepared first, and RT mixes were added to make the reaction. Program:
50 °C for 1 hour, 85 °C for 5 minutes, hold at 4 °C. 1 µL RNase H was added to each tube
and incubated at 37 °C for 1 hour. RT reactions were purified using AMPure XP beads.
Sequencing library was prepared in two steps. First, four junction PCR (jPCR) reactions
were performed for each treatment group.
One jPCR Reaction Program
cDNA 20 µL 98 °C 45 s
Primer (j-fwd + j-rev, 5 µM each) 5 µL 98 °C 15 s
KAPA HiFi 2X 25 µL 65 °C 30 s
72 °C 45 s
Go to Step 2 14 times
Total vol. 50 µL 72 °C 60 s
jPCR reactions were purified with AMPure beads and eluted in nuclease-free water. Then
five sequencing-ready PCR reactions were performed for each treatment group. Index
primers are from NEBNext Multiplex Oligos for Illumina (Index Primers Set 1, NEB cat.
E7335). Index Primer 2,4,6, and 12 were used for treatment group DMSO, SHP1705,
MG132, and COMBO respectively.
38
One seq-ready PCR Reaction Program
DNA from cleaned jPCR 20 µL 98 °C 45 s
Universal PCR Primer (µM) 2.5 µL 98 °C 15 s
Index Primer (10 µM) 2.5 µL 65 °C 30 s
KAPA HiFi 2X 25 72 °C 45 s
Go to Step 2 9 times
Total vol. 50 µL 72 °C 60 s
Sequencing-ready library was purified using SPRI beads (Beckman cat. B23318).
3.2.12 STARR-seq data analysis
Library complexities were calculated using the preseq program (v.2.0.0).109 After peak
calling, motif counts were done with the countPWM function in the Biostrings package
in R with threshold set to 80%. Normalized abundance of single motifs was calculated by
the following formula:
���������� ��������� = ������ �� ������ ∗ ���� ������
������ �� ����� ∗ ���������� �� �������
and normalized to DMSO group to obtain relative normalized abundance. Logistic
regression was done using the glm function in R.
3.2.13 Other statistical methods
P-values for qPCR and CellTiter-Glo were calculated by one-way ANOVA test. P-values for
luciferase reporter assay for promoter activities are calculated using unpaired t-test.
39
3.3 Results and Discussion
3.3.1Knockdown of BMAL1 and CLOCK disrupts the proliferation of metastatic
mesenchymal stem-like TNBC cells
To test the dependency of TNBC cells on BMAL1 and CLOCK, we performed a shRNA
screening on a panel of cells including TNBC cells across diQerent molecular subtypes
(Table 3-1) and two non-cancer cell lines (immortalized human mammary gland epithelial
cell MCF10A and human lung fibroblast cell IMR-90).
Table 3-1. TNBC cell line panel screened for BMAL1 and CLOCK function
Cell line TNBC Subtype Tumor Source Mutations
1 MDA-MB-231 MSL Metastasis, pleural
effusion
BRAF, CDKN2A, KRAS, NF2,
TP53, PDGFRA
2 MDA-MB-157 MSL Metastasis, pleural
effusion NF1, TP53
3 MDA-MB-436 MSL Metastasis, pleural
effusion BRCA1, TP53
4 MDA-MB-453 LAR Metastasis, pleural
effusion PI3K, CDH1, PTEN
5 Hs578T MSL Primary CDKN2A, HRAS, TP53
6 BT549 M Primary PTEN, RB1, TP53
7 HCC70 BL Primary PTEN, TP53
8 HCC1143 BL Primary TP53
9 MCF10A Immortalized
Epithelial Non-tumor n.a.
10 IMR90 Lung Fibroblast Non-tumor n.a.
40
From the initial screening, only mMSL TNBC cells showed visibly hampered proliferation
(Figure 3-1). We further quantified cell proliferation with CellTiter-Glo assay and confirmed
their reduced proliferation following knockdown (KD) (Figure 3-2). This result is consistent
with previous reports that BMAL1 and CLOCK are essential in stem-like cancer cells in
glioblastoma and AML.53,54
Supplementary Fig. 2
BMAL1 CLOCK
BMAL1 CLOCK
BMAL1 CLOCK
BMAL1 CLOCK
BMAL1 CLOCK
BMAL1 CLOCK
BMAL1 CLOCK
BMAL1 CLOCK
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱✱✱
✱✱✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱✱
✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱✱✱
✱✱✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱
✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱
✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱✱✱
✱✱✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱
✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
ns
ns
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱
✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
shCONT
shBMAL1#689
shBMAL1#97
✱✱✱✱
✱✱✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 shCONT
shCLOCK#1053
shCLOCK#1494
✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
✱✱
✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5 ✱✱✱✱
✱✱✱✱
shCONT
shBMAL1#689
shBMAL1#97
0.0
0.5
1.0
1.5
Relative Expression
✱
✱✱
shCONT
shCLOCK#1053
shCLOCK#1494
0.0
0.5
1.0
1.5
✱
✱✱
BMAL1 CLOCK BMAL1 CLOCK
41
Figure 3-1. Initial screening of proliferation phenotype after genetic knockdown of BMAL1 and
CLOCK across the panel of TNBC cell lines and two non-cancerous cell lines. Proliferation is
qualified by Crystal Violet staining in 6-well plates. P-values are calculated by one-way ANOVA.
Figure 3-2. Quantified cell proliferation of MDA-MB-231 and MDA-MB-157 cells after knockdown of
BMAL1 and CLOCK by CellTiter-Glo after 2 and 4 days of seeding. P-values are calculated by one-way
ANOVA.
To understand the mechanism of repressed proliferation by the knockdown of BMAL1 and
CLOCK, we performed RNA sequencing on all KD cells and compared the change of
Hallmark pathways (Figure 3-3). We consider pathways that are enriched in at least three
out of four groups the most relevant ones to BMAL1 and CLOCK KD. Interestingly,
epithelial-mesenchymal transition (EMT) is enriched. Because only mMSL cells are
sensitive to BMAL1 and CLOCK KD, this result further suggests that the EMT and the related
stem cell feature are potential indicators of BMAL1 and CLOCK functioning as oncogenic
genes and potential targets for therapy. On the other hand, all the enrichments with
negative NES are cell cycle-related (E2F targets, G2-M check point, and MYC target genes).
Because circadian rhythm and cell cycle are well coordinated process and B/C are
important regulators of cell cycle, this result suggests that BMAL1 and CLOCK may
contribute to cell cycle directly in the transcription level, although more detailed
mechanistic confirmation would be needed.
42
Figure 3-3. Gene set enrichment analysis on hallmark pathway gene sets after knockdown of
BMAL1 and CLOCK in MDA-MB-231 cells. Pathways that are enriched in at least 3 out of 4 groups
are highlighted in text boxes. Pathways with positive and negative normalized enrichment scores
are plotted separately.
3.3.2 CRY2 stabilizer SHP1705 represses the negative arm genes in the core circadian
network
To harness the tumor-supporting function of BMAL1 and CLOCK as potential therapeutic
targets, we proceed to investigate if their activity can be targeted by small molecules in
TNBC. Currently there is a lack of direct small-molecule modulators for BMAL1 and
CLOCK, we thus used several small molecules that strengthen the negative arms of the
clock network and in turn repress BMAL1 and CLOCK function. (Figure 1-)
To test the small molecules’ eQects on repressing BMAL1/CLOCK (B/C) activity and cell
proliferation, we first compared their IC50 (Figure 3-4) and their eQect on the expression of
canonical B/C target genes (DBP, CRY1, and PER1, Figure3-5). REV-ERB agonist SR2906575
and CK2 inhibitor GO28982 have relatively low IC50, but the expression of B/C target genes
was not suppressed. This indicates that in these TNBC cell lines that we tested, their
viability eQects are more likely due to mechanisms other than modulating B/C activity. On
the other hand, CRY2 stabilizer SHP170576,110 can faithfully repress the expression of DBP,
43
CRY1, and PER1, but has less of an eQect on cell proliferation than the other small
molecules. Because our interest is in targeting the activities of B/C, we sought to
understand and expand the function of SHP1705 via manipulation of the transcriptional
potential of the BMAL1::CLOCK complex.
Figure 3-4. IC50 values of CRY2 stabilizer SHP1705, REV-ERB agonist SR29065, and CK2 inhibitor
GO289 on three mMSL TNBC cell lines.
0.25 0.5 1 2 4 8 16 32 64 128
0
50
100
150
GO289
Conc. µM
Relative Viability
MDA-MB-231
MDA-MB-157
MDA-MB-436
44
Figure 3-5. Expression level of canonical BMAL1 and CLOCK target genes assayed by quantitative
RT-PCR. P-value was calculated by unpaired t-test.
To confirm the eQect of SHP1705 on B/C, we performed bulk RNA-seq on MDA-MB-231
cells treated with SHP1705 for either 8 or 24 hours. SHP1705 exhibited a relatively small
eQect on the mRNA transcription landscape resulting in a total of 1451 diQerentially
expressed (DE) genes, of which only a handful had more than a 1.5-fold change (Figure 3-
6). However, it was highly selective towards genes in the core circadian regulatory network
and repressed the negative-arm genes for at least 24 hours. (Figure 3-6). Gene set
enrichment analysis (GSEA) and gene ontology (GO) analysis confirmed that SHP1705
repressed core circadian gene network and had minor eQects on hallmark pathways in
TNBC, reflecting its target specificity (Figure 3-7 and Table 3-2). This focused selectivity of
SHP1705 on core circadian gene network may not be suQicient to inhibit the broader
function of B/C in mMSL TNBC cells and may explain the limited eQects that SHP1705 has
on cell proliferation as a single agent.
45
Figure 3-6. Volcano plot of DE genes after 8 or 24 hours of SHP1705 treatment on MDA-MB-231
cells. Core circadian genes are marked in read circle.
Figure 3-7. (Left) GSEA analysis on the circadian rhythm gene set showing significant enrichment.
(Right) GO term enrichment showing significant results on E-box binding and circadian rhythms terms in
the Cell Component and KEGG pathway terms, respectively.
Table 3-2. GSEA analysis on hallmarks pathways after 8 hours of SHP1705 treatment.
46
3.3.3 CRY stabilizer and proteasome inhibitors synergize to repress BMAL1 and CLOCK
activity and proliferation in mMSL TNBC cells
To improve upon the eQects observed with SHP1705 treatment, we sought for other
complementary modulators of B/C transcriptional activity. It was previously reported that
the proteasome is a direct regulator of B/C transcriptional activity.111 In this model, once
B/C initiates a round of transcription burst, the dimer becomes inactive on the DNA strand,
preventing new rounds of active dimer from binding. Proteasomal degradation of B/C is
necessary for new rounds of transcriptional bursting (Figure 3-9).
47
Figure 3-8. Reported mechanism of proteasome function in BMAL1 and CLOCK transcription
activity. In each cycle of transcription burst, inactive BMAL1 and CLOCK need to be removed by the
proteasome from DNA in order for a new round of active BMAL1 and CLOCK binding to initiate
another round of transcription burst.
We reasoned that if CRY2 and the proteasome both regulate B/C-mediated gene
transcription programs, their eQects would be synergistic. Indeed, when we tested two
proteasome inhibitors, MG132 and carfilzomib (CFZ), on all three mMSL cell lines they all
showed significant synergy in inhibiting cell proliferation (Figure 3-9).
Figure 3-9. (Top) SHP1705 synergizes with proteasome inhibitor MG132 and a clinically available
proteasome inhibitor carfilzomib to repress MDA-MB-231 cell proliferation. Results across the full
concentration spectrum were shown on the right and Bliss synergy score was calculated by
SynergyFinder. (Bottom) Synergy of SHP1705 with the two proteasome inhibitors on two other
mMSL TNBC cell lines.
In summary, we found that in mMSL TNBC cells, CRY2 stabilizer SHP1705 can selectively
suppress B/C activity in regulating core circadian network genes and synergize with
proteasome inhibitors to trigger reduced cell proliferation.
48
3.3.4 Combination of SHP1705 and MG132 inhibits the circadian transcription program
To understand and leverage the mechanism of this synergy, we performed RNA-seq on
single- and dual-drug-treated cells. We first validated the sequencing data by confirming
that CRY1, PER1, and DBP expressions are repressed by SHP1705 as tested by quantitative
PCR (Figure 3-10).
Figure 3-10. Alignment counts of CRY1, PER1, and DBP genes in RNA-seq data. Gene counts
recapitulate results from qPCR and thus validates the RNA sequencing results.
Venn diagrams showed that treatment with SHP1705 combined with either MG132 or CFZ
resulted in combination-specific sets of DE genes, which likely accounts for the observed
synergistic eQect (Figure 3-11, Top). At the dosage levels we used (100 nM MG132 and 10
nM carfilzomib, the highest concentrations that have no significant eQect on cell
proliferation by themselves), CFZ has a significant eQect on the transcriptome after 8 hours
of treatment, whereas MG132 shows almost no eQect (Figure 3-11, Bottom). CFZ was
optimized for stronger binding to the core of the enzymatic unit of the proteasome,
112 thus
we reasoned that even at a low dose, its eQect on general proteostasis can obscure its
eQect on transcription. By contrast, the eQect of the SHP1705 and MG132 combination is
more likely to be an immediate result of their disruption of the gene transcription
mechanism, because as single agents they both showed minor eQects on gene
49
transcription at the dosage level that we used. We therefore focused on analyzing the
SHP1705-MG132 combination moving forward.
Figure 3-11. (Top) Venn diagram showing the intersection of DE genes in MDA-MB-231 cells treated
with single or dual drugs. (Bottom) Volcano plot showing the diOerential expressed genes in MG132-
treated cells and carfilzomib-treated cells.
Gene set enrichment analysis on the drug combination data showed negatively enriched
E2F targets and the G2M checkpoint (Figure 3-12), which is consistent with the results in
KD analysis (Figure 3-3) and confirms that the function of BMAL1 and CLOCK are repressed
by the drug combination. These hallmark sets are not observed in CFZ-treated cells (Figure
3-12), indicating that the combination is not a mere enhancement of proteasome
inhibition.
50
Figure 3-12. GSEA results on hallmarks pathway gene sets in MG132-SHP1705 combinationtreated cells (Top) and carfilzomib-treated cells (Bottom).
Given that this drug combination directly targets part of the transcription mechanism, we
performed enrichment analysis on TF target genes to investigate the molecular
mechanism. To focus on genes that account for the synergistic eQect, we ran an
enrichment analysis on DE genes that appear exclusively in the combination group but not
in single drug-treated groups. E-box-binding factors are still significantly enriched in the
combination group, as in SHP1705-treated cells (Figure 3-13, left), suggesting the role of Eboxes in this eQect. The next most enriched TFs include NFY factors, YY1, ETS family, and
E2F family factors (Figure 3-13, right). Interestingly, the binding sites of these TFs are found
overrepresented in the promoters of CCGs, and they thus are thought as potential
combinatorial regulators of the circadian rhythmicity of gene transcription.113
51
Figure 3-13. Transcription factor enrichment analysis on DE genes from cells treated with SHP1705
(Left) or SHP1705-MG132 combination-specific (Right).
This result implies that the drug combination expanded the repression of core B/C target
genes by SHP1705 alone to a larger transcription program comprised of circadiancontrolled genes. Indeed, by intersecting DE genes in the combination group and an
annotated cycling gene set,114 we found that most (6875/9075, 76%) of the DE genes are
cycling genes, as predicted by our hypothesis (Figure 3-14).
Figure 3-14. Venn diagram showing overlap between DE genes induced by the SHP1705
combination and annotated cycling genes in primates.
3.3.5 SHP1705 and MG132 repress gene transcription through a cis-regulatory mechanism
We reasoned that the combination-exclusive genes in MG132-SHP1705 and CFZ-SHP1705
groups are core DE genes responsible for the synergy, and thus intersected the TF
52
enrichment results from these two groups (Figure 3-15, top). The refined TF list confirmed
the potential involvement of E-box-binding factors (MYC, MAX, and USF1/2) and the TFs
whose binding motifs are enriched in CCG promoters. Interestingly, we noticed that many
of these TFs bind to the same DNA motif, allowing us to further refine the list of TFs to a set
of binding motifs (Figure 3-15, bottom). Binding motifs of TFs are combined in CREs to
determine their functions, this observation thus points to a possibility that by using the
drug combination we repressed the activity of certain CREs whose functions are mediated
by these trans-acting factors. Supporting this hypothesis, the list of motifs we found in this
study largely overlaps with another list of motifs that are reported to comprise TNBCspecific super-enhancers.115 These results suggest that these motifs may constitute a
specific subtype of CREs.
53
Figure 3-15. (Top) Intersection of enriched TFs from combination-specific DE genes in the
MG132-SHP1705 and the Carfilzomib-SHP1705 group. The TFs that are shared by both
groups are colored based on their binding motifs. (Bottom) Refined list of motifs that represent
the binding sites of the enriched TFs.
To confirm that these motifs indeed are located near each other in the active CREs of
TNBC, we analyzed Histone 3 Lysine 27 acetylation (H3K27ac) ChIP-seq data of MDA-MB231 cells. By searching for clusters of these motifs, we found that these motifs indeed
collocate in CREs and can show specific arranging features such as alternating SP1 motif
and E-boxes in tandem. We also quantified significant co-occurrence of pairs of the motifs
in the clusters and found that E-boxes tend to co-occur with most other motifs (Bonferroni
adjusted p for significant level 0.05, Figure 3-16, left).
Since these motifs are collocated in the CREs, we reasoned that if they function
collaboratively, their binding TFs would have correlated expression levels as well. We
therefore tested the correlation of E-box-binding factors (EBTFs) and TFs that bind to the
hypothetical collaborative motifs, which we term hypothetical EBTF-Cofactors (EBTF-Co)
for simplicity. In the TCGA-BRCA dataset, CLOCK and ARNT showed a moderate correlation
with EBTF-Cos (Figure 3-16, middle). However, it is noteworthy that the statistical power
may be reduced because of the involvement of non-tumor tissues such as stromal and
immune cells in patient tumor samples. To alleviate the concern and focus on cancer cells,
we performed the same analysis in datasets of breast cancer cell lines from the Cancer
Cell Line Encyclopedia (CCLE).107 Interestingly, MAX, CLOCK, and ARNT showed strong
positive correlations with all EBTF-Co’s, whereas MYC, BMAL1, and HIF1A did not. (Figure
3-16, right) Among these heterodimer pairs of b-HLH TFs, MAX, CLOCK, and ARNT are b-
54
subunits whose expression level tends to be more stable and constitutive, whereas MYC,
BMAL1, and HIF1A are a-subunits that are more dynamically expressed.116 Specifically,
BMAL1 mRNA level is under strong circadian control, and MYC mRNA degrades quickly.
This dynamic nature of the a-subunits might compromise their statistical power. Therefore,
the correlation between EBTF b-units and EBTF-cofactors supports our hypothesis that
they collaboratively contribute to the CREs that contain their binding motifs.
Figure 3-16. (Left) Pair-wise co-occurrence analysis of EBTF-cofactors in the H3K27ac ChIP-seq
data from MDA-MB-231 cells. p-values were calculated using Fisher’s exact test and corrected for
multi-test using the Bonferroni method. (Middle and Right) Correlation analysis of the expression
level of E-box binding TF genes and co-factor genes in TCGA BRCA dataset (middle) and in CCLE
breast cancer dataset (right).
To test whether the activity of CREs was repressed by the drug combination, we cloned the
promoters of representative DE genes that are repressed only by the combination
treatment but not by single drug treatments (e.g. CDC20 and RUNX2). These promoters
contain the binding motifs of the EBTF-cofactors (Figure 3-17, left), and their activity was
repressed after 8 hours of combination drug treatment as measured by a luciferase
reporter assay (Figure 3-17, right). We also tested promoters of genes that are upregulated
by the dual drug treatment (HSPA5 and HMOX). Interestingly, these promoters also contain
multiple EBTF-Co motifs (Figure 3-18, left) and thus ought to be repressed according to our
hypothesis. Instead, their activities were upregulated (Figure 3-18, right).
55
Figure 3-17. Promoter sequence (left) and their activity of CDC20 and RUNX2 after drug treatment.
Figure 3-18. Promoter sequence (left) and their activity of HSPA5 and HMOX1 after drug treatment.
56
We noticed that the up-regulated genes were mostly heat shock-responsive genes, and, in
parallel, heat shock pathways were also significantly enriched in GSEA analysis (Figure 3-
19). Heat shock pathways respond to protein misfolding stress; thus the upregulated
promoter activity might result from a transient response to the stress on proteostasis.
Therefore, we monitored prolonged promoter activity by luciferase reporter assays and
found that after a transient upregulation for a few hours, the activity of these promoters
was eventually repressed (Figure 3-19, top panel). These results confirmed that the drug
combination repressed gene expression by inhibiting the activity of their promoters that
contain the motifs of EBTF-Cofactors (Figure 3-19, bottom panel).
Figure 3-19. (Top) GSEA analysis showed that Heat Shock Factor (HSF) activation and response to
heat stress pathways are enriched, conforming the activation of heat shock responsive
transcription programs. (Bottom) Prolonged luciferase activity assay showing that after a transient
activation, the activity of the heat shock-responsive promoters is eventually repressed by the
combination.
57
3.3.5 The combination of SHP1705 and MG132 suppresses a circadian transcription
program by inhibiting selected types of CREs
To define the features of the CREs that are repressed by the drug combination in an
unbiased manner, we performed Self-Transcribing Active Regulatory Region Sequencing
(STARR-seq) to test CRE activities in cells treated with single drugs or their combination
(Figure 3-20).117
Figure 3-20. STARR-seq experiment design. Genomic library from MDA-MB-231 cell genomic DNA
is inserted into the hSTARR-seq vector, and their CRE activity are quantitatively inferred as read
abundance. Peaks are called for each treatment group and motifs in each peak are counted to
generate peak-motif count matrices. Then motif analyses are implemented on the matrices.
We used fragmented genomic DNA of MDA-MB-231 cells as an input library (700 base pairlong on average, Figure 3-21, left). We achieved a complexity of more than 10 million
unique reads for all groups (Figure 3-21, right), which theoretically can cover twice of the
human genome (around 3.1 billion bp in length). The input copy number of each fragment
cannot be accurately defined in this experimental setting, therefore a precise quantitative
comparison of STARR activity across diQerent treatments cannot be reliably performed. We
thus selected the most highly enriched fragments by peak calling and performed a
qualitative comparison to define CRE types that are repressed by the treatment of drug
combination.
58
Figure 3-21. (Left) Genomic library size distribution. (Right) Library complexity of each treatment
group.
We first examined the global TFBS landscape of all STARR-seq peaks across diQerent
treatments (Figure 3-22). A total of 258 clustered non-redundant motifs (columns) were
counted through each peak (rows).118 The result shows that even with a relatively loose
stringency (80% of maximum match score for each position weight matrix), motifs are
relatively sparse in the peaks, and the drugs did not cause a visible change in the global
landscape of motifs in STARR peaks. The following analyses thus focus on the EBTF-Co
motifs.
59
Figure 3-22. Landscape of all TFBS in all STARR peaks across diOerent treatment showed that the
drug treatment did not alter the global CRE activity to a significant extent.
Firstly, we tested if the drugs changed the representation of each single motif by comparing
their abundance which is normalized to library complexity and peak number in each
treatment group. Indeed, the E-box and most EBTF-Cos were less represented in the
combination group compared to the DMSO group (Figure 3-23). This result suggests that
the overall activity of E-boxes and the other motifs are potentially repressed by the drug
combination. Interestingly, SHP1705 led to the over-representation of several motifs, but
the mechanism is unknown and can be explored in future studies.
60
Figure 3-23. Normalized motif counts of all EBTF-Co factors after diOerent treatments relative to
the control group (DMSO).
Next, we tested the pair-wise co-occurrence of each motif in diQerent treatment groups.
We first tested the co-occurrence of all 258 clustered motifs. In general, motifs tend to cooccur with only a subset of partners and show traceable patterns in a heatmap, although
we observe an inflation of co-occurrence in SHP1705 group (Figure 3-24, top). This result
supports that we can subtyping CREs based on their motif composition and test their
functionality. Comparing the combination to the DMSO group, we found several significant
co-occurrences that were lost, including E-box with SP, ETS, and YY1; SP with IRF and YY1;
NFY with JUN; IRF with PBX3 and JUN (Figure 3-24, bottom). These losses of significance
imply that the enhancer activities of peaks containing these motifs together are underrepresented in the combination group, thus constituting promoters that are likely to be
repressed by the drug combination.
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
DMSO
SHP1705 MG132
COMBO
0
1
2
4
Relative Normalized Abundace
EBOX SP E2F NFY ETS IRF SIX5 PBX3 YY RUNX JUN
61
Figure 3-24. Co-occurrence analysis on 258 non-redundant motif clusters (top) and on EBTFassociated motifs (bottom) in all treatment groups. Significant co-occurrences that were lost in the
combination group are highlighted by a white frame.
To catch more specific features of CREs that contain these motifs, we quantified the
numbers of motifs in STARR-seq peaks from each treatment group and performed
62
dimension reduction by principal component analysis (PCA, Figure 3-25). However, each
PC accounts for the same variance as single variables and no clear cluster can be defined
(Figure 3-25, middle). This might be a result of the large sample size and relatively low
dimension of variables. However, we observed that many peaks clustered into one almost
exact point on the PC plot. This is likely because they contain same types of motifs but
diQerent counts. According to the “TF collective” model of enhancer grammar, numbers of
each TF-binding motif in an enhancer exhibit certain degrees of flexibility for its activity.119 A
high-throughput screening of CRE activities of motifs showed that most motifs exhibit
relatively low activity as a single functional unit of a CRE, even many of their activity linearly
increase as the number increases.120 We thus assume that the combination of motifs plays
a more important role than the number of each type of motifs, and then focused on the
presence or absence of each motif and excluded their counts. The motif count matrices
thus can be collapsed into binary-entry matrices that indicate if the motif exists (“1”) in a
peak or not (“0”). This brings the 49,000 to 79,000 peaks in diQerent treatment groups down
to around 1064 “CRE types” defined by their motif components, which is less than the total
possible combinations of the 11 motifs (2048 theoretical ways of combinations).
63
Figure 3-25. PCA analysis on the motif counting matrix did not reveal meaningful subtypes of the
CREs based on their motif counts, and c. each PC accounts for no more variance than single
dimension.
Generally, each STARR-seq-defined CRE type contains at least 3 diQerent motifs, which is
consistent with the fact that eukaryotic CREs are usually controlled by multiple factors. We
defined combination-repressed CRE types in the DMSO group as types of more than a twofold reduction of normalized reads from all peaks that belong to that type. We found 188
types out of over 1000 that are repressed (Figure 3-26).
Figure 3-26. Summarized CRE “types” that are potentially repressed by the drug combination and
detailed motif distributions in CREs from selective types. Each row in the heatmap represents a
“type” of CRE defined by the existence of indicated motifs in it. Representative CREs belonging to
the repressed types are plotted on the side, numbers on each motif represent their location in the
CRE. Direction of the motif-representing color square represent the strand on which the motif is
located.
Additionally, most of the repressed types contain E-boxes (93/188), confirming our
hypothesis that the E-box is one of the major active motifs that is repressed by the drug
combination. We also trained a logistic model based on single motifs and found that IRF,
E2F, EBOX, and ETS are among the most important contributors to the repression by the
64
drug combination (Figure 3-27, left). In all, the repressed motifs may represent a
transcriptional mechanism that is yet to be revealed.
To understand more specific features of the repressed CRE types, we visualized motif
distributions in the peaks that belong to them by finding motif clusters.121 (Figure 3-26,
zoomed-in on the right) The numbers and locations of the motifs show great diversity, and
no obvious unifying feature were noticed, but they generally match the defined types.
Interestingly, Most E-boxes accompany another motif in proximity (<50 bp, Figure 5c),
implying that they might collaborate functionally, such as in DNA binding. To quantify the
potential eQect of motif combinations on the responsiveness to drug-combination
treatment, for each CRE we constructed a vector that indicates if each pair of motifs
cooccur in this CRE and performed logistic regression (Figure 3-27, right). All the pairs that
lost co-occurrence significance in Fisher’s test all appeared as positive contributors in
logistic regression, confirming the consistency of our result (Figure 3-27, red points). This
logistic model can correctly classify 87% of the defined repressed type, but also shows a
relatively high rate of false positives (50%). A more focused library will be needed to define
a more refined model to quantify CRE types that are potentially repressed by the drug
combination.
IRF
E2F
ETS
EBOX
JUN
NFY
SP
YY1
PBX3
SIX5
RUNX
-2
-1
0
1
2
3
4
Motifs sorted by importance
Regression Coefficient
-2
-1
0
1
2
3
Cooccurrences sorted by importance
Regression Coefficient
IRF_JUN
EBOX_ETS
IRF_PBX3
EBOX_YY1
SP_IRF SP_YY1NFY_JUN
EBOX_SP
65
Figure 3-27. (Left) Contribution of single motifs to the potential repression by drug combination.
(Right) Contribution of motif pairs to being a “combination-repressed type”. Pairs that lost
significance in co-occurrence analysis are colored in red.
In summary, by using unbiased high-throughput CRE screening we found combining a CRY
stabilizer with proteasome inhibitors suppresses the circadian gene transcription program
by inhibiting CREs of specific types defined by their distinct combinations of TFBSs.
Although more focused screenings and functional assays need to be done to validate and
quantify these types, the results provide a foundation that CREs can be manipulated by
small molecules and serve as a potential therapeutic modality in TNBC.
3.4 Discussions and Conclusion
Genes that control the circadian rhythm have emerged as important regulators of cancer,
and the growing availability of their small molecule modulators makes them attractive
targets for cancer treatment. However, the complexity of their biological functions in
cancer has only started to be revealed. As ubiquitous regulators of cell activity, circadian
genes can become oncogenic54 or tumor-suppressing65 in diQerent tumor types and tissue
contexts. Previously we reported that in glioblastoma stem cells, BMAL1 has an altered
binding landscape in the genome compared to noncancerous cells, resulting in new
binding sites that rewire the metabolic gene networks to fuel tumor growth and also drive
stemness.54 Others also reported the important roles of CLOCK in regulating the tumor
microenvironment and immune responses in glioblastoma.55 However, a deeper biological
model that connects their direct function, transcription, to physiological activities in
cancer requires elaboration.
66
Breast cancer is among the types of cancer that are better understood in terms of their
underlying biology and potential treatment options. To provide comparative insights, we
selected breast cancer as our model for investigating the oncogenic function of BMAL1 and
CLOCK, with a particular focus on the transcriptional activity, using small molecule
modulators as experimental tools. We first showed that circadian genes have clinical
relevance in the diagnosis and prognosis of breast cancer patients. Then we used RNAi to
screen TNBC cell lines that are derived from diQerent molecular subtypes. We found that
only mMSL cells rely on BMAL1 and CLOCK to proliferate. This result reinforced our current
knowledge that cancer cells with highly stem cell-like features may rely on BMAL1 and
CLOCK.53,54 Interestingly, GSEA analysis indicated that EMT processes are significantly
altered after BMAL1 and CLOCK KD. This implies potential role of BMAL1 and CLOCK in the
metastasis of breast cancer. We thus applied small molecules to target their
transcriptional activity in these TNBC cells.
Transcription factors are usually small and contain mostly intrinsically disordered regions,
thus it is extremely challenging to find selective small molecule ligands to bind them and
manipulate their functions. However, in cases where eQective drugs are available, they
exemplify cancer therapies with the best outcomes. For example, antagonizing estrogen
receptor (ER), which is a ligand-binding transcription factor, can achieve a 90% five-year
survival in ER-positive breast cancer.
17,18 Inhibiting HIF2A, a TF that belongs to the same bHLH-PAS family as BMAL1 and CLOCK, with the small molecule drug Belzutifan can
eQectively suppress multiple types of tumors associated with the von Hippel-Lindau
disease.71 Here we leveraged the negative arms of the circadian network to indirectly
67
achieve repression of BMAL1 and CLOCK activity. Probably due to the core circadian
network is rewired in these highly malignant TNBC cells, only the CRY stabilizer SHP1705
achieved faithful repression of core BMAL1 and CLOCK targets. To expand the repressive
eQect of SHP1705, we added inhibitors of the proteasome, which was previously reported
to be indispensable in maintaining BMAL1 and CLOCK-mediated transcription. This drug
combination indeed displayed synergy against cell proliferation and eQectively repressed
circadian cycling genes.
This synergy provides us both a leverage to study the transcriptional programs for which
BMAL1 and CLOCK are responsible in fueling tumor growth, and a potential therapeutic
strategy to suppress that program. In ER-driven breast cancer, ER serves as the master TF
that can strongly induce downstream transcriptional programs upon binding to its
activating ligand.
122,123 However. in mMSL TNBC, BMAL1 and CLOCK do not appear to
function as a “binary switch” as ER, because a CRY stabilizer faithfully repressed wellestablished B/C target genes but has little impact on the global transcriptome and failed to
recapitulate the strong phenotype observed following genetic knockdown of B/C.
To understand the disparity between these driver TFs of breast cancer, we tried to examine
their function from the DNA side, that is, the CREs through which the TFs function. ER’s
potent activity is mediated by the strong enhancer activities induced by it to drive
transcription.
124 These enhancers that are activated by singular kinds of factors are
common among developmental and lineage-determining TFs, but the similar concepts are
less applicable for constitutive housekeeping TFs like BMAL1 and CLOCK.125,126 In an
analysis of super enhancers (SEs) in TNBC, a group of TF binding motifs including E-box
68
was identified to constitute the SEs.115 Although currently we have no evidence to support
that this is a SE-mediated mechanism, this finding inspired us to hypothesize that strong
CREs in TNBC can be activated and maintained by multiple TFs in the absence of a singular
master driver.
Following this logic, we examined CRE activities after drug treatment and defined features
of CREs that are potentially repressed by the combination of CRY2 stabilizer and
proteasome inhibitor. These CREs comprise diQerent combinations of the binding motifs of
several constitutive TFs, such as SP1, E2Fs, NFYs, and YY1. These factors may
collaboratively activate CREs that drive tumor-fueling transcription programs. The
mechanism by which the eQects of small molecules are translated from these trans-acting
(TFs) to cis-acting factors remains to be studied. We hypothesize that it is achieved by
interfering with the dynamics of the CREs since both drugs disrupt the turnover of the TFs
bound to the CREs. Nonetheless, our results revealed a potential mechanism by which we
can target selective CREs based on their regulatory properties. This model provided a new
perspective of modulating gene transcription programs with greater precision in cancer,
opening a door for defining cancer drivers and biomarkers based on cis-regulatory
programs, and tailoring corresponding therapies.
Meanwhile, this model is still primitive because of several limiting factors. First, our current
library lacks suQicient focus to provide a more detailed and determinative description of
the features of CREs that are repressed by the drugs. Second, dependencies of the active
CREs on core promoter features cannot be assayed by the current STARR-seq setup
because the vector uses a TATA-box-dependent promoter, whereas E-box-containing CREs
69
also often work with CG-rich promoters that lack a TATA-box.
127 Third, there is a deficiency
in both our knowledge of the intricate biology, especially the dynamics, of CREs and the
corresponding computational methodologies. In the future, more comprehensive
quantifications of CRE features, such as motif composition and position dependency,
needs to be done with more focused library screening and more precise computational
methods to better define the target CREs of the drugs. Mechanistic studies are also needed
to strengthen the biological rationale for targeting cis-element activities through
manipulating their partner trans-acting factors. Finally, to translate this model into preclinical and clinical studies, optimizing proteasome inhibitors for in vivo stability might be
required. This optimization should aim to maintain the global proteostasis of cells while
eQectively targeting transcription.
In summary, this study identified the master circadian clock genes BMAL1 and CLOCK as
potential targets in the mMSL type of TNBC. Furthermore, we proposed a new biological
model for targeting CREs with small molecules. These results oQer new opportunities for
understanding driving mechanisms, discovering new biomarkers, and designing
therapeutics for TNBC.
70
Chapter 4. EGects of CK2 Inhibitors on the Microenvironment of
Glioblastoma
4.1 Introduction, Objective, and Rationale
CLOCK is reported to be an important regulator in glioblastoma that modulates multiple
part of the TME, including microglia recruitment, angiogenesis, and other immune cell
activities. Given the importance of the TME in tumor growth and treatment, in this chapter
we investigate the eQect of the small molecules that target the core circadian genes on the
TME component of glioblastoma.
A major challenge to study the TME of glioblastoma is the lack of appropriate models for
testing. Study of TME, especially the components in the immune system, usually requires
syngeneic animal models where native immunity against the tumors are present. However,
such model is currently not easily accessible. In this chapter, we tested several
bioengineered models to study the eQect of small molecules on the TME component of
glioblastoma.
4.2 Materials and Methods
4.2.1 Cell Culture
Glioblastoma stem cells (GSCs) T3565 and T387, non-malignant neural cells NM263 and
NM290 are gifts from the laboratory of Dr. Jeremy Rich at the University of Pittsburgh School
of Medicine. Glioblastoma stem cell MGG31 is a gift from Harvard Medical School. All cells
are cultured in Neuralbasal A media (Gibco Cat. 10888022) containing B-27 supplement
(Gibco Cat. A3285801) in a humidified incubator containing 5% CO2. Culture media is
71
changed every 2-3 days and cells are passaged when neural spheres formed by GBM or
non-malignant cells are 100 µM in diameter. Accutase cell detachment solution (Sigma
Cat. SCR-005) was used to dissociate neural spheres at room temperature for passaging.
Microglia cell line HMC-3 was purchased from ATCC (Cat. CRL-3304) and cultured in MEM
media containing 10% FBS.
4.2.2 Small molecule information
CX4945 and SGC-CK2-1 was purchased from Selleck Chemicals. GO289 was a gift from Dr.
Tsuyoshi Hirota’s lab in Nagoya University.
4.2.3 shRNA gene knockdown
The production of shRNA lentivirus and gene knockdown experiments are performed as
described in section 3.2.2.
4.2.4 Media conditioning
To perform chemotaxis on microglia cells, cell conditioned media was produced. In brief,
48 hours after lentiviral transduction for shRNA, neural spheres were harvested and
dissociated. 50,000 cells were seeded in each well of a 6-well plate and cultured for 48
hours. The cells were then spun down at 1000 rpm for 3 minutes and the conditioned
media was harvested and stored in -80 C until usage.
4.2.5 Chemotaxis assay
Chemotaxis was performed using µ-slide chemotaxis purchased from ibidi (Ibidi Cat.
80326). On the afternoon of the first day of experiment, 2000 HMC cells were seeded into
the middle chamber (migration channel) of the chemotaxis slide, the slide was then put in a
72
petri dish with wet paper to keep humidity and incubated for 6 hours or overnight. When
HMC cells are adhered to the migration channel, conditioned media were added to the leftand right-hand side of the chemokine chamber. The slide was then imaged in a Nikon
microscopy supplied with 5% CO2 for 12 hours. The time-series image was analyzed using
MatLab to trace the movement of each cell, and their displacement was quantified.
4.2.4 Neural sphere invasion assay
Cultrex Basement Membrane Extract 2 (BME2) (Biotechne cat. 3432-010-01) was used for
the culture of neural spheres in 3D. Cultured glioblastoma cells were harvested from
culture and dissociated into single cells. For each 3D neural sphere culture, 10,000 cells
were used and collected as a pellet in a 1.7 mL Eppendorf tube. BME2 was thawed on ice
and added to the cell pellet (60 µL per 10,000 cells), cells were resuspended by thorough
pipetting. 50 µL BME2 containing cells were seeded in the middle of the wells of a 24-well
plate and incubated in a 37 C incubator for 1 hour for the BME to solidify. Then culture
media was added to the cells, fresh media was exchanged every 3 days.
4.3 Results
4.3.1Microfluidic chemotaxis assay shows reduced microglia recruitment by GSC cells
after knockdown of CLOCK
To validate the role of CLOCK in glioblastoma for recruiting microglia cells, we established
a chemotaxis assay based on a microfluidic chip to test the recruitment of microglia cells.
First, we tested if GSC-conditioned media will trigger transcriptional changes of immuneregulatory genes in the microglia cell line HMC3. We found that after 48 hours of
73
conditioning, the culture media of GSC can induce changes of important regulators of the
TME in HMC3 cells such as VEGFA, CCL2, and IL-1B (Figure 4-1). This suggests that
conditioned media can potentially be used for functional assay of microglia.
Figure 4-1. Conditioned media by GSC cells change the expression level of TME-regulatory genes in
microglia cells.
We then used shCLOCK-KD GSC cells to condition the media and compare its eQect on
microglia migration to shCONT GSC conditioned media. The results showed that the
microglia cells migrated more towards the control media versus media conditioned by
shCLOCK-KD GSC. This confirmed that CLOCK aids in recruiting microglia in glioblastoma
and the chemotaxis assay of HMC3 is suitable for testing microglia recruitment in
glioblastoma. (Figure 4-2)
74
Figure 4-2. Chemotaxis assay showing that after CLOCK knockdown, the GSC cells show reduced
recruitment of microglia cells compared to control group.
4.3.2 The eQect of small molecules on the proliferation and TME-related gene expression in
glioblastoma cells
CLOCK has been reported to regulate the TME of glioblastoma through the genes OLFML3
and LGMN. We first use quantitative RT-PCR to test if the expression level of these genes is
changed by small molecules that target the circadian genes. We also included several
other genes that are master regulators of TME-related cellular processes, such as VEGFA,
IL6, TGF, and CCL2. Interestingly, GSC cells T3565 and T387 express trace level of OLFML3
(Cycle number > 35 in quantitative RT-PCR), while all three cell lines express. Consistent
with results in TNBC cells, SHP1705 can reliably repress the expression of canonical target
genes of BMAL1 and CLOCK. However, it showed no overt eQect on the expression level of
TME-related genes we tested (Figure 4-3). By contrast, GO289 repress the expression of
75
TME related genes and interestingly MYC. This result shows us that CK2 inhibitors have the
potential of modulating TME component of glioblastoma.
Figure 4-3. EOect of small molecules targeting core circadian genes on the expression level of
TME-related genes after 24-hours of treatment.
We then choose to test current available CK2 inhibitors in GSC cells. To define the
sensitivity of GSC cells to CK2 inhibitors, we characterized their IC50 value using CellTiter
Glo assay (Figure 4-2). The three drugs show diQerent IC50 values towards the nonmalignant and GSC cells. Interestingly, GO298 exhibits selective inhibitive eQects towards
the two GSC cell lines, whereas the other two drugs inhibit all cells to a similar extent.
4.3.3 CK2 inhibitors repress the invasion of GSC cells into the matrix
CK2 was reported to promote the invasion of glioblastoma cells. We tested the eQect of
GO289 on repressing the extension of invasive ends in a 3D culture assay (Figure4-4). As
76
the concentration of GO289 increases, the length of the invasive sprouts is significantly
reduced.
Figure 4-4. Microscopic image showing that the invasion sprouts of glioblastoma spheres are
significantly shortened after GO289 treatment.
4.4 Discussion and Conclusion
In this chapter, we optimized bioengineered assays to characterize the TME properties of
glioblastoma to test the eQect of drugs. We validated that knockdown of CLOCK reduces
the recruitment of microglia cells. We also tested the eQect of diQerent small molecules in
regulating the TME components of glioblastoma. We showed that GO289 can reduce the
invasion of GSC cells into the cell matrix. These results provided potential new ways of
establishing accessible model systems to study the TME of glioblastoma, especially for
addressing the needs of drug testing. We also found that modulating CK2 could be a
potential therapeutic strategy for reducing invasion of glioblastoma cells into the
surrounding area.
77
Chapter 5. E-box-binding Transcription Factors in Cancer
5.1 Introduction
In the previous parts we proposed a model of targeting CREs based on their composition of
transcription factor binding motifs, with a focus on E-box. However, the relationships
between TFs that binds to the same motif and the mechanism of how dynamics of the
activities of such CREs are largely unknown. Experimental study of the molecular
mechanism is beyond the scope of this dissertation, but in this chapter, we propose a
thorough review on current knowledge of E-box binding TFs and their roles in cancer as a
prelude of future studies.
An E-box is a regulatory motif of DNA, with the consensus sequence 5’-CANNTG-3’, that is
found abundantly in most eukaryotic genomes. Transcription factors (TFs) can bind to Eboxes in the promoter and enhancer region of genes through their basic helix-loop-helix
(bHLH) domain or zinc finger domain to regulate their expression. E-box-binding TFs
(EBTFs) regulate genes that are diverse in function. During development EBTFs determine
the lineage commitment of skeletal muscle, cardiovascular and neuronal tissues, as well
as hematopoiesis. In homeostasis they regulate many housekeeping genes and essential
physiological processes, such as the cell cycle, circadian rhythm, and metabolism.
Therefore, it is not surprising that EBTFs have fundamental functions in maintaining
homeostasis and are deeply involved in tumorigenesis.
Several families of EBTFs are widely studied in cancer biology. The MYC family of proteins
feature prominently, as around 28% of tumors harbor at least one amplification of a MYC
78
paralog, making it one of the most dysregulated oncogenic genes in human cancer.128
Hypoxia-inducible factor (HIF) proteins are also well known EBTFs, and the hypoxic
hallmark of solid tumors has attracted much attention to this E-box-binding family of
genes, which have also been shown to regulate many other processes in tumor
development.129 Further EBTFs include those that regulate the epithelial-mesenchymal
transition (EMT). It is generally accepted that these TFs are associated with stem cell
features of cancer cells.130,131 More lately, our lab and others have shown that the master
circadian E-box-binding regulators, BMAL1 and CLOCK, also have important functions in
several cancer types.43,53–55 These examples emphasize the importance of EBTFs in cancer
biology.
Despite all the progress that has been made, the exact molecular roles of EBTFs in cancer
is far from clear. In particular, a basic understanding of how these TFs select and
participate in the transcription processes mediated by PolII is still lacking. In recent years,
we have witnessed great progress in understanding the PolII transcriptional machinery in
detail and the biology of regulatory elements in DNA. Such new knowledge provides
unprecedented opportunities to rethink EBTFs in their most native role as DNA-binding
proteins. By doing so we might be able to better understand this family of proteins and
develop better strategies to target the TFs in cancer.
In this chapter, we review current understandings of the structure and molecular biology of
the EBTF families that have been shown to play important roles in tumorigenesis. We
discuss their mutual regulation to gain some insights into how these proteins are
coordinated during tumorigenesis and tumor suppression, and we summarize the
79
common processes they convergently regulate. At last, we propose that targeting the
whole EBTF network in specific cancer types could be eQective in suppressing multiple
hallmarks of cancer simultaneously and have potential as a cancer therapeutic strategy.
5.2 Important EBTFs in cancer
5.2.1 MYC family proteins
MYC was first discovered as a homolog of the viral oncogene v-myc in multiple chick
retroviruses; thus the gene was named cellular-MYC (c-MYC) to specify its endogeneity.
The MYC family has three members: the most prominent c-MYC, MYCN, which was initially
found to be associated with neuroblastoma, and MYCL associated with small cell lung
cancer, hence the names.132 The three MYC proteins have relatively limited sequence
consensus, but they all share the entirely conserved bHLH-Leucine Zipper (bHLH-LZ)
domain that binds to the E-box, and six highly conserved MYC boxes (MYCL lacks MB3a)
that are known for interacting with other proteins.133 (Figure 5-1, blue block) MYC will be
used to refer to c-MYC in this article. Upon heterodimerization with its partner, such as
MAX, MYC preferentially bind to the canonical E-box sequence 5’-CACGTG-3’.134,135
The role of MYC in human tumorigenesis was exemplified by the translocated MYC coding
sequences downstream from the immunoglobulin heavy chain enhancer in Burkitt
lymphomas.136–138 Since then, MYC is found to be one of the most dysregulated, usually
over-activated, oncogenic gene in human cancers.
128 It is classified as a tumors-driving
master transcription factor (MTF) in certain cancer types.139 Overexpression of MYC alone
is suQicient to trigger a cancerous phenotypic change in cultured cells, and to induce de
80
novo tumorigenesis in multiple mouse models.140,141 The importance of MYC is also
underscored by the fact that repression of MYC can result in fast regression of tumors in
animal models, making it a promising target for tumor therapy.135 Despite its pivotal role in
tumorigenesis and the great attention it attracted, the exact behavior of MYC is still far
from clear.
The fundamental of MYC biology is that MYC function diQers when expressed at high
levels, as in many tumor cells, versus at relatively low physiological levels.
133 MYC
expression is ubiquitous but is delicately regulated to be kept at a low level in normal
tissue. The turnover of MYC proteins is fast with a half-life of around 30 minutes.142 When in
low abundance, MYC mostly binds to E-boxes and their close variants, whereas when
overexpressed, it binds to more non-specific binding sites.143 This feature might be a result
of the intrinsic disordered properties of the MYC protein, which allows it to dynamically
interact with multiple partners in modest aQinity. Therefore, in high concentrations, MYC
specificity is easily overridden by a mass-action drive, leading to superfluous binding.134
The consequence of MYC binding is complex. There are currently several models
describing the mechanism.
Classically, MYC is thought to be a pleiotropic transcription factor that activates, rather
weakly, the transcription of genes through binding to the E-boxes in their promoters, as a
heterodimer with its canonical cofactor MAX, which is also a bHLH-LZ protein.144 This
model implies a group of “target genes” that are regulated by MYC. Attempts to recognize a
set of MYC target genes using diQerent large-scale analyses has resulted in sets with
surprisingly small overlaps.133 This disparity might be a result of the abundance of E-boxes
81
in the genome and their diQerent open states in diQerent cells, since MYC is generally
considered a non-pioneer transcription factor, meaning that it only binds to chromatin
regions that are already accessible but cannot open a closed chromatin, as was clearly
shown in iPSC studies.145 To some extent, there is consensus over MYC target genes,
including the HALLMARK MYC Target gene sets proposed by Liberzon et al. in the
molecular signature database (MSigDB).146 Such core common target gene sets have been
useful as indicators of MYC activity in hypotheses-generating cases, but they should not be
used to preclude potential genes regulated by MYC, especially in cancerous contexts
where chromatin accessibility is largely remodeled and mutations in regulatory elements
are common, resulting in de novo binding sites of TFs.
Apart from being an activator, MYC has also been proposed as a repressor of gene
expression.147 The most studied repressive mechanism of MYC is through its interaction
with two other proteins, MIZ1148 and SP1149, to recruit co-repressors. It has been proposed
that the repressive function of MYC is of comparable importance to the activating
function,
150 although genomic-level correlation analysis implies rather weak eQects of the
repressive function of TFs in general.151 Finally, it is noteworthy that MYC can also regulate
RNAPlI and RNAPIII-mediated transcription of ribosomal RNA and tRNA,
152,153 but this
function is out of the scope of this review.
As a weak activator model is insuQicient to explain its broad participation in various
physiological processes of the cell and its strong tumorigenic eQect, later a “general gene
amplifier” model was proposed by two simultaneous papers and introduced a new view of
MYC.154,155 According to this model, MYC can act as an amplifier that increases the overall
82
RNA production of the whole cell. This model can explain some observations in MYCdriven tumorigenesis, but still oversimplifies MYC function since elevated RNA production
is neither suQicient nor necessary for tumorigenesis, and it fails to explain the complicated
up- and down-regulation of genes after its levels change.
Given the diQerent behaviors of MYC at distinct levels in the cell, a gene-specific aQinity
model has also been proposed.144,156 According to this model, promoters of diQerent genes
require diQerent levels of MYC protein to activate their transcription. This is possible due to
the relatively low aQinity of MYC-MAX binding. This model provides an explanation of the
paradox of broad DNA-binding and specific gene regulation by MYC, but still lacks the
ability to unify MYC function in diQerent tumors and ignores the broad involvement of MYC
in multiple processes of RNAPII-mediated transcription.
Recent progress on characterizing the protein interactome of MYC has shed new lights on
understanding the function of MYC and further expands the role of MYC function in cancer.
The interactome of MYC was revealed by mass-spectrometry analysis of
immunoprecipitated MYC or through BioID screening.
157,158 Functional assays combined
with selective depletion of certain MYC boxes has also revealed specific functions of
diQerent MYC boxes. These studies have helped to define a core group of MYC-associated
proteins.144,159 These MYC interactors mark the broadness of MYC function since they are
involved in various fundamental cellular processes, such as chromatin topology and
remodeling, the cell cycle, general transcription, and ubiquitination. The interactome also
reveals more fundamental functions of MYC in participating in general transcription
mechanisms, including the formation of the preinitiation complex, initiation, pausing,
83
elongation, and splicing.159 This aspect of MYC function reinforces the essentiality of MYC
in tumorigenesis.144
These newly revealed mechanisms urge that more basic structural understanding and
regulatory element logic are key to dissect the role of MYC in cancer.
144
Figure 5-1. Structural overview of EBTFs. Functional domains of EBTFs. Proteins can bind to Eboxes through a bHLH domain or a zinc finger domain. MYC family genes contain an extra leucine
zipper domain downstream of bHLH. The three MYC family genes share highly conserved MYCboxes (MB), except for MYCL lacking the MB3a. Other regions of the proteins have low degrees of
conservation. MYC also contains a PEST domain (not shown) that might contribute to its fast
turnover. MAX is a relatively small protein with a defined bHLH-LZ domain, but functions of the
other regions of the protein are not well understood. Apart from MYC, MAX can also dimerize with
itself or other bHLH TFs such as MAD and MLX. HIF family TFs feature the bHLH-PAS family TFs,
which have a PAS-A and PAS-B domain immediately upstream of the bHLH. The PAS domain
provides additional control of protein dimerization and might have a responsive function to
environmental cues. A PAS-associated COOH-terminal (PAC) occurs C-terminal to the PAS motifs
and is proposed to contribute to the PAS domain fold. The a-subunits of HIF proteins has an
oxygen-dependent degradation (ODD) domain that contains two conserved prolyl residue (402
84
ODD and 564 ODD) that can be hydroxylated and induce proteasome-mediated degradation. They
also have two transactivation domains (C-TAD and N-TAD) that facilitate target gene expression.
BMAL1 and CLOCK are also bHLH-PAS family TFs, and BMAL1 has a defined TAD in the N-terminus.
The ZEB and SNAI family bind E-boxes through their zinc fingers as monomers. ZEB1 and ZEB2
share up to 85% amino acid sequence homology in their zinc finger clusters but have low
conservancy elsewhere. They both have a homeobox which seems to not bind DNA. Other defined
domains are best known for interacting with other proteins. SNAI family proteins are featured by
their conserved SNAG domain which is initially found in SNAI and GFI family proteins and a zinc
finger cluster in the C- terminus. TWIST is a small bHLH protein that contains a characteristic
TWIST box and binds to DNA as dimers.
5.2.2 Hypoxia-Inducible Factors
The hypoxia-inducible factor (HIF) family TFs exemplify the bHLH-PAS family of proteins. In
general, these proteins are characterized by a PER-ARNT-SIM (PAS) domain, which
contains PAS-A and PAS-B located instantly upstream the bHLH domain.160,161 (Figure 5-1,
pink block) Similar to MYC and MAX, bHLH-PAS proteins also form heterodimers, with an
α-subunit serving as a stimuli-responder or regulator of tissue specificity and the β-subunit
expressed more stably and ubiquitously. The PAS domain serves as another layer of
dimerization control on top of bHLH for higher specificity, and the PAS-B domain can
sometimes serve as a sensing domain that can bind to small molecules in the environment
or sensory/regulatory proteins.162 bHLH-PAS TFs have a more defined structure than MYC,
which is rather disordered. In the case of HIF proteins, there are three α-subunits, HIF1, 2,
and 3-α. They can all dimerize with the ubiquitous β-subunit HIF1β (also known as
ARNT).163 Dimerization determines the binding to specific E-box variants. The consensus
binding motif of the HIFα-ARNT dimer is 5’-A/G-CGTG-3’, which is an E-box variant usually
referred to as the hypoxia responsive element (HRE).163 HIFs mainly serve as activators of
gene transcription once bound to DNA. A hypoxia ancillary sequence 5’-CAGGT-3’ that is
located only several nucleotides downstream from the HRE has been proposed to be
85
necessary for HIF activation of VEGF and EPO – two well-documented HIF-target genes –
but this sequence lacks the structural basis for HIF to bind and is likely to be dispensable
for other genes.164
In normoxic conditions, HIF-1α is continuously expressed but undergoes fast hydroxylation
mediated by prolyl-4-hydroxylase (PHD) at conserved proline residues. Hydroxylated HIFs
will bind to the von Hippel Lindau (VHL) E3 ligase and be polyubiquitinated, then undergo
degradation in the proteasome. As oxygen level goes down, HIF α-units are stabilized and
dimerize with the β-unit, then bind to DNA and activate the transcription of target genes.
Gene activation via HIF1A/2A is associated with two transactivation domains (TADs), with
the N-TAD located in the oxygen-dependent degradation domain and the C-TAD at the Cterminus. The C-TAD domain interacts with CBP and p300, recruiting them to the HRE
motif of target genes, which modify the local chromatin, and interact with the core
transcription machinery to activate gene transcription. Other cofactors of HIF1α/2α
include PKM2 (which builds a direct link to the Warburg eQect), and a CDK8-mediator
which promotes pause release of RNAPII.165,166 It is noteworthy that ARNT itself can act as a
coactivator of other factors without HIF-α.167 Multiple studies have attempted to identify a
set of HIF target genes, using both experimental and computational methods.168,169 Similar
to MYC, and potentially for the same reason, eQorts to define target genes of HIFs in
diQerent types of cells has resulted in a small intersection set which can serve as a core
group of HIF-regulated genes.
The most prominent role of HIF in cancer is its regulation of metabolism in response to the
hypoxic tumor microenvironment. But like MYC and other EBTFs, HIFs are also found to be
86
involved in many other aspects of tumorigenesis, including angiogenesis, the immune
response, epigenetic regulation, the epithelial-mesenchymal transition (EMT), etc. An
example of HIF as a key driver of tumorigenesis is in clear cell renal cell carcinoma
(ccRCC), where mutation in VHL is observed in most cases and leads to aberrant
accumulation of HIFs.170 HIF1 and HIF2 are both involved in this type of cancer and have
complicated interactions, exemplifying how EBTFs from the same family can coordinate to
fuel tumorigenesis.
5.2.3 BMAL1 and CLOCK circadian clock proteins
The core mammalian circadian regulators BMAL1 and CLOCK are another example of
bHLH-PAS proteins. (Figure5-1, green block) The level of BMAL1 mRNA and protein
oscillates with an approximate 24-hour period as a result of a tightly regulated feedback
loop, whereas CLOCK levels stay relatively stable. BMAL1 and CLOCK form a heterodimer
and bind to the canonical E-box sequence to activate the transcription of target genes,
including their own repressors such as the cryptochromes CRY1/2 and the period genes
PER1/2/3. PER and CRY proteins can form a complex to repress the transcription mediated
by BMAL1-CLOCK, thus forming a negative feedback loop to induce circadian oscillation of
gene expression. Two additional layers of feedback control of BMAL1-CLOCK function
exist, mediated by the nuclear receptors REV-ERBα, REV-ERBβ and ROR, which make up a
tripartite feedback mechanism of circadian gene expression, which is reviewed in detail by
Takahashi.
171
In mice, the BMAL1-CLOCK dimer activates transcription of target genes in the morning.
Then as the PER and CRY protein levels accumulate in the late afternoon, they translocate
87
into the nucleus to interact with BMAL1-CLOCK to repress the transcription mediated by
the dimer. PER and CRY are then targeted and degraded by proteasome, leading to
reactivated BMAL1 and CLOCK to start a new transcription cycle in the morning.171 BMAL1-
CLOCK activation involves chromatin interaction and modification. Like HIFs, BMAL1 and
CLOCK also interact with p300 and CBP to acetylate histones for transcription. CLOCK
itself has been shown to have histone acetyltransferase activity and can acetylate H3K9
and H3K14.172
Unlike MYC and HIF, BMAL1-CLOCK has been proposed to have pioneer properties and
can open closed chromatins.173 But in a physiological context, this function seems to have
specific requisites for certain cofactors, which can be tissue-specific,
174 therefore the
actual binding sites of physiological BMAL1-CLOCK still depend on specific contexts, such
as in diQerent organs. Again, defining a set of BMAL1-CLOCK target genes by intersecting
sets in diQerent contexts results in a small group of genes primarily regulated by BMAL1
and CLOCK.
113 This gene set is sometimes referred to as clock-controlled genes (CCGs),
but note that CCGs are defined by their 24-hour rhythmic expressions and comprise
diQerent genes depending on the context in which they are defined.
A disrupted circadian rhythm at an organismal level has long been marked as a potential
risk factor for cancer. However, the molecular function of BMAL1 and CLOCK in tumors
has only been studied rather recently. Part of the reason is that mutations in BMAL1 and
CLOCK are not commonly observed in cancer, implying that they themselves do not
commonly function as mutated drivers of tumor initiation. Nonetheless, BMAL1, CLOCK,
and other core clock genes have been shown to be widely dysregulated at the
88
transcriptional level across cancer types.42 This is also true for MYC and HIFs, implying
transcriptional mechanism of driving tumorigenesis might exist.
Recent studies have discovered pivotal roles for BMAL1 and CLOCK in multiple types of
cancers. For example, in glioblastoma (GBM) stem cells, acute myeloid leukemia (AML),
and hepatocellular carcinoma (HCC), BMAL1 is essential for the proliferation of tumor
cells.53,54 Knock-down of BMAL1 can significantly reduce the growth of tumors both in vitro
and in vivo. In GBM, BMAL1 has been shown to gain thousands of new binding sites
compared to normal neural stem cells and is rewired to support tumor specific
metabolism of both glucose and fatty acids.54 These results echo the case of MYC where
EBTF functions vary significantly between tumorous and physiological conditions.
It is noteworthy that, at the cellular level, malignancy does not necessarily disrupt the
circadian rhythm of the cell, since cancer cells can either have strong circadian rhythms or
be totally arhythmic.54,87 We recommend that in a tumorigenic context, distinctions should
be made between the function of the circadian TFs in tumor cells and the actual circadian
rhythm of cells and organisms.
5.2.4 EMT Transcription Factors
Importantly, the key transcription factors governing the process of epithelial-mesenchymal
transition (EMT) are all E-box binding proteins. EMT was first discovered as an essential
process during certain stages of embryo development such as gastrulation.175 In cancer
cells EMT is featured by upregulation of mesenchymal markers such as vimentin (VIM), and
downregulation of epithelial markers such as E-cadherin. EMT in cancer was initially
89
studied in relation to its role in metastasis.176 Although a wide consensus on EMT has not
been reached yet, 4,72,73now it becomes generally accepted that tumor cells can have
intermediate hybrid E/M states spanning a continuous E-M spectrum, and a more hybrid
state is associated with more aggressive stem cell properties.177–180
Multiple families of TFs can induce EMT in cancer, including the SNAIL family, SNAI1 and
SNAI2, bHLH-containing proteins TWIST1 and TWIST2, and the zinc-finger E-box binding
homeobox family, ZEB1 and ZEB2. They all bind to E-boxes to induce an EMT program in
cells, but the specific functions of diQerent proteins are non-redundant.180 SNAI1 has a
zinc finger domain that consists of four zinc finger motifs and can bind to the E-box variant
5’-CAGGTG-3’. The SNAG domain can compete with H3 to prevent lysine-9 from being
demethylated, hence activating gene expression.181 The ZEB proteins have two zinc finger
clusters that bind to 5’-CAGGTG/A-3’ and show higher aQinity to promoters that have two
E-boxes with variable distances in between, such as in the case of CDH1.182 TWIST binds to
the E-box as a homo- or heterodimer and can act as both a repressor and activator of gene
transcription. The diQerent binding preferences of the TFs forms the basis of their distinct
functions.182
EMT-TFs can regulate the expression of a set of common genes and their own specific
targets as either repressors or activators. Their most prominent common function is to
repress the expression of CDH1 through binding to the E-boxes in the promoter region of
the gene. Other common target genes include the interleukins and TGF-beta superfamily
genes. Currently the most prominent function of the EMT-TFs is their regulation of cancer
stem cell-related features such as drug resistance, phenotypic plasticity, immune evasion,
90
etc.180 Importantly, EMT-TFs can function in non-epithelial types of cancer such as
glioblastoma.183 Therefore, researchers have suggested that instead of focusing solely on
the EMT program, more attention should be placed in understanding the specific functions
of the diQerent EMT-TFs.130
5.2.5 Other E-box Binding TFs reported in cancer
Upstream stimulatory factors (USF) 1 and USF2 are ubiquitously expressed transcription
factors that both have a bHLH-LZ domain that binds to the E-box as a heterodimer or
homodimer.184 They also contain a USF-specific region (USR) upstream of bHLH that is
important for E-box dependent transactivation. USFs are transcription activators that have
a small group of defined target genes. USF1 is associated with familial combined
hyperlipidemia and was found to bind the promoter of genes that regulate lipid and
cholesterol metabolism, which is often dysregulated in cancer cells.185 USF2 can compete
with MYC to antagonize the function of MYC.186 USF1 also interacts with p53 and regulates
its function.187
Other bHLH-PAS proteins such as AHR and NPAS in cancer are reviewed in ref 162.
5.3 Regulatory features of E-box-containing regulatory elements
Although EBTF families feature distinct functions, certain features of E-box-regulated
genes are commonly observed. In an early study that analyzed the promoters of CCGs, it
was found that some other motifs are overrepresented in addition to E-boxes, including
those of SP1, ZF5, NRF1, and EGR, which represent CG-rich motifs.113 This implies that
EBTFs might recruit general TFs under the assistance of SP1. However, our own data
91
showed that tandem E-boxes without CG-rich regions can eQectively enhance
transcription, meaning that CG-rich motifs are not necessary for E-box function
(unpublished). Other overrepresented motifs include NFY and E2F.113 In mechanistic study
on the EBTF sterol regulatory element-binding protein (SREBP) family member SREBP1,
SP1 and NFY are reported to be partner factors.188 Interestingly, in an analysis that aimed to
identify overrepresented motifs in bidirectional promoters (defined as promoters of less
than 1kb length and flanked by two protein-coding genes that are transcribed in opposite
directions), E-box, E2F, NRF1, NFY, and CG-rich motifs were also reported.127 These
reports emphasize the importance of co-motifs in determining the function of E-boxes.
Conversely, the TATA-box is usually absent in these bidirectional promoters as well as
housekeeping genes.127 Although he TATA-box and E-box are not mutually exclusive in
promoters, genes containing both motifs implement highly specialized functions such as
regulating certain developmental programs.189,190 (Figure 5-2A,i) As any functional
promoters or enhancers comprise multiple TF-binding sites, these features imply that Eboxes and other motifs have specific functions in regulating transcription and their distinct
combinations encode specific types of transcriptional regulation.
Little is known about the transcriptional regulatory properties of E-boxes. One of their
important functions is to convergently coordinate the temporal control of gene expression
through oscillators. (Figure 5-2A, ii) Oscillator is a general mechanism of dynamic
transcriptional control that can be achieved by feedback loops.191,192 The most prominent
example is the 24-hour circadian rhythm implemented by the tripartite feedback loop as
described above. The circadian clock is intimately interlocked with the cell cycle, which is
92
another oscillator that is dominated by the dynamics of MYC. The molecular details of the
coupling of these two oscillators and their dynamics and system properties are extensively
studied with experiments and mathematical modeling.193–197 (Figure 5-2,B) Because EBTFs
are usually regulated by E-boxes too, they can form interlocked feedbacks and provides a
space for temporal control of various period lengths. The bHLH TF HES1 exemplifies a 2-
hour ultradian oscillator through a self-feedback loop via an E-box in its promoter.198 The
core inflammatory TF NF-κB is also an example of intrinsic oscillatory gene regulated by
EBTFs.199
Another feature of EBTFs including MYC and BMAL1 is they fit in the Kamikaze model of
transcription, where ubiquitin-dependent proteolysis is required for RNAPII elongation,
and newly synthesized activators need to be loaded for a new round of transcription
burst.
111,200,201 This feature might serve as a mechanism driving the temporal control of cell
activities by EBTFs, but the molecular details and the kinetics of this type of transcription is
largely unknown. (Figure 5-2A, iii)
Another key function of E-boxes might be to maintain the robustness of expression of
essential genes, since many of the genes containing E-boxes are ubiquitously expressed
and regulate basic physiological activities of the cell. Supporting this hypothesis, genes
driven by E-box-enriched bidirectional promoters are expressed in a higher frequency than
the average of all human genes.127 Furthermore, in Drosophila where CpG islands are
absent in promoters, E-boxes are often found in promoters of housekeeping genes.202 This
intrinsic robustness of E-box-containing genes underlies the importance of E-boxes in cell
homeostasis.202 (Figure 5-2A, i)
93
Figure 5-2. Functional features of E-box elements in transcriptional regulation. (A) (i) E-boxes are
used to generate oscillations of gene expression through negative feedback loops. The negative
regulator in the loop can be an EBTF itself, such as in the case of HES1 which features a 2-hour
ultradian rhythm, or target genes of EBTFs, such as PER and CRY in the case of BMAL1, which
generates a 24-hour rhythm. ii. E-boxes determine diOerent functions of the promoter by working
with diOerent partner motifs. For example, in TATA-less promoters that have high CG-content, Eboxes participate in keeping the genes continuously/robustly expressed, whereas in some
developmental genes an E-box works together with the TATA-box to initiate downstream
developmental programs. iii. EBTFs can serve as a temporal control node in the turnover of
transcription burst. The physiological function of such control is unclear.(B) Phase lock of circadian
clock and cell cycle. Mathematical modeling and experimental validation revealed that the two
oscillators of circadian rhythm and cell cycle exhibit multiple phase-locked states that exhibit
robustness against molecular fluctuations.
94
In recent years our understanding of the PolII-mediated gene transcription mechanism has
been greatly expanded upon, and its role as a potential therapeutic target in cancer has
been explored.
86,203,204 This progress has also revealed new functions of EBTFs in regulating
multiple steps of transcription. MYC is a prominent example. Firstly, evidence has shown
that MYC can facilitate the formation of the preinitiation complex of PolII by interacting
with TATA-box binding protein (TBP) and potentially modulating the energetic landscape of
TFIID during preinitiation complex (PIC) formation.205 GTF2F1, which is a component of
TFIIF, binds directly to MYC through MB0 and serves another way through which MYC
participates in PIC assembly.206 During initial PolII elongation, MYC has been shown to
facilitate mRNA capping by recruiting RNGTT and RNMT.207 During productive elongation,
MYC can recruit positive transcription elongation factor b (p-TEFb) and enable CDK9 to
phosphorylate Ser2 on the CTD of PolII and allow PolII to continue with productive
elongation.208,209 A recent report showed that SPT5, another key regulator of elongation, is
recruited my MYC.157 Other EBTFs have also been reported to participate in these steps of
RNAPII transcription.
5.4 Mutual regulation of diPerent EBTFs in cancer
As EBTFs all bind to E-boxes and share common regulatory features, it is not surprising that
they can mutually regulate each other, both within each protein family and across diQerent
protein families (Figure 5-3). This can occur via competitive binding to DNA, direct binding
to each other, or regulation of the turnover of each other. These interactions connect
EBTFs into a dense regulatory network.
95
Figure 5-3. Physiological functions of EBTFs and their mutual regulation to form a coordinated
network. (A) In homeostasis EBTFs each have their prominent functions, including control of the
cell cycle by MYC, oxygen sensing by HIF, and control of the circadian rhythm by BMAL1 and
CLOCK. The colored blocks highlight the primary physiological function of each EBTF family. MYC
tightly regulates the expression of E2F target genes to control the cell cycle. HIF is continuously
expressed in cells and rapidly degraded by the proteasome under normal oxygen and ROS levels.
When oxygen levels are low and HIF proteins are less hydroxylated, they are stabilized and bind to
DNA to activate downstream gene expression. BMAL1 and CLOCK dimerize and bind to DNA to
activate clock-controlled genes, which include their own suppressor CRYs and PERs. CRY and PER
translocate into the nucleus and form a repressive complex to inhibit BMAL1 and CLOCK
transcriptional activity, forming a delayed negative feedback loop. When they are degraded, a new
cycle of BMAL1 and CLOCK transcriptional activity is activated. Reported mutual regulations across
diOerent EBTFs are exemplified at the interface of each color block corresponding to the two
families. As they bind to the same DNA motif, regulate each other, and contain E-boxes in their own
promoters, their functions unavoidably converge into a network to coordinate the essential
processes of the cell.
96
5.4.1 MYC-HIF interactions
MYC and HIF closely interact with each other. HIF1α can inhibit MYC activity through
various mechanisms. It can not only bind directly to MAX and interfere with MYC-MAX
dimer activity, but also activate expression of other MYC competitors such as MAD and
MXI1.210 HIF1α also competes with MYC for DNA binding, which is exemplified in the
promoter region of p21.211 Additionally, HIF1α has been reported to promote the
proteasomal degradation of MYC.212 Paradoxically, MYC has been mostly reported to
promote the activity of HIF1α. This is exemplified by decreased HIF1α levels after MYC
knock-down in multiple myeloma cells, and stabilization of HIF1α proteins after MYC
overexpression.213Mechanistically, MYC can reduce the binding of HIF1α to the VHL
complex and decrease its degradation.214 In addition, MYC increases mitochondrial
OXPHOS and ROS production, which inhibits PHD activity in non-hypoxic conditions.215 The
paradox also lies in the functional consequence of normal MYC and HIF1α activity. MYC
usually promotes the function and biogenesis of mitochondria, whereas HIF1α represses
them by activating FOXO3a which consequently represses mitochondrial gene expression
and induces BNIP3, triggering mitochondria degradation through autophagy.216
However, in cancer cells MYC and HIF1α are not incompatible since many cancer types
have both MYC and HIF1α in high levels. The seeming conflict can be explained by the
deregulation of MYC levels. When at high levels, MYC can still maintain its activity
stoichiometrically and override the inhibitory eQect of HIF1α.217 USP29 has been shown to
maintain HIF1α and MYC levels at the same time to promote tumorigenic metabolism.218 In
such cases, MYC and HIF1α may cooperatively tailor a gene expression program that takes
97
advantage of the pro-tumorigenic aspect of each protein to fuel tumor growth. For
example, both proteins activate genes for glucose import and glycolysis such as LDHA and
HKII, which contributes to the Warburg eQect in cancer.219 This exemplifies how EBTFs
coordinate in a network to fuel cancer progression.
Such cooperation is also observed for HIF2α, although unlike HIF1α, HIF2α is better known
to promote MYC activity in cancer. HIF2α enhances MYC function by stabilizing the MYCMAX dimer in clear cell renal carcinoma cells and colorectal carcinoma cell lines.220 In
turn, MYC has been reported to activate HIF2α transcription by directly binding to its
promoter in T cell leukemia and maintaining a pool of cancer stem cells.221 By contrast, in
physiological endothelial cells, it is reported that HIF2α represses MYC expression.222
5.4.2 MYC-BMAL1/CLOCK interactions
The circadian clock proteins also closely interact with MYC through various mechanisms.
Two groups have shown that MYC (both MYC and MYCN) can inhibit BMAL1 and CLOCK
function in cancer cells and disrupt the circadian dynamics of the core clock
molecules.223,224 Consequently, periodic glutamine metabolism in cancer cells is altered.
MYC disruption of BMAL1 can occur through upregulation of REV-ERBs, which inhibits the
transcription of BMAL1, and/or through direct inhibition of BMAL1 transcription by binding
to the promoter as an inhibitory MYC-MIZ complex.223,224 It is noteworthy that these findings
were determined in cancer cells with relatively low MYC levels and intact circadian
rhythms. But like with HIF1α, high MYC levels and circadian oscillations are not
incompatible since high levels of MYC, and intact oscillations have also been observed
simultaneously in the same GBM cell line.54
98
On the other hand, MYC is a clock-controlled gene itself. CRY2 can cooperatively bind MYC
with FXBL3, promoting its ubiquitylation by SCF-FBXL3 and consequent degradation by the
proteasome.
225 This mechanism is also context-dependent because in mouse spleen, Cry2
knockout had no eQect on Myc levels and Cry1/Cry2 double knockout repressed Myc
levels.
226 Per2 mutation and Bmal1 deletion in lung tumors has also been shown to lead to
an increase in Myc levels, although specific detail is not clear.40 Another report using a
mouse model suggests an indirect mechanism of Myc control by Bmal1 through catenin or
Ctnnb1 as an intermediate eQector. 226
5.4.3 HIF-BMAL1 interactions
Another well studied pair of E-box binding proteins that have been shown to regulate each
other is the bHLH-PAS TFs HIF1α and BMAL1. Under physiological conditions, two
independent groups using diQerent models showed that hypoxic responses mediated by
HIF1α are under circadian control, and the hypoxic gene expression pattern is disrupted
when BMAL1 is knocked out.227,228 This is consistent with the presence of the E-box in the
HIF1α promoter. In turn, HIF1α can participate in the regulation of circadian rhythm
mediated by BMAL1.228 Pharmacological stabilization of HIF results in a lengthened period
and dampened amplitude of Per2 and BMAL1 rhythmicity. HIF1α also has a positive eQect
on the function of BMAL1 to activate transcription, as is shown in both studies.227,228
Interestingly, BMAL1 and HIF1α can be co-immunoprecipitated in Co-IP experiments,
suggesting that they can at least form a complex together or even dimerize to regulate gene
transcription.228 ChIP-seq experiments of both factors in the same osteosarcoma cell line
showed that BMAL1 and HIF1α share a large portion of binding sites on chromatin, which
99
constitutes approximately a third of all HIF1α targets and a quarter of BMAL1 targets.
These results further underscore their close relationship in regulating gene expression.
However, their relationship is not well studied in cancer. Hypoxia-induced HIF activity is
proposed to promote the disruption of the circadian rhythm in hepatocellular carcinoma,
but a more detailed mechanism still needs to be revealed.229 Correlations between
hypoxia/circadian clock and radiation resistance has been noted in glioma, yet
mechanistic studies remain to be done.230
Other regulators of the circadian clock, such as PER and CRY, also interact with HIFs but
will not be elaborated upon here.231–233
5.4.4 EMT TFs interactions
The interactions between EMT-TFs and bHLH TFs are relatively less studied directly,
although they share multiple phenotypic commonalities. The most understood interaction
is with HIFs, which are mainly accounted for by their transcriptional regulation of each
other and exemplified by the induction of EMT-TFs by HIF proteins. In multiple cell types,
HIF overexpression or hypoxia is suQicient to induce EMT.234 HIF1α can directly bind to the
HRE motifs in the promoter region of TWIST1, SNAI1, SNAI2 and ZEB1 to activate their
expression. Indirect regulations of EMT-TFs by HIFs are often observed, too. For example,
HIF1 can also activate histone modifiers such as HDAC3 to promote SNAI1 activation
indirectly.235 Other intermediate genes include WDR5, lncRNA, FoxM1, ILK and
PAFAH1B2.236–238
100
It is observed that high MYC levels and EMT often co-occur in cancer, and they can
contribute to the same characteristics of later stage tumors. It has been reported that
over-expression of MYC can induce EMT in lung cancer and melanoma cells through SNAI1
and ZEB.239,240 MYC also facilitates TGFβ-induced EMT as a coactivator of the SMAD
complex.241 This exemplifies physiological antagonistic factors can cooperate in certain
tumors, because in physiological conditions, TGFβ represses MYC expression and inhibits
cell proliferation, in turn MYC suppresses the activation of TGFβ-induced genes.
The interaction between circadian regulators and EMT factors has only been noted
recently.242 It has been reported lately that the EMT process in cancer is gated by the
circadian rhythm in the cell.243 Plus, BMAL1 has been shown to facilitate the EMT in
colorectal cancer.244
5.4.5 Interactions within the same family
In addition to the inter-family mutual regulation, an intra-family interaction layer also
exists. Most E-box-binding genes have some extent of redundancy. In physiological
conditions, diQerent members of the same family often share a large portion of target
genes and implement similar functions in diQerent tissues or at diQerent developing
stages. However, in cancer cells they often have independent or even antagonistic
functions when they are simultaneously expressed in the same cells.
HIF family members provide prominent examples of such relationships.245 Pioneering work
showed that in RCC cells where VHL function is defective, HIF2 fuels tumor growth partly
by activating cyclin D1, and has suppressive interactions with HIF1.
246 In mouse xenograft
101
models, HIF2 overexpression significantly enhanced tumor growth in vivo, whereas HIF1
overexpression suppressed tumor progression. Consistently, HIF1 overexpression lowered
HIF2 protein levels, and vice versa. They also showed that the DNA binding function of HIF2
is responsible for its repressive activity towards HIF1.246 The antagonistic functions of HIF1
and 2 are also marked in tumor associated stromal cells and have a reversed eQect on
angiogenesis in the tumor microenvironment.247 Despite all the examples, HIF1 and 2 are
not always mutually antagonistic, but can also collaborate to meet diQerent needs of the
cancer cells.248
Along similar lines, ZEB1 has been shown to participate in the initialization and progression
of melanoma cells. By contrast, ZEB2 suppresses the onset and metastasis of melanoma
in mouse.249,250 SNAI1 and SNAI2 have also been shown to have opposite eQects on the
expression of phospholipase D (PLD), which has been proposed as a prognostic marker of
breast cancer. PLD also has opposite eQects on the expression of SNAI1 and SNAI2. The
authors thus proposed a feedback loop model to explain the mutually antagonistic eQect
of the two factors on each other.251 In a diQerent ovarian tumor model where SNAI1 and
SNAI2 were also found to be mutually exclusive, SNAI1 was found to bind to E-boxes in the
promoter region of SNAI2 and recruit HDAC to repress SNAI2 expression.252More examples
of EMT-TF mutual regulation can be found in ref.180
All together, these examples show important features of the intimately connected EBTF
network. First, the regulatory relationships among the TFs in normal tissues are largely
rewired in cancer, and the new networks depend on tumor type and their evolutionary
trajectory, thus are diverse. This explains the controversy that a certain EBTF is oncogenic
102
in some tumor but is tumor-suppressive in others. Another feature is that in the rewired
networks EBTFs are still closely related, because of their intrinsic DNA-binding specificity.
On the one hand, targeting strategies on network levels might be needed to inhibit the
tumor-fueling EBTF network. On the other hand, such network provides more actionable
nodes to targeting certain EBTF or the whole network. Hypothetically, the flexibility of this
network might also provide evolutionary spaces for tumor cells to develop plasticity.
5.5 Perspectives of EBTFs in cancer
Because of the shared binding specificity and functional interconnectivity of the EBTFs,
they also convergently regulate phenotypical hallmarks of tumors.
5.5.1 Tumor initiation
Although all EBTFs discussed above are often dysregulated in cancer, they are rarely
mutated,42,128,130 and the mechanism of how they “drive” tumorigenesis is still largely
unknown. The rarity of their mutations implies the importance of the functional intactness
of these proteins in cancer development.
MYC is the earliest and most documented oncogenic TF. Even a small disturbance of MYC
homeostasis can induce abnormal phenotypic changes in cells.128 MYC can facilitate the
progression of the cell cycle by upregulating genes that promote the passing of
checkpoints. Interestingly, pan-cancer analysis showed that MYC amplification is mutually
exclusive with many canonical oncogenic drives such as PIK3CA, PTEN, APC, and BRAF.128
This result implies that MYC has its specific mechanisms of driving tumorigenesis, likely
through transcriptional regulation of the cell cycle.
103
Other EBTFs are not recognized as general cancer drivers so far in knock-out or overexpression-based in vivo tumor development assays. However, this does not exclude their
potential oncogenic role since these models might miss some necessary background or
cofactors, such as de novo enhancers gained through mutations in non-coding regions or
epigenetic changes in the genome.
5.5.2 Metabolism
Cancer cells usually require specific metabolic programs to meet their needs for
continuous proliferation. The most prominent consequence of EBTFs in cancer is their
ability to rewire the metabolic program of the cells, as a large proportion of metabolic
genes contain E-boxes in their promoter region.
MYC, BMAL1-CLOCK and HIFs can all regulate genes that are responsible for glycolysis.
Common gene targets include the GLUT family216,253, which controls glucose intake into the
cell, and most of the enzymes involved in glycolysis. LDHA is also a well-documented gene
regulated by MYC, HIF, and BMAL1.254,255 These enzymes together may cooperate to fuel
the Warburg eQect.256
The TCA cycle turnover is also altered in cancer cells to support cancer progression.
Prominently, MYC activates glutamine transporters and feeds more glutamine into the TCA
cycle, resulting in the glutamine-addicted metabolic feature of many MYC-driven
cancers.257,258 MYC, HIFs, and BMAL1 can all regulate the source and level of acetyl Co-A
entry into the TCA cycle and regulate lipid and cholesterol metabolism.216,259,260 Another
pivotal TCA metabolite as a common gene target of EBTFs is α-KG, which is of great
importance because it is a key node connecting metabolism with histone modification,
104
marking the importance of epigenetic regulation by these factors, and with fatty acid
synthesis through ACACA, which is also a shared target gene.260 It is also not surprising
that all three factors can directly regulate the synthesis, elimination, and fusion dynamics
of mitochondria.212,261,262
It is noteworthy that BMAL1 and CLOCK are the master regulators of organismal
metabolism in response to sleeping and feeding and coordinate metabolism across
tissues and organs. This function is reviewed elsewhere.259
5.5.3 Immune evasion and inflammation
It is critical for tumor cells to evade the surveillance of the immune system, and during
tumorigenesis malignant cells evolve multiple mechanisms to suppress the immune
reaction against them. All the E-box binding proteins broadly participate in both innate and
adaptive immune regulation. 129,263–265
The recruitment of macrophages and other myeloid cells is the first level of immune
regulation. MYC, HIFs, and BMAL1-CLOCK can all regulate cytokines responsible for their
recruitment, including CCL family chemokines and interleukins. EBTFs can cooperate in a
tumor to promote a conducive microenvironment. For example, MYC and TWIST have been
shown to collaboratively support a pro-metastatic phenotype of macrophages through
regulating the secretion of CCL2 and IL13. 266
Immune checkpoint mediated by PD-L1 is another critical mechanism used by cancer
cells to evade the immune response. MYC can regulate PD-L1 through either direct binding
to its promoter or through post-transcriptional mechanisms in multiple types of
cancers.267–269 PD-L1 is also reported to be a direct target of HIF1,270 whereas BMAL1
105
regulates PD-L1 expression in an indirect way through lactate metabolism in
macrophages.271 Regulation of PD-L1 has also been studied in EMT contexts.272
NF-κB seems to be a central mediator of EBTF balance in immune regulation. MYC itself is
a target of NF-κB.273 BMAL1 can dimerize with RelB and block a subunit of the NF-κB
transcription complex.274 CLOCK can acetylate the RelA subunit and GRs to regulate their
DNA binding activity.274,275 Twist 1 can also interact with RELA. 276
5.5.4 Angiogenesis and other tumor microenvironments
EBTFs also remodel other components of the tumor microenvironment, including
extracellular matrix (ECM) components and promoting angiogenesis. Cancer-specific
angiogenesis is an important feature of solid tumors and its potential as a therapeutic
target has been underscored by the success of recent clinical trials involving antiangiogenic therapy. VEGF is a central promoting factor of angiogenesis and has been
shown to be directly regulated by HIF and BMAL1.277 VEGF is also reported to be closely
regulated by MYC.278,279 BMAL1 has been shown to be associated with drug resistance of
colorectal cancer cells via its regulation of VEGF.280 Other coordinating factors of
angiogenesis have also been reported to be under EBTF control.
Recent advances in mechanobiology revealed the important role of ECM components and
corresponding signaling pathways in tumorigenesis.281,282 High ECM stiQness is a driving
force of tumorigenesis and itself can result in an abnormal chromatin state.283 Collagen
and integrin are the most studied ECM signaling-related molecules in cancer and have
been shown to be regulated by EBTFs. The most prominent regulators are the EMT-TFs,
which can directly regulate the type and amount of collagen genes produced by cells, and
106
contribute to stem cell features of cancer cells.131,284 MYC can regulate genes enriched in
the ECM, cell adhesion and cell junction gene sets and regulate invasiveness.279 BMAL1-
CLOCK has been reported to regulate the secretory pathway of collagens and maintain
their homeostasis.285 HIFs also have well-documented functions in regulating ECM
components by regulating collagen prolyl and lysyl hydroxylase and integrins.286
5.5.5 Cancer Stem Cells
Although the CSC concept still lacks a uniform definition across tumor types, some
common features are recurrently observed in certain cancers such as glioblastoma, AML,
breast cancer, HCC, etc. CSCs defined in these cancers usually have high heterogeneity,
drug- and immune-resistance, and ability to self-renew. Interestingly, all the EBTFs
discussed in this review are widely reported to be associated with CSCs.53,54,131,287,288 These
examples imply that transcriptomic features might be able to uniformly define CSCs and
guide targeting strategies.
Summarizing all these functions, we propose a network perspective of E-box biology in
cancer (Figure 5-4A) that bridges the fulfillment of phenotypic changes of cancer cells in
diQerent levels to meet their progressive needs. We also stratified their functions in line
with the ten cancer hallmarks to highlight their specific involvements (Figure 5-4B).
107
Figure 5-4. EBTFs Contributing to the hallmarks of cancer. (A) From a systems point of view,
transcription factors are nodes that form the complex transcription regulatory network, which
integrate all the information from external and internal signals. Then they decide which genes are
transcribed and generate the gene expression program of the cell. The expressed genes then carry
on the function to help tumor cells gain the hallmarks to progress. EBTFs form a subset of the
whole network and carry out certain cellular functions. (B) Summary of EBTFs reported to
contribute to the cancer hallmarks. Colored block indicates that the family is reported to
contribute to the hallmark. Because the EBTF network regulates essential cellular activities, their
function in tumors also contributes more to the pathological changes that cancer cells need to
fulfill their ultimate goal of unceasing proliferation. Current results support the strategy of targeting
the EBTFs to simultaneously eliminate the functional hallmarks thus halt tumor growth.
108
5.6 Targeting EBTFs in cancer
TFs was once thought to be “undruggable” targets because of their intrinsically disordered
structure. But recent years new advances in pharmacochemistry provided promising new
toolboxes for targeting TFs through various mechanisms such as inducing targeted protein
degradation, disrupting protein-protein interaction, and indirect targeting of TF modulators
and collaborators.289,290
MYC has been the most appealing yet challenging target among EBTFs for its centrality in
cancer. The most prominent strategy is to target MYC-MAX dimerization. Early examples
include OmoMyc, which is a 90 amino acid MYC mutant that bind to MYC and MAX to
disrupt their dimerization.291 Initially OmoMyc was only thought of as a tool because of its
size, but recently in vivo data showed its potential as a therapeutic.291 Small molecules that
disrupt MYC function has also been successfully developed and show favorable
pharmacokinetics and tolerability in preclinical studies.292 Multiple other mechanisms have
also been explored, such as targeting MYC transcription, translation, and DNA-binding,
etc.293
HIFs have attracted great interest as a therapeutic target in cancer for many years with
multiple tested mechanisms.294 Recently, a small molecule named belzutifan that binds to
the PAS-B domain of HIF2α showed exceptional eQicacy in VHL-associated RCC in clinical
trial.295 The results led to the first-in-kind approval from FDA to treat several cancer types
associated with VHL. The success of belzutifan proved the feasibility of faithfully targeting
bHLH-PAS TFs with small molecules and the therapeutic potential of these TFs.
109
Our lab and collaborators have developed several small-molecule-sets to target the
circadian network and the activity of BMAL1 and CLOCK, including stabilizers of
cryptochrome with precise isoform selectivity,78,79 REV-ERB agonists,73 and novel inhibitors
of casein kinase II (CK2).82 We also showed the potential of these molecules as cancer
therapeutics in pre-clinical models of multiple cancers.53,54
5.7 Outlook
We progressively reviewed important EBTFs in cancer, their shared binding motif and
target gene, close mutual regulation, convergent functions in homeostasis and cancer,
and established a network view of the biology of these TFs. Synthesizing these factors in a
unified model provide some important implications.
First, because EBTFs lie in a central node that fuels many hallmarks of cancer, targeting
this node provides the chance of shutting down multiple hallmarks simultaneously. This is
exemplified by the biology of MYC, which established a “coalition model” where MYC
interacts with a wide variety of proteins, which cooperate to achieve a collaborative
transcriptional program. Thus, it is proposed by the MYC-studying community that instead
of targeting individual functions of MYC in diQerent hallmarks, it is much more eQective to
“chop the MYC tree” to halt cancer cells from progressing.
159
On the other hand, however, this convergent view also imposes major challenges on
studying the biology. First the functions of the fundamental connecting node, the E-boxes,
remain largely unknown, making it hard to dissect the molecular mechanism of the EBTF
network. In addition, studying the properties of a network requires quantitative modeling,
but acquiring data for establishing the model necessitates delicate experimental design.
110
Luckily, well-characterized small molecules become available recently and provide handy
tools for this purpose.
Paving the ways for drugs that target EBTFs to clinic will be an important field of study. This
will involve biomarker discovery for these targets, recognition of potential benefits, and
careful design of pre-clinical and clinical trials, all of which requires better understanding
of the biology of EBTFs. A promising first step would be combining EBTF-targeting drugs
with current therapeutics. Because of the broad regulation of cancer hallmarks by EBTFs,
there’s a higher chance that these drugs will synergize with current therapeutics to
improve outcome. Successful examples include a MYC small molecule inhibitor, which is
shown to synergize with anti-hormone therapy to inhibit prostate cancer and breast cancer
cells,296 and several other studies that evaluated the eQicacy of HIF inhibitors in
combination therapies.
In summary, we depicted a network-based reasoning diagram for proposing and testing
new hypotheses and strategies to target EBTFs in cancer. This model will help to account
for the many discrepancies that have been encountered when trying to find a unifying
function for one particular factor across diQerent cancers. As chemical tools are becoming
more available to regulate activities of EBTFs, we believe targeting EBTFs will be a
promising new strategy for cancer therapy.
111
Chapter 6 Conclusions and Future Work
This thesis investigated the roles of circadian clock genes in three aspects of cancer: roles
as biomarkers for cancer patients, their functions as drivers, and their role in regulating the
TME. The results opened many roads for future studies.
As potential biomarkers, the scoring system we proposed can be tested in more detail for
breast cancer patients going through diQerent therapies. Such studies will provide more
precise stratification of breast cancer patients for more precise outcome prediction and
potential treatment design and thus expand the toolbox for patient care.
By studying circadian genes as potential drivers of the progression of cancer cells, we
proposed a new paradigm of targeting circadian TF-mediated transcription programs
through a cis-regulatory mechanism. Our model introduced a primary way of
characterizing CREs targeted by our drug combination and more detailed quantitative
description of the CREs are needed. The results also induced more questions for future
studies such as: What are the CRE components that drive tumor growth in other types of
TNBC and other cancer types? Is it possible to target those CREs using small molecules?
Can this drug combination strategy be applied in animal models and eventually cancer
patients? The drug combination also provides a tool for studying the transcriptional
dynamics of this type of CREs, the function of proteasome in regulating such dynamics,
etc.
By introducing new accessible models for studying the TME of glioblastoma and testing the
eQect of CK2 inhibitors, our results prompt further development of such models for the use
112
of drug testing. Further investigation of CK2 inhibitors in regulating glioblastoma TME and
their further development for in vivo use are also needed in the future.
In summary, this thesis provided important results for future studying the roles of circadian
genes in cancer to advance cancer biology and treatment in various perspectives.
113
References
1. Dorn, C. R. The epidemiology of cancer in animals. California medicine vol. 107
Preprint at (1967).
2. Wyld, L., Audisio, R. A. & Poston, G. J. The evolution of cancer surgery and future
perspectives. Nature Reviews Clinical Oncology vol. 12 Preprint at
https://doi.org/10.1038/nrclinonc.2014.191 (2015).
3. Sullivan, R. et al. Global cancer surgery: Delivering safe, aQordable, and timely
cancer surgery. The Lancet Oncology vol. 16 Preprint at
https://doi.org/10.1016/S1470-2045(15)00223-5 (2015).
4. CoQey, J. C. et al. Excisional surgery for cancer cure: Therapy at a cost. Lancet
Oncology vol. 4 Preprint at https://doi.org/10.1016/S1470-2045(03)01282-8 (2003).
5. Low, P. S., Singhal, S. & Srinivasarao, M. Fluorescence-guided surgery of cancer:
applications, tools and perspectives. Current Opinion in Chemical Biology vol. 45
Preprint at https://doi.org/10.1016/j.cbpa.2018.03.002 (2018).
6. Burotto, M., Wilkerson, J., Stein, W. D., Bates, S. E. & Fojo, T. Adjuvant and
neoadjuvant cancer therapies: A historical review and a rational approach to
understand outcomes. Seminars in Oncology vol. 46 Preprint at
https://doi.org/10.1053/j.seminoncol.2019.01.002 (2019).
7. Muss, H. B., Biganzoli, L., Sargent, D. J. & Aapro, M. Adjuvant therapy in the elderly:
Making the right decision. Journal of Clinical Oncology vol. 25 Preprint at
https://doi.org/10.1200/JCO.2006.10.3457 (2007).
8. Anampa, J., Makower, D. & Sparano, J. A. Progress in adjuvant chemotherapy for
breast cancer: An overview. BMC Medicine vol. 13 Preprint at
https://doi.org/10.1186/s12916-015-0439-8 (2015).
9. Bukowski, K., Kciuk, M. & Kontek, R. Mechanisms of multidrug resistance in cancer
chemotherapy. International Journal of Molecular Sciences vol. 21 Preprint at
https://doi.org/10.3390/ijms21093233 (2020).
10. Min, H. Y. & Lee, H. Y. Molecular targeted therapy for anticancer treatment.
Experimental and Molecular Medicine vol. 54 Preprint at
https://doi.org/10.1038/s12276-022-00864-3 (2022).
11. Oh, D. Y. & Bang, Y. J. HER2-targeted therapies — a role beyond breast cancer. Nature
Reviews Clinical Oncology vol. 17 Preprint at https://doi.org/10.1038/s41571-019-
0268-3 (2020).
114
12. Wang, R. C. & Wang, Z. Precision Medicine: Disease Subtyping and Tailored
Treatment. Cancers vol. 15 Preprint at https://doi.org/10.3390/cancers15153837
(2023).
13. Dougan, M. & DranoQ, G. Immune therapy for cancer. Annual Review of Immunology
vol. 27 Preprint at https://doi.org/10.1146/annurev.immunol.021908.132544 (2009).
14. Li, L., Liu, S., Han, D., Tang, B. & Ma, J. Delivery and Biosafety of Oncolytic
Virotherapy. Frontiers in Oncology vol. 10 Preprint at
https://doi.org/10.3389/fonc.2020.00475 (2020).
15. Cesur-Ergün, B. & Demir-Dora, D. Gene therapy in cancer. Journal of Gene Medicine
vol. 25 Preprint at https://doi.org/10.1002/jgm.3550 (2023).
16. Baker, D. J., Arany, Z., Baur, J. A., Epstein, J. A. & June, C. H. CAR T therapy beyond
cancer: the evolution of a living drug. Nature 619, (2023).
17. Harbeck, N. & Gnant, M. Breast cancer. The Lancet 389, 1134–1150 (2017).
18. Waks, A. G. & Winer, E. P. Breast Cancer Treatment: A Review. JAMA - Journal of the
American Medical Association 321, 288–300 (2019).
19. Pernas, S., Tolaney, S. M., Winer, E. P. & Goel, S. CDK4/6 inhibition in breast cancer:
current practice and future directions. Therapeutic Advances in Medical Oncology
vol. 10 Preprint at https://doi.org/10.1177/1758835918786451 (2018).
20. Xiao, T., Ali, S., Mata, D. G. M. M., Lohmann, A. E. & Blanchette, P. S. Antibody–Drug
Conjugates in Breast Cancer: Ascent to Destiny and Beyond—A 2023 Review.
Current Oncology vol. 30 Preprint at https://doi.org/10.3390/curroncol30070474
(2023).
21. Lehmann, B. D. et al. Identification of human triple-negative breast cancer subtypes
and preclinical models for selection of targeted therapies. Journal of Clinical
Investigation 121, (2011).
22. Gong, Y. et al. Metabolic-Pathway-Based Subtyping of Triple-Negative Breast Cancer
Reveals Potential Therapeutic Targets. Cell Metab 33, (2021).
23. Jiang, Y. Z. et al. Genomic and Transcriptomic Landscape of Triple-Negative Breast
Cancers: Subtypes and Treatment Strategies. Cancer Cell 35, (2019).
24. Lathia, J. D., Mack, S. C., Mulkearns-Hubert, E. E., Valentim, C. L. L. & Rich, J. N.
Cancer stem cells in glioblastoma. Genes and Development vol. 29 Preprint at
https://doi.org/10.1101/gad.261982.115 (2015).
115
25. Rodríguez-Camacho, A. et al. Glioblastoma Treatment: State-of-the-Art and Future
Perspectives. International Journal of Molecular Sciences vol. 23 Preprint at
https://doi.org/10.3390/ijms23137207 (2022).
26. McKinnon, C., Nandhabalan, M., Murray, S. A. & Plaha, P. Glioblastoma: Clinical
presentation, diagnosis, and management. The BMJ vol. 374 Preprint at
https://doi.org/10.1136/bmj.n1560 (2021).
27. Neftel, C. et al. An Integrative Model of Cellular States, Plasticity, and Genetics for
Glioblastoma. Cell 178, (2019).
28. Bożyk, A., Wojas-Krawczyk, K., Krawczyk, P. & Milanowski, J. Tumor
Microenvironment—A Short Review of Cellular and Interaction Diversity. Biology vol.
11 Preprint at https://doi.org/10.3390/biology11060929 (2022).
29. Elhanani, O., Ben-Uri, R. & Keren, L. Spatial profiling technologies illuminate the
tumor microenvironment. Cancer Cell vol. 41 Preprint at
https://doi.org/10.1016/j.ccell.2023.01.010 (2023).
30. Xiao, Y. & Yu, D. Tumor microenvironment as a therapeutic target in cancer.
Pharmacology and Therapeutics vol. 221 Preprint at
https://doi.org/10.1016/j.pharmthera.2020.107753 (2021).
31. Roma-Rodrigues, C., Mendes, R., Baptista, P. V. & Fernandes, A. R. Targeting tumor
microenvironment for cancer therapy. International Journal of Molecular Sciences
vol. 20 Preprint at https://doi.org/10.3390/ijms20040840 (2019).
32. Patke, A., Young, M. W. & Axelrod, S. Molecular mechanisms and physiological
importance of circadian rhythms. Nature Reviews Molecular Cell Biology vol. 21
Preprint at https://doi.org/10.1038/s41580-019-0179-2 (2020).
33. Takahashi, J. S. Transcriptional architecture of the mammalian circadian clock. Nat
Rev Genet 18, 164–179 (2017).
34. Schernhammer, E. S., Kroenke, C. H., Laden, F. & Hankinson, S. E. Night work and
risk of breast cancer. Epidemiology (2006)
doi:10.1097/01.ede.0000190539.03500.c1.
35. Wegrzyn, L. R. et al. Rotating night-shift work and the risk of breast Cancer in the
nurses’ health studies. Am J Epidemiol (2017) doi:10.1093/aje/kwx140.
36. Straif, K. et al. Carcinogenicity of shift-work, painting, and fire-fighting. The lancet
oncology Preprint at https://doi.org/10.1016/s1470-2045(07)70373-x (2007).
37. Kettner, N. M. et al. Circadian Homeostasis of Liver Metabolism Suppresses
Hepatocarcinogenesis. Cancer Cell (2016) doi:10.1016/j.ccell.2016.10.007.
116
38. Lee, S., Donehower, L. A., Herron, A. J., Moore, D. D. & Fu, L. Disrupting circadian
homeostasis of sympathetic signaling promotes tumor development in mice. PLoS
One (2010) doi:10.1371/journal.pone.0010995.
39. Fu, L., Pelicano, H., Liu, J., Huang, P. & Lee, C. C. The circadian gene Period2 plays an
important role in tumor suppression and DNA damage response in vivo. Cell (2002)
doi:10.1016/S0092-8674(02)00961-3.
40. Papagiannakopoulos, T. et al. Circadian Rhythm Disruption Promotes Lung
Tumorigenesis. Cell Metab 24, 324–331 (2016).
41. Pariollaud, M. & Lamia, K. A. Cancer in the Fourth Dimension: What Is the Impact of
Circadian Disruption? Cancer Discov 10, 1455–1464 (2020).
42. Ye, Y. et al. The Genomic Landscape and Pharmacogenomic Interactions of Clock
Genes in Cancer Chronotherapy. Cell Syst 6, 314-328.e2 (2018).
43. Battaglin, F. et al. Clocking cancer: the circadian clock as a target in cancer therapy.
Oncogene 40, 3187–3200 (2021).
44. Sulli, G., Lam, M. T. Y. & Panda, S. Interplay between Circadian Clock and Cancer:
New Frontiers for Cancer Treatment. Trends Cancer 5, 475–494 (2019).
45. de Assis, L. V. M. et al. Expression of the circadian clock gene BMAL1 positively
correlates with antitumor immunity and patient survival in metastatic melanoma.
Front Oncol (2018) doi:10.3389/fonc.2018.00185.
46. Li, W. et al. Decreased circadian component Bmal1 predicts tumor progression and
poor prognosis in human pancreatic ductal adenocarcinoma. Biochem Biophys Res
Commun 472, 156–162 (2016).
47. Ramos, C. A. et al. A Non-canonical Function of BMAL1 Metabolically Limits ObesityPromoted Triple-Negative Breast Cancer. iScience 23, (2020).
48. Fekry, B. et al. Incompatibility of the circadian protein BMAL1 and HNF4α in
hepatocellular carcinoma. Nat Commun (2018) doi:10.1038/s41467-018-06648-6.
49. Taniguchi, H. et al. Epigenetic inactivation of the circadian clock gene BMAL1 in
hematologic malignancies. Cancer Res (2009) doi:10.1158/0008-5472.CAN-09-
0551.
50. Zhang, S. et al. Circadian clock components RORa and Bmal1 mediate the antiproliferative eQect of MLN4924 in osteosarcoma cells. Oncotarget (2016)
doi:10.18632/oncotarget.11807.
117
51. Yeh, C. M. et al. Epigenetic silencing of ARNTL, a circadian gene and potential tumor
suppressor in ovarian cancer. Int J Oncol (2014) doi:10.3892/ijo.2014.2627.
52. Qu, M., DuQy, T., Hirota, T. & Kay, S. A. Nuclear receptor HNF4A transrepresses
CLOCK: BMAL1 and modulates tissue-specific circadian networks. Proc Natl Acad
Sci U S A (2018) doi:10.1073/pnas.1816411115.
53. Puram, R. V. et al. Core Circadian Clock Genes Regulate Leukemia Stem Cells in
AML. Cell 165, 303–316 (2016).
54. Dong, Z. et al. Targeting glioblastoma stem cells through disruption of the circadian
clock. Cancer Discov 9, 1556–1573 (2019).
55. Chen, P. et al. Circadian regulator CLOCK recruits immune-suppressive microglia
into the GBM tumor microenvironment. Cancer Discov 10, 371–381 (2020).
56. Kinouchi, K. & Sassone-Corsi, P. Metabolic rivalry: circadian homeostasis and
tumorigenesis. Nature Reviews Cancer Preprint at https://doi.org/10.1038/s41568-
020-0291-9 (2020).
57. Yang, M. Y. et al. Up-regulation of PER3 Expression Is Correlated with Better Clinical
Outcome in Acute Leukemia. Anticancer Res 35, (2015).
58. Qi, C. et al. A role for the clock gene Per1 in prostate cancer. Cancer Res (2009)
doi:10.1158/0008-5472.CAN-08-4199.
59. Gery, S., Virk, R. K., Chumakov, K., Yu, A. & KoeQler, H. P. The clock gene Per2 links
the circadian system to the estrogen receptor. Oncogene (2007)
doi:10.1038/sj.onc.1210585.
60. Hwang-Verslues, W. W. et al. Loss of corepressor PER2 under hypoxia up-regulates
OCT1-mediated EMT gene expression and enhances tumor malignancy. Proc Natl
Acad Sci U S A (2013) doi:10.1073/pnas.1222684110.
61. Tong, H. et al. Nr1d2 accelerates hepatocellular carcinoma progression by driving
the epithelial-to-mesenchymal transition. Onco Targets Ther (2020)
doi:10.2147/OTT.S237804.
62. Tao, L. et al. Rev-erbα inhibits proliferation by reducing glycolytic flux and pentose
phosphate pathway in human gastric cancer cells. Oncogenesis (2019)
doi:10.1038/s41389-019-0168-5.
63. Sulli, G. et al. Pharmacological activation of REV-ERBs is lethal in cancer and
oncogene-induced senescence. Nature (2018) doi:10.1038/nature25170.
118
64. Yu, H. et al. Cryptochrome 1 Overexpression Correlates with Tumor Progression and
Poor Prognosis in Patients with Colorectal Cancer. PLoS One (2013)
doi:10.1371/journal.pone.0061679.
65. Sha, A. A. et al. The circadian cryptochrome, CRY1, is a pro-tumorigenic factor that
rhythmically modulates DNA repair. Nat Commun 12, 401 (2021).
66. Cadenas, C. et al. Loss of circadian clock gene expression is associated with tumor
progression in breast cancer. Cell Cycle (2014) doi:10.4161/15384101.2014.954454.
67. Mannic, T. et al. Circadian clock characteristics are altered in human thyroid
malignant nodules. Journal of Clinical Endocrinology and Metabolism (2013)
doi:10.1210/jc.2013-2568.
68. Sancar, A., Ozturk, N., Lee, J. H. & Gaddameedhi, S. Loss of cryptochrome reduces
cancer risk in p53 mutant mice. Proc Natl Acad Sci U S A (2009)
doi:10.1073/pnas.0813028106.
69. van der Watt, P. J., Roden, L. C., Davis, K. T., Parker, M. I. & Leaner, V. D. Circadian
oscillations persist in cervical and esophageal cancer cells displaying decreased
expression of tumor-suppressing circadian clock genes. Molecular Cancer Research
18, (2020).
70. Doruk, Y. U. et al. A CLOCK-binding small molecule disrupts the interaction between
CLOCK and BMAL1 and enhances circadian rhythm amplitude. Journal of Biological
Chemistry 295, (2020).
71. Fallah, J. et al. FDA Approval Summary: Belzutifan for von Hippel-Lindau Disease–
Associated Tumors. Clinical Cancer Research 28, (2022).
72. Noel, R. et al. Synthesis and SAR of tetrahydroisoquinolines as Rev-erbα agonists.
Bioorg Med Chem Lett 22, (2012).
73. Solt, L. A. et al. Regulation of circadian behaviour and metabolism by synthetic REVERB agonists. Nature 485, 62–68 (2012).
74. Dierickx, P. et al. SR9009 has REV-ERB–independent eQects on cell proliferation and
metabolism. Proc Natl Acad Sci U S A 116, (2019).
75. He, Y. et al. Structure-Activity Relationship and Biological Investigation of a REVERBα-Selective Agonist SR-29065 (34) for Autoimmune Disorders. J Med Chem 66,
14815–14823 (2023).
76. Hirota, T. et al. Identification of small molecule activators of cryptochrome. Science
(1979) 337, 1094–1097 (2012).
119
77. Lee, J. W. et al. Development of Small-Molecule Cryptochrome Stabilizer Derivatives
as Modulators of the Circadian Clock. ChemMedChem 10, 1489–1497 (2015).
78. Miller, S. et al. Isoform-selective regulation of mammalian cryptochromes. Nat Chem
Biol 16, 676–685 (2020).
79. Miller, S. et al. CRY2 isoform selectivity of a circadian clock modulator with
antiglioblastoma eQicacy. Proc Natl Acad Sci U S A 119, e2203936119 (2022).
80. Licciardello, M. P. & Workman, P. A New Chemical Probe Challenges the Broad
Cancer Essentiality of CK2. Trends in Pharmacological Sciences vol. 42 Preprint at
https://doi.org/10.1016/j.tips.2021.02.002 (2021).
81. Kim, H. et al. Identification of a novel function of CX-4945 as a splicing regulator.
PLoS One 9, (2014).
82. Oshima, T. et al. Cell-based screen identifies a new potent and highly selective CK2
inhibitor for modulation of circadian rhythms and cancer cell growth. Sci Adv 5, 1–16
(2019).
83. Wells, C. I. et al. Development of a potent and selective chemical probe for the
pleiotropic kinase CK2. Cell Chem Biol 28, (2021).
84. Roeder, R. G. 50+ years of eukaryotic transcription: an expanding universe of factors
and mechanisms. Nature Structural and Molecular Biology vol. 26 Preprint at
https://doi.org/10.1038/s41594-019-0287-x (2019).
85. Vervoort, S. J. et al. Targeting transcription cycles in cancer. Nat Rev Cancer 22, 5–24
(2022).
86. Wang, Y. et al. CDK7-Dependent Transcriptional Addiction in Triple-Negative Breast
Cancer. Cell 163, 174–186 (2015).
87. Anafi, R. C., Francey, L. J., Hogenesch, J. B. & Kim, J. CYCLOPS reveals human
transcriptional rhythms in health and disease. Proc Natl Acad Sci U S A 114, 5312–
5317 (2017).
88. Li, S. Y. et al. Tumor circadian clock strength influences metastatic potential and
predicts patient prognosis in luminal A breast cancer. Proc Natl Acad Sci U S A 121,
(2024).
89. Nakao, R., Okauchi, H., Hashimoto, C., Wada, N. & Oishi, K. Determination of
reference genes that are independent of feeding rhythms for circadian studies of
mouse metabolic tissues. Mol Genet Metab 121, (2017).
120
90. Ianevski, A., Giri, A. K. & Aittokallio, T. SynergyFinder 2.0: Visual analytics of multidrug combination synergies. Nucleic Acids Res 48, (2021).
91. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: A flexible trimmer for Illumina
sequence data. Bioinformatics 30, (2014).
92. Krueger, F. Trim Galore!: A wrapper tool around Cutadapt and FastQC to consistently
apply quality and adapter trimming to FastQ files. Babraham Institute (2015).
93. Andrews, S. FastQC - A quality control tool for high throughput sequence data.
http://www.bioinformatics.babraham.ac.uk/projects/fastqc/. Babraham
Bioinformatics (2010).
94. Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: Summarize analysis
results for multiple tools and samples in a single report. Bioinformatics 32, (2016).
95. Kim, D., Paggi, J. M., Park, C., Bennett, C. & Salzberg, S. L. Graph-based genome
alignment and genotyping with HISAT2 and HISAT-genotype. Nat Biotechnol 37,
(2019).
96. Liao, Y., Smyth, G. K. & Shi, W. FeatureCounts: An eQicient general purpose program
for assigning sequence reads to genomic features. Bioinformatics 30, (2014).
97. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2. Genome Biol 15, (2014).
98. Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based approach
for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102,
(2005).
99. Korotkevich, G., Sukhov, V., Budin, N., Atryomov, M. N. & Sergushichev, A. Fast gene
set enrichment analysis. bioRxiv. bioRxiv (2021).
100. Peterson, H., Kolberg, L., Raudvere, U., Kuzmin, I. & Vilo, J. gprofiler2 -- an R package
for gene list functional enrichment analysis and namespace conversion toolset g:
Profiler. F1000Res 9, (2020).
101. Kuleshov, M. V. et al. Enrichr: a comprehensive gene set enrichment analysis web
server 2016 update. Nucleic Acids Res 44, (2016).
102. Blighe K, Rana S & Lewis M. Publication-ready volcano plots with enhanced colouring
and labeling. Bioconductor (2022).
103. Franco, H. L. et al. Enhancer transcription reveals subtype-specific gene expression
programs controlling breast cancer pathogenesis. Genome Res 28, (2018).
121
104. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat
Methods 9, (2012).
105. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol 9, (2008).
106. Pagés, H., Aboyoun, P., Gentleman, R. & DebRoy, S. Biostrings: EQicient
manipulation of biological strings. R package version 2.58.0. Preprint at (2020).
107. Barretina, J. et al. The Cancer Cell Line Encyclopedia enables predictive modelling of
anticancer drug sensitivity. Nature 483, (2012).
108. Muerdter, F. et al. Resolving systematic errors in widely used enhancer activity
assays in human cells. Nat Methods 15, (2018).
109. Daley, T. & Smith, A. D. Predicting the molecular complexity of sequencing libraries.
Nat Methods 10, (2013).
110. Humphries, P. S. et al. Carbazole-containing amides and ureas: Discovery of
cryptochrome modulators as antihyperglycemic agents. Bioorg Med Chem Lett 28,
(2018).
111. Stratmann, M., Suter, D. M., Molina, N., Naef, F. & Schibler, U. Circadian Dbp
Transcription Relies on Highly Dynamic BMAL1-CLOCK Interaction with E Boxes and
Requires the Proteasome. Mol Cell 48, 277–287 (2012).
112. Goldberg, A. L. Development of proteasome inhibitors as research tools and cancer
drugs. Journal of Cell Biology 199, (2012).
113. Bozek, K. et al. Regulation of clock-controlled genes in mammals. PLoS One 4,
e4882 (2009).
114. Mure, L. S. et al. Diurnal transcriptome atlas of a primate across major neural and
peripheral tissues. Science (1979) 359, (2018).
115. Huang, H. et al. Defining super-enhancer landscape in triple-negative breast cancer
by multiomic profiling. Nat Commun 12, (2021).
116. Pan, Y., van der Watt, P. J. & Kay, S. A. E-box binding transcription factors in cancer.
Frontiers in Oncology vol. 13 Preprint at https://doi.org/10.3389/fonc.2023.1223208
(2023).
117. Arnold, C. D. et al. Genome-wide quantitative enhancer activity maps identified by
STARR-seq. Science (1979) 339, (2013).
118. Castro-Mondragon, J. A. et al. JASPAR 2022: The 9th release of the open-access
database of transcription factor binding profiles. Nucleic Acids Res 50, (2022).
122
119. Smith, G. D., Ching, W. H., Cornejo-Páramo, P. & Wong, E. S. Decoding enhancer
complexity with machine learning and high-throughput discovery. Genome Biology
vol. 24 Preprint at https://doi.org/10.1186/s13059-023-02955-4 (2023).
120. Sahu, B. et al. Sequence determinants of human gene regulatory elements. Nat
Genet 54, (2022).
121. Andersson, R. & Sandelin, A. Determinants of enhancer and promoter activities of
regulatory elements. Nat Rev Genet 21, 71–87 (2020).
122. Eeckhoute, J., Carroll, J. S., Geistlinger, T. R., Torres-Arzayus, M. I. & Brown, M. A celltype-specific transcriptional network required for estrogen regulation of cyclin D1
and cell cycle progression in breast cancer. Genes Dev 20, (2006).
123. Deroo, B. J. & Korach, K. S. Estrogen receptors and human disease. Journal of Clinical
Investigation vol. 116 Preprint at https://doi.org/10.1172/JCI27987 (2006).
124. Liu, Z. et al. Enhancer activation requires trans-recruitment of a mega transcription
factor complex. Cell 159, (2014).
125. Dejosez, M. et al. Regulatory architecture of housekeeping genes is driven by
promoter assemblies. Cell Rep 42, (2023).
126. Zabidi, M. A. et al. Enhancer-core-promoter specificity separates developmental and
housekeeping gene regulation. Nature 518, (2015).
127. Lin, J. M. et al. Transcription factor binding and modified histones in human
bidirectional promoters. Genome Res 17, 818–827 (2007).
128. Schaub, F. X. et al. Pan-cancer Alterations of the MYC Oncogene and Its Proximal
Network across the Cancer Genome Atlas. Cell Syst 6, (2018).
129. LaGory, E. L. & Giaccia, A. J. The ever-expanding role of HIF in tumour and stromal
biology. Nature Cell Biology vol. 18 Preprint at https://doi.org/10.1038/ncb3330
(2016).
130. Brabletz, T., Kalluri, R., Nieto, M. A. & Weinberg, R. A. EMT in cancer. Nature Reviews
Cancer vol. 18 128–134 Preprint at https://doi.org/10.1038/nrc.2017.118 (2018).
131. Lambert, A. W. & Weinberg, R. A. Linking EMT programmes to normal and neoplastic
epithelial stem cells. Nat Rev Cancer 21, (2021).
132. Massó-Vallés, D., Beaulieu, M. E. & Soucek, L. MYC, MYCL, and MYCN as therapeutic
targets in lung cancer. Expert Opinion on Therapeutic Targets vol. 24 101–114 Preprint
at https://doi.org/10.1080/14728222.2020.1723548 (2020).
123
133. Baluapuri, A., Wolf, E. & Eilers, M. Target gene-independent functions of MYC
oncoproteins. Nature Reviews Molecular Cell Biology vol. 21 255–267 Preprint at
https://doi.org/10.1038/s41580-020-0215-2 (2020).
134. Beaulieu, M. E., Castillo, F. & Soucek, L. Structural and Biophysical Insights into the
Function of the Intrinsically Disordered Myc Oncoprotein. Cells vol. 9 Preprint at
https://doi.org/10.3390/cells9041038 (2020).
135. Eilers, M. & Eisenman, R. N. Myc’s broad reach. Genes and Development vol. 22
2755–2766 Preprint at https://doi.org/10.1101/gad.1712408 (2008).
136. Hayward, W. S., Neel, B. G. & Astrin, S. M. Activation of a cellular onc gene by
promoter insertion in ALV-induced lymphoid leukosis. Nature 290, (1981).
137. Vennstrom, B., Sheiness, D., Zabielski, J. & Bishop, J. M. Isolation and
characterization of c-myc, a cellular homolog of the oncogene (v-myc) of avian
myelocytomatosis virus strain 29. J Virol 42, (1982).
138. Schmitz, R., Ceribelli, M., Pittaluga, S., Wright, G. & Staudt, L. M. Oncogenic
mechanisms in Burkitt lymphoma. Cold Spring Harb Perspect Med 4, (2014).
139. Reddy, J. et al. Predicting master transcription factors from pan-cancer expression
data. Sci Adv 7, (2021).
140. Sun, L. et al. Modelling liver cancer initiation with organoids derived from directly
reprogrammed human hepatocytes. Nat Cell Biol 21, (2019).
141. Haberle, V. & Stark, A. Eukaryotic core promoters and the functional basis of
transcription initiation. Nature Reviews Molecular Cell Biology vol. 19 Preprint at
https://doi.org/10.1038/s41580-018-0028-8 (2018).
142. Reflecting on 25 years of myc.
143. Sabò, A. et al. Selective transcriptional regulation by Myc in cellular growth control
and lymphomagenesis. Nature 511, (2014).
144. Baluapuri, A., Wolf, E. & Eilers, M. Target gene-independent functions of MYC
oncoproteins. Nature Reviews Molecular Cell Biology vol. 21 255–267 Preprint at
https://doi.org/10.1038/s41580-020-0215-2 (2020).
145. Soufi, A. et al. Pioneer transcription factors target partial DNA motifs on
nucleosomes to initiate reprogramming. Cell 161, (2015).
146. Liberzon, A. et al. The Molecular Signatures Database Hallmark Gene Set Collection.
Cell Syst 1, (2015).
124
147. Herkert, B. & Eilers, M. Transcriptional repression: The dark side of Myc. Genes and
Cancer vol. 1 Preprint at https://doi.org/10.1177/1947601910379012 (2010).
148. Schneider, A., Peukert, K., Eilers, M. & Hänel, F. Association of Myc with the zincfinger protein Miz-1 defines a novel pathway for gene regulation by Myc. in Current
Topics in Microbiology and Immunology vol. 224 (1997).
149. Gartel, A. L. et al. Myc represses the p21 (WAF1/CIP1) promoter and interacts with
Sp1/Sp3. Proc Natl Acad Sci U S A 98, (2001).
150. Tu, W. B. et al. MYC Interacts with the G9a Histone Methyltransferase to Drive
Transcriptional Repression and Tumorigenesis. Cancer Cell 34, 579-595.e8 (2018).
151. Aibar, S. et al. SCENIC: Single-cell regulatory network inference and clustering. Nat
Methods 14, (2017).
152. Gomez-Roman, N., Grandori, C., Eisenman, R. N. & White, R. J. Direct activation of
RNA polymerase III transcription by c-Myc. Nature 421, (2003).
153. Grandori, C. et al. c-Myc binds to human ribosomal DNA and stimulates
transcription of rRNA genes by RNA polymerase I. Nat Cell Biol 7, (2005).
154. Nie, Z. et al. c-Myc Is a Universal Amplifier of Expressed Genes in Lymphocytes and
Embryonic Stem Cells. Cell 151, 68–79 (2012).
155. Lin, C. Y. et al. Transcriptional amplification in tumor cells with elevated c-Myc. Cell
151, 56–67 (2012).
156. van Dang, C. et al. DiQerent promoter aQinities account for specificity in MYCdependent gene regulation. (2016) doi:10.7554/eLife.15161.001.
157. Baluapuri, A. et al. MYC Recruits SPT5 to RNA Polymerase II to Promote Processive
Transcription Elongation. Mol Cell 74, 674-687.e11 (2019).
158. Kalkat, M. et al. MYC Protein Interactome Profiling Reveals Functionally Distinct
Regions that Cooperate to Drive Tumorigenesis. Mol Cell 72, 836-848.e7 (2018).
159. Lourenco, C. et al. MYC protein interactors in gene transcription and cancer. Nature
Reviews Cancer vol. 21 579–591 Preprint at https://doi.org/10.1038/s41568-021-
00367-9 (2021).
160. Wu, D. & Rastinejad, F. Structural characterization of mammalian bHLH-PAS
transcription factors. Current Opinion in Structural Biology vol. 43 1–9 Preprint at
https://doi.org/10.1016/j.sbi.2016.09.011 (2017).
125
161. Kolonko, M. & Greb-Markiewicz, B. BHLH–PAS proteins: Their structure and intrinsic
disorder. Int J Mol Sci 20, (2019).
162. Bersten, D. C., Sullivan, A. E., Peet, D. J. & Whitelaw, M. L. BHLH-PAS proteins in
cancer. Nature Reviews Cancer vol. 13 827–841 Preprint at
https://doi.org/10.1038/nrc3621 (2013).
163. Dengler, V. L., Galbraith, M. D. & Espinosa, J. M. Transcriptional regulation by hypoxia
inducible factors. Critical Reviews in Biochemistry and Molecular Biology vol. 49 1–
15 Preprint at https://doi.org/10.3109/10409238.2013.838205 (2014).
164. Kimura, H. et al. Identification of hypoxia-inducible factor 1 ancillary sequence and
its function in vascular endothelial growth factor gene induction by hypoxia and nitric
oxide. Journal of Biological Chemistry 276, (2001).
165. Galbraith, M. D. et al. HIF1A employs CDK8-Mediator to stimulate RNAPII elongation.
Cell 153, (2013).
166. Luo, W. et al. Pyruvate kinase M2 is a PHD3-stimulated coactivator for hypoxiainducible factor 1. Cell 145, (2011).
167. Card, P. B., Erbel, P. J. A. & Gardner, K. H. Structural basis of ARNT PAS-B
dimerization: Use of a common beta-sheet interface for hetero- and
homodimerization. J Mol Biol 353, (2005).
168. Rashid, I. et al. HRE Finder: A Tool for Quarrying Hypoxia-Response Element in
Genomic Sequences. Int J Curr Microbiol Appl Sci 6, 1580–1586 (2017).
169. Benita, Y. et al. An integrative genomics approach identifies Hypoxia Inducible
Factor-1 (HIF-1)-target genes that form the core response to hypoxia. Nucleic Acids
Res 37, 4587–4602 (2009).
170. Baldewijns, M. M. et al. VHL and HIF signalling in renal cell carcinogenesis. Journal of
Pathology vol. 221 125–138 Preprint at https://doi.org/10.1002/path.2689 (2010).
171. Takahashi, J. S. Transcriptional architecture of the mammalian circadian clock. Nat
Rev Genet 18, 164–179 (2017).
172. Doi, M., Hirayama, J. & Sassone-Corsi, P. Circadian Regulator CLOCK Is a Histone
Acetyltransferase. Cell 125, (2006).
173. Menet, J. S., Pescatore, S. & Rosbash, M. CLOCK: BMAL1 is a pioneer-like
transcription factor. Genes Dev 28, (2014).
126
174. Qu, M., Qu, H., Jia, Z. & Kay, S. A. HNF4A defines tissue-specific circadian rhythms by
beaconing BMAL1::CLOCK chromatin binding and shaping the rhythmic chromatin
landscape. Nat Commun 12, (2021).
175. Lim, J. & Thiery, J. P. Epithelial-mesenchymal transitions: Insights from development.
Development (Cambridge) vol. 139 Preprint at https://doi.org/10.1242/dev.071209
(2012).
176. Diepenbruck, M. & Christofori, G. Epithelial-mesenchymal transition (EMT) and
metastasis: Yes, no, maybe? Current Opinion in Cell Biology vol. 43 Preprint at
https://doi.org/10.1016/j.ceb.2016.06.002 (2016).
177. Nieto, M. A., Huang, R. Y. Y. J., Jackson, R. A. A. & Thiery, J. P. P. EMT: 2016. Cell vol.
166 Preprint at https://doi.org/10.1016/j.cell.2016.06.028 (2016).
178. Simeonov, K. P. et al. Single-cell lineage tracing of metastatic cancer reveals
selection of hybrid EMT states. Cancer Cell 39, (2021).
179. Kröger, C. et al. Acquisition of a hybrid E/M state is essential for tumorigenicity of
basal breast cancer cells. Proc Natl Acad Sci U S A 116, 7353–7362 (2019).
180. Stemmler, M. P., Eccles, R. L., Brabletz, S. & Brabletz, T. Non-redundant functions of
EMT transcription factors. Nature Cell Biology vol. 21 Preprint at
https://doi.org/10.1038/s41556-018-0196-y (2019).
181. Lin, Y. et al. The SNAG domain of snail1 functions as a molecular hook for recruiting
lysine-specific demethylase 1. EMBO Journal 29, (2010).
182. Peinado, H., Olmeda, D. & Cano, A. Snail, ZEB and bHLH factors in tumour
progression: An alliance against the epithelial phenotype? Nature Reviews Cancer
vol. 7 Preprint at https://doi.org/10.1038/nrc2131 (2007).
183. Iser, I. C., Lenz, G. & Wink, M. R. EMT-like process in glioblastomas and reactive
astrocytes. Neurochemistry International vol. 122 Preprint at
https://doi.org/10.1016/j.neuint.2018.11.016 (2019).
184. Sirito, M. et al. Members of the USF family of helix-loop-helix proteins bind DNA as
homo- as well as heterodimers. Gene Expr 2, (1992).
185. Verhoeven, A. USF1 (upstream transcription factor 1). Atlas Genet Cytogenet Oncol
Haematol (2011) doi:10.4267/2042/44944.
186. Choe, C., Chen, N. & Sawadogo, M. Decreased tumorigenicity of c-Myc-transformed
fibroblasts expressing active USF2. Exp Cell Res 302, (2005).
127
187. Bouafia, A. et al. p53 Requires the Stress Sensor USF1 to Direct Appropriate Cell Fate
Decision. PLoS Genet 10, (2014).
188. Reed, B. D., Charos, A. E., Szekely, A. M., Weissman, S. M. & Snyder, M. Genomewide occupancy of SREBP1 and its partners NFY and SP1 reveals novel functional
roles and combinatorial regulation of distinct classes of genes. PLoS Genet 4, (2008).
189. Takeda, S. et al. Myogenic regulatory factors can activate TATA-containing promoter
elements via an E-box independent mechanism. Journal of Biological Chemistry 270,
(1995).
190. del Bene, F. et al. In vivo validation of a computationally predicted conserved Ath5
target gene set. PLoS Genet 3, (2007).
191. Droin, C., Paquet, E. R. & Naef, F. Low-dimensional dynamics of two coupled
biological oscillators. Nat Phys 15, (2019).
192. Yosef, N. & Regev, A. Impulse control: Temporal dynamics in gene transcription. Cell
vol. 144 Preprint at https://doi.org/10.1016/j.cell.2011.02.015 (2011).
193. Paijmans, J., Bosman, M., ten Wolde, P. R. & Lubensky, D. K. Discrete gene replication
events drive coupling between the cell cycle and circadian clocks. Proc Natl Acad
Sci U S A 113, (2016).
194. Heltberg, M. L., Krishna, S., KadanoQ, L. P. & Jensen, M. H. A tale of two rhythms:
Locked clocks and chaos in biology. Cell Systems vol. 12 Preprint at
https://doi.org/10.1016/j.cels.2021.03.003 (2021).
195. Gaucher, J., Montellier, E. & Sassone-Corsi, P. Molecular Cogs: Interplay between
Circadian Clock and Cell Cycle. Trends in Cell Biology vol. 28 Preprint at
https://doi.org/10.1016/j.tcb.2018.01.006 (2018).
196. Feillet, C. et al. Phase locking and multiple oscillating attractors for the coupled
mammalian clock and cell cycle. Proc Natl Acad Sci U S A 111, (2014).
197. Yan, J. & Goldbeter, A. Robust synchronization of the cell cycle and the circadian
clock through bidirectional coupling. J R Soc Interface 16, (2019).
198. Hirata, H. et al. Oscillatory expression of the BHLH factor Hes1 regulated by a
negative feedback loop. Science (1979) 298, (2002).
199. Zambrano, S., de Toma, I., PiQer, A., Bianchi, M. E. & Agresti, A. NF-κB oscillations
translate into functionally related patterns of gene expression. Elife 5, (2016).
128
200. von der Lehr, N., Johansson, S. & Larsson, L. G. Implication of the
ubiquitin/proteasome system in Myc-regulated transcription. Cell cycle
(Georgetown, Tex.) vol. 2 Preprint at https://doi.org/10.4161/cc.2.5.484 (2003).
201. Thomas, D. & Tyers, M. Transcriptional regulation: Kamikaze activators. Current
Biology vol. 10 Preprint at https://doi.org/10.1016/S0960-9822(00)00462-0 (2000).
202. Rach, E. A. et al. Transcription initiation patterns indicate divergent strategies for
gene regulation at the chromatin level. PLoS Genet 7, (2011).
203. Bradner, J. E., Hnisz, D. & Young, R. A. Transcriptional Addiction in Cancer. Cell 168,
629–643 (2017).
204. Kress, T. R., Sabò, A. & Amati, B. MYC: Connecting selective transcriptional control to
global RNA production. Nature Reviews Cancer vol. 15 593–607 Preprint at
https://doi.org/10.1038/nrc3984 (2015).
205. Wei, Y. et al. Multiple direct interactions of TBP with the MYC oncoprotein. Nat Struct
Mol Biol 26, (2019).
206. Kalkat, M. et al. MYC Protein Interactome Profiling Reveals Functionally Distinct
Regions that Cooperate to Drive Tumorigenesis. Mol Cell 72, 836-848.e7 (2018).
207. Lombardi, O., Varshney, D., Phillips, N. M. & Cowling, V. H. C-Myc deregulation
induces mRNA capping enzyme dependency. Oncotarget 7, (2016).
208. Kanazawa, S., Soucek, L., Evan, G., Okamoto, T. & Peterlin, B. M. c-Myc recruits PTEFb for transcription, cellular proliferation and apoptosis. Oncogene 22, (2003).
209. Eberhardy, S. R. & Farnham, P. J. Myc recruits P-TEFb to mediate the final step in the
transcriptional activation of the cad promoter. Journal of Biological Chemistry 277,
(2002).
210. Li, Y., Sun, X. X., Qian, D. Z. & Dai, M. S. Molecular Crosstalk Between MYC and HIF in
Cancer. Frontiers in Cell and Developmental Biology vol. 8 Preprint at
https://doi.org/10.3389/fcell.2020.590576 (2020).
211. Koshiji, M. et al. HIF-1α induces cell cycle arrest by functionally counteracting Myc.
EMBO Journal 23, (2004).
212. Zhang, H. et al. HIF-1 Inhibits Mitochondrial Biogenesis and Cellular Respiration in
VHL-Deficient Renal Cell Carcinoma by Repression of C-MYC Activity. Cancer Cell
11, (2007).
213. Zhang, J. et al. Targeting angiogenesis via a c-Myc/hypoxia-inducible factor-1αdependent pathway in multiple myeloma. Cancer Res 69, (2009).
129
214. Doe, M. R., Ascano, J. M., Kaur, M. & Cole, M. D. Myc posttranscriptionally induces
HIF1 protein and target gene expression in normal and cancer cells. Cancer Res 72,
(2012).
215. Lee, K. min et al. MYC and MCL1 Cooperatively Promote Chemotherapy-Resistant
Breast Cancer Stem Cells via Regulation of Mitochondrial Oxidative Phosphorylation.
Cell Metab 26, (2017).
216. Gordan, J. D., Thompson, C. B. & Simon, M. C. HIF and c-Myc: Sibling Rivals for
Control of Cancer Cell Metabolism and Proliferation. Cancer Cell vol. 12 Preprint at
https://doi.org/10.1016/j.ccr.2007.07.006 (2007).
217. Dang, C. v., Kim, J. W., Gao, P. & Yustein, J. The interplay between MYC and HIF in
cancer. Nature Reviews Cancer vol. 8 Preprint at https://doi.org/10.1038/nrc2274
(2008).
218. Tu, R. et al. USP29 coordinates MYC and HIF1α stabilization to promote tumor
metabolism and progression. Oncogene 40, (2021).
219. Dang, C. v. The interplay between MYC and HIF in the Warburg eQect. Ernst Schering
Foundation symposium proceedings Preprint at
https://doi.org/10.1007/2789_2008_088 (2007).
220. Gordan, J. D., Bertout, J. A., Hu, C. J., Diehl, J. A. & Simon, M. C. HIF-2α Promotes
Hypoxic Cell Proliferation by Enhancing c-Myc Transcriptional Activity. Cancer Cell
11, (2007).
221. Das, B. et al. MYC Regulates the HIF2a Stemness Pathway via Nanog and Sox2 to
Maintain Self-Renewal in Cancer Stem Cells versus Non-Stem Cancer Cells. Cancer
Res 79, (2019).
222. Zarrabi, A. J., Kao, D., Nguyen, D. T., Loscalzo, J. & Handy, D. E. Hypoxia-induced
suppression of c-Myc by HIF-2α in human pulmonary endothelial cells attenuates
TFAM expression. Cell Signal 38, (2017).
223. Altman, B. J. et al. MYC Disrupts the Circadian Clock and Metabolism in Cancer
Cells. Cell Metab 22, 1009–1019 (2015).
224. Shostak, A. et al. MYC/MIZ1-dependent gene repression inversely coordinates the
circadian clock with cell cycle and proliferation. Nat Commun 7, 1–11 (2016).
225. Huber, A. L. et al. CRY2 and FBXL3 Cooperatively Degrade c-MYC. Mol Cell 64, 774–
789 (2016).
226. Liu, Z. et al. Circadian regulation of c-MYC in mice. Proc Natl Acad Sci U S A 117,
(2020).
130
227. Peek, C. B. et al. Circadian Clock Interaction with HIF1α Mediates Oxygenic
Metabolism and Anaerobic Glycolysis in Skeletal Muscle. Cell Metab 25, (2017).
228. Wu, Y. et al. Reciprocal Regulation between the Circadian Clock and Hypoxia
Signaling at the Genome Level in Mammals. Cell Metab 25, (2017).
229. Yu, C. et al. Hypoxia disrupts the expression levels of circadian rhythm genes in
hepatocellular carcinoma. Mol Med Rep 11, (2015).
230. Shen, H., Cook, K., Gee, H. E. & Hau, E. Hypoxia, metabolism, and the circadian
clock: New links to overcome radiation resistance in high-grade gliomas. Journal of
Experimental and Clinical Cancer Research vol. 39 Preprint at
https://doi.org/10.1186/s13046-020-01639-2 (2020).
231. Kobayashi, M. et al. A circadian clock gene, PER2, activates HIF-1 as an eQector
molecule for recruitment of HIF-1α to promoter regions of its downstream genes.
FEBS Journal 284, (2017).
232. Dimova, E. Y. et al. The Circadian Clock Protein CRY1 Is a Negative Regulator of HIF1α. iScience 13, (2019).
233. Vaughan, M. E. et al. Cryptochromes Suppress HIF1α in Muscles. iScience 23,
(2020).
234. Tam, S. Y., Wu, V. W. C. & Law, H. K. W. Hypoxia-Induced Epithelial-Mesenchymal
Transition in Cancers: HIF-1α and Beyond. Frontiers in Oncology vol. 10 Preprint at
https://doi.org/10.3389/fonc.2020.00486 (2020).
235. Zhang, L. et al. Hypoxia induces epithelial-mesenchymal transition via activation of
SNAI1 by hypoxia-inducible factor -1α in hepatocellular carcinoma. BMC Cancer 13,
(2013).
236. Peng, P. H., Chieh-Yu Lai, J., Hsu, K. W. & Wu, K. J. Hypoxia-induced lncRNA RP11-
390F4.3 promotes epithelial-mesenchymal transition (EMT) and metastasis through
upregulating EMT regulators. Cancer Lett 483, (2020).
237. Wu, M. Z. et al. Interplay between HDAC3 and WDR5 Is Essential for HypoxiaInduced Epithelial-Mesenchymal Transition. Mol Cell 43, (2011).
238. Tang, C. et al. Transcriptional regulation of FoxM1 by HIF-1α mediates
hypoxia-induced EMT in prostate cancer. Oncol Rep 42, (2019).
239. Larsen, J. E. et al. ZEB1 drives epithelial-to-mesenchymal transition in lung cancer.
Journal of Clinical Investigation 126, (2016).
131
240. Lin, X. et al. C-myc overexpression drives melanoma metastasis by promoting
vasculogenic mimicry via c-myc/snail/Bax signaling. J Mol Med 95, (2017).
241. Smith, A. P. et al. A positive role for Myc in TGFβ-induced Snail transcription and
epithelial-to-mesenchymal transition. Oncogene 28, (2009).
242. Hadadi, E. & Acloque, H. Role of circadian rhythm disorders on EMT and tumourimmune interactions in endocrine-related cancers. Endocrine-Related Cancer vol.
28 Preprint at https://doi.org/10.1530/ERC-20-0280 (2021).
243. De, A. et al. Cancer stem cell generation during epithelial-mesenchymal transition is
temporally gated by intrinsic circadian clocks. Clin Exp Metastasis 37, (2020).
244. Zhang, Y. et al. Bmal1 knockdown leans epithelial–mesenchymal balance toward
epithelial properties and decreases the chemoresistance of colon carcinoma cells.
Int J Mol Sci 22, (2021).
245. Keith, B., Johnson, R. S. & Simon, M. C. HIF1 α and HIF2 α: sibling rivalry in hypoxic
tumour growth and progression. Nature Reviews Cancer vol. 12 Preprint at
https://doi.org/10.1038/nrc3183 (2012).
246. Raval, R. R. et al. Contrasting Properties of Hypoxia-Inducible Factor 1 (HIF-1) and
HIF-2 in von Hippel-Lindau-Associated Renal Cell Carcinoma. Mol Cell Biol 25,
(2005).
247. Fang, H. Y. et al. Hypoxia-inducible factors 1 and 2 are important transcriptional
eQectors in primary macrophages experiencing hypoxia. Blood 114, (2009).
248. HoeQlin, R. et al. HIF-1α and HIF-2α diQerently regulate tumour development and
inflammation of clear cell renal cell carcinoma in mice. Nat Commun 11, (2020).
249. Denecker, G. et al. Identification of a ZEB2-MITF-ZEB1 transcriptional network that
controls melanogenesis and melanoma progression. Cell Death Di]er 21, (2014).
250. Caramel, J. et al. A Switch in the Expression of Embryonic EMT-Inducers Drives the
Development of Malignant Melanoma. Cancer Cell 24, (2013).
251. Ganesan, R., Mallets, E. & Gomez-Cambronero, J. The transcription factors Slug
(SNAI2) and Snail (SNAI1) regulate phospholipase D (PLD) promoter in opposite ways
towards cancer cell invasion. Mol Oncol 10, (2016).
252. Sundararajan, V. et al. SNAI1 recruits HDAC1 to suppress SNAI2 transcription during
epithelial to mesenchymal transition. Sci Rep 9, (2019).
253. Dyar, K. A. et al. Muscle insulin sensitivity and glucose metabolism are controlled by
the intrinsic muscle clock. Mol Metab 3, (2014).
132
254. He, T. L. et al. The c-Myc–LDHA axis positively regulates aerobic glycolysis and
promotes tumor progression in pancreatic cancer. Medical Oncology 32, (2015).
255. Peek, C. B. Metabolic Implications of Circadian–HIF Crosstalk. Trends in
Endocrinology and Metabolism vol. 31 Preprint at
https://doi.org/10.1016/j.tem.2020.02.008 (2020).
256. Kim, S. H. & Baek, K. H. Regulation of cancer metabolism by deubiquitinating
enzymes: The warburg eQect. International Journal of Molecular Sciences vol. 22
Preprint at https://doi.org/10.3390/ijms22126173 (2021).
257. Wise, D. R. et al. Myc regulates a transcriptional program that stimulates
mitochondrial glutaminolysis and leads to glutamine addiction. Proc Natl Acad Sci U
S A 105, (2008).
258. Dong, Y., Tu, R., Liu, H. & Qing, G. Regulation of cancer cell metabolism: oncogenic
MYC in the driver’s seat. Signal Transduction and Targeted Therapy vol. 5 Preprint at
https://doi.org/10.1038/s41392-020-00235-2 (2020).
259. Panda, S. Circadian physiology of metabolism. Science vol. 354 Preprint at
https://doi.org/10.1126/science.aah4967 (2016).
260. Stine, Z. E., Walton, Z. E., Altman, B. J., Hsieh, A. L. & Dang, C. v. MYC, metabolism,
and cancer. Cancer Discovery vol. 5 1024–1039 Preprint at
https://doi.org/10.1158/2159-8290.CD-15-0507 (2015).
261. Puig, L. S., Valera-Alberni, M., Cantó, C. & Pillon, N. J. Circadian rhythms and
mitochondria: Connecting the dots. Frontiers in Genetics vol. 9 Preprint at
https://doi.org/10.3389/fgene.2018.00452 (2018).
262. Wahlström, T. & Arsenian Henriksson, M. Impact of MYC in regulation of tumor cell
metabolism. Biochimica et Biophysica Acta - Gene Regulatory Mechanisms vol. 1849
Preprint at https://doi.org/10.1016/j.bbagrm.2014.07.004 (2015).
263. Terry, S. et al. New insights into the role of EMT in tumor immune escape. Molecular
Oncology vol. 11 Preprint at https://doi.org/10.1002/1878-0261.12093 (2017).
264. Hergenhan, S., Holtkamp, S. & Scheiermann, C. Molecular Interactions Between
Components of the Circadian Clock and the Immune System. Journal of Molecular
Biology vol. 432 Preprint at https://doi.org/10.1016/j.jmb.2019.12.044 (2020).
265. Dhanasekaran, R. et al. The MYC oncogene — the grand orchestrator of cancer
growth and immune evasion. Nature Reviews Clinical Oncology vol. 19 Preprint at
https://doi.org/10.1038/s41571-021-00549-2 (2022).
133
266. Dhanasekaran, R. et al. MYC and twist1 cooperate to drive metastasis by eliciting
crosstalk between cancer and innate immunity. Elife 9, (2020).
267. Casey, S. C. et al. MYC regulates the antitumor immune response through CD47 and
PD-L1. Science (1979) 352, (2016).
268. Zou, J. et al. MYC inhibition increases PD-L1 expression induced by IFN-γ in
hepatocellular carcinoma cells. Mol Immunol 101, (2018).
269. Liang, M. Q., Yu, F. Q. & Chen, C. C-Myc regulates PD-L1 expression in esophageal
squamous cell carcinoma. Am J Transl Res 12, (2020).
270. Noman, M. Z. et al. PD-L1 is a novel direct target of HIF-1α, and its blockade under
hypoxia enhanced: MDSC-mediated T cell activation. Journal of Experimental
Medicine 211, (2014).
271. Deng, W. et al. The Circadian Clock Controls Immune Checkpoint Pathway in Sepsis.
Cell Rep 24, (2018).
272. Jiang, Y. & Zhan, H. Communication between EMT and PD-L1 signaling: New insights
into tumor immune evasion. Cancer Letters vol. 468 Preprint at
https://doi.org/10.1016/j.canlet.2019.10.013 (2020).
273. Kessler, D. J., Duyao, M. P., Spicer, D. B. & Sonenshein, G. E. NF-kB-like factors
mediate interleukin 1 induction of c-myc gene transcription in fibroblasts. Journal of
Experimental Medicine 176, (1992).
274. Bellet, M. M., Zocchi, L. & Sassone-Corsi, P. The RelB subunit of NFκB acts as a
negative regulator of circadian gene expression. Cell Cycle 11, (2012).
275. Millet, P., McCall, C. & Yoza, B. RelB: an outlier in leukocyte biology. J Leukoc Biol 94,
(2013).
276. Li, S. et al. TWIST1 associates with NF-κB subunit RELA via carboxyl-terminal WR
domain to promote cell autonomous invasion through IL8 production. BMC Biol 10,
(2012).
277. Jensen, L. D. & Cao, Y. Clock controls angiogenesis. Cell Cycle 12, (2013).
278. Mezquita, P., Parghi, S. S., Brandvold, K. A. & Ruddell, A. Myc regulates VEGF
production in B cells by stimulating initiation of VEGF mRNA translation. Oncogene
24, (2005).
279. Meškytė, E. M., Keskas, S. & Ciribilli, Y. Myc as a multifaceted regulator of tumor
microenvironment leading to metastasis. International Journal of Molecular
Sciences vol. 21 Preprint at https://doi.org/10.3390/ijms21207710 (2020).
134
280. Burgermeister, E. et al. Aryl hydrocarbon receptor nuclear translocator-like
(ARNTL/BMAL1) is associated with bevacizumab resistance in colorectal cancer via
regulation of vascular endothelial growth factor A. EBioMedicine 45, (2019).
281. Venning, F. A., Wullkopf, L. & Erler, J. T. Targeting ECM disrupts cancer progression.
Frontiers in Oncology vol. 5 Preprint at https://doi.org/10.3389/fonc.2015.00224
(2015).
282. Gensbittel, V. et al. Mechanical Adaptability of Tumor Cells in Metastasis.
Developmental Cell vol. 56 Preprint at https://doi.org/10.1016/j.devcel.2020.10.011
(2021).
283. Stowers, R. S. et al. Matrix stiQness induces a tumorigenic phenotype in mammary
epithelium through changes in chromatin accessibility. Nat Biomed Eng 3, (2019).
284. Scott, L. E., Weinberg, S. H. & Lemmon, C. A. Mechanochemical Signaling of the
Extracellular Matrix in Epithelial-Mesenchymal Transition. Frontiers in Cell and
Developmental Biology vol. 7 Preprint at https://doi.org/10.3389/fcell.2019.00135
(2019).
285. Chang, J. et al. Circadian control of the secretory pathway maintains collagen
homeostasis. Nat Cell Biol 22, (2020).
286. Petrova, V., Annicchiarico-Petruzzelli, M., Melino, G. & Amelio, I. The hypoxic tumour
microenvironment. Oncogenesis vol. 7 Preprint at https://doi.org/10.1038/s41389-
017-0011-9 (2018).
287. Lee, K. E. & Simon, M. C. From stem cells to cancer stem cells: HIF takes the stage.
Current Opinion in Cell Biology vol. 24 Preprint at
https://doi.org/10.1016/j.ceb.2012.01.005 (2012).
288. Wang, J. et al. c-Myc is required for maintenance of glioma cancer stem cells. PLoS
One 3, (2008).
289. Henley, M. J. & Koehler, A. N. Advances in targeting ‘undruggable’ transcription
factors with small molecules. Nature Reviews Drug Discovery vol. 20 669–688
Preprint at https://doi.org/10.1038/s41573-021-00199-0 (2021).
290. Bushweller, J. H. Targeting transcription factors in cancer — from undruggable to
reality. Nature Reviews Cancer vol. 19 Preprint at https://doi.org/10.1038/s41568-
019-0196-7 (2019).
291. Massó-Vallés, D. & Soucek, L. Blocking Myc to Treat Cancer: Reflecting on Two
Decades of Omomyc. Cells vol. 9 Preprint at https://doi.org/10.3390/cells9040883
(2020).
135
292. Han, H. et al. Small-Molecule MYC Inhibitors Suppress Tumor Growth and Enhance
Immunotherapy. Cancer Cell 36, (2019).
293. Llombart, V. & Mansour, M. R. Therapeutic targeting of “undruggable” MYC.
eBioMedicine vol. 75 Preprint at https://doi.org/10.1016/j.ebiom.2021.103756
(2022).
294. Ban, H. S., Uto, Y. & Nakamura, H. Hypoxia-inducible factor (HIF) inhibitors: a patent
survey (2016–2020). Expert Opinion on Therapeutic Patents vol. 31 Preprint at
https://doi.org/10.1080/13543776.2021.1874345 (2021).
295. Jonasch, E. et al. Belzutifan for Renal Cell Carcinoma in von Hippel–Lindau Disease.
New England Journal of Medicine 385, (2021).
296. Holmes, A. G. et al. A MYC Inhibitor Selectively Alters the MYC and MAX Cistromes
and Modulates the Epigenomic Landscape to Regulate Target Gene Expression. Sci.
Adv vol. 8 https://www.science.org (2022).
Abstract (if available)
Abstract
Circadian genes form a network that regulates the 24-hour rhythmic patterns of cellular and organismal activities. Recent studies have uncovered various connections between circadian genes and cancer outcomes, cancer cell progression, and the modulation of the tumor microenvironment, paving the way for further research into these genes to advance cancer biology and patient care. However, these studies also highlight the complex and context-dependent roles of circadian genes, necessitating detailed follow-up studies to fully understand and leverage their potential in cancer treatment.
To address these challenges, this thesis explores the roles of circadian genes from three perspectives: their potential as biomarkers in cancer patients, their role and targetability as therapeutic targets, and their capacity to modulate the tumor microenvironment. Firstly, we identified expression patterns of circadian genes in breast cancer that could serve as diagnostic and prognostic biomarkers. Secondly, we demonstrated that the core circadian genes BMAL1 and CLOCK are promising targets in triple-negative breast cancer and proposed a strategy to target their transcriptional programs through a cis-regulatory mechanism. Thirdly, we developed bioengineered models of the glioblastoma tumor microenvironment to test the effects of drugs targeting the circadian gene CK2. These studies enhance our understanding of circadian genes in cancer and suggest potential strategies for incorporating them into cancer treatment.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Microenvironmental and biomechanical regulation of mitochondrial membrane potential in cancer cells
PDF
Mechanistic modeling of angiogenic factors network and cancer therapy
PDF
Genome-scale modeling of macrophage activity in the colorectal cancer microenvironment
PDF
Predicting functional consequences of SNPs: insights from translation elongation, molecular phenotypes, and pathways
PDF
Investigating the complexity of the tumor microenvironment's role in drug response
PDF
Characterization of a new HER2 positive mouse model of metastatic breast cancer
PDF
Engineering immunotoxin and viral vectors for cancer therapy
PDF
The effect of inhibiting the NFkB pathway on myeloid derived suppressor cell cytokine production
PDF
Pathogenic variants in cancer predisposition genes and risk of non-breast multiple primary cancers in breast cancer patients
PDF
The role of heritability and genetic variation in cancer and cancer survival
PDF
Engineering a post-infarct human myocardium on a chip to reveal oxygen-dependent crosstalk in cardiac fibroblasts and myocytes
PDF
Identification of CBP/FOXM1 as a molecular target in triple negative breast cancer
PDF
Investigating the role of SASH1 gene located on chromosome 6 in ovarian cancer
PDF
Molecularly targeted micelle nanoparticles for cancer drug delivery and lymph node metastasis detection
PDF
Development of a colorectal cancer-on-chip to investigate the tumor microenvironment's role in cancer progression
PDF
Decreased levels of expression of transmembrane protein 56 (TMEM56) in breast cancer tissues
PDF
Molecular mechanisms of chemoresistance in breast cancer
PDF
Corisol's role in breast-to-brain metastasis
PDF
Exploration of the roles of cancer stem cells and survivin in the pathogenesis and progression of prostate cancer
PDF
Study of the role of bone morphogenetic proteins in prostate cancer progression
Asset Metadata
Creator
Pan, Yuanzhong
(author)
Core Title
Roles of circadian clock genes in cancer
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Biomedical Engineering
Degree Conferral Date
2024-12
Publication Date
10/18/2024
Defense Date
05/29/2024
Publisher
Los Angeles, California
(original),
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
breast cancer,circadian rhythm,glioblastoma,transcription
Format
theses
(aat)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Kay, Steve (
committee chair
), Finley, Stacey (
committee member
), Mumenthaler, Shannon (
committee member
), Roussos Torres, Evanthia (
committee member
)
Creator Email
michapan93@gmail.com,yuanzhop@usc.edu
Unique identifier
UC11399CJB5
Identifier
etd-PanYuanzho-13601.pdf (filename)
Legacy Identifier
etd-PanYuanzho-13601
Document Type
Dissertation
Format
theses (aat)
Rights
Pan, Yuanzhong
Internet Media Type
application/pdf
Type
texts
Source
20241018-usctheses-batch-1219
(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.
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
breast cancer
circadian rhythm
glioblastoma
transcription