Yale University
EliScholar – A Digital Platform for Scholarly Publishing at Yale
Yale Medicine Thesis Digital Library
School of Medicine
January 2019
Development Of Pancreatic Cancer Organoid
Model For Studying Immune Response In
Pancreatic Cancer
Jin Woo Yoo
Follow this and additional works at: https://elischolar.library.yale.edu/ymtdl
This Open Access Thesis is brought to you for free and open access by the School of Medicine at EliScholar – A Digital Platform for Scholarly
Publishing at Yale. It has been accepted for inclusion in Yale Medicine Thesis Digital Library by an authorized administrator of EliScholar – A Digital
Platform for Scholarly Publishing at Yale. For more information, please contact elischolar@yale.edu.
Recommended Citation
Yoo, Jin Woo, “Development Of Pancreatic Cancer Organoid Model For Studying Immune Response In Pancreatic Cancer” (2019).
Yale Medicine Thesis Digital Library. 3543.
https://elischolar.library.yale.edu/ymtdl/3543
Development of Pancreatic Cancer Organoid Models for
Studying Immune Response in Pancreatic Cancer
A Thesis Submitted to the
Yale University School of Medicine
in Partial Fulfillment of the Requirements for the
Degree of Doctor of Medicine
by
Jin Woo Yoo
2019
DEVELOPMENT OF PANCREATIC CANCER ORGANOID MODEL FOR STUDYING
IMMUNE RESPONSE IN PANCREATIC CANCER. Jin Woo Yoo, Prashanth R. Gokare,
Yevgeniya Foster, Brittany Fitzgerald, Nikhil S. Joshi, James J. Farrell. Section of
Gastroenterology, Department of Internal Medicine, Yale University, School of Medicine, New
Haven, CT.
The importance of immune system in pancreatic ductal adenocarcinoma (PDAC)
pathogenesis and therapy remains poorly understood largely due to the lack of effective model
systems. Cell lines are not physiologic as they cannot recapitulate the cancer stroma and lose
genetic heterogeneity over time. Genetically engineered mouse models of PDAC are more
physiologic than cell lines but lack neoantigens needed to mount T cell responses against tumor.
Organoid models of PDAC offer unique opportunity to study immune mechanisms in PDAC
since organoids can model complex layering of multiple cell types, creating a physiologically
relevant system that is highly tractable for genetic manipulation, co-cultures, and high
throughput assays. In this study, we sought to establish murine and human organoid models of
PDAC to investigate the biology of PDAC immune response, with the specific aims of
developing transplantable immunogenic murine PDAC organoid models for the study of antigen-
specific anti-tumor T cell responses and assembling a library of experimentally validated,
patient-derived PDAC organoid lines for pancreatic cancer precision medicine research.
To generate immunogenic murine organoid models of PDAC, pancreatic organoids were
isolated from “KP-NINJA” (KrasLox-STOP-Lox-G12D; P53flox/flox; inversion induced joined
neoantigen) mouse model that has been genetically engineered to express GFP-tagged T cell
neoantigens derived from lymphocytic choriomeningitis virus in an inducible fashion. Isolated
organoids were transformed in vitro using a lentiviral construct encoding Cre recombinase and
RFP reporter for expression of oncogenic KRAS and deletion of P53. A subset of transformed
organoids was additionally treated with an adenoviral construct encoding FLPo recombinase to
turn on neoantigen expression. Transformed organoids were combined with T cells in both in
vivo and in vitro setting to assess for impact on tumor growth. Patient-derived PDAC organoids
were generated using endoscopic ultrasound-guided fine needle biopsy (EUS-FNB) specimens,
surgical resection specimens, and tissues from patient-derived xenograft mouse models of
PDAC. Established human organoid lines were validated by Sanger sequencing, tumor formation
in vivo and immunohistochemistry of organoid-derived tumors.
Subcutaneous injection of transformed murine PDAC organoids formed tumors in mouse
that are histologically similar to early lesions found in human PDAC. Serial in vivo transfer of
these organoids by performing sequential rounds of organoid generation from tumors derived
from organoids formed progressively more advanced tumors. High level of neoantigen
expression in 100% of cells comprising murine PDAC organoids resulted in rejection of tumor
growth in mouse, while a low level of neoantigen expression restricted to 10% of cells permitted
tumor growth with increased immune infiltration. Expression of neoantigens in T cell-PDAC
organoid co-culture model systems promoted T cell infiltration of basement membrane matrix.
Additionally, we generated 30+ patient-derived PDAC organoid lines using EUS-FNB and
surgical specimens at Yale from 10/2017 to 5/2018.
We have successfully established murine and human organoid models of PDAC from
various tissues capturing discrete stages of PDAC progression. Our murine organoid models are
uniquely equipped to study antigen-specific T cell responses against tumor. Ongoing work
includes using CRISPR/Cas9-based lentiviral systems to define genes that impact anti-tumor T
cell responses and using patient-derived organoids for precision medicine research.
ACKNOWLEDGEMENTS
Work for this thesis was completed in the Joshi laboratory under the co-mentorship of
James J. Farrell, MD and Nikhil S. Joshi, PhD. Both Dr. Farrell and Dr. Joshi suggested
experiments and supervised the work done. Dr. Joshi developed the KP-NINJA mouse model
that was fundamental for the creation of immunogenic murine PDAC organoid models. Dr.
Farrell performed and provided all the endoscopic ultrasound-guided fine-needle biopsies for the
creation of patient-derived PDAC organoid lines. Prashanth Gokare, PhD collaborated with the
author on the development of three-dimensional co-culture system for murine pancreatic cancer
organoids and T cells and the creation of patient-derived PDAC organoids from surgical
resection specimens and their sequencing. Yevgeniya Foster, MD collaborated with the author on
creation of immunogenic murine PDAC organoid lines for characterizing immune responses in
vivo and immunohistochemical analysis of murine organoid-derived tumors. Brittany Fitzgerald
established the primary murine pancreatic cancer cell lines from KP-C mouse and collaborated
with the author on in vivo transfer of P14 mouse splenocytes and in vivo imaging for luciferase
detection. Marie Robert, MD provided surgical resection specimens for creation of patient-
derived pancreatic cancer organoids and interpretation of tumor histology. Ryan Sowell, PhD
from Kaech laboratory created the patient-derived xenograft mouse models, some of which were
used as source material for the creation of human PDAC organoids. All other experiments were
performed independently by the author. Dr. Farrell and Dr. Joshi reviewed and provided
comments on the manuscript. National Institute of Health-National Institute of Diabetes and
Digestive and Kidney Diseases Medical Student Research Fellowship (T35 grant), Yale
University School of Medicine Research Fellowship, and Richard Alan HirshField Memorial
Fellowship provided funding to support this work.
TABLE OF CONTENTS
LIST OF ABBREVIATIONS………………………………………………………………………………………….1
INTRODUCTION………………………………………………………………………………………………………….2
• Background
• Cell of Origin
• Genetic Landscape of Pancreatic Cancer
• Precursor Lesions
• Mutational Processes
• Tumoral Heterogeneity
• Molecular Subtyping of Pancreatic Cancer
• Deranged Signaling Pathways / Molecular Aberrations
• Tumor Microenvironment
• Metabolic Reprogramming
• Immune Response in Pancreatic Cancer is Unclear
• Pre-clinical Modeling of Pancreatic Cancer
STATEMENT OF PURPOSE……………………………………………………………………………………….18
METHODS…………………………………………………………………………………………………………………..19
• Acquisition of human specimens
• Isolation and culture of murine pancreatic organoids
• Isolation and culture of human PDAC organoids
• Isolation of primary murine PDAC cell lines
• Genetic manipulation of murine pancreatic organoids
• In vivo mouse assays
• Immunohistochemical analysis of tumors
• Sanger sequencing of organoids
• Development of organoid-T cell co-culture model systems
RESULTS…………………………………………………………………………………………………………………….26
• KP-NINJA mouse model provides substrate for creation of immunogenic murine
organoid models of PDAC
• In vitro transformed murine pancreatic organoids form tumors that are histologically
similar to early lesions found in human PDAC
• Serial in vivo transfer of transformed murine pancreatic organoids results in
progressively more advanced tumors
• Expression of neoantigens in murine PDAC organoids elicits effective immune response
in mouse
• Expression of neoantigens in murine PDAC organoids promotes T cell infiltration in T
cell-organoid co-culture model
• Assembly of human PDAC organoid library
1
DISCUSSION……………………………………………………………………………………………………………….33
REFERENCES……………………………………………………………………………………………………………..36
FIGURES……………………………………………………………………………………………………………………..38
TABLES……………………………………………………………………………………………………………………….48
LIST OF ABBREVIATIONS
CTGF
Connective tissue growth factor
EGF
Epidermal growth factor
ER
Estrogen receptor
ETC
Electron transport chain
EUS-FNB
Endoscopic ultrasound-guided fine needle
biopsy
FGF
Fibroblast growth factor
FLP
Flippase
FRT
Flippase recognition target
GFP
Green fluorescent protein
GM-CSF
Granulocyte-macrophage colony-stimulating
factor
GP
Glycoprotein
hENT1
Human equilibrative nucleoside transporter
HGF
Hepatocyte growth factor
HIF1α
Hypoxia-inducible transcription factor 1α
HR
Homologous recombination
IFN-γ
Interferon-γ
IGF1
Insulin-like growth factor 1
IHC
Immunohistochemistry
IL-1
Interleukin-1
IPMN
Intraductal papillary mucinous neoplasm
LCMV
Lymphocytic choriomeningitis virus
MCN
Mucinous cystic neoplasm
MDSC
Myeloid-derived suppressor cell
MMP
Matrix metalloproteinase
MMR
Mismatch repair
NF-κB
Nuclear factor-κB
NSG
NOD scid gamma
PanIN
Pancreatic intraepithelial neoplasm
PARP
Poly ADP-ribose polymerase
PDAC
Pancreatic ductal adenocarcinoma
PDGF
Platelet-derived growth factor
PDX
Patient-derived xenograft
PSC
Pancreatic stellate cell
2
RFP
Red fluorescent protein
rtTA
Reverse tetracycline-controlled transactivator
STAT3
Signal transducer and activator of transcription 3
TCA
Tricyclic acid
TCR
T cell receptor
TGFα
Transforming growth factor-α
TIMP
Tissue inhibitor of metalloproteinases
TNFα
Tumor necrosis factor-α
TRE
Tetracycline response element
TSLP
Thymic stromal lymphopoietin
VEGF
Vascular endothelial growth factor
I. INTRODUCTION
Background
Pancreatic ductal adenocarcinoma (PDAC; used interchangeably with pancreatic cancer
hereafter), the predominant form of pancreatic malignancy, is currently the fourth leading cause
of all cancer-related deaths in developed countries and is projected to become second only to
lung cancer by year 2024.(1) In 2015 worldwide, 367,000 patients were newly diagnosed with
pancreatic cancer, of whom 359,000 patients died due to pancreatic cancer-related causes within
the same year.(2) Although surgical resection is currently the only curative treatment for
pancreatic cancer, fewer than 20% of patients have resectable disease by the time their diagnosis
is made. The overall survival rate at 5 years is less than 7%, with most of the survivors at 5 years
belonging to the group of 10-20% of patients who undergo surgical resection of their tumors.(3)
Even for those patients undergoing surgery, 80% of them eventually relapse and die from
pancreatic cancer.
The exceptionally poor prognosis of pancreatic cancer can be attributed to several
factors.(2) First is its late diagnosis due to poor early detection, which is delayed by the absence
of clear or disease-specific symptoms and the lack of reliable biomarkers for effective screening.
Secondly, pancreatic cancer takes an aggressive course, with perineural and vascular invasions
3
and early distant metastases precluding a potentially curative surgical resection. Thirdly,
pancreatic cancer displays remarkable resistance to conventional modalities of cancer therapy,
including chemotherapy, radiotherapy as well as more recently developed molecularly targeted
therapies including immunotherapy. Finally, pancreatic cancer harbors complex tumor biology
with both intertumoral and intratumoral genetic heterogeneity, resulting in variable treatment
responses from patient to patient thus rendering a generalized approach to therapy difficult. A
comprehensive, mechanistic understanding of the pathophysiology underlying pancreatic cancer
is fundamental to overcoming these barriers.
Cell of Origin
The normal pancreas consists of two distinct functional components: endocrine and
exocrine. The endocrine component consists of glucagon-producing alpha cells and insulin-
producing beta cells that are anatomically organized into islets, and can give rise to a relatively
rarer form of pancreatic malignancies termed pancreatic neuroendocrine tumors, which have
been found to harbor mutational signatures clearly distinct from those of PDAC. These
signatures include inactivation of genes MEN1, ATRX and DAXX, derangements in the mTOR
signaling pathway, recurrent YY1 Thr372Arg missense mutations, and biallelic MUTYH
inactivating mutations.(4)
The exocrine component of the pancreas consists of digestive enzyme-secreting acinar
cells and bicarbonate-secreting ductal cells. Historically, ductal cells were thought to be the
unique source of PDAC, given their co-expression of epithelial markers, such as CK19. Recent
studies using genetically engineered mouse models of PDAC have shown that in fact both ductal
and acinar cells can give rise to PDAC precursor lesions by oncogenic KRAS activation.(4)
Furthermore, transient acinar-to-ductal metaplasia was observed in mouse models, with
4
reversible phenotypic and molecular changes that persisted in the presence of chronic
inflammation or oncogenic KRAS activation. Although there is also evidence for this
phenomenon in resected human PDAC surgical specimens, it has been argued that the
metaplastic lesions may be intraductal spread of pre-existing PDAC and/or its precursor lesions.
Genetic Landscape of Pancreatic Cancer
The genetic landscape of PDAC is characterized predominantly by mutations in four
major driver genes, listed in the order of decreasing frequency: KRAS, CDKN2A, SMAD4, and
TP53. Frequent alterations in these genes were first identified by candidate gene sequencing and
have since been corroborated repeatedly by multiple large exome and genomic sequencing
studies of PDAC.(5) Activating mutations of oncogene KRAS are seen in more than 90% of
PDACs, and inactivating mutations of tumor suppressor genes, CDKN2A, SMAD4 and TP53 in
50-80% of PDACs.(2) An additional 32 recurrent ‘passenger’ mutations – defined as those co-
occurring with driver mutations without conferring additional growth advantage – were also
identified, including but not limited to ARID1A, RNF43, TGFBR1, TGFBR2, MLL3, MKK4,
KDM6A, PREX2, RB1 and CCND1, at lower frequencies in approximately 10% of PDAC
tumors, highlighting the significance of tumoral heterogeneity (Table 1).(2, 4) It will be
important to fully characterize the functional significance of these passenger gene mutations as
they represent genetic differences among PDACs that may be exploited clinically.
Precursor Lesions
At least three histologically distinct precursor lesions of PDAC have been described so
far, consisting of pancreatic intraepithelial neoplasm (PanIN), and two types of mucinous cystic
lesions including intraductal papillary mucinous neoplasm (IPMN) and mucinous cystic
neoplasm (MCN). These precursor lesions are further characterized histologically and graded
5
according to their degree of dysplasia as lesions of low-grade versus high-grade dysplasia
(Figure 1).
Targeted sequencing of PanIN lesions along with their matched corresponding PDAC
surgical resection specimens demonstrated that the same four driver genes are mutated in PanIN
at very high frequencies as observed in PDAC. Comprehensive exome and whole genomic
sequencing studies also confirmed these findings, establishing PanIN as the canonical precursor
lesion of PDAC.(5) Similarly, shared mutations were also seen with mucinous cysts and their
matched corresponding PDACs. Targeted sequencing of IPMNs identified shared mutations in
genes GNAS and KRAS, and exome sequencing of IPMNs and MCNs identified shared
mutations in RNF43, indicating that cystic neoplasms represent additional precursor lesions of
PDAC that employ different progression pathways.(4)
Remarkably, mutational analysis comparing PanINs of different grades revealed a
positive correlation between the PanIN grade and the frequencies at which driver gene mutations
are found.(5) Furthermore, it revealed a sequential pattern in which mutations found to
accumulate in a predictive order following the PanIN grade. High-sensitivity methods to detect
KRAS mutations showed their involvement in more than 99% of all PanIN-1 lesions, suggesting
that oncogenic transformation of KRAS is most likely the initiating step in the development of
pancreatic cancer.(6) While KRAS mutations are found across all grades of PanINs and invasive
PDACs, the proportion of cells harboring the mutation increases with higher PanIN grade,
indicating a clonal expansion of cells carrying the mutation.(6) In addition to oncogenic KRAS,
inactivating mutations in CDKN2A can be seen in PanIN-2 and again at a higher frequency in
PanIN-3.(5) Similarly, SMAD4 and TP53 mutations are additionally found in PanIN-3 and in
6
invasive PDACs, with both SMAD4 and TP53 mutations occurring at higher frequencies in
invasive PDACs.
These findings may be explained by a linear progression model of pancreatic cancer
development, in which mutations are acquired in a gradual, step-wise pattern. By sequencing
primary PDACs and their matched metastatic tumors, it was estimated that the linear progression
from a nascent pancreatic cell acquiring an initiating driver gene mutation to the ultimate
development of invasive PDAC would take 10 or more years.(7) This notion is consistent with
the observation that nearly 33% of pancreata seen in autopsy series contain PanINs, suggesting
that PanINs are quite common and generally do not progress to an invasive cancer.(6) In
contrast, an alternative model termed chromothripsis proposes a punctuated evolution of
pancreatic cancer, in which catastrophic genomic events involving structural alterations cause
simultaneous inactivation of multiple driver genes in a single cell cycle. In support of this model,
whole genome sequencing of primary tumors demonstrated two-thirds of PDACs having
complex structural variations that, in a subset of cases, simultaneously inactivated multiple driver
genes.(8) In the same study, many tumors did not harbor the predicted sequence of mutations,
suggesting that these mutations may be acquired in a stochastic fashion consistent with a
chromothripsis model. Still, a third model that combines both linear progression and punctuated
evolution is entirely plausible. Distinguishing among these mechanistically distinct yet mutually
non-exclusive models has clinical importance, since under a linear progression model which
predicts a slow and gradual progression of disease, clinical efforts are best geared toward
improving methods for screening and early detection of pancreatic cancer, whereas under a
punctuated evolution model, an emphasis on enhancing systemic therapy is more appropriate.
7
Mutational Processes
To fully understand the pathophysiology of PDAC, it is essential to delineate the
mutational processes that are operative in the development of PDAC. Framing pancreatic cancer
in familiar evolutionary terms can facilitate a mechanistic understanding of how mutations arise
in the first place. In Darwinian evolution, mutations occur purely stochastically in dividing cells
at an expected somatic mutation rate of three single nucleotide variants per cell division.(6) In
the case of the pancreas which does not comprise of highly proliferative tissues, the probability
of a pancreatic cell acquiring an initiating driver gene mutation by random chance alone is
exceedingly low, and can be expected to largely depend on the total number of cell divisions
performed over the lifetime of the dividing cell. Not surprisingly, statistical analysis of various
types of human cancers, including PDAC, revealed a strong correlation between lifetime cancer
risk and the number of cell divisions performed by adult stem cells of a given organ.(9) This
finding lends support to the well-established finding that patient age is a major risk factor for the
development of PDAC. Indeed, most pancreatic cancer patients are diagnosed at beyond age 50,
with peak incidence occurring in the seventh and eighth decades of life.(2) However, the relative
contribution of intrinsic factors (e.g. stochastic mistakes taking place during DNA replication)
versus extrinsic factors (e.g. patient exposure to carcinogens or radiation) to lifetime risk remains
a point of contentious debate.
By whole-genome and RNA sequencing of resected PDAC surgical specimens, Connor
et. al identified four distinct mutational processes acting on the PDAC genome.(4, 10) Those
related to increasing age and number of cell divisions were the most prevalent, accounting for
approximately 70% of all mutational signatures observed. To lesser degrees, mismatch repair
(MMR) defects accounted for 2%, homologous recombination (HR) defects accounted for 11%,
8
and a process of unknown etiology termed “Signature 8” accounted for 15% of the mutational
signatures. Tumors with MMR and HR defects characteristically showed biallelic inactivation of
genes essential for the respective DNA repair processes, including MSH2, BRCA1, BRCA2 and
PALB2. Also, one allele was often lost in the germline, which explains the involvement of the
same genes in familial pancreatic cancers. Of note, tumors with MMR defects, owing to their
microsatellite instability, exhibited higher burdens of somatic mutations and increased
transcription of antitumor immune markers as determined by RNA sequencing, which may
translate to a greater responsiveness to immunotherapy.
Tumoral Heterogeneity
The complex genetic landscape of PDAC is complicated by significant tumoral
heterogeneity, which can be further categorized into intratumoral and intertumoral heterogeneity.
Intratumoral heterogeneity, which describes genetic heterogeneity that exists among cells
of a single tumor, is a well-recognized prognostic factor and an important cause of therapeutic
resistance in pancreatic cancer. The concept of intratumoral heterogeneity first became apparent
in lineage tracing studies of primary PDACs and matched metastatic tumors, which determined
that metastatic tumors arise from distinct subclonal outgrowths from the primary lesion, all of
which likely diverged from a single parental clone.(7) Intratumoral heterogeneity in a patient can
manifest in three forms: [1] subclonal heterogeneity within a primary tumor, where a founder
clone gives rise to various subclones by acquiring additional mutations, [2] subclonal
heterogeneity within a metastasis, where a metastasis-initiating cell gives rise to its descendant
subclones in a similar fashion, and [3] subclonal heterogeneity of metastasis-initiating cells
within a primary tumor, where metastasis-initiating cells share common ancestors but possess
distinct mutations that confer varying degrees of metastatic potential.(6)
9
Intertumoral heterogeneity describes genetic heterogeneity that exists among tumors of
same histological type occurring in different patients, and it has been well-described in
pancreatic cancer. To characterize the intertumoral differences systematically, several
classification systems have been proposed based on genomic, transcriptomic, and
immunohistochemical analyses.
Molecular Subtyping of Pancreatic Cancer
Waddell et al. classifies PDAC into four major subtypes based on patterns of structural
variation identified from their genomic analysis.(11) In their study, 20% of tumors had ‘stable’
genomes with fewer than 50 structural variants, 36% of tumors had ‘scattered’ structural events
with 50-200 variants, 14% of tumors had ‘unstable’ genomes with more than 200 structural
variants suggestive of defects in DNA maintenance, and lastly, 30% of tumors had a ‘locally
rearranged’ pattern with fewer than 50 structural variants localized to 1-3 chromosomes which
typically result from amplifications that encompass oncogenes or genomic catastrophes such as
in the case of chromothripsis. Interestingly, the ‘unstable’ subtype was predictive of platinum
and poly (ADP-ribose) polymerase (PARP) inhibitor responsiveness.
Transcriptomic studies of PDAC have also identified different molecular subtypes of
PDAC with prognostic and therapeutic implications, resulting in a number of classification
systems that differ based on the input material used and assumptions made for each study. Using
microarray expression analysis of microdissected epithelium, Collison et al. classifies PDAC into
three subtypes termed ‘classical’, ‘quasimesenchymal’ and ‘exocrine-like’.(12) Notably, the
classical subtype was predictive of therapeutic response to erlotinib, while the
quasimesenchymal subtype was negatively prognostic and predictive of therapeutic response to
gemcitabine. In a similar study, Bailey et al. analyzed transcriptomic data from bulk tissue
10
containing the tumor microenvironment, and identified an additional ‘immunogenic’ subgroup
based on presence of stromal immune cell populations.(13) Still, Moffitt et al. proposed a new
classification system by excluding transcripts from presumed normal pancreas from their
analysis, and identified two tumoral subtypes – ‘classical’ versus ‘basal-like’ – as well as two
stromal subtypes – ‘normal’ versus ‘activated’.(14) Tumors corresponding to ‘basal-like’
subtype and ‘activated’ stromal subtype were independently and additively negatively
prognostic. The basal type was also more responsive to chemotherapy on retrospective analysis.
Although large-scale genomic and transcriptomic analyses have greatly elucidated the
intertumoral heterogeneity of PDAC defining its molecular subtypes and established a
foundation for developing precision medicine, applying this knowledge clinically has been
limited by the common lack of access to complex tumor tissue biobanking and sequencing
platforms for most clinicians. To this end, Noll et al. asked whether immunohistochemical (IHC)
analysis, which is a far more accessible and technically feasible form of testing for clinicians at
large, could be used to subtype pancreatic tumors by protein expression, and determined two
IHC markers – HNF1A and KRT81 – for the differentiation of Collison subtypes.(15)
Specifically, HNF1A-positive tumors correlated to the exocrine-like subtype, KRT81-positive
tumors to quasimesenchymal subtype, and IHC-negative tumors to classical subtype. In addition,
their study identified CYP3A expression as a novel mechanism of drug resistance, found at
higher levels in exocrine-like tumors but inducible in all subtypes.
In 2009, Farrell et al. reported the predictive value of an IHC-based assay for guiding
precision medicine treatment of pancreatic cancer. In a phase III adjuvant therapy trial of 538
patients with early pancreatic cancer, the expression of human equilibrative nucleoside
transporter (hENT1) – a key mediator of cellular uptake of gemcitabine – measured by IHC
11
analysis of tumor microarrays was associated with increased overall survival and disease-free
survival in patients who received gemcitabine, but not in those who received 5-FU,
demonstrating hENT1 as a predictive biomarker for gemcitabine efficacy in patients with early
pancreatic cancer.(16)
Deranged Signaling Pathways / Molecular Aberrations
The full mutational landscape of pancreatic cancer is highly complex and diverse.
PDACs contain an average of 63 genetic alterations, the majority of which consists of infrequent
mutations found in fewer than 10% of PDACs.(2, 17) Nonetheless, many cases of these low-
frequency targets appear to be alternative perturbations of the same core signaling pathways that
are commonly deranged across all PDAC subtypes. By interrogating the exome of 24 PDACs,
Jones et al. determined 12 core signaling pathways consistent with the hallmarks of cancer
previously described by Hanahan and Weinberg, although the specific genes and the number of
genes altered in each pathway differed from patient to patient.(17, 18) Included among the
pathways were those affected by well-known driver genes, such as TP53 in DNA damage
response and SMAD4 in TGFβ signaling. Some pathways, such as RAS-ERK signaling and
DNA damage response, were predominated by a single frequently mutated gene, while others,
such as integrin signaling, regulation of invasion, homophilic cell adhesion and GTPase-
dependent signaling, involved many different genes. Biankin et al. further enriched our
knowledge of commonly deranged pathways by next-generation exome sequencing, shedding
light on the deregulation of axon guidance (SLIT and ROBO2), DNA damage repair (ATM) and
chromatin modification (EPC1) in PDAC, which were formerly unappreciated.(19)
Aberrant autocrine and paracrine signaling cascades ultimately promote pancreatic cancer
cell proliferation, migration, invasion, and metastasis.(2) Numerous cytokines, such as
12
transforming growth factor-α (TGFα), insulin-like growth factor 1 (IGF1), fibroblast growth
factors (FGFs) and hepatocyte growth factor (HGF), and their respective tyrosine kinase
receptors, lead to pathologic activation of multiple pathways that confer pancreatic cancer cell
mitogenic self-sufficiency. These signaling cascades also act to promote cancer cell migration
and invasion of both local and distant sites, leading to metastasis. Pancreatic cancer cell
proliferation is further enhanced by pathologic activation of anti-apoptotic and pro-survival
pathways, such as signal transducer and activator of transcription 3 (STAT3), nuclear factor-κB
(NF-κB) and AKT. Reactivation of genes involved in early development, such as WNT, SHH
and NOTCH, can also be seen in a subset of PDAC.
Pathway derangements in PDAC are numerous, and deconstructing their downstream
effects is further complicated by significant crosstalk between pathways creating synergistic
outcomes.(6) p53 normally cooperates with receptor SMADs to activate TGFβ-induced
transcription by forming complexes that bind separate cis-enhancer elements on a target gene
promoter. In PDAC, oncogenic KRAS interferes with TGFβ signaling by degrading SMAD4 and
inhibiting p53 by blockade of its amino-terminal phosphorylation. Furthermore, oncogenic
KRAS and mutant p53 form pathologic complexes that in turn inhibit p63, which normally acts
to oppose TGFβ-dependent cell migration, invasion and metastasis. Collectively, these findings
indicate that deranged pathways in pancreatic cancer exist not as independent processes but
rather as a complex tumorigenic network altering the systems biology of the cell.(6)
Tumor Microenvironment
A hallmark of PDAC is its abundant and dense collagenous stroma, which may account
for up to 90% of the total tumor volume. The tumor microenvironment of PDAC consists of a
highly complex assembly of diverse cell types, including pancreatic stellate cells (PSCs),
13
immune cells, endothelial cells and nerve fibers, which are influenced by the extracellular matrix
composed of matricellular proteins, fibrillar collagen, fibronectin, hyaluronic acid and a wide
range of cytokines, such as TGFβ, FGF, epidermal growth factor (EGF) receptor ligand, vascular
endothelial growth factor (VEGF) and connective tissue growth factor (CTGF).
There is now abundant evidence for the prominent role of pancreatic cancer-associated
stroma in tumor progression by actively promoting tumor growth, invasion and metastasis.
Recently, a protective effect of some of the stromal components contributing to a physical
containment of cancer cells has also been suggested. The dual function of PDAC stroma as both
a tumor promoter and a suppressor suggests that its pathogenic role may arise from a loss of
balance between epithelial cells and stroma. While normal extracellular matrix has the capacity
to restrain tumor growth through the histone demethylase JMJD1a, desmoplastic stroma consists
of aberrant matrix that is stiff with thickened collagen fibers and expresses p-MLC2 that
contributes to tumor progression.(20, 21) PSCs are major drivers of the desmoplastic reaction in
PDAC, wherein pancreatic tissue injury leads to PSC activation and trans-differentiation into α-
smooth muscle actin expressing myofibroblast-like cells secreting collagen-type I, matrix
metalloproteinases (MMPs) and tissue inhibitor of metalloproteinases (TIMPs) that remodel the
extracellular matrix. PSC activation can be triggered by various cytokines and stimuli, including
platelet-derived growth factor (PDGF), TGFβ1, FGF, EGF, tumor necrosis factor-α (TNFα),
interleukin-1 (IL-1), ethanol, endotoxins, hypoxia, pressure and oxidative stress, many of which
are produced by pancreatic cancer cells, endothelial and immune cells of the microenvironment.
Once established, PSC activation is maintained in an autocrine fashion. The resulting fibrous
stroma is a severely hypoxic, nutrient-deprived environment that promotes tumor aggressiveness
by activation of hypoxia-inducible factor-1a. In addition, activated PSCs directly promote
14
proliferation of cancer cells by secreting mitogenic factors such as stromal-derived factor-1,
PDGF, EGF, IGF-1 and FGF which activate MAPK- and AKT-signaling cascades.(22)
Another key feature of the PDAC microenvironment is its highly immunosuppressive
composition. Once the tumor is established, the tumor microenvironment is immunosuppressed
by several mechanisms, including an accumulation of regulatory T cells, M2 type tumor-
associated macrophages and myeloid-derived suppressor cells (MDSCs). Activated KRAS in
tumor cells directs the transcription of granulocyte-macrophage colony-stimulating factor (GM-
CSF), an inflammatory cytokine that promotes recruitment and trans-differentiation of myeloid
progenitor cells into MDSCs which in turn suppress the immune surveillance function of CD8+
T cells.(23) Tumor cells also stimulate the expression of IP-10 (CXCL10) in PSCs which attract
CXCR3+ regulatory T cells to the tumor milieu.(24) PSCs also secrete CXCL12 which attracts
CD8+ T cells away from the juxtatumoral stromal compartment, reducing their chance to interact
with cancer cells.(25) In addition, various cell types within the tumor microenvironment secrete
numerous cytokines that support the immunosuppressive phenotype, including IL-1b, IL-4, IL-5,
IL-6, IL-8, IL-10, IL-13, TNFα, TGFβ, FGF, PDGF, MMPs, thymic stromal lymphopoietin
(TSLP), interferon-γ (IFN-γ) and VEGF.(23) Ultimately, the PDAC microenvironment appears
to constitute a biological space of immune privilege where cancer cells are protected from
immune surveillance, as opposed to rendering T cells dysfunctional as mechanisms to bypass
mechanisms of T cell suppression can promote intratumoral infiltration of cytotoxic T cells and
uncover latent immune responses.(26, 27) Further research on the dynamic intersection of
pancreatic cancer and its tumor microenvironment is of great clinical importance as it will likely
provide answers to improving delivery of chemotherapy and developing effective
immunotherapy.
15
Metabolic Reprogramming
Successful pancreatic cancer cell survival and proliferation depends on its ability adapt to
a severely hypoxic and nutrient-deprived tumor microenvironment. Indeed, pancreatic cancer
cells are known to employ various metabolic changes through mechanisms that are mainly
driven by the expression of oncogenic KRAS and hypoxia-inducible transcription factor 1α
(HIF1α).(2) Oncogenic KRAS induces overexpression of glucose transporter 1, hexokinase 1 and
hexokinase 2, which significantly increases glucose uptake by pancreatic cancer cells. The
increased levels of glucose are funneled through aerobic glycolysis to provide substrates for ATP
production such as pyruvate as well as for the synthesis of nucleic acids, proteins, and fatty
acids. This process in PDAC is uncoupled from the tricyclic acid (TCA) cycle and electron
transport chain (ETC) via HIF1α-mediated induction of pyruvate dehydrogenase kinase 1, which
phosphorylates and inactivates pyruvate dehydrogenase, thereby limiting the conversion of
pyruvate to acetyl-CoA needed for the TCA cycle. The uncoupling of events results in increased
production of lactate, which in turn becomes an important nutrient for less hypoxic cancer cells,
and reduces the production of reactive oxygen species by ETC. Moreover, oncogenic KRAS
promotes macropinocytosis in cancer cells as a major mechanism for the uptake of extracellular
proteins to meet cellular requirements for glutamine and other amino acids. Similarly, HIF1α
activates the autophagy-lysosome system, a self-degrative process for cytoplasmic components
including organelles and macromolecules, to maintain intracellular energy supplies. In xenograft
mouse models of PDAC, pharmacologic inhibition of these processes substantially delayed
tumor growth.
16
Immune Response in Pancreatic Cancer is Unclear
Development of immunotherapies has revolutionized the treatment options for many
types of cancers, including but not limited to melanoma, renal and lung cancers. These therapies
rely on potentiating pre-existing tumor-specific T cells by blockade of immune checkpoints,
which are inhibitory pathways in place to maintain self-tolerance and modulate physiological
immune responses to minimize collateral tissue injury. The same pathways are exploited by
tumors to gain immune resistance against tumor-specific T cells. Some cancers, notably PDAC,
are refractory to immunotherapies, and it remains unclear why. The failure of numerous immune
checkpoint inhibitors to advance through clinical trials for treatment of PDAC created a
preconceived notion in the scientific community that PDACs are poorly immunogenic tumors.
However, an increasing number of studies have now shown prominent T cell infiltrates in the
vast majority of biopsies from PDAC patients and identified unique neoantigen qualities in long-
term survivors, indicating that a meaningful immune response in PDAC is achievable.(28, 29)
However, research in this area has been hampered by the lack of pre-clinical physiologic models
of PDAC that are suited to study anti-tumor immune response.
Pre-clinical Modeling of Pancreatic Cancer
“KP-C” (KrasLox-STOP-Lox-G12D; P53Lox-STOP-Lox-R172H/+; Pdx1-Cre) mice have been widely
used to investigate pancreatic cancer biology. Although this model has been greatly informative
regarding the genetic landscape of PDAC, it is ill-suited for the study of cancer immunology on
two levels. First, tumors develop aggressively in these mice, rapidly progressing to fatal
metastatic disease predominantly by 6 weeks of life. This creates a practical challenge in
investigating early disease when meaningful tumor-immune cell interactions may occur before
significant stromal development and/or the onset of other mechanisms of immune
17
suppression.(30) Secondly, pancreatic tumors that develop in these mice are poorly antigenic,
lacking neoantigen peptides which are critical for mounting anti-tumor T cell responses. In fact,
depletion of T cells in KP-C mice using anti-CD4 and anti-CD8 antibodies had no effect on the
progression of murine PDAC nor on the overall survival of these mice.(31) Thus, most
pancreatic cancer immunology studies have focused on murine and human PDAC cell lines,
which have their own limitations.(32) Namely, monolayer cell lines lack the structural
sophistication and functional differentiation of cells seen in vivo, and cannot recapitulate the
tumor microenvironment in mouse xenograft studies. Cell line-derived three-dimensional
spheroid cultures attempt to address this issue, but are difficult to propagate in spheroid form,
limiting longitudinal investigations. Furthermore, none of the cell-line derived models support
the growth of untransformed, non-neoplastic cells. Instead, they inevitably become monoclonal
over time by in vitro selection of the most aggressive clones, resulting in a loss of genetic
heterogeneity seen in primary tumors. Patient-derived xenograft (PDX) mouse models, which are
established by implanting a piece of surgically resected tissue from a patient under the dermis of
immunocompromised mouse hosts, are inherently more physiologic but are cost-prohibitive and
excessively time-consuming, commonly taking upwards of 6 months to generate sufficient sizes
of mouse colonies, which is outside clinically meaningful timeframes for any approach to
personalized medicine for most pancreatic cancer patients.
A recent breakthrough in translational pancreatic cancer research has been the
development of organoid models of pancreas using human and mouse pancreatic tissues for pre-
clinical modeling of PDAC. Organoids, comprising of complex clusters of multiple cell types
derived from the tissue of interest, can recapitulate the intricate spatial architecture of the
progenitor organ structure and perform functions of the organ such as secretion or contraction.
18
Since a robust method for production of self-renewing intestinal organoids was first reported in
2009, tumor organoid models have been widely adopted for multiple organ systems.(33) In 2015,
Boj et al. recently described methods for reliably generating human and mouse PDAC organoids
using surgical resection specimens as well as endoscopic ultrasound-guided fine needle biopsy
(EUS-FNB) specimens.(34) PDAC organoids derived in this manner could recapitulate the
natural history of human PDAC when orthotopically transplanted into immunocompromised
mice, forming early PanIN-like lesions that progressed to invasive pancreatic cancer with robust
stromal response. The ability to generate organoid cultures from FNB specimens is a major
advantage, since it enables investigators to capture the full spectrum of PDAC ranging from
early premalignant lesions to late metastatic cancers, as opposed to surgical resection specimens
which account for fewer than 20% of patients diagnosed with PDAC who are surgical
candidates. The organoid model is physiologic yet possesses all the desirable intrinsic properties
of an in vitro system. PDAC organoid cultures can be propagated in vitro for expansion of
starting material, which is often the limiting factor for tissue-consuming studies such as deep
sequencing, and cryopreserved indefinitely without losing genetic heterogeneity. They are highly
tractable, amenable to genetic manipulation and high-throughput assays. Moreover, in contrast to
PDX mouse models, organoid cultures can be established rapidly in sufficient quantities for
studies in just 2-4 weeks from the time point of acquiring patient tissues, permitting a
personalized approach to pancreatic cancer medicine to investigate patient-specific tumor
biology, evaluate prognosis and guide therapy in real time.
II. STATEMENT OF PURPOSE
In this study, we sought to develop murine and human organoid models of PDAC to
investigate the biology of pancreatic cancer immune response. Our aims were mainly two-fold:
19
1. Development of an immunogenic murine PDAC organoid model to study antigen-
specific anti-tumor T cell responses in both in vivo and in vitro setting.
2. Creation of a clinically annotated library of validated, patient-derived PDAC organoid
lines as tools for studying human pancreatic cancer immunology.
III. METHODS
Acquisition of human specimens
Human pancreatic cancer tissues were obtained from patients undergoing endoscopic
ultrasound-guided fine needle biopsy (EUS-FNB) or surgical resection at Yale New Haven
Hospital. Some of the surgical resection specimens were used to create patient-derived xenograft
(PDX) mouse models, which subsequently became available as a secondary source of patient-
derived tissues for generation of organoids. Tissues were determined to be tumoral or normal by
evaluation of on-site clinical pathologist. Written informed consent was obtained from all
patients prior to tissue acquisition. This study was reviewed and approved by the Institutional
Review Board of Yale University. All EUS-FNB specimens were provided by James Farrell who
also performed the biopsies. All surgical resection specimens were histologically evaluated and
provided by Marie Robert. PDX mouse models of PDAC were previously established by Ryan
Sowell in Kaech laboratory.
Isolation and culture of murine pancreatic organoids
Murine pancreatic organoids were generated using normal or pre-neoplastic pancreatic
tissues from C57BL/6 mouse and KP-NINJA (KrasLox-STOP-Lox-G12D; P53flox/flox; inversion induced
joined neoantigen) mouse, respectively. Detailed procedures for isolation and propagation of
murine pancreatic organoids were adapted from Boj et al., 2015 and Huch et al., 2016. Briefly,
mouse pancreas was dissected and minced into sub-millimeter pieces before enzymatic digestion
