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
/
The role of PAX8 in epithelial ovarian carcinoma
(USC Thesis Other)
The role of PAX8 in epithelial ovarian carcinoma
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE ROLE OF PAX8 IN EPITHELIAL OVARIAN CARCINOMA
By
Emily Kate Adler
A Dissertation Presented to the
Faculty of the Graduate School at
The University of Southern California
In Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in
Genetic, Molecular and Cellular Biology
December 2017
Copyright 2017 Emily Kate Adler
i
ACKNOWLEDGEMENTS
I could not possibly have done this without the expertise, collaboration, and support in
and outside of the lab. There are many people that have helped me along the way and here I
would like to formally thank them.
I would like to thank my mentor Simon Gayther for his leadership, guidance, patience,
support, and expertise. In addition I would like to thank him for his leading role in contributing
to the scientist I am. Lastly I am grateful for the freedom he gave me to pursue my path.
I would like to thank Kate Lawrenson for her countless time spent in guidance,
troubleshooting, feedback, support, and her invaluable contribution to my learning and research.
To my dissertation committee Chair Michael Stallcup, I would like to say thank you for
his mentorship, especially his generosity of time, support in and outside of the lab, and for his
feedback on my dissertation.
I would like to thank Peggy Farnham for her expertise, time and guidance on my work as
my dissertation committee member.
To Wendy Cozen I am grateful for her support during difficult times, being great to work
with in her course, and her time and service on my dissertation committee.
I would like to thank my collaborators for their invaluable analyses and contributions and
for validating that today’s research is a team effort. To R. Ivetth Corona de la Fuente and Dennis
Hazelett for their crucial analysis of my ChIP-seq data, to Paulette Mhawech Faceuglia for her
expertise and time in pathology, Heidi Sowter for sharing her clinical data and time, Siddhartha
ii
Kar for his data analysis in respect to PAX8 expression and ovarian cancer risk, and Norma
Rodriguez-Malave for her time and efforts in microarray validation.
I would like to thank my past committee members Gerhard Coetzee and Agniezka
Kobielak for their feedback, support, and expertise.
I would like to thank past lab members and colleagues Janet Lee, Tassja Spindler,
Melissa Delgado, Roxanne Manek, Doerthe Brüeggmann, Elham Pakzamir, Howard Shen, Susan
Ramus, Charles Nicolet, and Gillian Little for their friendship and support, guidance, critical
feedback, and the lending of their time.
To my family I am grateful for their unending love and support – my parents, sister,
brother-in-law, aunts, uncles, and cousins. There is no way to measure their support over the
years.
I thank my friends and colleagues Lisa Yan, Lacey Westphal, Ying Wu, Damian Wang,
Kristen Kuhl, Cheray Skillicorn, Peter Yates, Annie Yates, Ayesha Bhatia, Ranjani
Lakshminarasimhan, Haejung Won, and Chris Edlund for their love, troubleshooting, and
invaluable support over the years.
To Gorjana Bezmalinovic for her encouragement, expertise, and guidance in one of the
most demanding TAships in the department, and to my students for the pleasure of being their
teacher.
I would like to thank the PIBBS department including Ite Laird-Offringa, Joyce Perez,
Bami Andrada, Marisela Zuniga, Deborah Johnson, the staff at the Eric Cohen Student Health
Center, Yibu Chen and Meng Li at the Norris Medical Library Bioinformatics Core, Lora Barsky
iii
and the USC FACS Core, the Doheny Imaging Core, Dennis Trana and Lillian Young in the
USC Pathology Cores, the UCLA Neuroscience Genomics Core for microarray processing, and
Charles Nicolet and the USC Epigenetics Core team.
Last but not least I would like to acknowledge the funding support of the USC NIGMS
Training Grant.
iv
I dedicate my thesis to:
Both of my parents for my inherent love of science
My dad, Gene Adler, the forever scientist, forever my dad
My mom, Pat Adler, who stayed up late with me countless nights helping me write high school
English essays I’d begrudgingly and inevitably do the night before they were due
My sister, Kim Stanley, the rocket scientist-turned computer scientist and all around badass
And my aunt, Emeira Adler, for her expert advice and soul-lifting laughter
Thank you for your unending support
Forever yours,
Dr. Adler, PhD
v
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ……………………………………………………………………...i
DEDICATION ……………………………………………………………………………….….iv
LIST OF FIGURES AND TABLES ……………………………...………………………...…vii
LIST OF ABBREVIATIONS ……………………………...……………………………...……ix
CHAPTER 1: INTRODUCTION TO EPITHELIAL OVARIAN CANCER………….…......1
CHAPTER 2: THE ROLE OF PAX8 IN NORMAL OVARIES……………………………...9
INTRODUCTION……………………………………………………………………………..9
MATERIALS AND METHODS……………………………………………………….........15
RESULTS……………………………………………………………………………….........22
DISCUSSION………………………………………………………………………….….....30
EXPERIMENTAL AND ANALYTICAL CONTRIBUTIONS……………………………..37
REFERENCES………………………………………………………………………...……..38
CHAPTER 3: THE GENOMIC LANDSCAPE OF PAX8 IN EPITHELIAL OVARIAN
CARCINOMA…………………………………………………………………………………...48
INTRODUCTION…………………………………………………………………....…....…48
MATERIALS AND METHODS……………………………………………………….........50
RESULTS……………………………………………………………………………….........55
DISCUSSION……………………………………………………………….……..………...75
EXPERIMENTAL AND ANALYTICAL CONTRIBUTIONS……………………………..80
REFERENCES………………………………………………………………………...…......81
CHAPTER 4: CONCLUSIONS ON THE ROLE OF PAX8 IN EPITHELIAL OVARIAN
CARCINOMA…………………………………………………………………………...………88
LIST OF PUBLICATIONS BY THE AUTHOR……………………...………………………92
vi
APPENDIX………………………………………………………………………………………93
vii
LIST OF FIGURES AND TABLES
CHAPTER 2: THE ROLE OF PAX8 IN NORMAL OVARIES
TABLE 1. Published PAX8 immunohistochemistry of the normal ovarian surface
epithelium…………………………………………..…………………..………...15
FIGURE 1. PAX8 expression in normal ovarian surface epithelium…………….……….…23
FIGURE 2. PAX8 protein and mRNA expression is correlated in OSECs……………….....25
TABLE 2. Cellular morphology in PAX8 positive and negative OSECs……………….......26
TABLE 3. PAX8 immunohistochemistry in OSECs and CICs of normal ovaries…....…......27
TABLE 4. Epidemiological risk factors do not correlate with PAX8 expression status.........27
FIGURE 3. PAX8, calretinin, and E-cadherin expression in normal ovarian surface
epithelium….……………………………………………………………………29
FIGURE 4. PAX8 does not induce neoplastic transformation in OSECs…………………....31
CHAPTER 3: THE GENOMIC LANDSCAPE OF PAX8 IN EPITHELIAL OVARIAN
CARCINOMA
TABLE 1. PAX8 expression in EOC histotypes..…………………………………....…...…56
FIGURE 1. PAX8 gene expression in EOC histotypes and precursor cells…….....................57
FIGURE 2. In vitro analysis of PAX8 knockdown models….…………………………........59
FIGURE 3. In vivo analysis of PAX8 knockdown models………………………………......60
FIGURE 4. ChIP-seq binding regions are shared and different in HEYA8 and IGROV1…..61
FIGURE 5. PAX8 ChIP-seq binding sites for HEYA8 and IGROV1………………...…......62
FIGURE 6. Defining the PAX8 binding motifs and identifying candidate cooperating
transcription factors……………………………………………………………..64
TABLE 2. Candidate PAX8 cooperating transcription factors………………………………66
viii
TABLE 3. Significantly changing genes in EOC models following PAX8 knockdown….…68
TABLE 4. Genes commonly changing in HEYA8 and IGROV1 models following PAX8
knockdown……………………………………………………………………….69
FIGURE 7. Identification of PAX8 regulatory targets using TADs…...…..……….……..…70
TABLE 5. PAX8 putative enhancer regulatory targets………………………………………72
FIGURE 8. Pathway enrichment analysis……………………………………………………74
ix
LIST OF ABBREVIATIONS
AMP adenosine monophosphate
AMPK AMP-activated protein kinase
ARID1A AT-rich interaction domain 1A
BRCA1 breast cancer 1
BRCA2 breast cancer 2
BWA Burrows-Wheeler Aligner
CCOC clear cell epithelial ovarian carcinoma
cDNA complementary DNA
CIC cortical inclusion cyst
CMYC v-myc avian myelocytomatosis viral oncogene homolog
Ct threshold cycle
DMEM Dulbecco's Modified Eagle's medium
DMSO dimethyl sulfoxide
EnOC endometrial ovarian epithelial carcinoma
EOC epithelial ovarian carcinoma
FBS fetal bovine serum
FC fold change
FDR false discovery rate
FTE fallopian tube epithelium
FTSEC fallopian tube secretory epithelial cell
G2M gap 2 to mitotic phase
x
G418 geneticin
GAPDH glyceraldehyde 3-phosphate dehydrogenase
H3K27ac histone 3 lysine 27 acetylation
HGSOC high-grade serous ovarian carcinoma
HOMER Hypergeometric Optimization of Motif EnRichment
hTERT human telomerase reverse transcriptase
IDR irreproducible discovery rate
IOE immortalized ovarian surface epithelial cells
MACS2 Model-based Analysis of ChIP-Seq
MEM minimal essential medium
MOC mucinous ovarian carcinoma
NOSE-CM normal ovarian surface epithelial cell culture media
OSE ovarian surface epithelium
OSEC ovarian surface epithelial cell
p53 tumor suppressor 53 (protein)
pEGFP plasmid enhanced green fluorescent protein
P8OR PAX8 occupied region
PAX paired box
PCR polymerase chain reaction
PBS phosphate buffered saline
qPCR semi-quantitative PCR
RNA ribonucleic acid
RNase ribonuclease
xi
RIPA radioimmunoprecipitation assay
RT-qPCR reverse transcription qPCR
RPMI Roswell Park Memorial Institute medium
shPAX8 antisense-PAX8 shRNA
shRNA short hairpin RNA
shScr antisense scrabled-sequence shRNA
SOC serous ovarian cancer
STIC serous tubal intraepithelial carcinoma
STIL serous tubal intraepithelial lesion
TP53 tumor suppressor 53 (gene)
1
CHAPTER 1: INTRODUCTION TO EPITHELIAL OVARIAN CANCER
Morbidity and mortality
Cancer that arises in ovarian tissue is termed ovarian cancer. The majority (>90%)
of these are of an epithelial nature, and are thus termed epithelial ovarian carcinomas
(EOCs). EOC is the most lethal gynecological malignancy in the United States and the
western world, and is the most lethal cancer of the reproductive system. Historically the
primary tumor has been difficult to distinguish from the rest of the tumor mass because
patients were and still are most often diagnosed after EOC has spread throughout the
abdominal cavity. At this stage it is impossible to distinguish from where the tumor has
originated. However because the ovaries are involved so pervasively and usually
comprise the bulk of tumor mass, such cancers are known as ovarian carcinomas.
In 2017 there will be an estimated 14,080 deaths from EOC, with 22,440 new
estimated cases
1
. While relative 5-year survival was 47% overall, the rate for locally
staged EOCs was 92%, and for distantly staged EOCs survival was 29% (from 2006-
2012)
2
. From 2009-2013 the incidence rate was 11.6 and from 2010-2014 the death rate
was 7.4 per 100,000 women
3
. The estimated lifetime risk of developing EOC is
approximately 1 in 75, and the lifetime risk of dying of EOC is approximately 1 in 100
1
.
Early detection and screening
EOC staging is based on its spread throughout the abdominal cavity and beyond.
In stage I lesions are confined to the ovar(ies) or fallopian tube(s), by stage II they have
spread to other pelvic organs but not lymph nodes, in stage III have spread to the
2
abdominal lining and/or retroperitoneal lymph nodes, and in stage IV have spread outside
of the abdominal cavity. EOC grading takes into account tumor histology and can be low
grade, borderline, or high grade.
Only 15% of EOC patients are diagnosed when their disease is locally staged
(stage 1 and stage 2)
4
. Based on the drastic increase in survival when diagnosed locally
(92%) versus distantly staged (29%, stage 3 and stage 4), early detection could
significantly reduce disease morbidity and mortality
2
. EOC screening methods exist, but
lack specificity and sensitivity. Because the ovaries are located deep within the
abdominal cavity, they are difficult to visualize or access without invasive surgery.
Transvaginal ultrasound is often used, but it cannot distinguish between malignant and
benign abnormalities and masses. Furthermore the majority of ovarian masses found
using transvaginal ultrasound are not cancer
4
. This issue is further confused by the
common occurrence of non-cancerous cysts in the ovary. Blood can be tested for CA-125
(encoded by mucin-16), a glycoprotein shed by many EOCs. However CA-125 is not
expressed by all EOCs and is less commonly detected in early stage disease. In addition
CA-125 can often be elevated in benign conditions
5
. There is no current method of
detecting early stage disease without removing fallopian tubes alone or a combination of
fallopian tubes, and/or the ovaries, the uterus, ascitic fluid (abnormal fluid found in the
peritoneal cavity), and the peritoneal lining. Given the high potential to benefit early
detection, and the lack of adequate methods to do so, it is imperative to better understand
early disease so that current detection methods can be improved.
3
Risk factors, signs and symptoms
EOC etiology is not well understood. Some risk factors have been identified,
including a strong family history of breast or ovarian cancer, increased age, low
gravidity, and low parity, while oral contraceptive use is protective. While still unclear, it
is thought that a lower exposure to estrogen is protective. About 10% of the most
common EOC subtype is attributed to hereditary mutations, the most common genes
being BRCA1 and BRCA2.
There are often no disease signs or symptoms. If they occur, they usually present
during late stage disease, as early stage is usually asymptomatic. They can also indicate a
vast array of conditions and even normal physiology. Indicators may be abdominal
swelling, abdominal pain, increased abdominal girth, decreased appetite and/or feeling
full more quickly after eating. Such symptoms are also common and benign in women
after menopause, when EOC most frequently occurs (median age is 63) and has the worst
prognosis
1, 6
.
Epithelial ovarian carcinoma subtypes
Malignant EOCs are categorized into four main subtypes based on their
histological appearance. They are termed high-grade serous, clear cell, endometrioid, and
mucinous EOC. High-grade serous ovarian cancer (HGSOC) is the most malignant
and most common histotype, accounting for over 60% of EOCs. It is classified by its
serous morphology, and is often papillary in appearance. HGSOCs are characterized by
TP53 mutations (considered to be 100%)
7
, BRCA1 and/or BRCA2 mutations or epigenetic
alterations (30%), CMYC, and genomic instability as evidenced by high copy number
4
variation
8
.
The ovarian clear cell carcinomas (CCOC) comprise about 10% of EOCs. It is
the most aggressive subtype upon recurrence after chemotherapy. CCOC is characterized
by mutations in the chromatin remodeling complex ARID1A, as well as HNF1 and
PIK3CA mutations
9
. It is named after the appearance of many clear, glycogen-containing
cells and its histological resemblance to renal clear cell carcinoma.
Ovarian endometrioid carcinomas (EnOCs) account for 10% of EOCs and
have a glandular appearance, histologically resembling the endometrial lining of the
uterus. Common EnOC mutations are in ARID1A, PTEN, and CTTNB1 (B-catenin).
Hereditary non-polyposis colorectal cancer genes MSH2 and MLH1 increase the risk of
EnOC
10
.
Mucinous EOC (MOC) is so named due to the presence of mucin containing
cells. Common mutations are found in KRAS and HER2
11
. MOC accounts for 3% of
EOCs and is the least studied and understood of malignant EOC subtypes. There is no
consensus on the originating tissue of MOC.
Other EOC subtypes include unclassified epithelial tumors which are
histologically epithelial but contain features intermediate between two or more specific
histotypes (HGSOC, CCC, EEC, MOC)
12
, and undifferentiated carcinoma, characterized
by a poorly differentiated epithelial structure which is difficult to assign to any of the
known subtypes.
PAX8
One common factor among HGSOC, COC, MOC and EnOC is the
5
overexpression of the tissue-specific developmental transcription factor PAX8
13
. Paired
box 8 (PAX8) belongs to the PAX gene family, which contains nine members. All nine
PAX genes contain a DNA-binding paired box which binds DNA in a sequence specific
manner
14
. PAX genes are classified into four groups based on their structural sequences,
namely their full paired domain, a whole or partial homeodomain, and the presence or
absence of an octapeptide. PAX2, PAX5 and PAX8 belong to group II which contain an
octapeptide and a partial homeodomain
15
.
PAX genes have been implicated in a variety of cancers and show oncogenic
potential
16
. PAX8 knockdown has been shown to affect proliferation and anchorage
independent growth in ovarian cancer cells
17
, and knockdown was shown to induce
cellular senescence
18
. In a separate study, PAX8 knockdown was shown to cause cell
death by apoptosis
19
.
PAX8 is crucial during development and organogenesis. Adult expression is
restricted to only certain tissues, including the reproductive tract epithelia. PAX8 is a
tissue specific TF critical to the development of the thyroid, kidneys, and Müllerian
tract
20
. After development has taken place it is not expressed by most cells in the body,
however in cells in which PAX8 was developmentally active, it is responsible for cell-
type specific maintenance of a differentiated phenotype
21
.
Because the effects of PAX8 in EOC, its pathways, gene targets and its cistrome
are largely unknown, I set out to determine the significance, if any, of PAX8
overexpression in EOC precursor cells. I hypothesized that PAX8 expression can
transform ovarian surface epithelial cells. I also hypothesized that PAX8 is an oncogene
in malignant, PAX8-overexpressing EOC. To test these hypotheses I evaluated the effects
6
of exogenous PAX8 expression in EOC precursor cells, and whether PAX8 knockdown
affects tumorigenicity of EOC cells in vitro and in vivo. To better understand how PAX8
contributes to tumor initiation and progression, I identified its binding targets throughout
the genome in EOC and gene expression changes resulting from PAX8 knockdown.
REFERENCES
1. American Cancer Society. Cancer Facts and Figures; 2017.
2. Surveillance, Epidemiology, and End Results 18 registries. National Cancer
Institute. 2016.
3. National American Registries AoCC. 2016.
4. Goff BA ML, Muntz HG, Melancon CH. Ovarian carcinoma diagnosis. Cancer.
2000;89(10):2068-2065.
5. Kabawat SE, Bast RC, Welch WR, Knapp RC, Colvin RB. Immunopathologic
characterization of a monoclonal antibody that recognizes common surface
antigens of human ovarian tumors of serous, endometrioid, and clear cell. Am J
Clin Pathol. 1983;79(1):98-104.
6. Howlader N, Noone AM, Krapcho M, et al. SEER Cancer Statistics Review,
1975-2012: National Cancer Institute.
7. Vang R, Levine DA, Soslow RA, Zaloudek C, Shih Ie M, Kurman RJ. Molecular
Alterations of TP53 are a Defining Feature of Ovarian High-Grade Serous
Carcinoma: A Rereview of Cases Lacking TP53 Mutations in The Cancer
Genome Atlas Ovarian Study. Int J Gynecol Pathol. Jan 2016;35(1):48-55.
7
8. The Cancer Genome Atlas Research Network. Integrated genomic analyses of
ovarian carcinoma. Nature. 2011;474:609-615.
9. Wiegand KC, Shah SP, Al-Agha OM, et al. ARID1A Mutations in
Endometriosis-Associated Ovarian Carcinomas. N Engl J Med. Oct 14
2010;363(16):1532-1543.
10. Watson JM, Sensintaffar JL, Berek JS, Martinez-Maza O. Constitutive production
of interleukin 6 by ovarian cancer cell lines and by primary ovarian tumor
cultures. Cancer Res. Nov 01 1990;50(21):6959-6965.
11. Enomoto T, Weghorst CM, Inoue M, Tanizawa O, Rice JM. K-ras activation
occurs frequently in mucinous adenocarcinomas and rarely in other common
epithelial tumors of the human ovary. Am J Pathol. Oct 1991;139(4):777-785.
12. Serov SF, Scully RE, Sobin LH. Histological typing of ovarian tumors: World
Health Organization; 1973.
13. Ozcan A, Liles N, Coffey D, Shen SS, Truong LD. PAX2 and PAX8 expression
in primary and metastatic mullerian epithelial tumors: a comprehensive
comparison. Am J Surg Pathol. Dec 2011;35(12):1837-1847.
14. Kozmik Z, Kurzbauer R, Dörfler P, Busslinger M. Alternative splicing of Pax-8
gene transcripts is developmentally regulated and generates isoforms with
different transactivation properties. Mol Cell Biol. Oct 1993;13(10):6024-6035.
15. Balczarek KA LZ, Kumar S. Evolution of functional diversification of the paired
box (Pax) DNA-binding domains. - PubMed - NCBI. Mol Biol Evol.
1997;14(8):829-842.
8
16. Robson EJD, He S-J, Eccles MR. A PANorama of PAX genes in cancer and
development. Nature Reviews Cancer. 2006;6(1):52-62.
17. Di Palma T, Lucci V, de Cristofaro T, Filippone MG, Zannini M. A role for
PAX8 in the tumorigenic phenotype of ovarian cancer cells. BMC Cancer.
2014;14(1):292.
18. Li CG, Nyman JE, Braithwaite AW, Eccles MR. PAX8 promotes tumor cell
growth by transcriptionally regulating E2F1 and stabilizing RB protein.
Oncogene. 2011;30(48):4824-4834.
19. Cheung HW, Cowley GS, Weir BA, et al. Systematic investigation of genetic
vulnerabilities across cancer cell lines reveals lineage-specific dependencies in
ovarian cancer. Proc Natl Acad Sci U S A. 2011;108(30):12372-12377.
20. Plachov D, Chowdhury K, Walther C, Simon D, Guenet JL, Gruss P. Pax8, a
murine paired box gene expressed in the developing excretory system and thyroid
gland. Development. Oct 1990;110(2):643-651.
21. Pasca di Magliano M, Di Lauro R, Zannini M. Pax8 has a key role in thyroid cell
differentiation. Proc Natl Acad Sci U S A. 2000;97(24):13144-13149.
9
CHAPTER 2: THE ROLE OF PAX8 IN NORMAL OVARIES
This chapter has been adapted from published original research by
Emily Adler, Paulette Mhawech-Fauceglia, Simon Gayther, and Kate Lawrenson
titled PAX8 expression in ovarian surface epithelial cells
in Human Pathology, Volume 46, Issue 7, July 2015
INTRODUCTION
Epithelial ovarian carcinoma
Epithelial ovarian carcinoma (EOC) is a deadly disease marked by frequent late stage
diagnosis and poor survival rates. Marginal improvements in prevention, detection, diagnosis and
treatment stem from a poor understanding of EOC etiology. It is not well understood why most
women develop ovarian cancer. Furthermore early diagnosis is a significant clinical challenge
leading to frequent late stage diagnosis - when survival rates are very low. Since relative 5-year
survival rates are above 92% when EOC is detected before local spread, early diagnosis could
drastically improve patient mortality
1
. However it is not clear where EOC starts and how.
EOC is an umbrella term for several histological subtypes of different etiologies and,
likely, different cellular origins. High-grade serous ovarian carcinoma (HGSOC) is the most
common subtype, accounting for approximately two-thirds of all EOCs. Historically, HGSOCs
were thought to arise from ovarian surface epithelial cells (OSECs), the mesothelial-type
epithelium covering the ovary, and from the epithelial lining of cortical inclusion cysts (CICs),
which are derived from invaginations of the ovarian surface. The OSE is a simple mesothelium
10
of cells that can be flattened or columnar. The surface is loosely attached to the basement
membrane and is distant from the ovarian cortex (inner region of the ovary) by the tunica
albuginea (tissue directly underneath the OSE)
2
. This separation of the OSE from the ovarian
cortex may serve as a partial barrier to mitogens produced by the stroma. This is supported by
the tendency of metaplastic cells lining ovarian inclusion cysts to occur more frequently on the
stromal side of the cyst, and be less frequent near the OSE tunica albuginea
3, 4
.
Decades of research have indicated that HGSOCs may arise in these tissues
3, 5-7
. However
this hypothesis has been criticized because metaplastic OSECs are rarely found on the surface of
the ovary, and because ovarian carcinomas have been shown to have markedly different
histological features and marker expression when compared to normal OSECs
2, 5, 8
. More recent
evidence indicates that HGSOCs can originate from epithelial cells lining the fallopian tube
fimbriae, through a precursor lesion termed serous tubal intraepithelial carcinoma (STIC).
STICs are associated with 20-60% of sporadic HGSOCs, raising the possibility that the
remaining 40-80% of HGSOCs may arise from a different cell type
9-12
.
There are several lines of evidence supporting OSE origins of HGSOC. In a study
comparing EOC high- and low-risk women (BRCA1 and/or BRCA2 carriers vs BRCA normal),
prophylactically removed ovaries from high-risk women had significantly more dysplastic
changes than those from low-risk women. Two women had malignant neoplasia at the
microscopic or near-microscopic level
13
. In high-risk women 85% had two or more of the
following: OSE pseudostratification, papillomatosis; deep cortical invaginations of the OSE,
frequently with multiple papillary projections within small cystic spaces (microscopic papillary
cystadenomas); CICs, frequently with epithelial hyperplasia and papillary formations; cortical
11
stromal hyperplasia and hyperthecosis; or hilar cell (anatomic region of OSE) hyperplasia. 75%
of high-risk women had three or more of these changes
13
. Early lesions have been found in many
contexts in both the OSE and CICs
14
. Monozygotic twin sisters of EOC patients (the majority
had the HGSOC histotype), despite being unaffected, had dysplastic changes in their grossly
normal OSE resembling early signs of HGSOC when compared to controls. This included
papillary structures with a fibrous stalk protruding from the OSE, covered by cuboidal
epithelium, which sometimes contained psammoma bodies (concentric calcifications seen in
HGSOC). The unaffected twins also had a high number of CICs, significantly greater than the
number in control ovaries. The CICs in unaffected sisters very frequently had stratified
epithelium, pleomorphic nuclei, loss of polarity, and abnormal, coarse chromatin. The epithelia
of CICs have been shown to transition from normal, to dysplastic, to invasive carcinoma.
Through this transition cells have been shown to contain the same BRCA1 heterozygous deletion
and TP53 missense mutation, while the OSE contained both mutant and wild type BRCA1
populations. BRCA2 was similarly involved with TP53 deletion in CIC and its respective OSE
5
.
In addition, the same aneusomies have not only been found in the OSE, CICs, and HGSOCs of
the same ovary, but the number of aneusomic cells has also been shown to increase in transition
through these respective tissues
5, 15
. Most of these studies rely on genetically predisposed women
(BRCA carriers and/or EOC families) due to the difficulty in detecting early EOC in the general
population. However, women with sporadic unilateral EOC have similarly been found to have
serous dysplasia and an increased number of inclusion cysts in their unaffected ovary
16
.
Dysplastic OSECs have also been found adjacent to and/or transitioning into carcinoma on the
OSE
3, 17
. Molecular analysis of CIC epithelia has shown these cells have increased aneuploidy,
12
cell proliferation, and decreased apoptosis
5
. However these studies have been limited by both the
limited availability of tissue and the lack of representative models, both in vivo and in vitro.
Animal models of HGSOC origins
Sporadic and induced genetic models of HGSOC both show evidence of OSE origins.
The laying hen is the only known animal to spontaneously and regularly develop ovarian
carcinoma beside humans. What both animals have in common, and what distinguishes them
from other animals, is their incessant ovulation. This phenomenon positively correlates with
ovarian carcinoma incidence as age increases
18, 19
. Hens present with several ovarian tumor types
similar to humans, including epithelial adenocarcinoma, serous differentiation, ascites, and TP53
mutations, while rarely developing non-reproductive tract tumors
20, 21
. Their primary sites for
adenocarcinomas can be ovarian, oviductal (fallopian tube), or both; and early microscopic
lesions have been identified in the ovarian stroma
18, 19
.
Mouse models have also provided evidence of OSE origins of HGSOC. In mutant TP53
mice with Dicer and Pten conditionally knocked out in the oviducts and ovaries using the MSHII
promoter, mice developed HGOCS in their ovaries, independently of their oviducts
22
. In a
transgenic HGSOC model using a MSHII driven knockout of Dicer and Pten, without a mutant
TP53 background, tumors arose from the fallopian tube stromal cells, not the epithelium
23
.
MSHII is expressed throughout the reproductive tract and is therefore not an exclusively
fallopian and/or ovarian inducible expression model. While a stromal origin could offer an
explanation for fallopian origin in non-STIC containing patients, such lesions have not been
reported, even after extensive examination of fallopian tubes in asymptomatic high-risk women
9
.
13
Subtype-specific models of EOC using developmental genes have also been investigated.
Naora et al. developed Hox mutant mice in the OSE. Induction of the Müllerian developmental
and patterning TF Hoxa9 leads to formation of EOC with lineage-specific attributes of HGSOC.
In xenografts of spontaneously transformed mouse OSECs expressing other Müllerian lineage
patterning Hox genes (Hoxa10 and Hoxa11) endometrioid- and mucinous-like EOCs were
formed, respectively
24
.
A FTE derived HGSOC has also been generated in mice
23, 25
. In an inducible PAX8
promoter-driven model, reporter expression was observed in the endometrium and fallopian tube
epithelium, but not in the OSE. These mice developed tumors resembling HGSOC, which
initiated in the FTE
25
. However, HGSOC models with fallopian tube driven tumors do not
preclude the OSE as an HGSOC initiation site.
Fallopian tube origins of HGSOC
Evidence has emerged showing primary occult microscopic lesions in the FTE. By serial
sectioning the entire length of the tubes, in a protocol termed SEE-Fim (Sectioning and
Extensively Examining of the Fimbriaeted end) STICs (serous tubal intraepithelial
carcinoma) were discovered in 38% of high risk women (n= 5/13)
26
. Further studies have added
to this finding, indicating 20-60% of high risk women carry STICs
9-11, 27
. However, even after
extensive examination of the entire length of both fallopian tubes, not all HGSOCs have been
found to contain STICs
9
.
In a study examining 9 low risk women with HGSOC, 7 (78%) were
determined to originate in the FTE and/or peritoneum. The ovaries were not examined in this
study
28
. General and low risk population-based studies examining the presence of STICs have
14
shown 0-0.8% of women harbor STICs
29, 30
. It is not known what proportion of these lesions
would develop into HGSOC. Nonetheless it is clear that a significant proportion of HGSOC arise
from FTSECs and as a result of this research, for high-risk women carrying BRCA1 and/or
BRCA2 mutations, prophylactic removal of both fallopian tubes (bilateral salpingectomy) is now
common clinical practice and has recently been shown to significantly reduce (although not
completely eliminate) ovarian cancer risk in this high-risk setting.
PAX8 expression in OSECs is controversial
Several studies have shown that paired box 8 (PAX8) is highly expressed in most
HGSOCs (>80%)
31-36
, suggesting that overexpression of this transcription factor plays a critical
role in this tumor type. This is further supported by previous studies showing that PAX8
knockdown in ovarian carcinoma cell lines induces growth arrest, apoptosis, and decreases cell
proliferation and tumorigenesis
36-39
. PAX8 is expressed throughout the Müllerian tract in the
epithelia of the cervix, endometrium, and fallopian tube. The observation of PAX8 expression in
FTSECs and in many CICs but not OSECs could suggest that the epithelial lining of CICs is
derived from the fallopian tube and that most, if not all HGSOCs likely originate from FTSECs.
However, while it is widely considered that OSECs do not express PAX8, there are conflicting
reports evaluating PAX8 expression in OSECs
2, 31-34, 40, 41
. While evidence of PAX8 expression
by OSECs has been reported in smaller sample sizes (n > 8 in only two studies)
2, 7, 32, 34, 41
, this
finding has been controversial
42
because it did not agree with the findings of others (Table 1)
31,
33, 40
.
15
Table 1. Published PAX8 immunohistochemistry of the normal ovarian surface epithelium
Reference N positive ovaries / total % Positive
Tong et al. 2011 0 / 5 0 %
Bowen et al. 2007 0 / 8 0 %
Tacha et al. 2011 0 / 9 0 %
Li et al. 2011 2 / 48 4 %
Banet and Kurman 2015 2 / 20 10 %
Auersperg 2011 2 / 6 33 %
Ozcan et al. 2011a 5 / 8 63 %
Ozcan et al. 2011b* 5 / 8 63 %
* The staining was originally reported but not shown in this article. The staining was shown in
the authors’ follow up letter to the editor
43
.
It has been observed in our lab that some primary OSEC cultures stain positive for PAX8
expression (Kate Lawrenson, private communication), which is contradictory to the widespread
belief that PAX8 is a Müllerian cell marker and is not expressed by OSECs. Given the
importance of the cell of origin in cancer, and particularly among EOC subtypes, I set out to
determine the frequency of PAX8 expression in normal ovaries. I hypothesized that PAX8 was
expressed by normal ovaries, and that at normal levels PAX8 did not induce tumorigenicity.
MATERIALS AND METHODS
Tissue Culture
Normal OSECs were isolated from histologically normal ovaries from women
undergoing surgery for conditions not involving the ovaries (such as endometrial cancer or
fibroids). Samples were collected with the approval of the University College London Hospital
Ethics Committee and informed patient consent. OSECs were harvested by brushing the surface
16
of the ovary with a cytobrush, and cultured in normal ovarian surface epithelial cell culture
media (NOSE-CM)
28
. hTERT-immortalized OSECs overexpressing CMYC (IOE19
CMYC
) have
been previously described
27
and were cultured in NOSE-CM supplemented with 3 µg/ml
blasticidin S hydrochloride (Sigma Aldrich, St. Louis, MO). HGSOC cell lines (n = 5) were
cultured as follows: OVCA429 and PEO14 in RMPI with 10% FBS (Seradigm, Providence,
UT); OVMZ15 in DMEM with 10% FBS, 1% nonessential amino acids, and sodium pyruvate;
OVCA433 in Eagle MEM with 10% FBS; COV318 in DMEM with 10% FBS and 0.1M L-
asparagine. All cell lines used in this study were confirmed to be free of contaminating
Mycoplasma infections by mycoplasma-specific PCR.
Immunohistochemistry
Formalin-fixed paraffin-embedded tissue blocks and hematoxylin and eosin (H&E)
stained sections of normal ovaries (n = 27) and fallopian tubes (n = 7) were obtained from the
Los Angeles County Hospital archive with approval of the University of Southern California
Institutional Review Board. Samples came from women receiving gynecological surgery for
non-cancer related conditions (such as fibroids). Ovarian sections were H&E stained and
examined by a gynecological pathologist to confirm presence of the surface epithelium.
Formalin-fixed paraffin-embedded tissue sections (4 µm) were processed for
immunohistochemistry. Endogenous peroxidase was blocked with 0.3% hydrogen peroxidase for
5 minutes. Antigen retrieval was carried out in high citrate buffer for 3 minutes in a steam-
cooker. Sections were incubated overnight with PAX8 antibody (Biocare Medical, Concord, CA,
1:400 dilution)
29
. Serial sections of PAX8 expressing ovaries (n = 4) were incubated with
17
calretinin antibody (Dako, Carpinteria, CA), and E-cadherin antibody (Novocastra, Wetzlar,
Germany, ready to use). A subsequent reaction was performed with the biotin-free HRP enzyme
labeled polymer of the Envision plus detection system (Dako, Carpinteria, CA).
Diaminobenzidine was used as the chromogen, and hematoxylin was applied to counterstain.
Positive controls were normal fallopian tubes for PAX8 staining, mesothelioma for calretinin,
and normal fallopian tube for E-cadherin. As a negative control, normal goat serum was used
instead of the primary antibody, resulting in a lack of detectable staining. PAX8 staining was
nuclear and scored as negative, weak, moderate, or strong. Calretinin staining was cytoplasmic
and scored as positive or negative. E-cadherin was membranous and scored as positive or
negative. For all antibodies, weak, moderate and strong staining was scored as positive
expression.
Immunofluorescent staining
Cells were grown to ~80% confluence on glass coverslips, fixed in 4% paraformaldehyde
and permeabilized with 0.5% Triton-X. Non-specific binding was blocked with 10% fetal bovine
serum in DMEM. Primary antibodies (PAX8: Proteintech, Chicago, IL, 1:100 dilution;
calretinin: EMD Millipore, Billerica, MA, 1:100 dilution) were incubated for 1 hour. Goat-anti
rabbit Alexafluor 568 and rabbit-anti mouse Alexafluor 488 (Life Technologies, Carlsbad, CA)
were used as secondary antibodies. Nuclei were stained with Hoechst 33342 (Thermo Scientific,
Rockford, IL). Coverslips were mounted on glass slides and imaged on a Zeiss Axio Imager Z1
fluorescent microscope. PAX8 staining was scored as negative, weak, moderate, or strong.
18
Calretinin staining was scored as positive or negative. For all antibodies, weak, moderate and
strong staining was scored as positive expression.
Correlation Analysis of PAX8 Staining
Each comparison was stratified by PAX8 staining in OSECs at the ovarian surface (n =
27, positive or negative). Positive PAX8 staining was defined as strong, moderate, or weak
expression; and negative staining was defined as no detectable PAX8 expression. Each
association was calculated using a two-tailed Fisher’s exact test. Cellular morphology was scored
by a gynecological pathologist, PAX8 expression in CICs (n = 18) was positive or negative, and
the available clinical data was stratified by risk factor (low vs. high risk).
RNA expression analysis
Total RNA was harvested from all cells at ~80% confluence (Qiagen, Valencia, CA) and
cDNA was generated (Life Technologies, Carlsbad, CA). Primary cells were used at early
passages. cDNA was detected using real-time reverse transcription semi-quantitative PCR (RT-
qPCR) and Taqman gene expression probes (Life Technologies, Carlsbad, CA). RT-qPCR was
run on the ABI7900 (Applied Biosystems, Foster City, CA) standard program for 40 cycles.
Expression levels were determined using the ΔΔCt method. Ct values were normalized to the
average expression of β-Actin and GAPDH, and ΔΔCt values were calibrated to a single sample
(OSEC250). Negative controls for cDNA synthesis and RT-qPCR showed no detectable gene
expression for each probe (no template control lacked a cDNA template, and no RT control
19
lacked reverse transcriptase). Unpaired Wilcoxon rank sum tests were used to test for significant
differences in gene expression between cell line groups.
Generation of IOE19
CMYC.PAX8
cells
A model of CMYC overexpression has been previously described.
27
The pEGFP-hPAX8a
construct was generously provided by Peter Kopp
30
. IOE19
CMYC
cells were transfected using the
BioT reagent (Bioland Scientific, Paramount, CA) according to the manufacturer’s protocol and
positive cells were selected for and pooled using fluorescence associated cell sorting (FACSAria
II, BD Biosciences, Franklin Lakes, NJ) for GFP expressing cells and maintained with 1 mg/ml
G418 (Sigma Aldrich, St. Louis, MO) in culture, creating the stable IOE19
CMYC.PAX8
cell line.
The pEGFP-U6 construct was used in the same manner as a control (Addgene, Cambridge, MA)
to generate stable IOE19
CMYC.GFP
cells. PAX8 expression was confirmed by RT-qPCR as
described above. Stable construct expression was confirmed by PAX8 qPCR at early and late
passages.
Migration assay
Sub-confluent cells were serum and growth factor starved for 24 hours. Cells were seeded
into transwell inserts perforated with 8 micron pores. The chemoattractant, 10% fetal bovine
serum, was added to the well underneath each insert. After 16-18 hours, membranes were fixed
in methanol and the nuclei stained with Hoechst 33342. Membranes were mounted onto slides
and the number of migrated cells was counted using a fluorescent microscope. Statistical
20
significance was determined using a paired two-tailed Student’s T-test and values were
normalized to IOE19
CMYC.GFP
cells.
Invasion assay
The QCM ECMigration Cell Invasion Assay was used according to the manufacturer’s
protocol (EMD Millipore, Billerica, MA). Cells were starved in the same manner as the
migration assay, and the same chemoattractant was used. The membrane was pre-covered by
matrigel and a fluorescent dye was used to quantify the amount of cells that invaded into the
bottom chamber after 24 hours of exposure to the chemoattractant. Fluorescence was measured
using a Hidex Plate CHAMELEON plate reader (Turku, Finland) and MikroWin software
(Mikrotech, Overath, Germany, V2000) and normalized to IOE19
CMYC.GFP
cells. Statistical
significance was determined using a paired two-tailed Student’s T-test.
Anchorage independent growth assay
Cells were suspended in 0.3% agar diluted in culture media and allowed to solidify. After
4 weeks at 37°C and 5% CO
2
, cells were fixed and stained with p-iodonitrotetrazolium chloride
(Sigma Aldrich, St. Louis, MO). The total number of colonies was counted under a light
microscope. Colony forming efficiency was calculated as (number of colonies formed / total
number of cells x 100) and normalized to IOE19
CMYC.GFP
cells. Statistical significance was
determined using a paired two-tailed Student’s T-test.
21
Cell cycle analysis
Cells were grown to ~80% confluence, trypsinized, and washed in 1X PBS. Cells were
fixed as a single cell suspension in 70% ethanol added dropwise, and stained with 10 ug/ml
DAPI (4',6-diamidino-2-phenylindole) in 1X PBS. Cells were analyzed on a SORP LSRII flow
cytometer (BD Biosciences, Franklin Lakes, NJ) and the data was analyzed using FlowJo
software v7.6 (FlowJo, Ashland, Oregon). Cell cycle phase frequency was calculated using the
Dean-Jett-Fox model
44, 45
.
Proliferation in three-dimensional culture
Cells were grown in three-dimensional culture at a density of 5,000 cells per well in a 96
well plate. Plates were pre-coated with 1.5% polyhema (2-hydroxyethyl methacrylate) (Sigma-
Aldrich, St. Louis, MO) in 95% ethanol to prevent cellular adherence to the plate. After 2 days
PrestoBlue (Thermo Fisher Scientific, Waltham, MA) was added to live cells. Fluorescence was
measured after 3 and 19 hours in PrestoBlue using a Hidex Plate CHAMELEON plate reader
(Turku, Finland) and MikroWin software (Mikrotech, Overath, Germany, V2000). Significance
was determined using a paired two-tailed Student’s T-test.
Cellular senescence assay
Cells were grown for 2 days to ~80% confluence and fixed in 0.05% glutaraldehyde
(Sigma-Aldrich, St. Louis, MO). Cells were stained with X-gal (5-bromo-4-chloro-3-indolyl-β-
D-galactopyranoside) solution (1mg/ml X-gal in DMSO, 5mM K
4
Fe(CN)
6
·3H
2
O, 5mM
K
3
Fe(CN)
6
, 1M MgCl
2
in 1X PBS) (Sigma-Aldrich, St. Louis, MO) for 24 hours at 37°C in 5%
22
CO
2
. Blue cells were scored as positive for cellular senescence and counted using a light
microscope.
RESULTS
PAX8 protein expression in normal ovaries
PAX8 expression was evaluated in 27 histologically normal ovaries with OSECs present
on the surface, 18 of which also contained CICs. From the same patient cohort, seven fallopian
tubes were analyzed as positive controls for PAX8 expression. PAX8 was expressed in the
OSECs of 12 ovaries (44%), nine of which (33% of all ovaries) showed moderate to strong
staining (similar to expression levels in FTSECs). OSECs in the remaining 15 ovaries (56%)
showed no evidence of PAX8 staining. PAX8 was observed in the CICs of 15 ovaries (83%), 14
of which showed strong staining. Within the ovary, histiocytes stained strongly for PAX8, but
there was no expression in ovarian stromal cells. In two cases, ovaries were from patients
carrying deleterious BRCA1 mutations (the BRCA1 status of the remaining patients was
unknown). In one of these cases there was no detectable PAX8 expression on the ovarian
surface, and no CICs were present in the ovary. In the second case, OSECs showed weak PAX8
staining and strong staining was observed in the CIC epithelium. All seven fallopian tubes
showed strong PAX8 staining in FTSECs, and no staining was observed in ciliated or stromal
cells. Examples of PAX8 staining in OSECs, CICs and FTSECs are shown in Figure 1a, and are
summarized in Figure 1b.
23
Figure 1. PAX8 expression in normal ovarian surface epithelium. (a) PAX8 expression was
evaluated in 27 normal ovaries using immunohistochemistry. PAX8 expression ranged from
strong to negative in OSECs and was strong, moderate or negative in CIC epithelial cells.
FTSECs stained positive for PAX8. Examples of normal OSEC morphology on the ovarian
surface are indicated with a black arrowhead. Tissue sections are shown at 100x magnification,
negative control shown at 40x. (b) The range in staining intensity and proportion of tissues with
PAX8 expression in primary OSEC, CIC, and FTSEC tissue.
Abbreviations: OSEC = ovarian surface epithelial cell, CIC = cortical inclusion cyst, FTSEC =
fallopian tube secretory epithelial cell.
24
PAX8 mRNA expression in normal ovarian surface epithelial cells
To investigate PAX8 expression in OSECs by an independent method, mRNA expression was
measured in 66 early passage primary OSEC lines using RT-qPCR. PAX8 mRNA expression
was observed in 47/66 OSEC lines (71%); 18 OSEC lines (27%) showed moderate expression,
and 29 OSEC lines (44%) showed detectable but lower PAX8 expression (Figure 2a). Negligible
expression of other PAX family members (PAX2, PAX5 and PAX7) was found in these cell
lines indicating the probe was specific to the PAX8 transcript (Figure 2a). Immunofluorescent
staining in a proportion of OSEC lines (n = 7) indicated PAX8 mRNA and protein expression are
correlated (Figure 2b).
Correlations between PAX8 expression and OSEC morphology and marker expression
It has been postulated that PAX8 positive OSECs may be FTSECs ectopically located on
the surface of the ovary through endosalpingiosis. To evaluate this, the relationship between
cellular morphology and PAX8 expression in OSECs was examined. OSECs can be flattened and
elongated, resembling classical mesothelial-type cells, or columnar and/or cuboidal, resembling
differentiated epithelial monolayers. Of the 12 ovaries with PAX8 staining in OSECs, nine
(75%) had a flattened morphology typical of OSECs, one (8%) had a cuboidal morphology, one
(8%) exhibited mixed (flat and cuboidal) morphology, and the morphology of one specimen
(8%) was ambiguous. In the 15 ovaries negative for PAX8 staining, 11 (73%) had flattened
epithelium, two (13%) had cuboidal morphology, and two (13%) showed both flat and cuboidal
cells on the same ovary (Table 2). There was no association between the cellular morphology of
25
Figure 2. PAX8 protein and
mRNA expression is correlated
in OSECs. (a) Boxplot comparing
relative PAX gene expression in
primary OSEC cultures (black
triangles), normalized to B-Actin
and GAPDH and calibrated to a
single OSEC line (OSEC250).
HGSOC cell lines are shown as
red circles. (b) OSEC cultures
stained for PAX8 displayed PAX8
moderate expression (left panel
and bar), weak expression (middle
panel and bar), or negative
expression (right panel and bar).
Expression of PAX8 detected by
immunofluorescence correlated
with mRNA abundance. PAX8
moderate and weak mRNA
expression was significantly
different (P < 0.0001, one way
ANOVA). 84% of OSEC9 and
OSEC254 cells were PAX8
positive, and 48% of OSEC255
cells were weakly positive for
PAX8. All cultures were calretinin
positive by immunofluorescence
using an anti-calretinin antibody.
All images were taken at 200x
magnification. OSEC = ovarian
surface epithelial cell, HGSOC =
high-grade serous ovarian
carcinoma.
26
normal OSECs and PAX8 staining. In addition, no correlations were found between PAX8
staining and the presence of CICs (Table 3), or the following epidemiological risk factors: age,
oral contraceptive use, parity, gravidity, miscarriage, or menopausal status (Table 4).
Table 2. Cellular morphology in PAX8 positive and negative OSECs
Cellular Morphology
Cell type Flat Cuboidal Mixed Ambiguous P-value
OSEC PAX8 - 73% (11) 13% (2) 13% (2) 0
OSEC PAX8 + 75% (9) 8% (1) 8% (1) 8% (1)
1.00
Normal ovaries (n = 27) were immunohistochemically stained for PAX8 protein expression, and
the epithelial lining of the ovarian surface was scored for PAX8 expression and cellular
morphology. The percent of total samples is shown and the number of samples is shown in
parenthesis. Statistical significance was determined using a Fisher’s Exact Test comparing flat
and cuboidal cellular morphologies in the PAX8- and PAX8+ OSEC groups.
Abbreviations: OSECs = ovarian surface epithelial cells, ambiguous = unable to score
morphology.
The expression of calretinin, a mesothelial cell marker expressed in OSECs but not
FTSECs, and E-cadherin, which is expressed in FTSECs and, in some instances, OSECs
33
,
including deep clefts
34
, and along the ovarian fimbria
20
, was evaluated in PAX8-positve ovaries
(n = 4). Diffuse co-expression of PAX8, calretinin, and E-cadherin was observed in three
ovaries. In the fourth ovary there were two areas of focal PAX8 expression in OSECs that
stained negative for calretinin, but positive for E-cadherin expression; three areas of OSECs that
27
Table 3. PAX8 immunohistochemistry in OSECs and CICs of normal ovaries
Cortical Inclusion Cyst
OSE Not present PAX8 - PAX8 + P-value
OSEC PAX8 - 40% (6) 7% (1) 53% (8)
OSEC PAX8 + 31% (4) 15% (2)* 54% (7)*
1.00
Normal ovaries (n = 27) were immunohistochemically stained for PAX8 protein expression, and
the epithelial lining of the ovarian surface and CICs (n = 18) was examined for PAX8
expression. The percent of total samples is shown and the number of samples is shown in
parenthesis. * One patient sample had one CIC staining PAX8 negative and the other was
PAX8 positive. P-value shown from a Fisher’s Exact Test comparing PAX8- and PAX8+ OSEC
groups.
Abbreviations: CIC(s) cortical inclusion cyst(s), OSE = ovarian surface epithelium, OSECs =
ovarian surface epithelial cells.
Table 4. Epidemiological risk factors do not correlate with PAX8 expression status
Risk factor (higher risk group) Stratification P-value
Parity (nulliparous) 0-2 vs 3+ 0.11
Gravidity (nulligravida) never vs ever 0.23
Age (older) > 51 vs 51+ 0.26
Miscarriage (insufficient evidence*) never vs ever 0.66
Menopausal status (post) pre vs post 0.70
Oral contraceptive use (no) no vs yes 1.0
Normal ovaries (n = 27) were immunohistochemically stained for PAX8 protein expression, and
the epithelial lining of the ovarian surface was examined for PAX8 expression. The available
clinical data (stratified by risk factor, low vs high risk) was evaluated for correlation with PAX8
expression in the OSE (negative vs positive) using a two-sided Fisher’s Exact Test. The
available clinical data was evaluated using a two-sided Fisher’s Exact Test comparing PAX8
negative and positive OSE groups stratified by each risk factor (low vs high risk). For factors
with continuous data, the stratification with the lowest P-value is shown.
Abbreviations: CIC(s) = cortical inclusion cyst(s), OSE = ovarian surface epithelium, OSECs =
ovarian surface epithelial cells.
* One study has reported a positive correlation between a higher number miscarriages and
epithelial ovarian cancer, but in a small population and no correlation was found between never
and ever having miscarried
46
. Another study has reported no significant correlations
47
.
28
calretinin; and three areas of OSECs that co-expressed calretinin and E-cadherin. Areas of PAX8
and calretinin co-expression in one ovary and PAX8 and E-cadherin co-expression were
observed in OSECs from three ovaries. Examples of PAX8, calretinin, and E-cadherin staining
are shown in Figure 3. Early passage primary OSEC lines that were positive for PAX8 protein
and mRNA expression also stained positive for calretinin (Figure 2b).
The role of PAX8 in early stage transformation OSECs
The role of PAX8 in the early stage transformation of OSECs was evaluated. Given its
frequent overexpression in ovarian carcinomas I postulated that this overexpression promotes
carcinogenesis. PAX8 examined in an established stepwise model of neoplastic transformation of
OSECs created by overexpressing hTERT (to immortalize OSECs creating the IOE19 cell line)
followed by overexpression of the CMYC oncogene (IOE19
CMYC
cells)
27
. Briefly, hTERT
overexpression extends the lifespan of primary OSECs without inducing DNA copy number
changes or evidence of transformation, while CMYC overexpression partially transforms these
cells, as shown by their acquired ability to form colonies in anchorage independent growth
assays. To study the transformative ability of PAX8, a PAX8 cDNA construct was transfected
into IOE19
CMYC
cells to create IOE19
CMYC.PAX8
cells. PAX8 mRNA expression was increased
significantly in IOE19
CMYC.PAX8
cells compared to IOE19
CMYC
cells (p = 0.004), which mimicked
endogenous expression but was lower than the IOE19
CMYC.PAX8
levels typically observed in
HGSOCs (Figure 4a). Cellular transformation of IOE19
CMYC.PAX8
cells was evaluated using
several different phenotypic assays. Compared to IOE19
CMYC
cells, cells were significantly less
29
Figure 3. PAX8, calretinin, and E-cadherin expression in normal ovarian surface
epithelium. Antibody staining controls are shown in the top panel. (a) mesothelioma with
positive calretinin staining, (b) fallopian tube epithelium negative for calretinin and (c) positive
for E-cadherin expression (negative E-cadherin staining seen in stroma), (d) breast tumor
specimen positive for E-cadherin expression. (e, i, m, q) A section of normal ovary stained for
various proteins in serial sections (to their right). The black rectangle indicates the
magnified region of serial sections in corresponding serial sections to their right. Ovary shown
in (e) contains OSECs that are (f) calretinin positive, (g) PAX8 negative, and (h) E-cadherin
negative. Ovary shown in (i) contains OSECs that are (j) calretinin positive, (k) PAX8 negative,
and (l) E-cadherin positive. Ovary shown in (m) contains OSECs that are (n) calretinin positive,
(o) PAX8 positive, and (p) E-cadherin positive. Ovary in (q) contains OSECs that are (r)
calretinin negative, (s) PAX8 positive, and (t) E-cadherin positive. All tissue stained by DAB
immunohistochemistry.
Abbreviations: FT = fallopian tube, OV = ovary, OSECs = ovarian surface epithelial cells.
30
migratory (p = 0.028) (Figure 4b), but PAX8 did not induce changes in invasion, anchorage
independent growth, cell cycle distribution, proliferation in three-dimensional culture, or cellular
senescence (Figure 4c-g).
DISCUSSION
High-grade serous ovarian carcinoma is characterized by frequent late-stage presentation,
having no specific symptoms associated with early or often even late stage disease.
Understanding the molecular mechanisms that underlie cancer initiation and identifying the
precursor tissues of the specific subtypes of ovarian cancer is a promising path to develop novel
early-stage screening biomarkers for ovarian carcinoma as well as new therapies, which would
have a substantial impact on reducing disease morbidity and mortality.
The origin of the most common ovarian carcinoma subtype, HGSOC, remains a topic of
debate. For many years it was widely considered that HGSOC arises from OSECs, either on the
surface of the ovary or trapped within CICs where OSECs come into close contact with the
mitogenic microenvironment of the ovarian stroma
4
. Recent histopathological data
9-11
and
modeling studies
23, 25
have demonstrated that HGSOCs can arise from FTSEC precursors.
Nonetheless, the relative proportion of HGSOCs that initiate from FTSECs is not yet clear, and
31
Figure 4. PAX8 does not induce neoplastic transformation in OSECs. IOE19
CMYC
cells
stably expressing GFP (+GFP) or a PAX8-GFP (+PAX8) fusion protein were assayed for
cellular transformation. (a) IOE19
CMYC.PAX8
cells express higher levels of PAX8 than IOE19
CMYC
parental cells, and IOE19
CMYC.GFP
control cells, but less than HGSOC cells (OVCA433)(p <
0.0001, unpaired T-test). (b) IOE19
CMYC.PAX8
cells were less migratory than GFP control cells in
an in vitro transwell migration assay. (* p = 0.028) (c-f) IOE19
CMYC.PAX8
cells were not
significantly different in from IOE19
CMYC.GFP
in (d) anchorage independent growth in soft agar (e)
progression through the cell cycle (f) proliferation measured after 3 or 19 hours (g) senescence.
Error bars show standard deviation. Error bars show standard deviation and significance was
determined using a paired, two-tailed Student’s T-test.
Abbreviations: OSECs = ovarian surface epithelial cells, IOE19 = immortalized ovarian surface
epithelial cells.
a
b
c d
e
f
g
32
this hypothesis does not sufficiently explain or address the evidence supporting HGSOC
initiation from OSECs
3, 5-7
.
While PAX8 is overexpressed in >80% of HGSOCs, its mechanistic role in
tumorigenesis is not well known. PAX8 is expressed by FTSECs and until recently was widely
considered absent in OSECs. In this chapter, I show that PAX8 is expressed by 44 to 71% of
OSECs from the ovaries of pre-and post-menopausal women, in contrast to published reports that
state OSECs do not express PAX8
31, 33, 40
. This study represents the largest analysis of PAX8
expression in normal OSECs published to date (27 normal ovarian tissues and 66 primary OSEC
lines). A PAX8-specific monoclonal antibody
48
was used, limiting the likelihood of non-specific
staining. In contrast to previous reports where only PAX8 protein expression was evaluated, I
also measured PAX8 mRNA expression to address potential observer bias in visual scoring of
IHC staining intensity. The proportion of OSECs showing PAX8 mRNA expression (71%) was
higher than PAX8 protein expression in OSECs (44%). This may be sampling bias resulting
from the smaller sample size in the PAX8 IHC, or may be a result of sampling error due to
representative tissue sectioning biases. One 5 um section of each roughly almond-sized ovary
was examined, and because these sections often contain only small populations of OSECs, and
PAX8 expression varied along the OSE, it is possible that ovaries classified as PAX8 negative
may express PAX8 in regions that were not examined. This is also a possible explanation of why
different studies find very different expression of PAX8 in the OSE.
It has been argued that PAX8 expressing cells on the surface of the ovary may be
FTSECs that have ectopically relocated through endosalpingiosis (fallopian tissue implanted
outside the fallopian tubes) both because OSECs are not expected to express PAX8, and because
33
PAX8 positive cells on the ovarian surface can morphologically resemble FTSECs
40
. Some small
regions had focal co-expression of PAX8 and E-Cadherin in the absence of calretinin expression,
and it is possible that these cells may be fallopian in origin. However, my data do not support the
hypothesis that all PAX8 positive cells on the ovary are FTSECs. First, endosalpingiosis has
been reported in only around 10% of women
49
whereas I found PAX8 expression in 44 and 71%
of normal OSECs. Secondly, the morphology of OSECs expressing PAX8 did not resemble that
of FTSECS, which are columnar. Instead, most OSECs expressing PAX8 had a flattened,
cuboidal (mesothelial) morphology typical of OSECs, and no differences were found in
morphology between OSECs of ovaries that expressed PAX8 and those that did not. However,
most importantly, I found that OSECs expressing PAX8 also co-express calretinin. Calretinin is
a mesothelial cell marker expressed in OSECs but not FTSECs. PAX8 expression in OSECs was
most frequently found in combination with calretinin and E-cadherin (which is expressed in both
OSECs and FTSECs). Calretinin was also co-expressed with PAX8 in primary OSEC cultures.
This indicates that PAX8 expressing cells on the ovarian surface are likely to be OSECs and not
FTSECs.
A minimum of two oncogenes is usually required to transform normal cells
50
. CMYC is a
major driver of tumorigenesis in many cancers, and activates many downstream genes
51
. 59% of
HGSOCs harbor CMYC overexpression
52
, and OSECs can be partially transformed by CMYC
overexpression
53
. While overexpression of PAX8 occurs frequently in HGSOC, forced
expression of PAX8 in partially transformed OSECs at levels similar to primary OSECs did not
induce features of neoplastic transformation. The only significant phenotypic alteration observed
was decreased migration in PAX8 expressing cells, in agreement with the literature
54
. This
34
indicates that expression of PAX8 in OSECs is insufficient to induce transformation, and may
represent a non-pathological state even in cells partially transformed by CMYC overexpression.
These data may indicate that PAX8 overexpression occurs later in HGSOC development, with
multiple somatic alterations required for PAX8 to promote neoplasia. PAX8 expression in
OSECs may therefore be associated with normal physiological processes, such as ovulation, or
may simply be a part of the phenotypic heterogeneity that characterizes this cell type
55, 56
.
Because an increased PAX8 level failed to increase features of transformation, an
oncogenic role is not supported by this study. A dose-dependent effect is not indicated by these
results in IOE19
CMYC.PAX8
cells (or in IOE4
CMYC.PAX8
cells expressing PAX8 at levels in the range
of HGSOC cells, Kate Lawrenson, personal communication). To promote transformation it
instead may require a particular threshold amount of PAX8 to be met, or require a particular
genetic background and/or cellular context to produce oncogenic effects. It is also possible that
PAX8 can act as both an oncogene or tumor suppressor depending on cellular context. Evidence
for this phenomenon has been found in the AMPK (Adenosine monophosphate-activated protein
kinase) gene. When cellular energy levels are low, AMPK acts like an oncogene and promotes
cell survival. When energy levels are high, AMPK acts like a tumor suppressor, and suppresses
cell proliferation via activation of catabolism and inhibition of anabolism
57
. In this study PAX8
expression was analyzed in the context of relatively normal cells in relatively consistent culture
conditions. In such conditions PAX8 may serve to maintain homeostasis, as it is involved in
differentiation of normal cells
58, 59
. As cells continue to transform and become malignant, the
alterations in pathways and checkpoints may provide a permissive environment for the
expression of PAX8 to promote tumorigenesis.
35
One cellular context that may affect PAX8 is transcription factor cooperativity, i.e. the
presence and combination of various cofactors and transcription factors that bind PAX8 and
direct its binding to the genome. Transcription factor cooperativity is a key determinant of
genomic regulation
60
and can be altered in cancer
61
. In a cancer cell, altered cofactor
biochemistry can result in cofactor and transcription factor combinations not normally found in
that cell type. Through altered genome binding and/or epigenetic modification these complexes
can impact gene expression, effectively “high jacking” a TF and resulting in gene expression and
behaviors not normally seen in a particular cell type. This is supported by androgen receptor
reprogramming seen when comparing cancer to normal binding in prostate tissues
62
.
Genetic mouse models have used specific TFs to drive tissue-specific HGSOC
development
23, 25
. In an inducible PAX8 promoter-driven HGSOC model, reporter expression
was observed in the endometrium and fallopian tube epithelium, but not in the OSE
25
.
Expression was induced by two-week doxycycline treatment. Because it is unclear at what
age(s), in the presence of what environmental factor(s), or for how long PAX8 expression occurs
in human OSE, if PAX8 is expressed in murine ovaries, the age and duration of doxycycline
exposure may have been insufficient to capture PAX8 activity. It is also possible that PAX8 is
expressed in human OSE but not in mice as ortholog expression varies among species, especially
among tissue-specific genes
63-65
. Anatomical differences such as the murine bursa may also
impact the OSE microenvironment and function leading to differential expression between
humans and mice. Furthermore, major physiological differences in reproduction may also have
an effect. While humans consistently ovulate during reproductive years, mice ovulate in the
presence of males. This leads to a marked difference in the number of ovulations per lifetime,
36
which has been correlated with EOC risk
66
.
Further models and evidence supports that ovarian carcinoma can be generated from both
FTSECs and OSECs. Nikitin et al. used fluorescent label tracing to identify a population of cells
in the ovarian hilum, which is at the proximal end of the ovary closest to and connected to the
uterus, its OSE also continuous with the lumen of the fallopian tube, that retained its labeling
more intensely and for longer than surrounding cells. The hilar epithelial cells were shown to
populate the surrounding OSE, suggesting a pool of tissue stem cells in this region that
repopulates the OSE after the cyclical rupture and repair of ovulation
8
. This transitional zone of
epithelia, at the convergence of the ovarian and fallopian tube epithelium, and the mesothelium
lining the outside of the uterus, is an attractive candidate for carcinoma initiation because other
sites of transitional epithelium in the body have been pinpointed as origin sites of carcinomas.
This includes cervical cancer originating at the cervix, the squamo-columnar junction of the
uterine and vaginal lining, and esophageal adenocarcinoma, originating at the squamo-columnar
junction of the esophagus and stomach. There is no known squamous epithelium in the ovary,
fallopian tube, or uterus; they are covered by mesothelial, pseudostratified columnar, and
columnar epithelial cells, respectively. However there is also evidence for a natural transition in
luminal cell types between the ovary and fallopian tube – along their continuous epithelial
connection of the ovarian fimbriae – a long fimbriae at the distal, ovarian end of the fallopian
tube that is, unlike the rest of the numerous fallopian fimbria, connected to another organ – at the
surface of the ovary. Auesperg has shown that well known OSE and FTE markers gradually
change in expression along the transition from FTE to OSE epithelia and has proposed that like
the initiation sites at the squamo-columnar junctions of the cervix and esophagus, this
37
transitional zone is a “unifying hypothesis” for the site of HGSOC initiation
2
. This single
fimbriae, attached to an ovary the size of an almond, is often broken during tissue removal, and if
left intact, still difficult to capture in tissue sectioning. The ovaries and fallopian tubes used in
this study were separated before tissue embedding and sectioning, which is standard practice,
and therefore it was not possible to evaluate the ovarian fimbria or the surrounding OSE (there
are no landmarks to identify it) in this study.
Understanding the origins of HGSOC presents a major challenge to its clinical
management, and the gaps in knowledge surrounding disease origins limit the development of
early detection, prevention, and treatment strategies. HGSOC is usually diagnosed at a late stage,
and while therapeutic advances have marginally improved five-year survival rates over the last
decade, the disease continues to be associated with a poor prognosis. Following the evidence that
HGSOC can arise from FTSECs, many have proposed that salpingectomy should be considered
or even routinely adopted when performing hysterectomies in post-menopausal women
10
.
However, this research, together with other reports
7, 67
, continues to suggest the importance of
OSECs in the origins of HGSOC.
EXPERIMENTAL AND ANALYTICAL CONTRIBUTIONS
Tissue Culture: Emily Adler
Immunohistochemistry:
Tissue selection and PAX8 scoring: Paulette Mhawech-Facueglia
Immunohistochemical staining: USC Pathology Core
E-cadherin and calretinin scoring: Emily Adler
38
Combination of PAX8, E-cadherin, and calretinin scoring: Emily Adler
Immunofluorescent Staining: Emily Adler
Correlation analysis of PAX8 staining: Emily Adler
RNA expression analysis: Emily Adler
Generation of IOE
CMYC.PAX8
cells:
IOE
CMYC.PAX8
cells: Emily Adler
IOE
CMYC.GFP
cells: Emily Adler
Phenotypic experiments and analyses:
Migration assay: Emily Adler
Invasion assay: Emily Adler
Anchorage independent growth assay: Emily Adler
Cell cycle analysis: Emily Adler
Proliferation in three-dimensional culture: Emily Adler
Cellular senescence assay: Emily Adler
REFERENCES
1. America Cancer Society. Cancer Facts and Figures; 2017.
2. Auersperg N. The origin of ovarian carcinomas: a unifying hypothesis. Int J Gynecol
Pathol. Jan 2011;30(1):12-21.
3. Bell DA, Scully RE. Early de novo ovarian carcinoma. A study of fourteen cases.
Cancer. 1994;73(7):1859-1864.
39
4. Scully RE. Pathology of ovarian cancer precursors. Journal of Cellular Biochemistry.
1995;59(S23):208-218.
5. Pothuri B, Leitao MM, Levine DA, et al. Genetic analysis of the early natural history of
epithelial ovarian carcinoma. PLoS One. 2010;5(4):e10358.
6. Yates MS, Meyer LA, Deavers MT, et al. Microscopic and early stage ovarian cancers in
BRCA1/2 mutation carriers: building a model for early BRCA-associated tumorigenesis.
Cancer Prev Res (Phila). Mar 2011;4(3):463-470.
7. Auersperg N. Ovarian surface epithelium as a source of ovarian cancers: unwarranted
speculation or evidence-based hypothesis? Gynecol Oncol. Jul 2013;130(1):246-251.
8. Flesken-Nikitin A, Hwang C-I, Cheng C-Y, et al. Ovarian surface epithelium at the
junction area contains cancer-prone stem cell niche. Nature. 2013;495(7440):241-245.
9. Przybycin CG, Kurman RJ, Ronnett BM, et al. Are all pelvic (nonuterine) serous
carcinomas of tubal origin? Am J Surg Pathol. Oct 2010;34(10):1407-1416.
10. Gilks C, Miller D. Opportunistic salpingectomy for women at low risk for development
of ovarian carcinoma: The time has come. Gynecologic Oncology. 2013;129(3):443–444.
11. Kindelberger DW, Lee Y, Miron A, et al. Intraepithelial carcinoma of the fimbria and
pelvic serous carcinoma: Evidence for a causal relationship. Am J Surg Pathol.
2007;31(2):161-169.
12. Tang S, Onuma K, Deb P, et al. Frequency of serous tubal intraepithelial carcinoma in
various gynecologic malignancies: a study of 300 consecutive cases. Int J Gynecol
Pathol. Mar 2012;31(2):103-110.
40
13. Salazar H, Godwin AK, Daly MB, et al. Microscopic benign and invasive malignant
neoplasms and a cancer-prone phenotype in prophylactic oophorectomies. J Natl Cancer
Inst. Dec 18 1996;88(24):1810-1820.
14. Gusberg SB, Deligdisch L. Ovarian dysplasia. A study of identical twins. Cancer. Jul 1
1984;54(1):1-4.
15. Korner M, Burckhardt E, Mazzucchelli L. Different proportions of aneusomic cells in
ovarian inclusion cysts associated with serous borderline tumours and serous high-grade
carcinomas support different pathogenetic pathways. J Pathol. Sep 2005;207(1):20-26.
16. Mittal KR, Zeleniuch-Jacquotte A, Cooper JL, et al. Contralateral ovary in unilateral
ovarian carcinoma: a search for preneoplastic lesions. Int J Gynecol Pathol. Jan
1993;12(1):59-63.
17. Graham RM, Schueller EF, Graham JB. Detection of ovarian cancer at an early stage.
Obstet Gynecol. Aug 1965;26:151-156.
18. Fredrickson TN. Ovarian tumors of the hen. Environ Health Perspect. Aug 1987;73:35-
51.
19. Harris EA, Fletcher OJ, Anderson KE, et al. Epithelial cell tumors of the hen
reproductive tract. Avian Dis. Mar 2014;58(1):95-101.
20. Barua A, Bitterman P, Abramowicz JS, et al. Histopathology of ovarian tumors in laying
hens, a preclinical model of human ovarian cancer. Int J Gynecol Cancer. May
2009;19(4):531-539.
41
21. Hakim AA, Barry CP, Barnes HJ, et al. Ovarian adenocarcinomas in the laying hen and
women share similar alterations in p53, ras, and HER-2/neu. Cancer Prev Res (Phila).
Feb 2009;2(2):114-121.
22. Kim J, Coffey DM, Ma L, et al. The ovary is an alternative site of origin for high-grade
serous ovarian cancer in mice. Endocrinology. Jun 2015;156(6):1975-1981.
23. Kim J, Coffey DM, Creighton CJ, et al. High-grade serous ovarian cancer arises from
fallopian tube in a mouse model. Proc Natl Acad Sci U S A. Mar 6 2012;109(10):3921-
3926.
24. Cheng W, Liu J, Yoshida H, et al. Lineage infidelity of epithelial ovarian cancers is
controlled by HOX genes that specify regional identity in the reproductive tract. Nature
Medicine. 2005;11(5):531-537.
25. Perets R, Wyant GA, Muto KW, et al. Transformation of the fallopian tube secretory
epithelium leads to high-grade serous ovarian cancer in Brca;Tp53;Pten models. Cancer
Cell. Dec 9 2013;24(6):751-765.
26. Medeiros F, Muto MG, Lee Y, et al. The tubal fimbria is a preferred site for early
adenocarcinoma in women with familial ovarian cancer syndrome. Am J Surg Pathol.
Feb 2006;30(2):230-236.
27. Tang S, Onuma K, Deb P, et al. Frequency of Serous Tubal Intraepithelial Carcinoma in
Various Gynecologic Malignancies: A Study of 300 Consecutive Cases. International
Journal of Gynecological Pathology.2012;31(2):103-110.
42
28. Gilbert L, Basso O, Sampalis J, et al. Assessment of symptomatic women for early
diagnosis of ovarian cancer: results from the prospective DOvE pilot project. Lancet
Oncol. 2012;13(3):285-291.
29. Rabban JT, Garg K, Crawford B, et al. Early detection of high-grade tubal serous
carcinoma in women at low risk for hereditary breast and ovarian cancer syndrome by
systematic examination of fallopian tubes incidentally removed during benign surgery.
Am J Surg Pathol. 2014;38(6):729-742.
30. Nishida N, Murakami F, Higaki K. Detection of serous precursor lesions in resected
fallopian tubes from patients with benign diseases and a relatively low risk for ovarian
cancer. Pathol Int. Jun 2016;66(6):337-342.
31. Bowen NJ, Logani S, Dickerson EB, et al. Emerging roles for PAX8 in ovarian cancer
and endosalpingeal development. Gynecol Oncol. Feb 2007;104(2):331-337.
32. Ozcan A, Liles N, Coffey D, et al. PAX2 and PAX8 expression in primary and metastatic
mullerian epithelial tumors: a comprehensive comparison. Am J Surg Pathol. Dec
2011;35(12):1837-1847.
33. Tacha D, Zhou D, Cheng L. Expression of PAX8 in normal and neoplastic tissues: a
comprehensive immunohistochemical study. Appl Immunohistochem Mol Morphol.
2011;19(4):293-299.
34. Ozcan A, Shen SS, Hamilton C, et al. PAX 8 expression in non-neoplastic tissues,
primary tumors, and metastatic tumors: a comprehensive immunohistochemical study.
Mod Pathol. Jun 2011;24(6):751-764.
43
35. The Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian
carcinoma. Nature. 2011;474:609-615.
36. Cheung HW, Cowley GS, Weir BA, et al. Systematic investigation of genetic
vulnerabilities across cancer cell lines reveals lineage-specific dependencies in ovarian
cancer. Proc Natl Acad Sci U S A. 2011;108(30):12372-12377.
37. Li CG, Nyman JE, Braithwaite AW, et al. PAX8 promotes tumor cell growth by
transcriptionally regulating E2F1 and stabilizing RB protein. Oncogene.
2011;30(48):4824-4834.
38. Di Palma T, Lucci V, de Cristofaro T, et al. A role for PAX8 in the tumorigenic
phenotype of ovarian cancer cells. BMC Cancer. 2014;14(1):292.
39. Rodgers LH, hAinmhire EÓ, Young AN, et al. Loss of PAX8 in high-grade serous
ovarian cancer reduces cell survival despite unique modes of action in the fallopian tube
and ovarian surface epithelium. Oncotarget. 2016;7(22):32785-32795.
40. Li J, Abushahin N, Pang S, et al. Tubal origin of ovarian low-grade serous carcinoma.
Modern Pathology. 2011;24(11):1488-1499.
41. Banet N, Kurman RJ. Two types of ovarian cortical inclusion cysts: proposed origin and
possible role in ovarian serous carcinogenesis. Int J Gynecol Pathol. Jan 2015;34(1):3-8.
42. Tong GX, Hamele-Bena D. The differential expression of PAX2 and PAX8 in the
ovarian surface epithelium and fallopian tubal epithelium is an important issue. Am J
Surg Pathol. Jul 2012;36(7):1099-1100; author reply 1100-1092.
44
43. Ozcan A, Truong LD. PAX2 and PAX8 Expression in the Ovarian Surface Epithelium
and Inclusion Cysts.[Letter]. American Journal of Surgical Pathology. July 2012
2012;36(7):1100-1102.
44. Fox MH. A model for the computer analysis of synchronous DNA distributions obtained
by flow cytometry. Cytometry. Jul 1980;1(1):71-77.
45. Dean PN, Jett JH. Mathematical analysis of DNA distributions derived from flow
microfluorometry. J Cell Biol. Feb 1974;60(2):523-527.
46. Braem MG, Onland-Moret NC, Schouten LJ, et al. Multiple miscarriages are associated
with the risk of ovarian cancer: results from the European Prospective Investigation into
Cancer and Nutrition. PLoS One. 2012;7(5):e37141.
47. Dick ML, Siskind V, Purdie DM, et al. Incomplete pregnancy and risk of ovarian cancer:
results from two Australian case-control studies and systematic review. Cancer Causes
Control. Nov 2009;20(9):1571-1585.
48. Tacha D, Qi W, Zhou D, et al. PAX8 mouse monoclonal antibody [BC12] recognizes a
restricted epitope and is highly sensitive in renal cell and ovarian cancers but does not
cross-react with b cells and tumors of pancreatic origin. Appl Immunohistochem Mol
Morphol. Jan 2013;21(1):59-63.
49. Hesseling MH, De Wilde RL. Endosalpingiosis in laparoscopy. J Am Assoc Gynecol
Laparosc. May 2000;7(2):215-219.
50. Greenhalgh DA, Welty DJ, Player A, et al. Two oncogenes, v-fos and v-ras, cooperate to
convert normal keratinocytes to squamous cell carcinoma. Proc Natl Acad Sci U S A. Jan
1990;87(2):643-647.
45
51. Zeller KI, Jegga AG, Aronow BJ, et al. An integrated database of genes responsive to the
Myc oncogenic transcription factor: identification of direct genomic targets. Genome
Biol. 2003;4(10):R69.
52. Tashiro H, Miyazaki K, Okamura H, et al. c-myc over-expression in human primary
ovarian tumours: Its relevance to tumour progression. International Journal of Cancer.
1992;50(5):828-833.
53. Lawrenson K, Grun B, Benjamin E, et al. Senescent fibroblasts promote neoplastic
transformation of partially transformed ovarian epithelial cells in a three-dimensional
model of early stage ovarian cancer. Neoplasia. Apr 2010;12(4):317-325.
54. Ruiz-Llorente S, Carrillo Santa de Pau E, Sastre-Perona A, et al. Genome-wide analysis
of Pax8 binding provides new insights into thyroid functions. BMC Genomics.
2012;13:147.
55. Dyck HG, Hamilton TC, Godwin AK, et al. Autonomy of the epithelial phenotype in
human ovarian surface epithelium: changes with neoplastic progression and with a family
history of ovarian cancer. Int J Cancer. 1996;69(6):429-436.
56. Bhartiya D, Singh J. FSH–FSHR3–stem cells in ovary surface epithelium: basis for adult
ovarian biology, failure, aging, and cancer. Reproduction. 2015;149(1):R35-48.
57. Liang J, Mills GB. AMPK: a contextual oncogene or tumor suppressor? Cancer Res.
2013;73(10):2929-2935.
58. Magliano MP, Lauro RD, Zannini M. Pax8 has a key role in thyroid cell differentiation.
Proc Natl Acad Sci U S A. 2000;97(24)13144-13149.
46
59. Narlis M, Grote D, Gaitan Y, et al. Pax2 and Pax8 Regulate Branching Morphogenesis
and Nephron Differentiation in the Developing Kidney. J Am Soc Nephrol.
2007;18(4)1121-1129.
60. Stefflova K, Thybert D, Wilson MD, et al. Cooperativity and rapid evolution of cobound
transcription factors in closely related mammals. Cell. 2013;154(3):530-540.
61. Pomerantz MM, Li F, Takeda D, et al. The androgen receptor cistrome is extensively
reprogrammed in human prostate tumorigenesis. Nat Genet. Nov 2015;47(11):1346-
1351.
62. Mark M, Fugen P, David L, et al. The androgen receptor cistrome is extensively
reprogrammed in human prostate tumorigenesis. Nature Genetics. 2015;47(11):1346-
1351.
63. Yanai I, Graur D, Ophir R. Incongruent Expression Profiles between Human and Mouse
Orthologous Genes Suggest Widespread Neutral Evolution of Transcription Control.
Omics. 2004;8(1):15-24.
64. Zheng-Bradley X, Rung J, Parkinson H, et al. Large scale comparison of global gene
expression patterns in human and mouse. Genome Biol. 2010;11(12):R124.
65. Huminiecki L, Wolfe KH. Divergence of Spatial Gene Expression Profiles Following
Species-Specific Gene Duplications in Human and Mouse. Genome Research.
2004;14(10A):1870-1879.
66. Purdie DM, Bain CJ, Siskind V, et al. Ovulation and risk of epithelial ovarian cancer.
International Journal of Cancer. 2003;104(2):228-232.
47
67. Auersperg N. The origin of ovarian cancers--hypotheses and controversies. Front Biosci
(Schol Ed). 2013;5:709-719.
48
CHAPTER 3: THE GENOMIC LANDSCAPE OF PAX8
IN EPITHELIAL OVARIAN CARCINOMA
INTRODUCTION
PAX8 in epithelial ovarian carcinoma
PAX8 encodes a member of the paired box family of transcription factors (TFs) that is
essential for normal embryonic development of the Müllerian ducts
1
. Despite decades of research
in the kidney and thyroid, only recently has it been recognized as a key gene in epithelial
ovarian cancer (EOC) etiology. Most studies have focused on the most common EOC
histotype, high-grade serous ovarian cancer (HGSOC)
2-4
.
PAX8 is active in a tissue specific manner during development. Its expression is
maintained after development in the thyroid, kidneys, and Müllerian tract. PAX8 is responsible
for cell-type specific maintenance of a differentiated phenotype
5, 6
. It is often used as a marker
for these cells in pathology. Initial studies of PAX8 were due to its key role in congenital
hypothyroidism from loss of function mutations. PAX8 drives the expression of critical thyroid
hormones through promoter binding to thyroid-specific genes in follicular cells
7
. These are
essential to thyroid development and function. It was found using a PAX8 knock out murine
model that administration of the lacking thyroid hormones restored normal function.
Interestingly, PAX8 is overexpressed in thyroid carcinomas
8
.
PAX8 is also critical to kidney development during organogenesis, lineage specification,
mesenchymal to epithelial transition, and cellular proliferation and differentiation
5, 9-11
.
Congenital renal disorders are associated with a lack of PAX8 expression, renal epithelial cells
49
are maintained with PAX8 expression, and renal cell carcinoma is associated with
overexpression of PAX8
8-10, 12-14
.
PAX8 is essential to the development of the Müllerian tract. Pax8 -/- mice are infertile,
having abnormal oviducts (fallopian tubes), a non-functional uterus, and no vaginal opening. In
humans, after organogenesis and tissue development, PAX8 maintains a differentiated phenotype
in fallopian tube secretory epithelial cells
15
, endometrial and endocervical epithelial cells
16
, and
as shown in Chapter 2, ovarian surface epithelial cells are PAX8-positive in up to 71% of
women
17
. The major types of EOC overexpress PAX8. Up to 100% of HGSOCs, 100% of
CCOCs, 12% of EnOCs, and 14% of MOCs show high expression levels. Despite the common
involvement of this TF in EOCs, very little is known about how PAX8 contributes to their
etiology
8, 14, 18
.
Project Achilles identified PAX8 as a top gene that when silenced kills ovarian cancer
cells. Through a genome-wide RNA interference screen of 102 cell lines representing 18
different cancer types identified, siRNA to knock down PAX8 was associated with apoptotic
cell death in 25/25 (100%) of EOC cell lines
19
. This study provided evidence that PAX8 is
lineage-specifically overexpressed in EOC. This was based in part on the frequent amplification
of the PAX8 locus in EOC, which has also been found by The Cancer Genome Atlas
20
. Other
reports show that knock down or knock out of PAX8 in various EOC cell lines is associated with
cellular senescence, apoptosis, and decreased proliferation, as well as slower initiation into the
cell cycle
21,22
.
While PAX8 is overexpressed in >80% of EOCs, its biological role is poorly understood.
Despite being a TF, the EOC-specific cistrome for PAX8 has not yet been described and little is
known
about
its effects on genomic regulation or functional effects on EOC subtypes. The
50
advancement of scientific technology has allowed an unprecedented level of molecular profiling
including genome-wide gene expression and epigenetic modifications. These data can be
overlaid to provide detailed examination of gene regulation. Thus I sought to investigate the
oncogenic potential of PAX8 in EOC cells and map its epigenetic and genetic targets. A better
understanding of this TF could be crucial clinical intervention, and in turn reduce disease
morbidity and mortality.
MATERIALS AND METHODS
Immunohistochemistry
Immunohistochemistry (IHC) for PAX8 (catalogue number 10336, Proteintech) was
performed on sections of normal ovaries and on a tissue microarray (TMA) containing 231
ovarian carcinoma biopsies from patients with known outcome from the University of Derby.
Staining was confirmed on whole sections from another part of the tumour for ~10% of the TMA
biopsies. IHC staining intensity was used to assess PAX8 protein expression in each tissue (0 =
negative; 1 = low; 2 = medium; 3 = high). Human tissues were used with the approval of the
institutional review board at the University of Derby.
Cell Culture and PAX8 Knockdown
The ovarian cancer cells IGROV1 (luciferase labelled) and HEYA8 were cultured in
Dulbecco's Modified Eagle's medium (DMEM)(Caisson) and Roswell Park Memorial Institute
medium (RPMI) (Lonza), respectively. Media were supplemented with 10% fetal bovine serum
(Seradigm). Lentiviral supernatants containing individual short hairpin RNAs against PAX8 or
51
control shRNA (shScr) was generated by co-transfection of HEK293T cells with 3
rd
generation
packaging vectors and PAX8-shRNAs cloned into the pLKO.1 vector (Sigma Aldrich) (clone
IDSs: NM_003466.3-772s21c1 (shPAX8_1), NM_03346.3-1070s21c1(shPAX8_2),
NM_003466.2-784s1c1 (shPAX8_3), NM_003466.3-1917s21c1 (shPAX8_4)). HEYA8 and
IGROV1 cells were infected by overnight lentiviral transduction and positive cells selected using
800 (HEYA8) or 200 (IGROV1) ng/mL puromycin (Sigma-Aldrich). Cell line authentication
was performed on all cell lines using the Promega Powerplex 16HS Assay (performed at the
University of Arizona Genetics Core facility). All cultures were confirmed to be free of
Mycoplasma infections using a Mycoplasma specific PCR.
RT-qPCR
RNA was harvested using the Quick-RNA MiniPrep kit (catalogue number 1055, Zymo
Research) and reverse transcribed to cDNA using the MMLV reverse transcriptase (Promega).
Target gene expression was normalized to ACTB and GAPDH expression (catalogue number,
Life Technologies), and relative expression calculated using the ∆∆Ct method. Probe
information can be found in Appendix Table 1.
In vitro phenotypic assays
Growth curves were performed by plating 1-2 x10
5
cells into triplicate 60 millimeter
dishes and passaging and counting cells at regular intervals. Population doublings (PDs) were
calculated using the following formula: PD = log (final cell number/initial cell number)/log2, and
cumulative population doublings were plotted. For anchorage independent growth assays 1-3
52
x10
3
cells were suspended in culture media containing 0.33% Noble Agar (Sigma Aldrich),
plated on top of a base layer of culture media containing 0.6% Noble Agar. After 4 weeks
colonies were stained by overnight incubation at 37ºC with 1% p-iodonitrotetrazolium violet
(Sigma) dissolved in 100% methanol (VWR). Colonies were visualised and counted by phase
microscopy.
In vivo tumorigenicity assays
Xenograft assays were performed under the approval and guidance of the University of
Southern California Institutional Animal Care and Use Committee. 8 million cells were injected
intra-peritoneally into 6-week old nu/nu mice (Simonsen Laboratories). Animals were sacrificed
after 28 days and tumors were measured by digital caliper. Tumors were fresh frozen, embedded
in OCT and, sectioned with a cryostat, and placed on slides. H&E-stained tissue was examined
by a pathologist.
Microarray analyses
RNA was isolated in triplicate and gene expression microarray profiling performed using
the Illumina HumanHT-12 v4 Expression BeadChips. Arrays were run at the University of
Southern California Epigenome Core and University of California at Los Angeles Neuroscience
Genomics Core, using standard protocols. Data analysis was performed using Partek Genomic
Suite. For each cell line triplicate arrays for two independent knockdown clones were compared
to control shRNA and parental samples. Pathway analyses were performed using Ingenuity
Pathway Analysis software.
53
Chromatin immunoprecipitation sequencing
Chromatin immunoprecipitation sequencing (ChIP-seq) was performed based on the
methods of Schmidt et al.
23
. Cells were fixed in 1% formaldehyde for 10 minutes, and quenched
with glycine. Cells were harvested, lysed in a sarkosyl-containing buffer, and sonicated using
the Covaris E220evolution Focused-Ultrasonicator. 10 μg of an antibody raised against PAX8
(NBP1-32440, Novus, validation) or 5 μg of an antibody raised against H3K27ac (C15410196,
Diagenode, validation) was incubated with 100 μg and 4ug of chromatin, respectively, at 4 °C
overnight. Blocked magnetic Dynabeads (Life Technologies) were then added to the antibody-
lysate conjugates and incubated at 4°C for 4 hours with rotation. Beads were washed with RIPA
buffer and treated with RNase and proteinase K (both Qiagen). DNA was eluted from the beads
in Tris-EDTA buffer and cleaned up using the QIAquick PCR Purification kit (Qiagen). For each
cell line two independent immunoprecipitations (IPs) and one input sample were submitted for
library preparation and next-generation sequencing at the USC Epigenome Core Facility.
Sequencing was carried out on an Illumina NextSeq using 75 base pair single-end reads.
ChIP-seq Data analysis
ChIP-seq data were processed using MACS2
24
with p-value cutoff of 0.001. The smaller
of input or signal was linearly scaled to the same depth as the larger dataset. IDR version 2.0
pipeline with a threshold of p < 0.05 was used to control the irreproducible discovery rate
25
.
After mapping to hg19 with BWA
26
and removing duplicates, there were > 15 million high-
quality reads for all ChIP-seq replicates. SPP
27
was used to calculate the cross-correlation QC
metrics (Appendix Table 2) and MACS2 + IDR to obtain the rescue ratio and self-consistency
54
ratio for each cell line. SPP+IDR was used for peak calling and comparison with the
MACS2+IDR peaks. ~94% of the peaks called by MACS2+IDR pipeline were included in the
SPP+IDR pipeline, which gave confidence in using the MACS2+IDR peaks for discovery of the
PAX8 cistrome.
HEYA8 and IGROV1 have 21,563 (average length = ~2.4 Kbp) and 15,030 (average
length = ~2 Kbp) H3K27ac enriched regions, respectively. HOMER
28
was used to broadly
identify super enhancer regions using the H3K27ac ChIP-seq peaks for individual replicates.
Then, to avoid inclusion of non-reproducible peaks in this analysis, the MACS2+IDR H3K27ac
peaks that overlap the super enhancer regions identified by HOMER were merged, and the
MACS2+IDR/HOMER regions were used as the superenhancer regions for each cell line.
For motif discovery, the web application MEME-ChIP was utilized with all PAX8 peaks
for each cell line, individually
29
. To associate the differentially expressed genes (DEGs) and the
PAX8 ChIP-seq peaks, a set of topological association domains (TADs) was used from H1
human embryonic stem cells
30
.
Pathway enrichment analysis was performed with the web application Metascape
31
, using
a custom analysis including “Oncogenic signatures”, “GO Molecular Function”, “Canonical
Pathways”, “GO Biological Processes”, Hallmark Gene Sets” and “KEGG Pathway” with the
default parameters.
55
RESULTS
PAX8 expression in ovarian cancer histotypes
Immunohistochemistry was used to evaluate the expression of PAX8 protein in 231
primary tumors representing the four major histotypes of ovarian cancer - HGSOC, clear cell
ovarian cancer (CCOC), endometrioid ovarian cancer (EnOC) and mucinous ovarian cancer
(MOC). PAX8 was most highly expressed in HGSOCs (88%) and CCOCs (89%) (p = 0.007),
but was also expressed in over half of MOCs and EnOCs, suggesting that PAX8 is a pan-EOC
marker (Table 1). Highest PAX8 expression was associated with advanced stage carcinomas
compared to low stage carcinomas (p = 0.013). It did not correlate with tumor grade.
PAX8 mRNA levels were measured in EOC cells and normal EOC precursor cells: 72
primary ovarian cancer precursor cells (66 normal immortalized OSEC and 6 normal
immortalized FTSEC cultures) and 58 ovarian cancer cell lines derived from the histological
range of EOC histotypes. These data were consistent with the immunohistochemical analyses:
PAX8 was expressed at higher levels in EOC cell lines than normal precursor cells (p < 0.0001,
two-tailed unpaired T-test). The highest PAX8 expression was observed in HGSOC and CCOC
cell lines (Figure 1). There was significant variation in PAX8 expression across EOC histotypes
(p < 0.0002, one-way ANOVA), with PAX8 overexpressed in HGSOC (n=13), CCOC (n=15),
MOC (n=2), and NS/MIX cell lines (n=8) but not EnOC (n=3), ADENOC (n=7), or UNDIFF
cell lines (n=6). Taken together, these data suggest that PAX8 may function as an oncogene in
the development of most EOC histotypes.
56
Table 1. PAX8 expression in EOC histotypes
PAX8 staining N (%)
Clinical Feature Low/Absent Medium/High
P-value
†
Stage
1 17 (30) 39 (70) 0.013*
2 6 (21) 22 (79)
3 16 (14)* 101 (86)*
4 5 (20) 20 (80)
Grade
G1 2 (7) 25 (93) 0.072
G2 11 (24) 35 (76)
G3 30 (20) 121 (80)
Residual Disease
Optimal debulking 21 (22) 73 (78) 0.391
Sub-optimal debulking 22 (17) 106 (83)
Histological type
Serous 15 (12)* 109 (88)* 0.007*
Mucinous 6 (35) 11 (65)
Endometrioid 9 (29) 22 (71)
Clear Cell 2 (11)* 16 (89)*
Undifferentiated 10 (29) 24 (71)
Borderline 3 (43) 4 (57)
A significantly higher number of stage 3 and serous ovarian carcinomas express high/medium
levels of PAX8 (denoted by *).
†
The Pearson Chi Squared test was used to derive the p-values.
57
Figure 1. PAX8 gene expression in EOC histotypes and precursor cells. HEYA8 cells are
(high-grade) undifferentiated (papillary TP53 wild type) and IGROV1 are mixed (endometrioid,
clear cell and undifferentiated). *p < 0.05 compared to IOE and IFTE, two-tailed unpaired T-test.
Abbreviations: EOC, epithelial ovarian cancer; IOE, immortalized ovarian surface epithelial cells;
IFTE, immortalized fallopian tube secretory epithelial cells; HGSOC, high-grade serous; CCOC,
clear cell; EnOC, endometrioid; MOC, mucinous; ADENOC, adenocarcinoma only; UNDIFF,
undifferentiated and poorly differentiated; NS, subtype not specified; MIX, mixed histotype.
EOC
histotypes
EOC
precursors
*
*
*
*
58
PAX8 knockdown reduces anchorage dependent and independent growth
A model of EOC cells with stable PAX8 knockdown was generated using two
independent shRNAs in the HEYA8 and IGROV1 cell lines. These cells represent models of
high-grade ovarian adenocarcinoma (Figure 1 legend) but not a specific histological subtype
32-34
.
HGSOC cell lines COV318, PEO14, and UWB1.289 were transduced but would not generate
stable proliferating cell lines. A control line expressing a non-silencing ‘scrambled’ shRNA
hairpin (shScr) was also generated for each line. In HEYA8 there was a 74% and 58% reduction
in PAX8 expression for the shPAX8_1 and shPAX8_2 transduced lines, respectively (p > 0.05)
(Figure 2a). In IGROV1 a 70% and 79% reduction was achieved in PAX8 expression for
shPAX8_3 and shPAX8_4, respectively (p > 0.05) (Figure 2b). In both cell lines negligible
changes in PAX8 expression were observed in shScr controls. There was marked reduction in
PAX8 protein expression in shPAX8 transduced cell lines compared to shScr and parental cells
(Figure 2c and 2d).
Phenotypic effects were observed in the PAX8 EOC knockdown models. shPAX8 cells
were associated with a significant reduction in anchorage-independent growth compared to shScr
(HEYA8 shPAX8_1, p = 0.003; HEYA8 shPAX8_2, p = 0.003; IGROV1 shPAX8_3, p =
0.0008; IGROV1 shPAX8_4, p = 0.0007) (Figure 2e and 2f). Significant increases in population
doubling times in all PAX8 knockdown models were also observed compared to shScr cell lines
(HEYA8 shPAX8_1, p = 0.004; HEYA8 shPAX8_2 p = 0.004; IGROV1 shPAX8_3, p = 0.008;
IGROV1 shPAX8_4, p = 0.011) (Figure 2 g and 2h).
59
Figure 2. In vitro analysis of PAX8 knockdown models. PAX8 was stably knocked down
using short hairpin RNAs (shRNAs). PAX8 gene expression in knockdown and control lines in
(a) HEYA8 and (b) IGROV1 ovarian cancer cell lines. Knockdown of PAX8 protein was
confirmed by Immunofluorescent staining in (c) HEYA8 and (d) IGROV1 cells. Nuclei were
counterstained with Hoescht DNA stain. 200X magnification. (e) Anchorage independent growth
assays in HEYA8 and (f) IGROV1 PAX8 knockdown models. Anchorage dependent growth
assays in (g) HEYA8 and (h) IGROV1 models. Data shown are mean ± standard deviation, and
are representative of at least three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001,
two-tailed paired T-test. In panels (g) and (h) T-tests values (two tailed, paired) for knockdown
lines compared to shScr lines are indicated.
60
PAX8 knockdown impairs tumorigenicity in vivo
The effects of PAX8 knockdown on in vivo tumorigenicity in HEYA8 models was
investigated. Abdominal distention was observed in parental and shScr-expressing HEYA8 cells
4 weeks after intraperitoneal injection of tumor cells, indicative of significant tumor burden
(Figure 3a and 3b). Necroscopic examination of the abdominal cavity revealed tumors spread
throughout the peritoneal cavity in all 4 HEYA8 and in 4/5 HEYA8+shScr injected mice.
Pathological examination of tumors formed in HEYA8 and HEYA8+shScr mice classified these
tumors as high-grade poorly differentiated adenocarcinomas, consistent with the known
pathology of the HEYA8 cancer cell line
32
. In contrast, HEYA8+shPAX8-1 injected mice
demonstrated modest tumor growth compared to controls. None of the HEYA8+shPAX8_2 mice
(n=3) showed any evidence of tumor growth at the time of sacrifice (Figure 3c). Subsequent
pathological analyses of HEYA8+shPAX8_2 mice confirmed the absence of tumor deposits in
tissue harvested from these mice, resembling the phenotype of mice injected with vehicle alone
(n=3).
Figure 3. In vivo analysis of PAX8 knockdown models. Data shown are representative of
tumor burden (n = 4). (a) In vivo growth of HEYA8 PAX8 models, representative images of
peritoneal cavities. Tumors are indicated with an asterisk. (b) Reproductive tracts in xenograft
models, tumors are indicated with arrowheads. O, ovary; T, oviduct (murine fallopian tube); U,
uterine horn; shScr, non-targeting control shRNA. (c) Quantitative analysis of largest tumor
volume. *p < 0.05 versus parental cells. Two tailed unpaired T-test.
61
Characterizing the PAX8 cistrome
To catalogue PAX8 occupied regions (P8ORs) of DNA in EOC, chromatin
immunoprecipitation followed by next generation sequencing (ChIP-seq) was done with HEYA8
and IGROV1 cell lines. In parallel ChIP-seq for H3K27ac was performed to identify when
PAX8 binding coincided with active chromatin (Figure 4).
2923 sites (average length 566 bp) were identified throughout the genome where PAX8
was bound in HEYA8 cells and 3029 PAX8 binding sites (average length = 673 bp) in IGROV1
cells (Figure 5). 653 PAX8 binding sites (22%) overlapped between the two cell lines (Figure
5c). PAX8 binding sites were significantly enriched in promoter regions for both cell lines (p <
0.05, Fisher’s exact test). Approximately 6% (n=182) and 16% (n=478) of all PAX8 binding
Figure 4. ChIP-seq binding regions are shared and different in HEYA8 and IGROV1.
Browser tracks in igv showing genomic regions where ChIP-seq peaks were the same for both
replicates in: both cell lines for PAX8 (purple box) and H3K27ac (black box) and different for
each cell line for PAX8 (blue box) and H3K27ac (green box).
62
Figure 5. PAX8 ChIP-seq binding sites for HEYA8 and IGROV1. (a) Distribution of PAX8
binding sites across promoter, gene body and intergenic regions of the human genome for
PAX8 binding sites in HEYA8 (orange), IGROV1 (blue) and the intersection (the common sites)
of the two sets (yellow). (b) Distribution of the distance of the PAX8 binding sites in HEYA8
(orange) and IGROV1 (blue) respect to the closest transcription start site (TSS) normalized by
the number of PAX8 binding sites in HEYA8 and IGROV1, respectively. (c) Venn diagram
showing the number of PAX8 binding sites in HEYA8 (orange) and IGROV1 (blue) and intersect
(overlap of orange and blue). (d) Venn diagram showing the number of differentially expressed
genes in PAX8-knockdown versus control (absolute fold change>1.2 and false discovery rate
(FDR) < 0.05) for HEYA8 (orange) and IGROV1 (blue), which were used to identify PAX8
regulatory targets.
63
sites fell in promoter regions in HEYA8 and IGROV1 cells, respectively (Figure 5a and 5b).
Although more than 60% of PAX8 binding sites were found outside promoters or gene bodies,
no evidence was found for enrichment in intergenic regions, even when excluding repeat-rich
regions. There were some significant differences in the PAX8 cistrome between the two cell
lines: 60% (n=1829) of PAX8 binding sites in IGROV1 overlapped with H3K27ac positive
regions compared to only ~25% (n = 731) of P8ORs in HEYA8 cells. This was despite the
finding that there were 2.5 times more H3K27ac positive regions observed in HEYA8 cells.
To further evaluate the PAX8 cistrome, superenhancers were identified in the two cell
lines - defined as enhancers >10kb that are highly enriched in H3K27ac and lineage-specific
transcription factors
35
. 391 superenhancers were observed (average length = 126K bp) in
HEYA8 and 523 in IGROV1 (average length = 71K bp) based on the H3K27ac ChIP-seq data.
HEYA8 P8ORs were significantly enriched at superenhancers, with an overlap of approximately
326 (11%) PAX8 binding sites in these regions (odds ratio 1.82; p < 0.01).
Characterizing the PAX8 binding motif
PAX proteins contain a PAIRED DOMAIN that is further broken down into the PAI and
RED subdomains. These domains are responsible for PAX8 binding to DNA. MEME-ChIP suite
was used to characterize a PAX8 binding motif from the PAX8 ChIP-seq data. The same PAX-
like motif was found in both HEYA8 and IGROV1 cell lines (Figure 6). This motif correlates
with a PAX8-dsDNA homology model that was generated using the PAX5-dsDNA complex
(PDB id: 1k78, sequence identity = 85%) as a template. The homology model had high structure
similarity to this template (RMSD = 0.2) as well as another PAX8 homolog, PAX6-dsDNA
(PDB id: 6pax; sequence identity = 72%; RMSD = 1.6). The protein side-chain/DNA-base
64
Figure 6. Defining the PAX8 binding motif and identifying candidate cooperating
transcription factors. (a) Binding motifs identified by MEME-ChIP using the H3K27ac positive
PAX8 binding sites in the enhancer set for HEYA8 and IGROV1. (b) Pax5 binding motif (from
Jaspar
36
) was selected as the closest binding motif to the primary PAX8 binding motif based on
the adjusted p-value across four sets of PAX8 binding sites used for motif discovery (all (A),
superenhancer (S), promoters (P), and enhancer (E) PAX8 peaks) for HEYA8 (orange) and
IGROV1 (blue). Significantly enriched motifs within ±250 bp of the summits of PAX8 peaks in (c)
HEYA8 and (d) IGROV1 cells. Other than PAX-like motifs, only motifs identified in both cell lines
are shown (with the exception of TEAD1/3, which was unique to IGROV1). The grey line at 0 bp
indicates the summit of the PAX8 peak.
65
contacts observed in the homology model extend 17 base pairs. First, the PAI subdomain makes
contact with 5 continuous bases in the major groove and one base in the minor groove, followed
by the linker that makes contact with 3 non-continuous bases in the minor groove. At the C
terminus of the PAIRED DNA binding domain, the RED subdomain makes contact with 4
continuous bases via the major groove. Since it is known that protein side-chain/DNA-major
groove contacts provide higher specificity than contacts in the minor groove, the expected motif
would have a region of medium to low specificity (provided by the linker) in the middle, flanked
by a couple of highly-specific bases (provided by the PAI and RED subdomains). This is the
exact pattern exhibited by the identified PAX-like motif (Figure 6a).
Identifying PAX8 cooperating transcription factors
To identify candidate cooperating transcription factors that may cooperate with PAX8 in
regulating gene expression, PAX8 peaks that overlap H3K27ac were divided into three non-
overlapping sets: (S) PAX8 binding sites that overlap superenhancer regions, (P) PAX8 binding
sites that overlap promoter regions defined as 1000 bp upstream and 100 bp downstream of the
transcription start site of a gene, and (E) PAX8 binding sites that neither overlap superenhancers,
nor promoters, but are H3K27ac positive (i.e. typical enhancers). PAX motifs were significantly
enriched in all classes of PAX8 binding sites in both cell lines (Figure 6b, Table 3). Found in
both cell lines was enrichment of Arnt::Ahr motifs for P8ORs overlapping superenhancers or
enhancers, but not promoters (in HEYA8 superenhancers, p = 4.37x10
-6
; in HEYA8 typical
enhancers, p = 9.98x10
-8
; in IGROV1 superenhancers, p = 1.19x10
-6
; in IGROV1 typical
enhancers p = 9.76x10
-11
).
66
Table 2. Candidate PAX8 cooperating transcription factors
HEYA8 IGROV1
Motif/
Factor All
Super
enhancer Promoter Enhancer All
Super
enhancer Promoter Enhancer
PAX5 -75.64 -122.14 -75.64 -237.09 -84.48 -156.36 -84.48 -276.91
PAX2 -76.06 -140.02 -76.06 -242.56 -90.26 -202.7 -90.26 -253.45
PAX1 -42.29 -85.93 -42.29 -121.7 -38.96 -114.76 -38.96 -138.49
PAX6 -38.46 -121.77 -38.46 -133.79 -16.11 -94.49 -16.11 -127.34
PAX9 -35.48 -58.45 -35.48 -83.5 -26.75 -76.52 -26.75 -95.52
TEAD3 0 0 0 0 0 -14.77 0 -25.26
Arnt::Ahr 0 -12.34 0 -16.12 0 -13.64 0 -23.05
TEAD1 0 0 0 0 0 -22.46 0 -11.55
Zbtb3_
primary
0 0 0 -7.92 0 0 0 -14.5
GMEB2 0 0 0 -6.17 0 0 0 -14.07
Klf4 0 0 0 0 0 0 0 -11.27
Zfp161_
secondary
0 0 0 0 0 0 0 -9.9
JUND 0 -5.81 0 -5.13 0 -8.61 0 0
SP1 0 0 0 0 0 0 0 -8.61
Sp4_
secondary
0 0 0 0 0 0 0 -8.25
FOS 0 -5.52 0 -5.23 0 0 0 -8.21
JUNB 0 -5.02 0 -5.18 0 -8.2 0 0
HIF1A::AR
NT
0 0 0 0 0 -7.76 0 -6.05
Klf1 0 0 0 0 0 0 0 -7.47
KLF5 0 0 0 0 0 0 0 -6.28
Nfe2l2 0 0 0 -6.2 0 0 0 0
Max_
secondary
0 0 0 0 0 0 0 -5.71
NFIA 0 0 0 0 0 0 0 -5.43
Klf7_
primary
0 0 0 0 0 0 0 -5.19
KLF14 0 0 0 0 0 0 0 -5.12
NFYA 0 0 0 0 0 0 0 -5.07
67
Log adjusted p-values of motifs found within 50 bases of PAX8 binding site summits of HEYA8
and IGROV1 using all, superenhancer, promoter and enhancer-associated PAX8 binding sites.
(Figure 6c and 6d, Table 3). In addition JUND and JUNB motifs were enriched in
superenhancers in both cell lines (in HEYA8 JUND, p = 3.00x10
-3
; JUNB, p = 6.60x10
-3
; in
IGROV1, JUND p = 1.82x10
-4
; JUNB, p = 2.75x10
-4
). A recent report identified TEAD as an
important PAX8 cooperating transcription factor
37
. A TEAD-like motif flanked the PAX-like
motif in IGROV1 but not in HEYA8 (TEAD3 enrichment in IGROV superenhancers, p =
1.76x10
-10
, in IGROV typical enhancers, p = 9.46x10
-6
; TEAD1 enrichment in IGROV
superenhancers, p = 3.85x10
-7
, in IGROV typical enhancers, p = 1.07x10
-11
).
Identifying PAX8 target genes
Global gene expression analysis was performed to identify differential gene expression
following PAX8 knockdown in HEYA8 and IGROV1. 1,055 and 1,293 differentially expressed
genes (DEGs) were found in HEYA8 and IGROV1 models respectively (FDR<0.05; Appendix
Table 3, validation) using a liberal fold-change cutoff of 1.2. The top genes in each cell line are
in Table 3 with the common DEGs to both cell lines are listed in Table 4. Gene expression
changes were integrated with PAX8 binding sites throughout the genome to map PAX8 binding
sites to target genes. Topological association domains (TADs) were used from human
embryonic stem cells to annotate the DEGs. TADs are megabase-scale regions, largely stable
across different cell types, of increased local interaction frequency
30
. Using these annotations, 3
categories were created for associations with PAX8 ChIP-seq: (1) direct regulatory targets,
defined as DEGs that have a PAX8 binding site in their promoter region; (2) putative regulatory
68
targets, defined as DEGs that have a PAX8 binding site within the same TAD but lacking PAX8
binding in the promoter; and (3) indirect regulatory targets, defined as DEGs where there is no
PAX8 binding site within the same TAD or the promoter.
Table 3. Top ten significantly changing genes in EOC models following PAX8
knockdown
HEYA8 IGROV1
Gene Name Fold Change P-value Gene Name Fold Change P-value
Downregulated Genes Downregulated Genes
DCDC2 -2.8 8.82E-07 SPON1 -13.4 3.50E-10
SLC7A5 -2.6 1.09E-03 KRT24 -12.0 4.32E-12
CDH5 -2.1 1.12E-05 MMP7 -6.8 4.77E-06
FAM167A -2.1 1.25E-05 APBB1IP -3.3 2.06E-07
C8ORF13 -2.0 2.25E-05 LOC644612 -3.3 5.00E-04
H2AFY2 -2.0 1.23E-03 THY1 -3.2 1.84E-08
IL11 -2.0 2.37E-04 C1ORF85 -3.1 8.82E-09
C13ORF15 -2.0 7.73E-09 STK32B -3.0 2.16E-05
PAX8 -1.9 4.57E-06 SAMD5 -3.0 4.90E-06
SCD -1.9 6.61E-06 CYP4F11 -2.8 5.45E-06
Upregulated Genes Upregulated Genes
IL1B 2.6 3.92E-05 LOC149501 8.0 1.19E-09
LOC644350 2.6 7.50E-07 KRT18P13 8.2 2.23E-10
SCG5 2.7 1.92E-06 MGC42367 8.4 8.26E-10
STC1 2.7 2.51E-05 LOC399965 8.8 6.92E-11
HNRPLL 2.7 5.59E-11 EMP1 9.1 3.39E-11
IL8 3.3 3.18E-05 ANKRD1 10.8 4.06E-09
LPXN 3.5 3.06E-07 LOC644743 11.0 1.35E-11
IL13RA2 3.9 4.50E-14 KRT8 12.2 1.74E-11
LOC728285 4.2 8.81E-09 LOC647954 12.2 5.67E-12
TGFBI 6.2 5.51E-09 CDH17 20.4 6.37E-10
69
Table 4. Genes commonly changing in HEYA8 and IGROV1 models following PAX8
knockdown
HEYA8 IGROV1
Gene Name Fold Change P-value
Fold
Change P-value
ANKRD1 1.6 2.02E-03 10.8 4.06E-09
VASN 1.7 4.02E-07 6.0 5.38E-11
PRSS23 1.5 2.10E-04 3.9 6.86E-10
C20orf75 -1.7 9.06E-05 -2.8 8.22E-05
STC1 2.7 2.51E-05 1.8 2.18E-05
HIST2H2AA3 2.0 8.41E-07 1.7 2.24E-06
HIST2H2AA4 1.9 1.79E-06 1.6 8.88E-07
AJAP1 -1.9 3.82E-10 -1.8 2.17E-09
HIST1H2BK 1.9 2.20E-05 1.8 1.33E-05
SAT1 1.8 4.47E-07 1.7 2.75E-05
Ranked by absolute fold change.
There were 3,062 TADs with an average length of 852.2 Kbp (range 80 Kbp - 4.44 Mbp),
of which 1,483 in HEYA8 and 1,413 in IGROV1 contained PAX8 binding sites. About two
thirds (2,012) of TADs had a PAX8 peak in at least one of the cell lines, with 884 (29%) shared
between HEYA8 and IGROV1 models. More than half of the TADs containing PAX8 binding
sites - 862/1483 (58%) in HEYA8 and 732/1413 (52%) in IGROV1 - contained just one or two
PAX8 binding sites (Figure 7a). No correlation was found between the length of the TAD and
the number of PAX8 binding sites. About a quarter of TADs, 714 TADs in HEYA8 and 830 in
IGROV1, contained PAX8 regulated DEGs and most TADs harboring DEGs (566 and 619 for
HEYA8 and IGROV1) contained at most one DEG (Figure 7b). The maximum number of DEGs
identified within the same TAD was 9 in HEYA8 and 6 in IGROV1 and there was no correlation
between the length of the TAD and the number of DEGs. 434 and 549 TADs contained at least
one P8OR plus one DEG in HEYA8 and IGROV1 cell models, respectively (Figure 7c-d). The
70
Figure 7. Identification of PAX8 regulatory targets using TADs. (a) Distribution of the
number of PAX8 peaks per TAD. (b) Distribution of the number of differentially expressed genes
(DEGs) per TAD. Venn diagrams showing the number of TADs with at least one PAX8 peak or
at least one DEG for (c) HEYA8 and (d) IGROV1. (e-f) Differentially expressed genes divided
into three non-overlapping sets: direct regulatory targets, which have at least one PAX8 peak in
the promoter (at the bottom); putative enhancer regulatory targets, which have at least one
PAX8 peak within the same TAD (in the middle); and indirect regulatory targets, which do not
have a PAX8 peak within the same TAD (on the top); for HeyA8 (e) and IGROV1 (f).
71
mean distance in base pairs to the nearest DEG within the same TAD was 309 Kbp (sd=357
Kbp) in HEYA8 and 235 Kbp (sd=275 Kbp) in IGROV1.
From the 1,055 DEGs for HEYA8, 16 direct regulatory targets were identified based on
PAX8 binding to the promoter, and 575 putative enhancer regulatory targets based on the
presence of a PAX8 binding site within the same TAD (Figure 7e). From the 1,293 DEGs for
IGROV1, 46 direct regulatory targets and 755 as putative enhancer regulatory targets identified
(Figure 7f). Somewhat surprisingly, the presence of a H3K27ac mark at a PAX8 binding site was
not correlated with loss of gene expression following PAX8 knockdown (see annotation bar in
Figure 7e and 7f, left). 54 genes were identified that are both differentially expressed following
PAX8 knockdown and have PAX8 binding at an enhancer in the same TAD in both cell lines
(Table 5).
Using both direct and putative enhancer regulatory targets for HEYA8 (589 DEGs) and
IGROV1 (801 DEGs), a pathway enrichment analysis was conducted using Metascape
31
. 5
commonly enriched pathways in both HEYA8 and IGROV1 cell lines were found: DNA
replication, response to LPS (lipopolysaccharide), TNFA (tumor necrosis factor alpha) signaling
via NFKB (nuclear factor-kappaB), extracellular matrix organization and anatomical structure
morphogenesis. The top 20 enriched gene sets and pathways are shown in Figure 8.
72
Table 5. PAX8 putative enhancer regulatory targets
HEYA8 IGROV1
Gene Symbol P-value Fold change P-value Fold change
ANKRD1 2.02E-03 1.57 4.06E-09 10.79
VASN 4.02E-07 1.74 5.38E-11 6.04
IL1B 3.92E-05 2.57 1.10E-04 1.26
PAX8 4.57E-06 -1.95 6.51E-08 -2.35
CLIC3 1.19E-04 1.25 1.58E-03 2.07
HIST2H2AA3 8.41E-07 2.00 2.24E-06 1.73
AJAP1 3.82E-10 -1.94 2.17E-09 -1.77
SLC26A2 1.50E-03 -1.29 2.51E-06 -1.91
HIST1H2BK 2.20E-05 1.90 1.33E-05 1.85
GCLM 5.78E-06 -1.33 3.94E-05 -1.87
HIST2H2AA4 1.79E-06 1.86 8.88E-07 1.56
MXD4 3.62E-04 1.73 1.10E-04 1.35
ELFN2 7.37E-05 -1.50 7.67E-06 -1.72
RRAS2 6.42E-04 -1.45 5.91E-08 -1.69
CTSL2 1.65E-04 -1.41 6.56E-09 -1.61
SLC25A15 6.35E-07 -1.59 4.85E-06 -1.61
HIST2H2AC 1.35E-06 1.59 3.62E-04 1.50
ATOX1 6.52E-05 1.26 8.64E-07 1.52
CASP1 3.96E-08 1.52 5.99E-04 1.24
S100A13 7.89E-06 -1.51 2.82E-04 -1.36
EVI5L 8.13E-06 1.26 9.11E-07 1.46
CEP55 5.11E-05 -1.45 2.16E-03 -1.21
UBA7 7.76E-05 1.44 7.35E-06 1.29
CHAF1A 5.70E-05 -1.24 1.21E-04 -1.43
SKA3 9.17E-04 -1.36 7.51E-04 -1.41
SH3BGRL3 5.28E-08 1.41 3.79E-04 1.23
PTPRF 8.77E-10 1.41 1.32E-05 1.35
S100A6 5.02E-04 1.34 4.37E-05 1.41
NRP1 4.26E-06 1.36 1.09E-05 1.39
EHD2 4.79E-04 1.29 6.48E-04 1.39
73
XPO4 2.85E-03 -1.36 2.78E-05 -1.37
RHOC 7.17E-07 1.37 1.26E-04 1.24
KIAA1539 2.13E-07 1.29 4.36E-04 1.36
JARID2 1.27E-03 1.28 8.05E-06 1.36
CASP4 2.59E-03 1.21 3.46E-05 1.36
TROAP 1.02E-03 -1.31 2.81E-03 -1.36
DUSP28 1.25E-03 1.31 7.97E-05 1.34
SEPW1 6.62E-05 1.33 1.63E-03 1.30
FEN1 9.88E-06 -1.33 8.67E-05 -1.23
ACAA2 1.45E-03 -1.21 5.36E-05 -1.33
MPHOSPH8 7.94E-04 -1.30 4.22E-04 -1.31
POLR3K 4.38E-04 -1.27 1.83E-04 -1.31
FOXM1 2.30E-06 -1.29 1.24E-04 -1.26
NUP210 2.99E-03 -1.20 9.35E-05 -1.29
POLD1 7.56E-08 -1.28 1.92E-03 -1.23
CHTF18 3.49E-04 -1.26 2.28E-03 -1.28
CDCA5 1.41E-03 -1.24 6.80E-05 -1.28
DSCC1 2.08E-05 -1.28 4.92E-05 -1.28
FRAT2 8.96E-04 -1.26 1.82E-03 -1.20
GPRIN1 1.16E-04 -1.25 3.79E-04 -1.25
ZNF259 3.22E-04 -1.24 7.71E-05 -1.25
FANCG 1.99E-03 -1.21 1.79E-04 -1.25
OIP5 1.46E-04 -1.22 2.65E-03 -1.25
M6PRBP1 1.99E-05 1.22 1.95E-03 1.20
74
Figure 8. Pathway enrichment analysis. (a) Top 20 pathways and gene sets enriched using 9
direct and 462 putative enhancer regulatory targets for PAX8 in HEYA8. (b) Top 20 pathways
and gene sets enriched using 46 direct and 739 putative enhancer PAX8 regulatory targets in
IGROV1. DNA replication (green), TGFB UP.V1 DN (orange), extracellular matrix organization
(purple) and NFKB signaling (pink) overlap between HEYA8 and IGROV1 top 20 enriched
pathways.
75
DISCUSSION
Characterizing the TFs deregulated during tumorigenesis has provided key insights into
disease etiology, disease origins and therapeutic targeting for many tumor types. A better
understanding of transcriptional networks can be obtained by identifying the factors responsible
for higher-order deregulation of gene expression. Moreover, there is potential to target tumor
cells by inhibiting a factor that simultaneously deregulates tens to hundreds of proto-oncogenes
and tumor suppressor genes. Examples of such factors include the androgen receptor, a TF
involved in normal prostate development which becomes reprogrammed during prostate
carcinogenesis
38
. The androgen receptor is targeted by anti-androgenic therapies used for the
treatment of prostate cancer
39
. Similarly, the estrogen receptor is targeted by tamoxifen for the
treatment of estrogen receptor positive breast cancer.
In contrast, TFs deregulated in epithelial ovarian carcinoma are poorly characterized, and
there are no targeted treatments for EOCs. Only a handful of TFs and chromatin remodelers are
known to be deregulated in a histotype-specific manner: Wilms tumor 1 is a well-known marker
of serous tumors, and ARID1A is somatically mutated in ~50% of clear cell and ~30% of
endometrioid tumors
40
. A detailed mechanistic understanding of the putative master regulators
driving EOC development is currently lacking.
In this study I examined PAX8, a TF that is overexpressed by the major EOC histotypes,
but is most often studied in only one subtype (HGSOC). This study supports previous
observations that PAX8 has a strong effect on cellular proliferation of EOC cells. These effects
were observed in vitro during anchorage dependent growth, and were even further pronounced in
anchorage independent growth using two independent shRNAs in two cell lines. These data may
explain the drastic reduction in tumor burden in mice following intraperitoneal injection of
76
PAX8 knockdown cells, which is supported by the literature
41
. There were several DEGs related
to cell proliferation with reduced expression after PAX8 knock down as well cellular pathways.
To examine the molecular effects of PAX8 in EOC, I examined gene expression and
regulation. Genome-wide binding regions for this factor have only been identified in rat thyroid
cells
42
and only for a small number of cell lines in HGSOC
43
. No studies have been published for
any other EOC subtypes, which make up about one third of EOCs. To characterize PAX8
binding genome-wide within the context of chromatin state, PAX8 ChIP-seq data was integrated
with maps of global histone activation generated by ChIP-seq for H3K27ac in two EOC cell
lines. To link P8ORs to target genes, differential gene expression following PAX8 knockdown
was used.
To associate DEGs and TF binding sites, many studies assign TF ChIP-seq peaks to
DEGs within a specified genomic window. However the selection of the window size can be
arbitrary and the results might not be robust. Since the genome is subdivided into topological
association domains that represent regions of high interaction frequency, TAD boundaries were
used to link the potential interaction of PAX8 binding sites to PAX8 target genes. While PAX8
is indicated to regulate target gene expression via promoter binding, a previous observation was
supported - that most PAX8 binding occurs outside of promoter regions
42, 43
. Non-promoter
PAX8 binding sites were explored in more detail using the H3K27ac ChIP-seq data. It was found
that PAX8 occupancy was enriched at superenhancers, regions of dense H3K27ac signal that are
commonly found at genes involved in differentiation, development and tumorigenesis
43
.
Intriguingly, a lack of enriched motifs for other cofactors suggests that PAX8 may tend to
function as a solitary TF at promoters, or alternatively, PAX8 may be a promiscuous TF in
regard to association with other cofactors, interacting with a multitude in a gene-specific manner.
77
At enhancers and superenhancers, a select handful of TF motifs preferentially occurred
within close proximity to P8ORs. Of particular interest are the Arnt::Ahr and JUND/JUNB
motifs. JUND/JUNB, as well as TEAD1/TEAD3 motifs flank the summit of PAX8 ChIP-seq
peaks in IGROV1. This may indicate co-binding of these two factors to adjacent DNA, whereas
the Arnt::Ahr motif is enriched within the summit, suggesting an alternative mechanism of co-
regulation. This could indicate that PAX8 and Arnt::Ahr act in a DNA-binding complex. While
functional assays would be required to verify cooperation between PAX8 and these novel
cofactors, there are potentially significant translational implications of these findings. If an anti-
PAX8 therapy was developed, a combination therapy targeting both PAX8 and its preferred
cofactors might have the most potent effects. Even in the absence of a PAX8 inhibitor, inhibiting
its cofactors could possibly be sufficient to inactivate oncogenic PAX8 signaling and impair
tumor growth even more potently than as seen by PAX8 knockdown alone in in vivo.
Functionally, these are promising co-targets. While the Arnt::Ahr heterodimer pathway is
pathway is not well studied in EOC; in breast, prostate and oral squamous carcinomas AHR is
implicated in cell migration, invasion and metastasis
44-46
and has been proposed as a novel
therapeutic target
47, 48
. JUND/JUNB is part of the AP-1 TF complex, which has been implicated
in a broad array of cancer-relevant phenotypes including proliferation and apoptosis
49
.
Interestingly JUNB can interact with BRCA1
50
, which when mutated or silenced in the germline
confers a high risk of developing HGSOC.
It is interesting to note the level of knockdown achieved here using long term culture
compared to siRNA treatments. Cheung et al. were able to deplete PAX8 to undetectable levels
and kill EOC cells. The system used in my study was designed for longer term experiments than
can be achieved with siRNA transfection, and thus likely selects for cells with a level of PAX8 in
78
which cells are able to survive and proliferate. Cells with very low or no PAX8 may have been
killed due to this deficiency in the early stages in the generation of the cell line models.
There are two major caveats to the analyses in this study. Since there are no TAD maps
for EOC, TADs defined in a human embryonic stem cell line (H1) were used because TADs are
thought to be highly conserved across cell types
30
. However EOCs, in particular HGSOCs, are
highly genomically unstable. They contain many chromosomal rearrangements that could disrupt
TAD boundaries, removing or creating enhancer-target gene interactions. TAD maps and
targeted or genome-wide chromosome conformation capture data (such as 4C or Hi-C)
performed in the HEYA8 and IGROV1 cell lines would be needed to confirm the predicted
physical interactions between enhancer-associated PAX8 binding sites and the target genes
defined here. The cell line models I used were well suited for my experiments: they have high
expression of PAX8, tolerate PAX8 knockdown long-term, have the ability to form tumors in
mice, and are general models of EOC. However there may likely be some histotype-specific
differences in the PAX8 cistrome that would not have been captured in these analyses.
Consequently these results are generically applicable to all EOC histotypes, but not to each
histotype individually. Thus far one study has profiled P8ORs in EOC by using a small number
of HGSOC cell lines
37
and my data are supported by their findings. Because there are no studies
examining any other EOC histotypes, my study is highly valuable.
To better characterize the PAX8 binding motif, the motif predicted by the ChIP-seq
analyses was related to the predicted structure of the PAX8 protein in complex with DNA. PAX5
and PAX6 have been crystalized in complex with DNA
51, 52
, while PAX8 has not. Due to its
similarity to PAX8, the PAX5/DNA complex was used as a template to generate a PAX8/DNA
homology model. This revealed two subdomains within the PAX8 DNA-binding domain which
79
contact the major groove of DNA with greater specificity, joined by a linker domain with
markedly lower DNA binding specificity. The locations of the most critical bases in the motif
were predicted by the protein structure, giving added weight to the predicted motif. Despite
identifying the same motif in the two cell lines used, there was a high degree of cell-type
specificity in the genomic locations of the P8ORs and PAX8 DEGs. The overlap was between
13%-22%, similar to the 26% of PAX8 binding sites in HGSOC cell lines that intersected the
P8ORs in the FTSEC cell lines
37
. As seen with many development-related oncogenes, several
developmental pathways were found: anatomical structural morphogenesis was common to both
cell lines, while tissue morphogenesis, positive regulation of programmed cell death,
mesenchymal cell development, TAP63 signaling, Notch signaling, and epithelial cell
differentiation were cell-type specific. Supporting the in vitro results, proliferation related
pathways were identified: epithelial cell proliferation was found in IGROV1 cells, while DNA
replication, G2M checkpoint, mitotic cell cycle process, DNA-directed DNA polymerase
activity, regeneration, and KRAS signaling were found in either cell line. Interestingly, 8
immune related pathways were found (cellular response to LPS, IKK/NFKB signaling, TNFA
signaling via NFKB, complement system, TLR2 signaling, phagocytosis, IL1A production, NLR
signaling), two of which were shared (cellular response to LPS and TNFA signaling via NFKB)
between both cell lines. The immune system was also found to be associated with PAX8 in
another study
42
.
In conclusion, in EOC PAX8 is involved in cell proliferation and anchorage independent
and dependent growth of epithelial ovarian cancer cells in vitro, and suppresses tumor growth in
vivo. This study establishes a genome-wide map of PAX8 direct and indirect regulatory targets
for the first time in epithelial ovarian cancer cells. While it is clear that the PAX8 cistrome
80
exhibits a high degree of cell-type specificity, analyses of cellular pathways and cofactors
converged on common molecular targets and partners that provide a better understanding of
epithelial ovarian cancer etiology and may represent important and much needed therapeutic
targets for EOC.
EXPERIMENTAL AND ANALYTICAL CONTRIBUTIONS
Tissue Culture: Emily Adler
Generation of PAX8 knockdown cells: Emily Adler
Immunohistochemistry and scoring: Heidi Sowter and the University of Derby
RT-qPCR: Emily Adler
In vitro phenotypic assays:
Anchorage independent growth assay: Emily Adler
Proliferation assay: Emily Adler
In vivo tumorigenicity assays:
Intraperitoneal injection: Emily Adler and Nathan Lee
Tumor measurement and tissue embedding: Emily Adler, Kate Lawrenson, and Nathan
Lee
Tissue sectioning: Emily Adler
Tissue staining: USC Pathology Lab
Tumor grading: Paulette Mhawech-Facueglia
Microarray analysis:
Cell culture and RNA isolation: Emily Adler
81
Microarray processing: USC Epigenome Center and UCLA Neuroscience Genomics
Core
Microarray analysis: Emily Adler
Microarray validation RT-qPCR: Norma Rodriguez-Malave
Chromatin immunoprecipitation: Emily Adler
Chromatin immunoprecipitation library preparation and sequencing: USC Epigenome Center
Chromatin immunoprecipitation sequencing analysis: R. Ivetth Corona de la Fuente and Dennis
Hazelett
Pathway analysis: R. Ivetth Corona de la Fuente and Dennis Hazelett
REFERENCES
1. Mittag J, Oehr W, Heuer H, et al. Expression and thyroid hormone regulation of annexins
in the anterior pituitary. J Endocrinol. Dec 2007;195(3):385-392.
2. Laury AR, Hornick JL, Perets R, et al. PAX8 reliably distinguishes ovarian serous tumors
from malignant mesothelioma. Am J Surg Pathol. May 2010;34(5):627-635.
3. Mhawech-Fauceglia P, Wang D, Samrao D, et al. Pair Box 8 (PAX8) protein expression
in high grade, late stage (stages III and IV) ovarian serous carcinoma. Gynecol Oncol.
Oct 2012;127(1):198-201.
4. Perets R, Wyant GA, Muto KW, et al. Transformation of the fallopian tube secretory
epithelium leads to high-grade serous ovarian cancer in brca;tp53;pten models. Cancer
Cell. Dec 2013;24(6):751-765.
82
5. Plachov D, Chowdhury K, Walther C, et al. Pax8, a murine paired box gene expressed in
the developing excretory system and thyroid gland. Development. Oct 1990;110(2):643-
651.
6. Pasca di Magliano M, Di Lauro R, Zannini M. Pax8 has a key role in thyroid cell
differentiation. Proc Natl Acad Sci U S A. 2000;97(24):13144-13149.
7. Zannini M, Francis-Lang H, Plachov D, et al. Pax-8, a paired domain-containing protein,
binds to a sequence overlapping the recognition site of a homeodomain and activates
transcription from two thyroid-specific promoters. Mol Cell Biol. Sep 1992;12(9):4230-
4241.
8. Laury AR, Perets R, Piao H, et al. A comprehensive analysis of PAX8 expression in
human epithelial tumors. Am J Surg Pathol. Jun 2011;35(6):816-826.
9. Narlis M, Grote D, Gaitan Y, et al. Pax2 and pax8 regulate branching morphogenesis and
nephron differentiation in the developing kidney. J Am Soc Nephrol. Apr
2007;18(4):1121-1129.
10. Grote D, Souabni A, Busslinger M, et al. Pax 2/8-regulated Gata 3 expression is
necessary for morphogenesis and guidance of the nephric duct in the developing kidney.
Development. Jan 2006;133(1):53-61.
11. Imberti B, Tomasoni S, Ciampi O, et al. Renal progenitors derived from human iPSCs
engraft and restore function in a mouse model of acute kidney injury. Sci Rep.
2015;5:8826.
12. Hu Y, Hartmann A, Stoehr C, et al. PAX8 is expressed in the majority of renal epithelial
neoplasms: an immunohistochemical study of 223 cases using a mouse monoclonal
antibody. J Clin Pathol. 2012;65(3):254-256.
83
13. Halt K, Vainio S. Coordination of kidney organogenesis by Wnt signaling. Pediatr
Nephrol. Apr 2014;29(4):737-744.
14. Clayton EF, Ziober A, Yao Y, et al. Malignant tumors with clear cell morphology: a
comparative immunohistochemical study with renal cell carcinoma antibody, Pax8,
steroidogenic factor 1, and brachyury. Ann Diagn Pathol. 2017;17(2):192-197.
15. Bowen NJ, Logani S, Dickerson EB, et al. Emerging roles for PAX8 in ovarian cancer
and endosalpingeal development. Gynecol Oncol. Feb 2007;104(2):331-337.
16. Ozcan A, Shen SS, Hamilton C, et al. PAX 8 expression in non-neoplastic tissues,
primary tumors, and metastatic tumors: a comprehensive immunohistochemical study.
Mod Pathol. Jun 2011;24(6):751-764.
17. Adler E, Mhawech-Fauceglia P, Gayther SA, et al. PAX8 expression in ovarian surface
epithelial cells. Hum Pathol. Jul 2015;46(7):948-956.
18. Ates Ozdemir D, Usubutun A. PAX2, PAX8 and CDX2 Expression in Metastatic
Mucinous, Primary Ovarian Mucinous and Seromucinous Tumors and Review of the
Literature. Pathol Oncol Res. 2016;22(3):593-599.
19. Cheung HW, Cowley GS, Weir BA, et al. Systematic investigation of genetic
vulnerabilities across cancer cell lines reveals lineage-specific dependencies in ovarian
cancer. Proc Natl Acad Sci U S A. Jul 2011;108(30):12372-12377.
20. Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian
carcinoma. Nature. Jun 2011;474(7353):609-615.
21. Palma TD, Filippone MG, Pierantoni GM, et al. Pax8 has a critical role in epithelial cell
survival and proliferation. Cell Death & Disease. 2013-07-01 2013;4(7):e729.
84
22. Magliano MP, Lauro RD, Zannini M. Pax8 has a key role in thyroid cell
differentiation.Proc Natl Acad Sci U S A. 2000;97(24):13144-13149.
23. Schmidt D, Wilson MD, Spyrou C, et al. ChIP-seq: using high-throughput sequencing to
discover protein-DNA interactions. Methods. Jul 2009;48(3):240-248.
24. Zhang Y, Liu T, Meyer CA, et al. Model-based analysis of ChIP-Seq (MACS). Genome
Biol. 2008;9(9):R137.
25. Li QH, Brown JB, Huang HY, et al. Measuring Reproducibility of High-Throughput
Experiments. Annals of Applied Statistics. Sep 2011;5(3):1752-1779.
26. Li H, Durbin R. Fast and accurate long-read alignment with Burrows-Wheeler transform.
Bioinformatics. Mar 2010;26(5):589-595.
27. Kharchenko PV, Tolstorukov MY, Park PJ. Design and analysis of ChIP-seq experiments
for DNA-binding proteins. Nat Biotechnol. Dec 2008;26(12):1351-1359.
28. Heinz S, Benner C, Spann N, et al. Simple combinations of lineage-determining
transcription factors prime cis-regulatory elements required for macrophage and B cell
identities. Mol Cell. May 2010;38(4):576-589.
29. Machanick P, Bailey TL. MEME-ChIP: motif analysis of large DNA datasets.
Bioinformatics. Jun 2011;27(12):1696-1697.
30. Dixon JR, Selvaraj S, Yue F, et al. Topological domains in mammalian genomes
identified by analysis of chromatin interactions. Nature. Apr 2012;485(7398):376-380.
31. Tripathi S, Pohl MO, Zhou Y, et al. Meta- and Orthogonal Integration of Influenza
"OMICs" Data Defines a Role for UBR4 in Virus Budding. Cell Host Microbe. Dec
2015;18(6):723-735.
85
32. Buick RN, Pullano R, Trent JM. Comparative properties of five human ovarian
adenocarcinoma cell lines. Cancer Res. Aug 1985;45(8):3668-3676.
33. Domcke S, Sinha R, Levine DA, et al. Evaluating cell lines as tumour models by
comparison of genomic profiles. Nat Commun 2013;4:2126.
34. Benard J, Da Silva J, De Blois MC, et al. Characterization of a human ovarian
adenocarcinoma line, IGROV1, in tissue culture and in nude mice. Cancer Res. Oct
1985;45(10):4970-4979.
35. Hnisz D, Abraham BJ, Lee TI, et al. Super-enhancers in the control of cell identity and
disease. Cell. Nov 2013;155(4):934-947.
36. Stormo GD. Modeling the specificity of protein-DNA interactions. Quant Biol. Jun
2013;1(2):115-130.
37. Elias KM, Emori MM, Westerling T, et al. Epigenetic remodeling regulates
transcriptional changes between ovarian cancer and benign precursors. JCI Insight. Aug
2016;1(13).
38. Pomerantz MM, Li F, Takeda DY, et al. The androgen receptor cistrome is extensively
reprogrammed in human prostate tumorigenesis. Nat Genet. Nov 2015;47(11):1346-
1351.
39. Ewing CM, Ray AM, Lange EM, et al. Germline mutations in HOXB13 and prostate-
cancer risk. N Engl J Med. Jan 12 2012;366(2):141-149.
40. Wiegand KC, Shah SP, Al-Agha OM, et al. ARID1A mutations in endometriosis-
associated ovarian carcinomas. N Engl J Med. Oct 2010;363(16):1532-1543.
41. Di Palma T, Lucci V, de Cristofaro T, et al. A role for PAX8 in the tumorigenic
phenotype of ovarian cancer cells. BMC Cancer. 2014;14(1):292.
86
42. Ruiz-Llorente S, Carrillo Santa de Pau E, et al. Genome-wide analysis of Pax8 binding
provides new insights into thyroid functions. BMC Genomics.2012;13:147.
43. Elias KM, Labidi-Galy SI, Vitonis AF, et al. Prior appendectomy does not protect against
subsequent development of malignant or borderline mucinous ovarian neoplasms.
Gynecol Oncol. 2013;132(2):328-333.
44. Hanieh H, Mohafez O, Hairul-Islam VI, et al. Novel Aryl Hydrocarbon Receptor Agonist
Suppresses Migration and Invasion of Breast Cancer Cells. PLoS One.
2016;11(12):e0167650.
45. Huang S, Guo Y, Jacobi A, et al. Aromatic Hydrocarbon Receptor Suppresses Prostate
Cancer Bone Metastasis Cells-Induced Vasculogenesis of Endothelial Progenitor Cells
under Hypoxia. Cell Physiol Biochem. 2016;39(2):709-720.
46. Kimura Y, Kasamatsu A, Nakashima D, et al. ARNT2 Regulates Tumoral Growth in
Oral Squamous Cell Carcinoma. J Cancer. 2016;7(6):702-710.
47. Hall JM. The Aryl-hydrocarbon Receptor (AhR) as a Therapeutic Target in Human
Breast Cancer. Journal of Steroids & Hormonal Science. 2014;5(3).
48. Powell JB, Goode GD, Eltom SE. The Aryl Hydrocarbon Receptor: A Target for Breast
Cancer Therapy. J Cancer Ther. Sep 2013;4(7):1177-1186.
49. Shaulian E, Karin M. AP-1 as a regulator of cell life and death. Nature Cell Biology.
2002-05-01 2002;4(5).
50. Hu YF, Li R. JunB potentiates function of BRCA1 activation domain 1 (AD1) through a
coiled-coil-mediated interaction. Genes Dev. Jun 2002;16(12):1509-1517.
87
51. Xu HE, Rould MA, Xu W, et al. Crystal structure of the human Pax6 paired domain–
DNA complex reveals specific roles for the linker region and carboxy-terminal
subdomain in DNA binding. Genes Dev. May 1999;13(10):1263-1275.
52. Xu W, Rould MA, Jun S, et al. Crystal structure of a paired domain-DNA complex at 2.5
A resolution reveals structural basis for Pax developmental mutations. Cell.
1995;(80)4:639-650.
88
CHAPTER 4: CONCLUSIONS ON THE ROLE OF PAX8
IN EPITHELIAL OVARIAN CARCINOMA
PAX8 is a master transcription factor regulator of the thyroid and contributes
oncogenically to thyroid and kidney carcinomas. While its normal function in the cell(s) of
origin in HGSOC and the other EOC subtypes is still under investigation, I have shown that
PAX8 is expressed in the normal ovarian surface epithelium, along with the established normal
fallopian tube epithelium. The widespread expression of PAX8 in the Müllerian tract and its
overexpression in EOC led me to hypothesize that overexpression was both an early event in
EOC and contributed to oncogenesis.
To investigate if PAX8 could drive tumorigenicity in ovarian surface epithelial cells, I
exogenously expressed PAX8 in hTERT immortalized OSE cells that were partially transformed
by overexpression of CMYC. PAX8 expression did not drive tumorigenicity in IOE19
CMYC.PAX8
cells. While the expression of PAX8 in IOE19
CMYC.PAX8
cells does not indicate PAX8 that PAX8
contributes to or is an early event in oncogenesis in OSECs, the possibility cannot be ruled out.
PAX8 may require different or additional carcinogenic events to promote HGSOC.
In Chapter 3 I showed that PAX8 overexpression was associated with characteristics of
cancer cells - accelerated population doubling time, anchorage independent growth, and
tumorigenicity. This supports that PAX8 contributes to oncogenesis. Interestingly, the level of
PAX8 expression was not associated with the size of effect on oncogenic phenotypes, suggesting
there is not a dose-dependent effect of PAX8 expression.
89
Using ChIP-seq, 22% of PAX8 binding sites were common in HEYA8 and IGROV1 cell
lines. PAX8 binding in EOC cells was found to be frequently enriched in promoters in both cell
lines, and enriched in putative enhancer regions and superenhancers in HEYA8. The PAX8 motif
was similar in the two independent EOC cell lines, which corresponded with a PAX8 DNA-
binding homology model. In these cell lines PAX8 is associated with the differential expression
of >1000 genes, of which only 16% were shared. Outside of PAX motifs, only 25% of motifs
near PAX8 summits were shared. Additionally, while about half of the TADs in each cell line
contained P8ORs, only one third of them were shared.
9 of the top 20 cellular pathways identified were shared by HEYA8 and IGROV1 cells. 8
were immune related, 2 of which were common to both cell lines. Unsurprisingly, 7 pathways
were related to development, of which 1 was shared by HEYA8 and IGROV1; and in support of
the finding that PAX8 knockdown was associated with a substantial decrease in cellular
proliferation, 7 pathways were proliferation-related, of which only 1 was shared. The small
overlap in DEGs and PAX8 binding sites but related set of pathways altered suggests that PAX8
has a highly cell-type specific regulatory network that focuses on a similar set of pathways
involved in several cellular processes.
Cell type specificity was a reoccurring finding when investigating PAX8 occupancy and
differential gene expression between the two independent EOC cell lines. The frequent
differences between them suggest that the cell-type specificity common in the normal function of
PAX8 during development and in adulthood to continues in EOC. The cellular
microenvironment, genetic background, and availability of transcriptional co-regulators may
90
play a large role in this specificity. This is a possible explanation of why PAX8 expression in
IOE19
CMYC.PAX8
cells did not show increased transformation.
It is possible that the PAX8 DNA binding motif found in Ch3 in EOC cells may succumb
to promiscuity compared to its binding in normal cells. Just as a threshold of PAX8 expression
may not have been met in IOE19
CMYC.PAX8
cells to promote cellular transformation, increasing
levels in EOC may allow PAX8 to bind more sites throughout the genome due to its biochemical
abundance. Neither chapter refutes this threshold hypothesis.
In order to investigate this further, the ability of EOC cells to survive without PAX8
could be assessed by knocking it out using the CRISPR (Clustered
regularly
interspaced
short
palindromic
repeats) system. Alternatively, using siRNA against PAX8 could show at what
level EOC cells could survive without PAX8. In long term studies cell lines could be established
with stepwise levels of PAX8 to test if there is a particular threshold amount that is associated
with the phenotypic effects observed in this study. Rescue experiments in which PAX8 is added
back to cells could support that PAX8 is the driver of these effects.
Since PAX8 is overexpressed in the major histotypes of EOC, and as seen in this study
there can be a high amount of cell-type specificity between EOC cells, it would be valuable to
study the effects of PAX8 in each histotype. To decrease the likelihood of the influence of
different genetic backgrounds, several different cell lines should be used for each histotype.
PAX8 appears to act as an oncogene in cells that rely on it during development and for
their maintenance of differentiation, such as the EOC cell lines characterized here. It is unknown
if PAX8 could act as an oncogene and/or transform other cell types. It would be interesting to
overexpress PAX8 in various epithelial cell types susceptible to transformation, such as luminal
91
cells of the lung, colon, and breast. As PAX8 may require a particular genetic background or
microenvironment to transform cells, it would be prudent to test its effects in normal and
partially transformed cells.
As was shown in chapter 3, PAX8 appears to regulate gene expression more often at
enhancer elements rather than promoter elements. To test this, CRISPR could be used to knock
out the functional sequence of putative enhancers, followed by measuring the expression level of
each putative enhancer’s correlated DEG. It also appears that PAX8 may act with cooperating
transcription factors. ChIP-seq for these factors could be overlaid with PAX8 binding sites to test
this finding.
Understanding the transformation of cancer cells of origin into early and late stage
disease is critical to screening, prevention, and treatment in the clinic. Despite widespread
overexpression of PAX8 in EOC, there is a lack of basic understanding of whether and how it
contributes to oncogenesis. In this study I have shown that PAX8 is associated with
tumorigenesis in EOC. I have also shown that PAX8 is expressed in the ovarian surface
epithelium. These data are important in consideration of current treatment of EOC. In the early
stages this knowledge can help guide current clinical decisions made in regard to how to prevent
and/or cure the disease by the current practice of removal of the ovaries, fallopian tubes, and
uterus. This has a substantial effect on quality of life, especially in patients diagnosed during
fertile years. In addition, the data shown in Chapter 3 could help guide targeted therapy design
on a molecular level to treat patients with several subtypes of EOC. In sum this study contributes
to the much needed advancement of the clinical treatment of this deadly and poorly understood
disease.
92
PUBLICATIONS BY AUTHOR
1. Adler, E, Mhawech-Fauceglia, P, Gayther, SA, et al. PAX8 expression in ovarian
surface epithelial cells. Hum Pathol. 2015;46(7):948-956.
2. Kar, SP, Adler, E, Tyrer, J, et al. Enrichment of putative PAX8 target genes at serous
epithelial ovarian cancer susceptibility loci. Br J Cancer. 2017;116(4):524-535.
93
APPENDIX
Table 1. Gene expression probes
Gene Probe Notes
PAX8 Hs01015257_g1
ACTB Hs01060665_g1
GAPDH Hs02758991_g1
AJAP1 Hs00982497_m1
ANKRD1 Hs00173317_m1
HIST2H2 Hs00358508_s1 Detects histone cluster 2
PRSS23 Hs00970839_s1
STC1 Hs00174970_m1
VASN Hs01936449_s1
Table 2. Cross correlation QC metrics
Cell line Factor Replicate NSC RSC Qtag
HEYA8 H3K27ac Rep1 1.04 1.04 1
HEYA8 H3K27ac Rep2 1.13 1.22 1
HEYA8 PAX8 Rep1 1.02 1.01 1
HEYA8 PAX8 Rep2 1.02 0.95 0
IGROV1 H3K27ac Rep1 1.04 0.85 0
IGROV1 H3K27ac Rep2 1.09 1.06 1
IGROV1 PAX8 Rep1 1.04 0.96 0
IGROV1 PAX8 Rep2 1.02 0.63 0
94
Table 3. Validation of PAX8 target genes
Microarray RT-qPCR
HeyA8 IGROV1 HeyA8 IGROV1
Gene Name
FC P-value FC P-value FC P-value FC P-value
AJAP1
-1.9 3.82E-10 -1.8 2.17E-09 -30.1 0.04
-16.7 0.02
ANKRD1
1.6 2.02E-03 10.8 4.06E-09
2.5
0.08 21.8
0.05
HIST2H2AA3
2.0 8.41E-07 1.7 2.24E-06
HIST2H2AA4
1.9 1.79E-06 1.6 8.88E-07
HIST2H2AC
1.6 1.35E-06 1.5 3.62E-04
-1.2 0.28 -1.4 0.22
PRSS23
1.5 2.10E-04 3.9 6.86E-10 1.6 0.04 5.7 0.04
STC1
2.7 2.51E-05 1.8 2.18E-05 6.0 0.13 1.6 0.20
VASN
1.7 4.02E-07 6.0 5.38E-11 1.5 0.19 4.7 0.07
Genes identified in the microarray analyses were independently validated by RT-qPCR. All
except histone H2A subunits were validated. Two-tailed paired Student's T-test.
Abstract (if available)
Abstract
Cancer that arises in ovarian tissue is termed ovarian cancer. The majority (>90%) of these are of an epithelial nature, and are thus termed epithelial ovarian carcinomas (EOCs). EOC is the most lethal gynecological malignancy in the United States and the western world, and is the most lethal cancer of the reproductive system. Historically the primary tumor has been difficult to distinguish from the rest of the tumor mass because patients were and still are most often diagnosed after EOC has spread throughout the abdominal cavity. At this stage it is impossible to distinguish from where the tumor has originated. However because the ovaries are involved so pervasively and usually comprise the bulk of tumor mass, such cancers are known as ovarian carcinomas. ❧ This thesis discusses the role of PAX8 in ovarian cancer.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Integrative genomic and epigenomic analysis of human cancer
PDF
Insights from mouse models on the origin of ovarian cancer and its predisposing mechanisms
PDF
Investigating novel avenues for effective treatment of ovarian cancer
PDF
Understanding the role of BRCA1 and Aurora A kinase in polyploidy development in ovarian carcinoma precursor cells
PDF
Functional analysis of a prostate cancer risk enhancer at 7p15.2
PDF
The role of glucose-regulated proteins in endometrial and pancreatic cancers
PDF
Investigating the role of SASH1 gene located on chromosome 6 in ovarian cancer
PDF
Functional role of chromatin remodeler proteins in cancer biology
PDF
Mitotic regulation in ovarian epithelial tumors approaching in vitro crisis
PDF
Mechanism of secretion and function of heat shock protein-90 (Hsp90) family genes
PDF
POLR2B and its contributions to cancer cell growth
PDF
Lymphatic cell environment promotes sustained KSHV lytic replication and viral maintenance
PDF
The cancer stem-like phenotype: therapeutics, phenotypic plasticity and mechanistic studies
PDF
Targeting tumor-initiating stem-like cells through metabolic and epigenetic regulations
PDF
Effects of chromatin regulators during carcinogenesis
PDF
Targeting glioma cancer stem cells for the treatment of glioblastoma multiforme
PDF
Predicting bladder cancer behavior by molecular expression profiling
PDF
The effect of tumor-mediated immune suppression on prostate cancer immunotherapy
PDF
Understanding and controlling mitotic errors leading to aneuploidy in early ovarian cancer development
PDF
The role of DNA methylation in early detection and progression of pancreatic cancer
Asset Metadata
Creator
Adler, Emily Kate
(author)
Core Title
The role of PAX8 in epithelial ovarian carcinoma
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Genetic, Molecular and Cellular Biology
Publication Date
10/16/2017
Defense Date
07/12/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
OAI-PMH Harvest,ovarian cancer,ovarian carcinoma,PAX8
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Stallcup, Michael (
committee chair
), Cozen, Wendy (
committee member
), Farnham, Peggy (
committee member
)
Creator Email
ea0908@gmail.com,emilykadler@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-447154
Unique identifier
UC11263159
Identifier
etd-AdlerEmily-5839.pdf (filename),usctheses-c40-447154 (legacy record id)
Legacy Identifier
etd-AdlerEmily-5839.pdf
Dmrecord
447154
Document Type
Dissertation
Rights
Adler, Emily Kate
Type
texts
Source
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 a...
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
Tags
ovarian cancer
ovarian carcinoma
PAX8