University of Southern California Dissertations and Theses

Mechanistic studies of novel small molecule anti-cancer agents using next generation sequencing

(USC Thesis Other)

Mechanistic studies of novel small molecule anti-cancer agents using next generation sequencing

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content MECHANISTIC STUDIES OF NOVEL SMALL MOLECULE
ANTI-CANCER AGENTS USING NEXT GENERATION SEQUENCING

by
Yuting Kuang

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHARMACEUTICAL SCIENCES)

August 2015

ii

Copyright 2015 Yuting Kuang
DEDICATION
I dedicate my dissertation to my loving family whom I am forever indebted to for
the sacrifices they made to help me pursue my dreams. They have served as a source of
unconditional motivation, strength and love that no amount of gratitude and appreciation
can ever suffice.

iii

ACKNOWLEDGEMENTS
First, I would like to thank my mom, Dongqing Ye, for being a role model that I
always look up to, and for her endless love, encouragement and inspiration that have
supported me all the way through. I also want to thank my boyfriend, Tao Bi, for his
unwavering love, support and company in all these years.
I am immensely thankful to my mentor, Dr. Nouri Neamati for his guidance and
support. He has offered me valuable expertise in the field of drug design and discovery
that has been indispensable to my research. I would like to thank him for providing me
various opportunities to expand my knowledge and experience in the doctoral training.
I would like to extend my gratitude to all my colleagues and friends in the
Neamati lab (Dr. Soma Samanta, Dr. Shuzo Tamura, Dr. Yoshinari Miyada, Dr.
Tiangong Lu, Dr. Anahita Kyani, Dr. Bikash Debnath, Dr. Suhui Yang, Joseph Modak,
Wesam Mehanna, Shuai Hu, Andrea Shergalis, Christine Cuthbertson, Dr. Divya
Pathania, Dr. Kavya Ramkumar, Dr. Helen Ha, Dr. Shili Xu, Dr. Melissa Millard and Si
Li). It has been such a blessing to work with each and every one of them. I appreciate
their friendship and willingness to always lend a helping hand.
I would also like to thank my committee members, Dr. Bangyan Stiles, Dr. Roger
Duncan and Dr. Keigo Machida for their time, guidance and support. I thank Dr. Mats
Ljungman from University of Michigan for his expertise in next generation sequencing. I
would like to extend my thanks to Dr. Anthony El-Khoueiry and Dr. Pietro Tavern from
Astex, Inc. for their expertise and support.

iv

I thank Dr. Julio Camarero, Dr. Mario Sechi, Dr. Jia Zhou and Aetex, Inc for
sharing their compounds, expertise and resources in our long-term collaborations.
I am grateful to USC School of Pharmacy for providing me with great training
and learning opportunities as a graduate student, and for supporting me to finish my
research as visiting scholar at University of Michigan. I would also like to thank UMICH
Translational Oncology Program for the cooperative research environment and College of
Pharmacy for helping to make things possible.
Special thanks to my friends (Weiyan Xie, Ye Li, Tuo Zhan, Yaoyao Zhu,
Jianghan Qu and Wen Yu) for always being on my side.

v

ABSTRACT

A major challenge for preclinical evaluation of novel therapeutics lies in target
identification and elucidation of the mechanisms of action. The novel next generation
sequencing technique is now becoming available for broad application in the drug
discovery field to assist more in-depth and unbiased studies for traditional therapeutics
and novel compounds.
In this work, the bromouridine-labeled RNA sequencing (Bru-seq) technique was
employed for mechanistic studies of anti-cancer therapeutics. Capturing nascent RNA
transcription, Bru-seq allows for discovery of pharmacological changes posed by small
molecule treatments at the gene transcription level. The global gene transcription profile
represents a unique signature for the compound of interest. With the application of
published databases and bioinformatics tools including IPA, GSEA, CMAP and
Oncomine, bioinformatics analysis of the signature can assist discovery of major cellular
responses and pathways involved in the therapeutic activity. Bru-seq can also identify
potential biomarkers. Discovery of pathways and biomarkers can further be validated in
in vitro and in vivo models.
Following similar workflow, three anti-cancer therapeutic candidates have been
evaluated in this study, and demonstrate promising translational potential.
Combination treatment of the DNA demethylating agent SGI-110 and the
chemotherapeutic oxaliplatin exhibits significant synergistic effect in hepatocellular
carcinoma models. Through GSEA analysis of Bru-seq data, we identified inhibition of

vi

Wnt, EGF/IGF signaling as major contributors to activity of the combination. DNMT1
and survivin have been identified as biomarkers for future clinical evaluation.
Structural optimization of the quinazolinedione (QD) ROS modulator QD232
resulted in discovery of the potent anti-cancer compound QD325, which shows ROS-
dependent cytotoxicity in pancreatic cancer models. Global transcription analysis of
QD325- and QD232-treated cells suggested rapid activation of Nrf2-mediated oxidative
stress and unfolded protein response, which serve as the major mechanisms of action for
these ROS modulators. HO-1, CHOP and GRP78 have been identified as biomarkers for
QD compounds.
The novel class of N-(8-quinolinyl) nicotinamides (QNs) also shows substantial
anti-cancer activity in pancreatic cancer models, but with a slower reaction profile.
Studies of transcription signature after 24 h treatment identified significant upregulation
of genes GDF15, ATF3, DDIT3, HSPA5, WIPI1, GABARAPL1 and MAP1LC3B,
which suggests involvement of stress signaling and autophagy in HJC anti-cancer
activity.
Successful application of Bru-seq for studies of these distinct chemical identities
suggests the versatility and generality of this method, supporting its general use in
preclinical evaluation of future therapeutics.

vii

TABLE OF CONTENT

DEDICATION .................................................................................................................. ii

ACKNOWLEDGEMENTS ............................................................................................ iii

ABSTRACT ....................................................................................................................... v

LIST OF TABLES ............................................................................................................ x

LIST OF FIGURES ........................................................................................................ xii

CHAPTER 1
Using next generation sequencing as an effective tool for
preclinical evaluation of novel anti-cancer agents ......................................................... 1

1.1
Target identification and mechanistic studies of small molecule therapeutic candidates ... 1

1.2
Next generation sequencing ................................................................................................ 5

1.3
Bru-seq ................................................................................................................................ 6

1.4
Knowledge-based approaches for the analysis of sequencing data .................................... 8

1.5
Validation of sequencing and bioinformatics discoveries ................................................ 21

1.6
Proposed work flow for preclinical evaluation of small molecule anti-cancer drugs ....... 22

CHAPTER 2
Deciphering the synergistic effect of SGI-110 and oxaliplatin in
hepatocellular carcinoma ............................................................................................... 25

2.1
Hepatocellular carcinoma (HCC) ...................................................................................... 25

2.2
Deregulated methylation as cancerous features ................................................................ 25

2.3
Demethylating agent SGI-110 .......................................................................................... 27

2.4
Advantage of oxaliplatin as an combinatory agent in HCC .............................................. 28

2.5
SGI-110 sensitizes HCC cells to oxaliplatin treatment ..................................................... 29

viii

2.6
SGI-110 and oxaliplatin exhibit significant anti-cancer effect in in vivo HCC model ..... 37

2.7
Bru-seq identified major signaling regulations by SGI-110 and oxaliplatin treatment .... 42

2.8
Validation of biomarkers DNMT1, ephrin-B2 and survivin in HCC cell lines and
HCC tumors ................................................................................................................................ 57

2.9
Bru-seq identified EFNB2 as the most upregulated gene by SGI-110 treatment in
SNU-398 ..................................................................................................................................... 62

2.10
Conclusions ....................................................................................................................... 65

2.11
Materials and methods ...................................................................................................... 66

CHAPTER 3
Mechanistic studies of ROS modulators in pancreatic cancer ....... 71

3.1
Pancreatic cancer ............................................................................................................... 71

3.2
Modulation of redox balance as promising strategy for treatment of pancreatic cancer .. 72

3.3
Optimization of ROS modulators lead to discovery of QD325 ........................................ 73

3.4
QD325 exhibits ROS dependent cytotoxicity in pancreatic cancer models ..................... 75

3.5
QD325 exhibits anti-tumor activity in in vivo pancreatic cancer models ......................... 79

3.6
Nrf2-mediated oxidative stress and unfolded protein response are identified as major
mechanisms of action by using Bru-seq ..................................................................................... 85

3.7
Validation of biomarkers HO-1, NQO1, CHOP and GRP78 in pancreatic cancer cell
lines and PDAC tumors .............................................................................................................. 91

3.8
Inhibition of mitochondria gene transcription by QD325 and QD232 ............................. 98

3.9
Conclusions ..................................................................................................................... 102

3.10
Materials and methods .................................................................................................... 103

CHAPTER 4
Discovery of novel N-(8-quinolinyl) nicotinamides (QNs) for
the treatment of pancreatic cancer .............................................................................. 109

4.1
Phenotypic screen for cancer drug discovery ................................................................. 109

4.2
Structure activity relationship of QN compounds in pancreatic cancer cell lines .......... 110

ix

4.3
QN523 shows significant cytotoxicity in pancreatic cancer cell lines ............................ 119

4.4
QN523 exhibits anti-cancer activity in in vivo pancreatic cancer xenograft model ....... 121

4.5
Bru-seq analysis identified stress signaling and autophagy as major cellular
responses to treatment with QN523 ......................................................................................... 124

4.6
Discovery of compounds showing similar activity with QN523 .................................... 137

4.7
Validation of biomarkers GDF15, ATF3, DDIT3, HSPA5, WIPI1, GABARAPL1
and MAP1LC3B in pancreatic cancer cell lines ...................................................................... 144

4.8
Conclusions ..................................................................................................................... 147

4.9
Materials and methods .................................................................................................... 148

CHAPTER 5
Conclusions and future directions ................................................... 150

5.1
Workflow for mechanistic studies of novel compounds using next generation
sequencing ................................................................................................................................ 150

5.2
Limitation of the current workflow for preclinical evaluation ....................................... 152

5.3
Techniques as potential additions to the current workflow ............................................ 154

BIBLIOGRAPHY ......................................................................................................... 159

x

LIST OF TABLES

Chapter 1…………………………………………………………………………………1
Table 1-1. Knowledge bases for analysis of gene expression profile

Chapter 2………………………………………………………………………………..25
Table 2-1. IC
50
values (µM)
[1]
of anti-cancer drugs in HCC cell lines
Table 2-2. Top 30 gene sets positively associated with SGI-110 treatment
Table 2-3. Top 30 gene sets negatively associated with SGI-110 treatment
Table 2-4. Top 30 gene sets positively associated with oxaliplatin treatment
Table 2-5. Top 30 gene sets negatively associated with oxaliplatin treatment
Table 2-6. Top 30 gene sets positively associated with combination treatment
Table 2-7. Top 30 gene sets negatively associated with combination treatment

Chapter 3………………………………………………………………………………..71
Table 3-1. Structure and cytotoxicity of QD compounds in MiaPaCa-2, Panc-1
and BxPC-3 cells by MTT assay
Table 3-2. Cytotoxicity of QD compounds in gemcitabine resistant MiaPaCa-2 cells
and normal pancreatic cells by MTT assay

Chapter 4………………………………………………………………………………109
Table 4-1. IC
50
values of HJC compounds in pancreatic cancer cell lines
Table 4-2. IC
50
values of HJC compounds in cancer cell lines
Table 4-3. Top 20 gene sets upregulated with QN523 treatment

xi

Table 4-4. Top 20 gene sets downregulated with QN523 treatment
Table 4-5. Top 20 genes upregulated by QN523 treatment
Table 4-6. Top 20 genes downregulated by QN523 treatment
Table 4-7. Top 20 compounds affecting HSPA5 expression in NextBio
Table 4-8. Top 20 compounds affecting DDIT3 expression in NextBio
Table 4-9. Top 20 compounds affecting ATF3 expression in NextBio
Table 4-10. Top 20 compounds affecting GDF15 expression in NextBio
Table 4-11. Top 20 compounds affecting WIPI1 expression in NextBio
Table 4-12. Top 20 compounds affecting GABARAPL1 expression in NextBio
Table 4-13. Top 20 compounds affecting MAP1LC3B expression in NextBio
Table 4-14. Top 20 compounds correlating with QN523 transcription profile in
connectivity map

xii

LIST OF FIGURES

Chapter 1…………………………………………………………………………………1
Figure 1-1. Strategies for drug discovery
Figure 1-2. Bromouridine-labeled RNA sequencing
Figure 1-3. Bioinformatics analysis workflow
Figure 1-4. Proposed workflow for preclinical identification of small molecule anti-
cancer compounds

Chapter 2………………………………………………………………………………..25
Figure 2-1. DNMT1 and DNMT3A are overexpressed in hepatocellular carcinoma
Figure 2-2. 5-Aza-CdR (decitabine) and SGI-110 show similar cytotoxic activity in
hepatocellular carcinoma cell line Hep-3B
Figure 2-3. SGI-110 is cytotoxic to hepatocellular carcinoma cell lines
Figure 2-4. SGI-110 pretreatment sensitizes SNU-398 cells to oxaliplatin treatment
Figure 2-5. Combination of SGI-110 and oxaliplatin induces cell death and
inhibition of cell proliferation in SNU-398 cells
Figure 2-6. Single agent SGI-110 and its combination treatment with oxaliplatin
inhibit tumor growth in SNU-398 liver cancer xenograft model
Figure 2-7. SGI-110 and its combination treatment with oxaliplatin did not exert
systemic toxicity in vivo
Figure 2-8. Enrichment plot of the gene set MISSIAGLIA_REGULATED_BY_
METHYLATION_UP in the pre-ranked gene list from SGI-110 treatment

xiii

Figure 2-9. Bru-Seq reveals inhibition of Wnt signaling by SGI-110 and its
combination with oxaliplatin
Figure 2-10. Inhibition of cancer related genes by SGI-110 and its combination
with oxaliplatin
Figure 2-11. Inhibition of IGF and EGF signaling by the combination of SGI-110
with oxaliplatin
Figure 2-12. Oxaliplatin treatment downregulates DNMT1 and survivin levels in
SGI-110 pre-treated cells
Figure 2-13. Oxaliplatin treatment does not affect levels of cleaved caspase 3 or
PARP significantly in SGI-110 pre-treated SNU-398 cells
Figure 2-14. Oxaliplatin treatment does not affect DNMT1 or survivin levels in
SGI-110 pre-treated SNU-387 cells
Figure 2-15. Oxaliplatin treatment downregulates DNMT1 and survivin levels in
SGI-110 pre-treated tumors
Figure 2-16. EFNB2 expression is upregulated by SGI-110 and its combination
treatment with oxaliplatin
Figure 2-17. EFNB2 promoter is highly methylated in liver cancer
Figure 2-18. Induction of ephrin-B2 protein levels is only observed in SNU-398 cells

Chapter 3………………………………………………………………………………..71
Figure 3-1. QD compounds induce ROS accumulation in MiaPaCa-2 cells
Figure 3-2. Cytotoxicity of QD compounds correlates with ROS induction
Figure 3-3. Cytotoxicity of QD compounds is scavenged by NAC in MiaPaCa-2 cells

xiv

Figure 3-4. QD325 inhibits tumor growth of MiaPaCa-2 xenograft without systemic
toxicity
Figure 3-5. QD325 inhibits tumor cell proliferation in MiaPaCa-2 xenograft
Figure 3-6. QD325 is well tolerated with gemcitabine treatment in NOD/SCID mice
Figure 3-7. Top 15 canonical pathways regulated by QD232 or QD325 treatment as
revealed by IPA analysis of Bru-seq data
Figure 3-8. Top 30 canonical pathways affected by QD compound treatments as
shown with IPA (z score)
Figure 3-9. Top 30 canonical pathways affected by QD compound treatments as
shown with IPA (p value)
Figure 3-10. Top 50 gene sets upregulated by QD compound treatments as shown
with GSEA
Figure 3-11. Top 50 gene sets downregulated by QD compound treatments as shown
with GSEA
Figure 3-12. Transcription of oxidative stress responsive genes NQO1 and HMOX1
was upregulated by QD232 or QD325 treatment in MiaPaCa-2 cells dose-dependently
Figure 3-13. Transcription of unfolded protein response target genes DDIT3 and
HSPA5 was upregulated by QD232 or QD325 treatment in MiaPaCa-2 cells
Figure 3-14. QD compounds induce protein expression of target genes for oxidative
stress and unfolded protein response
Figure 3-15. Protein levels of NQO1, HO-1, CHOP and GRP78 in MiaPaCa-2
xenograft
Figure 3-16. QD compounds inhibit transcription of mitochondrial genome at HSP2

xv

and LSP promoters
Figure 3-17. QD compounds inhibit COXIII protein expression in MiaPaCa-2 cells
Figure 3-18. QD compounds downregulate mitochondrial DNA content

Chapter 4………………………………………………………………………………109
Figure 4-1. QN523 is cytotoxic in pancreatic cancer cell lines
Figure 4-2. QN523 inhibits tumor growth of MiaPaCa-2 xenograft without
systemic cytotoxicity
Figure 4-3. QN523 inhibits tumor cell proliferation in MiaPaCa-2 xenograft
Figure 4-4. QN523 induces stress responses in MiaPaCa-2 cells as revealed by
Ingenuity Pathway Analysis
Figure 4-5. QN523 induces apoptosis and stress responses in MiaPaCa-2 cells as
revealed DAVID analysis
Figure 4-6. QN523 induces transcription of stress responsive genes
Figure 4-7. QN523 induces transcription of autophagy related genes
Figure 4-8. Proposed model for mechanisms of action of QN523
Figure 4-9. Venn diagrams for comparison of compounds regulating potential
marker genes (compound correlation score > 50)
Figure 4-10. Compounds exhibiting similar transcription signatures with QN523
Figure 4-11. QN523 induces protein expression of stress markers dose-dependently
Figure 4-12. QN523 induces protein expression of autophagy markers
dose-dependently

xvi

Chapter 5………………………………………………………………………………150
Figure 5-1. Methods used for preclinical evaluation of novel small molecule
compound

1

CHAPTER 1 Using next generation sequencing as an effective tool for preclinical
evaluation of novel anti-cancer agents
1.1 Target identification and mechanistic studies of small molecule therapeutic
candidates
A major challenge in current drug discovery lies in identifying cellular targets and
elucidating mechanisms of action for bioactive small molecule therapeutic candidates. In
the past decades, the development in high throughput screening (Sundberg, 2000; Mayr et
al., 2009) and expanding collections of synthetic small molecules prepared by new
organic chemistry strategies (Nielsen et al., 2008; CJ et al., 2012) has accelerated the
discovery of bioactive small molecules. However, technologies needed for biological
characterization of the new chemical entities have not kept pace.
Approaches used for drug discovery can be classified into two types: the forward
approach that resembles forward genetics that starts from selective phenotype
intervention; and the reverse approach that resembles reverse genetics that begins with a
validated target of interest (Fig. 1-1) (Titov et al., 2012; Schenone et al., 2013).
The reverse approach has been widely applied for discovery of small molecule
inhibitors, where the proposed target is validated through demonstrating its relevance
with the disease of interest (Kauselmann et al., 2012). Many potent small molecule
inhibitors have been discovered with this highly efficient method. Out of 113 first-in-
class drugs approved by the FDA from 1999 to 2013, 78 of them were discovered
through target-based approaches (Eder et al., 2014). However, considering the overly

2

simplified one-on-one interaction in the screening process, the proposed impacts of such
small molecule regulators need to be carefully characterized in the disease models where
the in vivo conditions are better recapitulated. Chemical interactions with the presumed
target do not necessarily translate into phenotypic intervention, and significant off-target
effect might also take place in the presence of other potential binding partners. In fact,
lack of efficacy or unacceptable toxicity has led to failure of many programs in the
pharmaceutical companies where the reverse approach has been adopted (Paul et al.,
2010; Arrowsmith, 2011). Insufficient biological validation is considered a major cause
for such problems (Arrowsmith, 2011; Prinz et al., 2011).

Figure 1-1. Strategies for drug discovery

These considerations have increased the interests in cell- or organism-based
phenotypic assays for drug discovery, where the cellular context of protein function is
Strategies)for)cancer)drug)discovery)
Phenotype)
!
Biological)
Characteriza9on!
Target!
Reverse)approach)
(Target=based)screening)!
Forward)approach)
(Phenotypic)screening)!
• Lack!of!efficacy!
• Off-target!effect!
• MOA!
• Target!

3

well preserved and presents as a complete system (Swinney et al., 2011; Zheng et al.,
2013). In the field of oncology, development of new technology and increasing
understanding of cancer biology enable generation of mechanistically informed
phenotypic models that show great promise for cancer drug discovery (Moffat et al.,
2014). The forward approach can directly identify compounds achieving desired
biological effect, and more importantly, it allows for discovery of new therapeutic targets
and mechanisms of action without imposing predetermined hypothesis. But in this case,
the precise protein target and mechanisms of action responsible for the resulting
phenotypes remains to be determined.
Major approaches for target identification include direct affinity-based
biochemical methods, and genetic or genomic methods (Titov et al., 2012; Schenone et
al., 2013).
Affinity-based biochemical method allows discovery of the protein target based
on drug-target interaction. This method often requires labeling of compounds as affinity
probes and subsequent treatment with cell lysate containing the potential target. In the
discovery of circadian clock regulator KL001, the hit compound from circadian
phenotypic screening was conjugated to agarose for target identification. Proteins that
bind to immobilized KL001 and release upon high concentration of the free compound
were taken as candidate targets. Analysis of the candidate pool with liquid
chromatography-tandem mass spectrometry (LC-MS) identified CRY1 as target of
KL001, and the finding was validated with CRY1 antibody immuno-blotting (Hirota et
al., 2012). Chemical labeling of drug candidates has also been explored with the benefit
in allowance for validating bioactivity of the conjugated molecule. Using the active

4

biotinylated derivative of spliceostatin A (SSA), proteins binding with SSA were
enriched by pull down assay using streptavidin beads. The candidate proteins were then
isolated with SDS-page and analyzed with LC-MS, leading to identification of splicing
factor SF3b as the target (Kaida et al., 2007). For the covalent anti-cancer compound
PACMA31, conjugation with fluorescent probe BODIPY gave rise to visualization of
compound-protein complex on 2-D gel, leading to identification of protein disulfide
isomerase as its primary target (Xu et al., 2012). In addition, Lomenick et al. established
the method called DARTS (drug affinity responsive target stability) that takes advantage
of a reduction in the protease susceptibility of the target protein upon binding to facilitate
identification of cellular targets for small-molecule drugs (Lomenick et al., 2009).
DARTS represents an advantageous method for target identification by pinpointing the
direct binding partner of the small molecule basing on their binding affinity without
labeling the compound, thus preserving to the maximal extent the native state of drug-
target reaction. However, the application of such affinity-based assays is dependent on
the nature of the target proteins, and might be limited by sensitivity of the detection
method for identifying targets of low abundance.
Genetic manipulation has been an indispensible part of target identification,
especially in validation of unknown targets. RNAi has been used to generate both cell
line and murine models for target validation in drug discovery (Heck et al., 2004; Sachse
et al., 2005). Short hairpin RNA (shRNA) and small interfering RNA (siRNA) genome-
wide screens have been used to look for new therapeutic targets, as well as to identify
cellular targets of small molecules. However, the off-target effect of siRNA has been an
issue for such genetic studies (Buehler et al., 2012; Marine et al., 2012). The newly

5

developed CRISPR-Cas9 system provides a more reliable method for gene knockout or
knock-in mutation through generating tailored nuclease Cas9 guided by sequence specific
single (synthetic) guide RNA (sgRNA) (Hsu et al., 2014; Moore, 2015). Using the
CRISPR/Cas9 genome editing method, Neggers et al. generated a homozygous mutant
XPO1 Cys528Ser cell line which becomes resistant to the potent anticancer compound
selinexor, and validated XPO1 as the prime target of selinexor with cysteine 528 as one
of the primary residues for drug selectivity (Neggers et al., 2015).
As a more unbiased method, gene expression technology has played an important
role in profile-based inference of small molecule targets and mechanisms. Without
chemical modification of the compound or genetic perturbation of the system, gene
expression profiling is able to provide snap shots of the global gene expression changes
induced by the small molecule in a native system, giving rise to understanding of the
complete regulatory network behind the resulting phenotype. In a proof-of-concept study
using RNA sequencing, comparison on transcription profiles of drug resistant clones for
BI 2536 identified its physiological target as Polo-like kinase 1 (PLK) and the indirect
resistance mechanisms (Wacker et al., 2012). As a high-content informative approach,
this profile-based method is becoming indispensible in drug discovery for dissecting
cellular events responsible for therapeutic benefits.

1.2 Next generation sequencing
Next generation sequencing refers to the non-Sanger modern sequencing
techniques developed in the past decade, including Illumina sequencing

6

(www.illumina.com), Roche 454 sequencing (www.454.com), Ion torrent sequencing
(https://www.lifetechnologies.com) and SOLiD sequencing (http://www.applied
biosystems.com). With the high throughput technology platform, efficiency and
resolution of sequencing is highly improved, making sequencing a readily available tool
for general research.
In the field of cancer research, dramatic advances in next-generation sequencing
technology and computational approaches in massive data analysis enabled
unprecedented insight into genetic, genomic, and epigenomic alterations, as well as
transcriptome changes associated with cancer (Stratton et al., 2009; Vogelstein et al.,
2013; Nakagawa et al., 2015). Such comprehensive studies lead to significantly improved
understanding of the cancer biology, diagnosis and treatment.

1.3 Bru-seq
Accessibility of the microarray platform has opened doors for transcriptome
analysis in the field of cancer research and drug discovery, facilitating the study of the
transcriptome of different biological conditions. However, limitation of probe number
represents a major drawback for this profiling approach. Despite the flexibility of a
customized probe-set for advanced microarrays, application of the technique is limited by
the biased selection of gene targets. Decreasing cost of next generation sequencing makes
RNA-seq a more appealing approach for transcriptome study, which provides impartial
information of gene expression. With increased resolution, RNA-seq allows precise

7

studies on all RNA species including but not limited to mRNA, RNA slice variants,
microRNA and non-coding RNA.
Traditional RNA-seq looks at the levels of mature mRNA and other RNA species,
which are the results of RNA synthesis, splicing and degradation. While such processes
collectively contribute to mRNA levels and hence protein translation, the effect on
transcriptional regulation is masked when looking at the mixture of newly synthesized
and processed RNA.
Bromouridine-labeled RNA sequencing was developed by Paulson et al. to
address this problem (Paulsen et al., 2013; Paulsen et al., 2014). Introduced into the cell
culture system in the last 30 minutes of treatment, bromouridine instead of uridine is
incorporated into RNA, thus serving a label on the newly synthesized RNA molecules.
Upon sample collection, total RNA is extracted and subjected to isolation process using
bromouridine antibody. The bromouridine-labeled RNA synthesized in the last 30
minutes of treatment is isolated from the total RNA population, reverse transcribed,
expanded, and further processed for Illumina HiSeq sequencing (Fig. 1-2). Application of
Bru-seq allows detection of the nascent RNA levels, directly representing the gene
transcription process under upstream regulation.

8

Figure 1-2. Bromouridine-labeled RNA sequencing

1.4 Knowledge-based approaches for the analysis of sequencing data
With the application of Bru-seq, we will have access to full RNA species after
mapping all the reads to the reference human genome (hg19). In the current study, we are
interested in the gene expression profile and thus the aspects of other RNA species are
not discussed here.
There are about 22,000 annotated genes in the reference human genome. Reads
per kilobase per million mapped reads (RPKM) are used as normalized readout for every
gene (Mortazavi et al., 2008). For specificity, genes with mean RPKM < 0.5 are filtered
out since such low levels of expression may not be biologically relevant, and the
3!
Transcriptome!as!biological!signature:)sensi9ve,)highly)accessible!
Bromouridine)
(30!min)!
Treatment)
RNA)
extrac9on)
Pull)down)with)
Bromouridine)Ab)
cDNA)library)
prepara9on)
Illumina)
HiSeq)
Mapping)
to)hg19)
0
10
20
30 Control
Treatment
RPKM
BrU=labeled)RNA)sequencing)(Bru=seq) !
Adapted!from!Paulsen!M.!et.al.!PNAS.!2013!

9

calculated fold changes might be misleading for such genes. Also genes with size smaller
than 300 base pairs are filtered out considering reliable resolution during the sequencing
conditions. In general, a list with about 9,000 expressed genes will be generated as
representation of the transcriptome.
Genes with high fold changes are considered important as effectors or responders
for the drug treatment. Knowledge-based search for top regulated genes will provide
overall information about cellular functions, previous studies, and potential relationship
with cancer, facilitating characterization of such sensitive effectors to the treatment.
(Table 1-1)
However, more comprehensive analysis is required to elucidate the biological
meaning of the gene list so as to understand mechanisms of action of the treatment.
Beyond the genes themselves, there are questions we would like to ask about the
expression profiles obtained with drug treatment as a whole. Firstly, what pathways or
functional groups of genes are affected and what would be the outcomes? Secondly, are
there any known treatments that share a similar transcription profile?
For the first question, understanding identities of highly differentially expressed
genes and grouping the genes by their function is important. Extracting biologically
meaningful information is a challenging process.
Knowledge bases focusing on functional annotation, cellular network and
pathways are highly informative for the elucidation of gene lists and interpretation of
gene expression profiles. The Ingenuity Knowledge Database is one such curated
database with different algorithms for pathway identification, causal analysis and
prediction for cellular outcomes. Database for Annotation, Visualization and Integrated

10

Discovery (DAVID) is another sample that integrates gene ontology and biological
themes. Application of such databases expedites identification of cellular networks and
biological functions behind the gene lists.

Table 1-1. Knowledge-based analysis of gene expression profile

Websites Description License
Gene oriented analysis
Oncomine http://www.oncomine.com
Coexpression or
differential analysis in
different cancer types
(subtypes) vs.
corresponding normal
tissue
Yes
NextBio http://www.nextbio.com
Provide information of
correlated compounds,
knockdown studies,
clinical studies and
advanced literature
search
No for
basic
functions
Transcription profiling
IPA http://www.ingenuity.com/products/ipa
Pathway and function
analysis
Yes
DAVID http://www.ingenuity.com/products/ipa
Discovery of common
biological themes,
with focus on gene
ontology
No
GSEA http://www.broadinstitute.org/gsea/index.jsp
Gene set enrichment
on pre-ranked gene list
No
CMAP https://www.broadinstitute.org/cmap/
Similarity in gene
expression patterns to
known perturbagens
No

As for the second question, hierarchical clustering has traditionally been used for
identifying similarity among different transcription profiles (Hughes et al., 2000; Waring

11

et al., 2001). By comparing the patterns of transcription profile among samples, treatment
with similar gene expression patterns could be clustered and thus provide information
about similarity. However, similar effect could be masked by genetic differences between
cell lines, limiting the comparison within samples. Also hierarchical clustering requires
that the data come from the same experimental platform, thus placing a hurdle to
application of such techniques to samples tested in different systems or conditions.
To address the pitfalls of hierarchical clustering, knowledge-based and respective
analytical tools have been established for comparison of gene patterns. Analysis
algorithm focused on groups of genes (gene set enrichment analysis) and overall
transcription patterns (connectivity map) provides solution to the question.
Combined application of knowledge bases (Fig. 1-3) largely facilitates analysis of
transcriptional signatures, and provides thorough understanding of the drug-induced
cellular effects.

12

Figure 1-3. Bioinformatics analysis workflow

1.4.1 Gene orientated analysis
Oncomine is a curated collection of genomic data from clinical cancer samples,
representing 19 different types of cancer (Rhodes et al., 2004). The current Oncomine
v4.5 features 729 datasets and 91,866 samples, with both DNA sequencing and mRNA
expression data from both microarray and sequencing. 14,588 normal samples are also
included in the database. Oncomine allows for differential analysis in cancer vs. normal
samples in different types of cancer as well as cancer subtypes, giving rise to an
understanding of the potential role of the gene of interest in the clinical setting. It serves
as a useful tool for target validation in the context of cancer.
Gene$oriented*
analysis*
NextBio!
Oncomine!
Background!
informa/on!
Expression!in!
cancer!(subtypes)!
vs.!normal!!
Profile*
analysis*
IPA/DAVID!
GSEA!
CMAP!
Pathway!and!
causal!networks!
Enriched!!
gene!sets!
Transcrip/onal!
similarity!with!
known!cmpds!
Expressed*genes*
(~9000)!
Gene*size*>*300*bp*
RPKM*>*0.5*
All*genes*
(~22000)!
Calculate*fold*change*
(treatment/control)!
Rank*genes**
by*fold*changes*
(up!!!down)!
Preranked*gene*list*
(~9000)!
TranscripPonal*signature!

13

NextBio Research provides access to a curated database with close to 20,000
studies, which allows analysis of gene function in major disease areas (Kupershmidt et
al., 2010). Query with the gene or compound of interest will retrieve all correlated studies
and summarized in the form of most correlated tissues, diseases, compounds, gene
perturbations, literature and clinical trials. It also provides general information of the
gene serving as an online portal.
Analysis with Oncomine and NextBio facilitates functional characterization of the
potential cellular effectors in molecular studies and clinical oncology settings, and
supports the design of potential validation studies.

1.4.2 Ingenuity Pathway Analysis (IPA)
IPA is the leading software for pathway analysis, which extracts pathway
information from the transcription profile and facilitates interpretation of gene expression
data. It has also built up a distinct algorithm for predictive causal analytics, which
provide further explanation and prediction of cellular networks by integrating previously
observed cause-effect relations reported in literature, with focuses on directions of effects
rather than mere associations. Based on the manually curated collection of nearly 5
million experimental findings from biomedical literature or third-party databases, the
Ingenuity Knowledge Base serves to support analysis of canonical pathways, prediction
of likely upstream regulators of observed expressional changes, and inference of impacts
of biological functions and diseases (Chindelevitch et al., 2012; Kramer et al., 2014).

14

Besides elucidation of gene expression data, the IPA knowledge base has been
growing in the past years and it has the ability to integrate complex omics data at multiple
levels including metabolomics, proteomics, transcriptome and microRNA. New features
for RNA isoform analysis (mouse and human) have been added recently for further
interpretation of RNA-seq data.
IPA analysis has been used widely for elucidation of gene expression profiles for
a variety of samples. For the novel antineoplastic series of propynoic acid carbamoyl
methyl-amides (PACMAs), IPA analysis of microarray suggested initiation of Nrf2-
mediated oxidative stress by the compound in MDA-MB-435 cells, which was further
validated by observation of PACMA-induced mitochondrial superoxide species
accumulation (Yamada et al., 2011). For the survivin inhibitor YM-155, gene expression
profiling after treatment in SK-NEP-1 cells revealed significant regulation of death
receptor signaling, TNFR1 signaling and induction of apoptosis, confirming the highly
specific anti-cancer activity of the small molecule (Tao et al., 2012).
In an effort to decipher mechanisms of tumorigenesis in human pancreatic ductal
epithelial cells, stable expression of mutant K-ras, Her2, p16/p14shRNA and
Smad4shRNA were introduced in immortalized human pancreatic ductal epithelial
(HPDE) cell line. Analysis of gene expression array by IPA revealed that deregulation of
integrin-linked kinase signaling and the cell cycle were the most significant changes
underlying tumorigenic transformation driven by PDAC signature alterations (Chang et
al., 2013).
IPA has also been applied to clinical studies. For APR-246, a p53 activating agent
that can restore transcription activity of unfolded or mutant p53, IPA analysis of gene

15

expression profiles in treated tumor biopsies revealed changes in genes regulating
proliferation and cell death, suggesting the potential clinical efficacy of APR-246 in the
first-in-human study in refractory hematologic malignancies and prostate cancer patients
(Lehmann et al., 2012).
In our current application of IPA, we apply biologically meaningful filters to the
gene expression profile to collect the most representative genes for a given condition.
Typically, fold changes of gene transcription levels are used as observations. Using cut
off values of 1.5 or 2 fold, which is generally considered biologically relevant, the list of
up and downregulated genes are created and used as input for the analysis. Output of IPA
includes top canonical pathways, top upstream regulators, top diseases and bio functions,
top toxicity functions and top networks. Items are ranked by p-values in each category.
For our purpose, the top canonical pathways are of highest relevance, and the genes
contributing to enrichment of the pathways are thus further investigated.
DAVID provides similar function as IPA, with clustering focuses on gene
ontology terms. It also supports visualization of gene lists in the BioCarte and KEGG
pathways (Huang da et al., 2009b; Huang da et al., 2009a). With differences in
computational algorithm, analysis by DAVID does not take into consideration the extent
of differential expression. It shows the identity of the implicated genes, but the up and
downregulated gene lists need to be analyzed separately for each attributes of cellular
function.

16

1.4.3 Gene set enrichment analysis (GSEA)
Gene set enrichment analysis was introduced by the Broad Institute in 2005. This
idea of analyzing differential expression for groups of genes, as opposed to individual
genes, has brought the analysis of gene expression data a step forward.
Gene sets are groups of genes that share common biological function,
chromosomal location or regulation. Collection of annotated gene sets has been ever
growing, from 1,325 in the initial database to 10,348 gene sets in the current Molecular
Signatures Database (MSigDB, v5.0) (Subramanian et al., 2005). Gene sets are defined
based on prior biological knowledge. The current catalog is divided into 8 major
collections: hallmark gene sets, positional gene sets, curated gene sets, motif gene sets,
computational gene sets, gene ontology gene sets, oncogenic signatures and immunologic
signatures.
As described previously, genes from the expression profile are often ordered
according to their differential expression between experimental classes. The common
approach is to filter and focus on the genes at the top or bottom of the ranked list, as with
the analysis using IPA. However, sole focus on highly differentially expressed genes
might miss important effects on pathways, where the functional flux might be
dramatically altered when all the members in the pathway experience minor changes in
the same direction. Also since biological regulations are highly context dependent, the list
of significantly regulated genes might not necessarily overlap when the same treatment is
tested in different cell lines, under different conditions, thus making it difficult to
compare the experiments with limited numbers of genes.

17

To overcome such challenges, GSEA was developed to evaluate transcription
profiles at the level of gene sets, so as to determine the differentially expressed trends of
the group of genes rather than absolute expression of an individual gene. Every
experimental sample is represented by the total ranked list of expressed genes (from most
upregulated to most downregulated). By identifying the location of members of certain
gene sets on the query ranked list, GSEA determines whether the gene set tends to appear
towards the top or bottom of the ranked list. Correlation is indicated should the gene set
be enriched at either direction. By adopting such pattern-based statistical analysis, GSEA
allows discovery of minor, but consistent changes in the expression profile.
Outputs of GSEA include normalized enrichment score (NES) and false discovery
rate (FDR) corresponding to each NES. Enrichment score (ES) reflects the degree to
which a gene set is overrepresented at the top (positive) or bottom (negative) of the query
ranked list, and is normalized to NES to allow comparison among gene sets of different
sizes. Enrichment plots of the certain gene set visualize such comparison.
Flexibility empowers GSEA for a broad range of analysis. In an effort to identify
the resistance drivers for MET/anaplastic lymphoma kinase (ALK) inhibitors (e.g.
crizotinib) in advanced ALK-rearranged NSCLC, Wilson F. et al. discovered that
overexpression of P2Y purinergic receptors could confer resistance to ALK inhibition. A
P2Y signature was then defined by overexpressing P2Y in H3122 cell line, along with
signatures of other putative resistance drivers, and added to the oncogenic gene
signatures collection of MSigDB as part of the working reference. Gene signatures of the
previously determined drivers EGFR, HER2 and the newly identified driver P2Y are all

18

enriched in crizotinib-resistant tumors, validating the resistance mechanisms identified in
the in vitro screen system (Wilson et al., 2015).
GSEA was also used to identify potential targets for high-risk acute lymphoblastic
leukemia (ALL). Query with gene expression profile of rapidly engrafted patient ALL
cells (short time to leukemia vs. long time to leukemia in NOD/SCID mice model)
identified genes involved in mTOR signaling to be associated with early patient relapse.
Administration of mTOR inhibitors resulted in decreased leukemia proliferation in vitro
and delayed reoccurrence of post-treated leukemia in vivo, establishing the novel
therapeutic strategy for high-risk ALL (Hasan et al., 2015).

1.4.4 Connectivity map (CMAP)
Connectivity Map is a database built by the Broad Institute. The goal of
Connectivity Map is to describe all biological status – established by either genetic
perturbation or small molecule treatments – by genomic signature, and compare such data
with pattern-matching tools. The mRNA expression level was selected as biological
signature for its low cost, availability for high throughput format and high complexity to
provide rich description. The mammalian cell culture was chosen as the model system for
generalizable comparison among different perturbagens (Lamb et al., 2006; Lamb, 2007).
Different from the previous databases, which are built on curated signature
collections from public deposit or biomedical literature, Connectivity Map is a designed
database with direct in-house experimental data, allowing minimal experimental
variations in the data.

19

In the first-generation connectivity map, 164 small-molecule perturbagens
including FDA approved drugs and non-drug bioactive tool molecules were tested in the
highly characterized breast cancer cell line MCF7. The study included molecules with the
same targets (histone deacetylase inhibitors) to determine whether such compounds share
similar mRNA signatures, and molecules with the same clinical indications
(antidiabetics) to determine whether therapeutic classes can be established beyond direct
mechanisms of action. Molecules of distinct groups of targets (e.g. modulators of
transcription factors or inhibitors of extracellular receptors) were tested to study whether
a close regulation of gene transcription is necessary for establishment of mRNA
signature. Drug molecules that require chronic and systemic administration for
therapeutic effects (antipsychotics) were also tested to evaluate the potential of
establishing biologically relevant signatures for such different classes with an in vitro
model. Perturbagens were typically tested at 10 µM for 6 hours in MCF7 cells in
treatment and vehicle control pairs. Tests with a subset of perturbagens were extended,
including multiple concentrations, treatment durations or cell lines. Each treatment
condition (with its paired vehicle control) is termed as an instance under the certain
perturbagen. Comparison of different instances allows for the understanding of sensitivity
to concentration, treatment duration and cellular context for the biological signatures as
well as for the statistical comparison tool. The current Connectivity catalog contains
6,100 instances representing 1,309 diverse bioactive small molecules.
The reference gene-expression profile of each instance in the Connectivity
Database is compared to their paired vehicle control and represented as rank-ordered
gene lists. The query signature is represented by an upregulated gene list and a

20

downregulated gene list, and compared to each reference profile in the database. If the
query up list is near the top of the reference, or the query down list is near the bottom of
the reference, a positive connectivity score (positive connectivity) will be yielded. If the
query lists are in the opposite ends of the reference, a negative connectivity score
(negative connectivity) will be yielded. The connectivity score ranges from +1 to -1 for
each reference expression profile (instance), where 0 indicates no connectivity. Ranking
instances by their connectivity score with the query signature, top correlations - either
instances with similar or opposite effects - can then be identified. Mean score is
calculated for each perturbagen. Specificity score is calculated by comparing the score of
the query signature to those of signatures from MSigDB. MSigDB signatures having
higher connectivity scores are counted against specificity. The more of such signatures,
the higher the specificity scores would be, and less specific the instance is with the query
sample.
The proof-of-concept studies showed that connectivity could be successfully
established among compounds of the same targets or the same medical indications
regardless of the targets’ functional nature. Application of Connectivity Map identified
HSP90 as a target for gedunin, an under-characterized compound that abrogates AR
activity in prostate cancer with unknown mechanisms (Hieronymus et al., 2006; Lamb et
al., 2006). For the small molecule HIF2a inhibitors 40, 41 and 76, which decrease HIF2a
expression under conditions of normoxia and hypoxia (Zimmer et al., 2008), similarity in
transcription signatures with the anti-inflammatory cytokine 15-deoxy-12,14-
prostaglandin J
2
(PGJ
2
) were identified by Connectivity Map, leading to the discovery of
PGJ
2
mediated HIF2a regulation (Zimmer et al., 2010).

21

Connectivity Map was also successfully used to connect disease states with gene
expression signatures. Comparison of expression profiles of bone-marrow leukemic cells
from dexamethasone sensitive or resistant acute lymphoblastic leukemia patients
generated a dexamethasone sensitivity signature. Query through Connectivity Map
revealed strong correlation with mTOR inhibitor rapamycin, which is later shown to
sensitize glucocorticoid-resistant lymphoid CEM-c1 cells to dexamethasone treatment
through modulation of antiapoptotic MCL1 (Lamb et al., 2006; Wei et al., 2006). Here
the application of Connectivity Map led to the elucidation of resistant mechanisms as
well as rapid re-disposition of existing drugs. Also it was suggested as a great strategy for
identifying promising combination therapies to overcome therapeutic resistance.

1.5 Validation of sequencing and bioinformatics discoveries
Computational tools provide significant information for understanding the
molecular responses to drug treatment. However, the gene expression profile might not
reflect the corresponding protein expression levels and functional activity (Wang et al.,
2015), which are results of complex regulation processes including translation, post-
translational modifications and protein degradation.
Also the data for next generation sequencing are mostly generated in single cell
lines with limited treatment conditions (dose and time point), which provide snapshots of
the transcription profile but might not represent every major molecular effect in a general
disease model. Also the signaling models constructed using these bioinformatics

22

approaches might not reflect dynamic signaling flow or the internal feedback regulations
in temporal order and spatial relations.
To further confirm the discovery with computational studies, validation with
experiments in greater extent or with published data is necessary. Considering the nature
of novel drug candidate, there is no relevant published data for in silico validation in most
cases, thus experimental validation is required.

1.6 Proposed work flow for preclinical evaluation of small molecule anti-cancer
drugs
In an effort to discover and characterize new anti-cancer small molecules, it is
beneficial to optimize and streamline the preclinical evaluation process, aiming to
manage limited resources and project progress. In the current study, we propose the
application of next-generation-sequencing as an effective tool for dissecting mechanisms
of action for small molecule anti-cancer drugs in the preclinical evaluation stage. A
proposed workflow is shown in figure 1-4.

23

Figure 1-4. Proposed workflow for preclinical identification of small molecule anti-
cancer compounds

Small molecule drug candidates can be identified through a variety of screening
processes. Cytotoxicity screening with diverse compound library in selected cancer cell
lines can be used as an example for phenotypic screening in oncology. The lead
compound is tested in cell culture models for its anti-cancer activity and selectivity for
cancer cells over normal cells. At this stage, the optimization process can also take place
using cytotoxicity and selectivity as phenotypic readouts.
Optimized compounds can then be taken into in vivo models for evaluation of
potential toxicity and anti-cancer activity, so as to determine whether the compound
Workflow)for)preclinical)evalua9on)
Cell)culture)models)
• Cytotoxity)
• Op9mal)treatment)condi9ons)
Xenogra^)models)
• In#vivo#an9=cancer)ac9vity)
• Poten9al)toxicity)
AcHve!
Next)genera9on)
sequencing)
Bioinforma9cs)analysis)
• Transcrip9onal)signature)
• Cellular)responders)
• Poten9al)mechanisms)of)ac9on)
Valida9on)in)cell)
culture/tumor)samples)
• Biomarkers)
• Mechanisms)
• Targets)
Efficacious!without!major!toxicity!
Poor!solubility/efficacy/tolerance!
InacHve!
Compound)op9miza9on)
/project)cancella9on))

24

serves as a promising drug candidate for cancer therapeutics, or shows any potential
advance over the current standard of treatment.
Mechanistic studies can begin after confirming the biological activity of the
candidate compound. Bru-seq will be performed to obtain the transcription profile
induced by compound treatment at optimal conditions determined in cell culture models.
The signatures are then analyzed by bioinformatics tools for identification of cellular
responders and potential mechanisms of action. Results from the analysis serve to
generate a testable hypothesis for subsequent validation. Going back from the computer
to the bench, bioinformatics discoveries are then tested in cell culture models and in vivo
tumor samples, so as to characterize biomarkers that might correlate with anti-cancer
activity. Targets that are responsible for cellular functions and networks can be further
validated using traditional pharmacology.
We hypothesize that application of such workflow will be beneficial in preclinical
characterization of novel anti-cancer small molecules, and have applied this process to
three groups of compounds of distinct mechanisms of action to evaluate the method.

25

CHAPTER 2 Deciphering the synergistic effect of guadecitabine (SGI-110) and
oxaliplatin in hepatocellular carcinoma
2.1 Hepatocellular carcinoma (HCC)
Liver cancer claimed 746,000 lives worldwide in 2012 (Ferlay J, 2013). With
poor survival statistics and a growing rate of incidence due to an increase in HCV
infections and other liver diseases in the population, liver cancer has now become the
fifth leading cause of cancer-related deaths in the United States (Siegel et al., 2014).

Eighty percent of all liver cancer cases are hepatocellular carcinoma (HCC). Patients
diagnosed at advanced stage of the disease are not eligible for potential curative treatment
or transarterial chemoembolization, leaving systemic treatment as the major remaining
therapeutic option. Sorafenib, a VEGFR, PDGFR and Raf kinase inhibitor, is the only
FDA approved drug since 2007 for use as palliative treatment for these patients (Forner et
al., 2012). Treatment with sorafenib has been shown to improve median survival and
time to progression by 3 months in HCC patients compared to placebo (Llovet et al.,
2008). Unfortunately, drug resistance and adverse events have limited its applicability in
the clinic. New effective treatments for HCC are in urgent need.

2.2 Deregulated methylation as cancerous features
Overexpression of DNA methyltransferases DNMT1 and DNMT3A is a
characteristic of HCC (Fig. 2-1) (Wurmbach et al., 2007; Fan et al., 2009; Roessler et al.,
2010). Knocking down DNMT1 significantly inhibits HCC cell proliferation (Fan et al.,
2009), further implicating an oncogenic role of DNMT1. In line with overexpression of

26

DNMTs, DNA hypermethylation in the promoter regions of tumor suppressor genes like
CDKN2A (p16) and CDH1 (E-cadherin) has been associated with HCC. In addition,
microarray analyses and next generation sequencing have identified a series of DNA
methylation-regulated biomarkers specific for HCC (Nishida et al., 2012; Shitani et al.,
2012). Treatment with the DNA demethylating agent decitabine restores transcription of
many tumor suppressor genes silenced by promoter hypermethylation and inhibits cell
proliferation (Suh et al., 2000; Neumann et al., 2012; Zhang et al., 2012). Taken together,
these results provide the impetus for the therapeutic targeting of DNMTs in HCC.
Figure 2-1. DNMT1 and DNMT3A are overexpressed in hepatocellular carcinoma. A)
mRNA expression levels of DNMT1 in normal liver tissue and hepatocellular carcinoma.
B) mRNA expression levels of DNMT3A in normal liver tissue and hepatocellular
carcinoma. Analysis was done using the Oncomine database (www.oncomine.org).
Boxes represent the interquartile range (25
th
– 75
th
percentile). Whiskers represent the
minimum to maximum range. The bars denote the median.

27

Accumulating pre-clinical data has suggested epigenetic therapy as an appealing
strategy to target HCC. However, when decitabine was tested in several clinical trials in
solid tumors such as colon and lung cancers (Graham et al., 2009; Fan et al., 2014), it
was not effective as a single agent. Compelling preclinical data have prompted phase II
trials of combination treatment with decitabine and carboplatin in relapsed ovarian cancer
patients. When decitabine was given at 90 mg/m
2
, combination-induced neutropenia
became a major issue that lead to closure of the study (Glasspool et al., 2014). However,
low dose (10 mg/m
2
) decitabine treatment successfully re-sensitized heavily pretreated
ovarian cancer to carboplatin, achieving a 35% objective response rate and progression-
free survival of 10.2 months among 17 patients (Matei et al., 2012), suggesting great
potential for low dose hypomethylating agent combination treatment in solid tumors.

2.3 Demethylating agent guadecitabine (SGI-110)
Guadecitabine (SGI-110) is a dinucleotide comprised of deoxyguanosine and the
DNA demethylating agent decitabine (2-deoxy-5’-aza-cytidine), an FDA approved agent
for myelodisplastic syndrome (MDS). When activated, decitabine is incorporated into
DNA and the presence of nitrogen at the 5 position of the pyrimidine leads to formation
of covalent DNA-protein adducts with DNMTs (Jones et al., 1980; Song et al., 2012).
DNMT proteins bound to decitabine are degraded, resulting in a downregulation of total
DNMT protein levels and a reduction in the hypermethylation phenotype. Unfortunately,
decitabine is rather chemically unstable in vivo. Catalyzed by cytidine deaminase (CDA),
2-deoxy-5’-aza-cytidine is rapidly converted into the inactive metabolite 2-deoxy-5’-aza-

28

uridine. Importantly, SGI-110 is a dinucleotide of deoxyguanosine and decitabine to
protect the latter from CDA inactivation. SGI-110 is formulated as a pharmaceutically
stable subcutaneous injection formulation that yields longer half-life and more extended
decitabine exposure than decitabine IV infusion (Yoo et al., 2007; Chuang et al., 2010;
Tellez et al., 2014). Such desirable features make SGI-110 a clinically appealing
demethylating drug. The development of the second-generation hypomethylating agent,
SGI-110, represents an improvement over decitabine. SGI-110 treatment provides a
longer exposure window than decitabine with lower maximal concentrations. These
features may make it more amenable to combine with myelosuppressive agents. Our
preclinical studies with SGI-110 in the six HCC cell lines and the xenograft model
demonstrated significant antitumor activity of SGI-110 and synergism when used in
combination with oxaliplatin. These preclinical data provide a strong rationale for the
further testing of SGI-110 in clinical trials.

2.4 Advantage of oxaliplatin as an combinatory agent in HCC
Combination treatments have advantages over single agent applications in that
they attack multiple targets making it less likely for the tumor to develop resistance, and
allow anti-cancer agents to be used at lower doses, reducing adverse events. The platinum
compound oxaliplatin has been evaluated in combination with gemcitabine (GEMOX)
(Louafi et al., 2007) or with leucovorin and fluorouracil (FOLFOX4) (Qin et al., 2013) in
HCC. Although no significant survival benefit was observed in the FOLFOX4 phase III
trial, the efficacy reported in the GEMOX phase II trial and the favorable safety profile

29

shared by both studies suggest the potential of oxaliplatin-based treatment for HCC
patients. The combination of SGI-110 and other platinum compounds including cisplatin
(Fang et al., 2014) and carboplatin (Wang et al., 2014) showed encouraging anti-tumor
activity in ovarian cancer, implying great therapeutic potential of such combinations as
anti-cancer treatment. However, the combination of SGI-110 and oxaliplatin has never
been tested.

2.5 SGI-110 sensitizes HCC cells to oxaliplatin treatment
2.5.1 SGI-110 inhibits HCC cell proliferation
To evaluate the potency of SGI-110 in HCC, we used six cell lines with different
genetic backgrounds as in vitro models. Decitabine and SGI-110 showed similar
cytotoxicity in Hep-3B cells (Fig. 2-2). After 72 h treatment, SGI-110 at 10 µM inhibited
cell proliferation by no more than 27% in the six cell lines (Fig. 2.3A, C). However,
significant inhibition of cell proliferation was observed when cells were treated with SGI-
110 for 72 h and placed in fresh cell culture media for 7 days (Day 12 of the experiment).
Four of the HCC cell lines Hep-3B, SNU-398, SNU-449 and Hep-G2 were more
sensitive to SGI-110, with IC
50
values lower than 500 nM (Fig. 2-3A). SGI-110 also
significantly inhibited colony formation in these four cell lines at low mM (Fig. 2-3B).
The SNU-475 and SNU-387 cells having a doubling time of over 60 h (Park et al., 1995)
are more resistant to SGI-110 treatment. In long-term treatment, SGI-110 showed an IC
50

value of 54 µM in SNU-475 and 63 µM in SNU-387 in MTT and 10 µM in colony
formation assays (Fig. 2-3C and D). Since SGI-110 acts by incorporating into DNA and

30

modifying methylation patterns during DNA synthesis, the long doubling time of SNU-
475 and SNU-387 might be responsible for their lower sensitivity. SNU-475 is more
sensitive to SGI-110 treatment in colony formation assay than in MTT assay. Since cells
were seeded at lower density in colony formation assay to assess long-term anti-
proliferative effect, this result suggests that density of cells affects drug sensitivity of
SNU-475 cells, and that SGI-110 activity might be dependent on inhibition of cell
proliferation. As the direct target of SGI-110, DNMT1 protein level is a robust marker for
assessing treatment efficacy. DNMT1 protein levels were depleted within 24 h of SGI-
110 treatment (100 nM) in both SNU-398 and Hep-G2 cells, and started to recover after
SGI-110 removal, confirming that SGI-110 directly targets DNMT1 for protein
degradation (Fig. 2-3E). In conclusion, SGI-110 targets DNMT1 and inhibits HCC cell
proliferation in a long-term manner.
Figure 2-2. 5-Aza-CdR (decitabine) and SGI-110 show similar cytotoxic activity in
hepatocellular carcinoma cell line Hep-3B. A) Dose-response curves for 5-Aza-CdR and
SGI-110 in Hep-3B for acute cytotoxicity (72h), where cells were treated with SGI-110
for 72 h and subjected to MTT assay. B) Dose-response curves for 5-Aza-CdR and SGI-
110 in Hep-3B for long-term cytotoxicity (72h + 7d), cells were treated with SGI-110 for
72 h and left to recover in fresh cell culture media for seven days. MTT assay was then
performed to assay the number of live cells under each treatment condition. Inhibition of
cell proliferation was calculated against PBS-treated controls. Data points are shown as
Mean ± SD from three independent experiments.

31

Figure 2-3. SGI-110 is cytotoxic to hepatocellular carcinoma cell lines. A) Dose-
response curves for SGI-110 in 4 sensitive HCC cell lines. For acute cytotoxicity (72 h),
cells were treated with SGI-110 for 72 h (fresh drug added every 24 h) and subjected to
MTT assay. For long-term cytotoxicity (72 h + 7d), cells were treated with SGI-110 for
72 h as above and changed to fresh cell culture media for seven days. MTT assay was
then performed to evaluate the number of live cells under each treatment condition.
Inhibition of cell proliferation was calculated against PBS-treated controls. Data points
are shown as Mean ± SD from three independent experiments. B) Colony formation
assay for SGI-110 in 4 sensitive HCC cell lines. Cells were treated with SGI-110 for 72 h
and left in culture in fresh media until colonies formed in PBS-treated controls. Colonies
were stained with crystal violet and imaged. C) Dose-response curves for SGI-110 in 2
resistant HCC cell lines from MTT assay as described in panel A. D) Colony formation
assay for SGI-110 in 2 resistant HCC cell lines as described in panel B. E) In SNU-398
cells and Hep-G2 cells, DNMT1 levels were reduced by SGI-110 treatment in a time
Figure 1
SNU-387
0.001 0.01 0.1 1 10 100
-20
0
20
40
60
80
100
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
SNU-398
0.001 0.01 0.1 1 10
-20
0
20
40
60
80
100
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
10
Hep-3B
Control
SGI-110
SNU-398
Hep-G2
SNU-449
SNU-387
SNU-475
Control
SGI-110
Control
SGI-110
Control
SGI-110
Control
SGI-110
Control
SGI-110
1 0.1 0.01 0.001
10 1 0.1 0.01 0.001
A B
C D
IC
50
= 21 nM IC
50
= 251 nM
IC
50
= 433 nM IC
50
= 287 nM
IC
50
= 63.4 µM IC
50
= 53.9 µM
Hep-G2
0.001 0.01 0.1 1 10
-20
0
20
40
60
80
100
SGI-110 (72h+7d)
SGI-110 (72h)
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
Hep-G2
0.001 0.01 0.1 1 10
-20
0
20
40
60
80
100
SGI-110 (72h+7d)
SGI-110 (72h)
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
Hep-3B
0.001 0.01 0.1 1 10
-20
0
20
40
60
80
100
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
SNU-449
0.001 0.01 0.1 1 10
-20
0
20
40
60
80
100
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
SNU-475
0.001 0.01 0.1 1 10 100
-20
0
20
40
60
80
100
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
Hep-G2
0.001 0.01 0.1 1 10
-20
0
20
40
60
80
100
SGI-110 (72h+7d)
SGI-110 (72h)
SGI-110 (µM)
Inhibition of Cell Proliferation (%)
(µM)
(µM)
E
DNMT1
Treatment duration (h)
Time after SGI-110 removal (h)
- 24 48 72 72 72 72 72 72 -
- - - - 24 48 72 96 168 -
PCNA
SNU-398
(SGI-110 100 nM)
170 kDa
34 kDa
170 kDa
34 kDa
DNMT1
Treatment duration (h)
Time after SGI-110 removal (h)
- 24 48 72 72 72 72 72 72 72 72
- - - - 24 48 72 96 120 144 168
PCNA
Hep-G2
(SGI-110 100 nM)
SGI-110 (µM) SGI-110 (µM)
SGI-110 (µM) SGI-110 (µM)
SGI-110 (µM) SGI-110 (µM)

32

dependent manner. DNMT1 levels recovered after SGI-110 removal. Cells were treated
with SGI-110 at 100 nM for up to 72 h. Cell culture media containing SGI-110 were then
removed and cells were changed into fresh media. Samples were collected at indicated
time points from 24 h to 168 h after SGI-110 removal.

2.5.2 Pretreatment with SGI-110 sensitizes HCC cell lines to oxaliplatin
The platinum compound oxaliplatin is selected as the candidate for combination
with SGI-110 in this study based on its encouraging activity and safety profile in HCC as
a combinatory agent. Both oxaliplatin and the current standard of care, sorafenib, showed
higher IC
50
values in SNU-475, SNU-387 cell lines than SNU-398 and Hep-G2 (Table 2-
1). Pretreatment with SGI-110 sensitized HCC cells to oxaliplatin treatment in
preliminary studies. The combination was hence further tested in three schedules: 72 h
SGI-110 pretreatment, where oxaliplatin was given either with SGI-110 (schedule 1),
immediately after SGI-110 removal (schedule 2), or 72 h after SGI-110 removal
(schedule 3) (Fig. 2.4A). In the colony formation assay, the best synergy was observed
with combination treatment using schedule 2. Following this schedule, low-dose
oxaliplatin single treatment only inhibited colony formation by 15%, while pretreatment
with SGI-110 at 50 nM increased the inhibition to 54%, and pretreatment with SGI-110
at 100 nM further increased the inhibition to 94% (Fig. 2-4B).
We further tested the combination effect of SGI-110 and oxaliplatin in two SGI-
110 sensitive cell lines (SNU-398, Hep-G2) and two SGI-110 resistant cell lines (SNU-
475, SNU-387) using MTT assay performed on Day 12 following schedule 2 (Fig. 2-4A).
Pretreatment with SGI-110 sensitized SNU-398, Hep-G2 and SNU-475 cells to
oxaliplatin treatment, achieving over 50% inhibition of cancer cell proliferation with
significant synergism (Fig. 2-4C and D), but was not optimal in the other less sensitive

33

cell line SNU-387. The synergistic effect was further confirmed with colony formation
assay (Fig. 2-4E), suggesting the utility of combination of SGI-110 and oxaliplatin as an
effective therapy for HCC.

Day$1$ 2$ 3$ 4$ 5$ 6$ 7$ 8$ 9$ 10$ 11$ 12$
1"
2"
3"
Treatment"schedules"
1" 2" 3"
ctrl$ S$50$nM$ S$100$nM$
Ox$1$μM$
S$50$$
Ox$1$$
S$100$
Ox$1$
SGI;110$treatment$(S)$
OxaliplaBn$treatment$(Ox)$
A
Control
SGI-110 50 nM
Ox 1 uM
S + Ox
Ox 1 uM
S + Ox
Ox 1 uM
S + Ox
0
200
400
600
800
1000
Number of Colonies
p=0.043 p=0.003
Control
SGI-110 100 nM
Ox 1 uM
S + Ox
Ox 1 uM
S + Ox
Ox 1 uM
S + Ox
0
200
400
600
800
1000
Pretreatment
Schedule 1
Schedule 2
Schedule 3
Number of Colonies
p=0.003 p=0.004
B
Figure 2-1
C
D
SNU-398 Hep-G2 SNU-475 SNU-387
SNU-398
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.5
1.0
1.5
2.0
SGI-110 100 nM
SGI-110 200 nM
SGI-110 50 nM
Fa - Inhibition of Cell Proliferation
Combination Index (CI)
SNU-387
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.5
1.0
1.5
2.0
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Fa - Inhibition of Cell Proliferation
Combination Index (CI)
SNU-475
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.5
1.0
1.5
2.0
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Fa - Inhibition of Cell Proliferation
Combination Index (CI)
Hep-G2
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.5
1.0
1.5
2.0
SGI-110 100 nM
SGI-110 200 nM
SGI-110 50 nM
Fa - Inhibition of Cell Proliferation
Combination Index (CI)
Oxaliplatin at 0.25, 0.5, 1, 2 µM Oxaliplatin at 1, 2, 4, 8 µM
0.1 1 10
0
20
40
60
80
100
No SGI-110
SGI-110 50 nM
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Oxaliplatin (µM)
Inhibition of Cell Proliferation (%)
0.1 1 10
0
20
40
60
80
100
No SGI-110
SGI-110 50 nM
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Oxaliplatin (µM)
Inhibition of Cell Proliferation (%)
0.1 1 10
0
20
40
60
80
100
No SGI-110
SGI-110 50 nM
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Oxaliplatin (µM)
Inhibition of Cell Proliferation (%)
0 0 0 0
0.1 1 10
0
20
40
60
80
100
No SGI-110
SGI-110 50 nM
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Oxaliplatin (µM)
Inhibition of Cell Proliferation (%)
0.1 1 10
0
20
40
60
80
100
No SGI-110
SGI-110 50 nM
SGI-110 100 nM
SGI-110 200 nM
SGI-110 400 nM
Oxaliplatin (µM)
Inhibition of Cell Proliferation (%)
Oxaliplatin (µM) Oxaliplatin (µM) Oxaliplatin (µM) Oxaliplatin (µM)
E
0
SNU-398
Control
1
2
4
0.5
0
25 50 100 200
SGI-110 (nM)
Oxaliplatin (µM)
0
Hep-G2
Control
1
2
0.5
0
25 50 100 200
SGI-110 (nM)
Oxaliplatin (µM)
0.25

34

Figure 2-4. SGI-110 pretreatment sensitizes SNU-398 cells to oxaliplatin treatment. A)
IC
50
values of sorafenib and oxaliplatin in HCC cell lines. B) Schematic of three
treatment schedules. SGI-110 was given in the first 72 h (fresh drug added every 24 h),
and oxaliplatin was added at different time points after SGI-110 pretreatment. C)
Representative images and quantitation of colony formation assay for the 3 treatment
schedules. SNU-398 cells were coated on Day 0 and treated according to schedules 1, 2,
and 3 as indicated. Cells were kept in culture until colonies were observed in controls.
Colonies were stained with crystal violet and imaged. Number of colonies at each
treatment condition was quantified with Image J. Results are shown as Mean ± SD (n=3).
P-values were calculated using Student’s t-test. D) Dose-response curves for oxaliplatin
in 2 sensitive (SNU-398, Hep-G2) and 2 less sensitive (SNU-475, SNU-387) HCC cell
lines pretreated with SGI-110. Cells were treated with SGI-110 for 72 h followed by
oxaliplatin for 72 h, and placed in fresh cell culture media. Considering different
sensitivity of the cell lines, SGI-110 was given at 50, 100, 200, and 400 nM in all cell
lines, while oxaliplatin was given at 0.25 – 4 µM in SNU-398 and Hep-G2, and 0.5 – 8
µM in SNU-475 and SNU-387. MTT assay was performed on Day 12 to assess the
number of live cells under each treatment condition. Inhibition of cell proliferation was
calculated against controls. Data points were shown as Mean ± SD from three
independent experiments. E) Combination Index (CI) of SGI-110-oxaliplatin
combination treatment at different concentrations (non-constant ratio) was calculated
using Chou-Talalay method. CI < 1 indicates synergistic effect of two compounds. F)
Colony formation assay for SGI-110 and oxaliplatin in SNU-398 and Hep-G2 cells. Cells
were treated with SGI-110 for 72 h followed by oxaliplatin for 72 h, and kept in culture
until colonies were observed in PBS-treated control. Colonies were stained and imaged.
Red lines indicate wells with significant synergistic effect from the combination
treatment.

Table 2-1. IC
50
values (µM)
[1]
of anti-cancer drugs in HCC cell lines

[1]
Values are shown as Mean ± S.D. from three independent MTT assay experiments.

Pretreatment with apoptosis inhibitor Z-VAD-fmk and/or necroptosis inhibitor
necrostatin-1 partially rescues SNU-398 cells from SGI-110/oxaliplatin induced anti-
Sensitive Resistant
SNU-398 Hep-G2 SNU-475 SNU-387
Sorafenib 6.5 ± 1.1 13.1 ± 1.6 >30 25.0 ± 2.7
Oxaliplatin 2.5 ± 0.4 0.6 ± 0.1 15 ± 3.6 >30
SGI-110 0.25 ± 0.03 0.43 ± 0.04 53.9 ± 4.2 63.4 ± 5.1

35

cancer activity in MTT assay. (Fig. 2-5) The protection by these two inhibitors suggests
that cell death (apoptosis and necroptosis) is involved in the activity of SGI-110 and
oxaliplatin; however, the anti-cancer activity is not completely dependent on cell death.
Both cytotoxicity and inhibition of cell proliferation play a role in the synergistic effect of
SGI-110 and oxaliplatin.

36

Figure 2-5. Combination of SGI-110 and oxaliplatin induces cell death and inhibition of
cell proliferation in SNU-398 cells. A) Pretreatment with apoptosis inhibitor Z-VAD-fmk
or/and necroptosis inhibitor Necrostatin-1 protects SNU-398 from oxaliplatin-induced
cytotoxicity, but does not provide full protection. Cells were pretreated with inhibitor at
SGI-110 (nM)
Oxaliplatin (µM)
Z-VAD-fmk
(50 µM)
Necrostatin 1
(50 µM)
Pretreatment of inhibitors
given 1h before SGI-110
Pretreatment of inhibitors
given 1h before oxaliplatin
0 25 50 100 200
0 0.0% 6.4% 6.2% 0.7% 3.0%
0.25 -3.3% 2.2% 1.4% 6.9% 3.7%
0.5 -0.6% -1.3% 21.5% 19.0% 6.6%
1 3.5% 1.5% 19.7% 16.9% 5.0%
2 11.9% 38.4% 22.9% 13.3% 4.9%
4 46.4% 24.5% 7.2% 3.5% 3.7%
0 25 50 100 200
0 0.0% 10.9% 4.4% -2.9% -4.7%
0.25 8.0% 5.8% 3.9% -7.2% 1.8%
0.5 -0.4% 7.5% 13.8% 11.4% 5.1%
1 1.8% 5.7% 37.4% 17.1% 4.3%
2 13.7% 34.7% 31.5% 12.9% 4.3%
4 37.7% 17.5% 9.5% 3.2% 2.7%
0 25 50 100 200
0 0.0% 6.2% 1.5% -6.3% -3.1%
0.25 4.3% 4.0% -4.3% -8.6% 3.1%
0.5 -1.0% 4.2% 5.2% 19.0% 6.7%
1 -1.5% 14.4% 28.3% 16.5% 4.8%
2 13.4% 42.1% 37.1% 14.7% 5.9%
4 44.2% 20.2% 11.0% 4.1% 2.8%
0 25 50 100 200
0 0.0% 3.4% -0.5% 1.1% 17.2%
0.25 0.9% 0.6% -0.3% -0.8% 19.7%
0.5 0.7% 2.1% 6.5% 1.5% 11.4%
1 5.3% 1.2% 12.4% 2.4% 2.6%
2 -7.4% 5.9% 7.3% -1.4% -0.6%
4 3.8% 1.0% -0.4% -1.3% -0.1%
0 25 50 100 200
0 0.0% 1.4% 1.1% 4.3% 8.7%
0.25 2.3% 5.2% 4.8% 11.6% 6.8%
0.5 2.4% -3.1% 0.1% 11.9% 3.9%
1 -2.3% -3.3% 13.7% 6.6% 4.1%
2 -5.4% 10.0% 8.2% 1.0% -2.0%
4 3.1% 0.5% -1.8% -0.7% -1.6%
46.4%
0.0%
-8.6%
Protection from cytotoxicity (%)
Z-VAD-fmk
(50 µM)
&
Necrostatin 1
(50 µM)
0 25 50 100 200
0 0.0% 4.1% 4.4% 6.7% 24.5%
0.25 -1.1% 6.0% 3.5% 8.1% 22.2%
0.5 7.3% 2.8% 15.4% 10.1% 14.1%
1 5.2% 0.7% 4.5% 5.4% 3.7%
2 -5.8% 4.8% 1.4% 1.2% -1.7%
4 -0.9% -1.2% -2.2% -3.3% -3.5%
Oxaliplatin (µM)
SGI-110 (nM)
Oxaliplatin (µM) Oxaliplatin (µM)
Oxaliplatin (µM) Oxaliplatin (µM)
1 2 3 4
-20
0
20
40
60
80
100
Oxaliplatin (Ox)
Z-VAD-fmk
pretreatment + Ox
Oxaliplatin (µM)
Inhibition of cell proliferatiion (%)
1 2 3 4
-20
0
20
40
60
80
100
Oxaliplatin (µM)
Inhibition of cell proliferatiion (%)
Oxaliplatin (Ox)
Z-VAD-fmk
pretreatment + Ox
1 2 3 4
-20
0
20
40
60
80
100
Oxaliplatin (Ox)
Z-VAD-fmk
pretreatment + Ox
Oxaliplatin (µM)
Inhibition of cell proliferatiion (%)
1 2 3 4
-20
0
20
40
60
80
100
Oxaliplatin (Ox)
Z-VAD-fmk
pretreatment + Ox
Oxaliplatin (µM)
Inhibition of cell proliferatiion (%)
No SGI-110
100 nM SGI-110 200 nM SGI-110
50 nM SGI-110
Pretreatment of Z-VAD-fmk (50 µM) given 1h before oxaliplatin
A
B
C
Figure S3
Oxaliplatin only
DMSO pretreatment
12
25
50
100
200
12
25
50
100
200
12
25
50
100
0
20
40
60
80
100
Z-VAD-fmk
(µM)
Necrostatin-1
(µM)
Z-VAD-fmk
& Necrostatin-1
1:1 combination(µM)
Oxaliplatin (4 µM)
Inhibition of cell proliferatiion
with Oxaliplatin at 4 µM (%)

37

12-200 µM or 1:1 combination of both inhibitors at 12-100 µM each for 1 h. Oxaliplatin
was then added to a final concentration of 4 µM. After 72 h treatment, cells were left to
recover in fresh media for 4 d and subjected to MTT assay. Number of viable cells from
test wells was compared with DMSO treated controls. Data points are shown as Mean ±
SD from three independent experiments. Different concentrations of inhibitors were used
to determine the optimal dose for maximal protection. B) Pretreatment with apoptosis
inhibitor Z-VAD-fmk or/and necroptosis inhibitor Necrostatin-1 protects SNU-398 from
SGI-110/oxaliplatin induced cytotoxicity, but does not provide full protection. Cells were
treated with SGI-110 at 25-200 nM for 72 h, then with oxaliplatin at 0.25-4 µM for 72 h
in fresh media. Apoptosis inhibitor Z-VAD-fmk or/and necroptosis inhibitor Necrostatin-
1 were given at 50 µM either 1h before SGI-110 or 1h before oxaliplatin treatment. After
treatment, cells were left to recover in fresh media for 4 d and subjected to MTT assay.
Number of viable cells from test wells was compared with DMSO or inhibitor treated
controls. Treatment with inhibitors at 50 µM does not induce basal protection or
cytotoxicity compared with DMSO controls. Protection from cytotoxicity was calculated
by subtracting data of inhibitor pretreated samples from data of non-pretreatment samples
at the same SGI-110/oxaliplatin conditions. Data points are shown as Mean from three
independent experiments. C) Dose-response curves for oxaliplatin with 72 h priming of
SGI-110 and 1 h pretreatment of apoptosis inhibitor Z-VAD-fmk. Cells were treated with
SGI-110 for 72 h. After removal of SGI-110, cells were pretreated with Z-VAD-fmk at
50 µM for 1 h, then oxaliplatin was added to the treatment at different concentrations.
After treatment, cells were left to recover in fresh media for 4 d and subjected to MTT
assay. Inhibition of cell proliferation was calculated against controls. Data points are
shown as Mean ± SD from three independent experiments.

2.6 SGI-110 and oxaliplatin exhibit significant anti-cancer effect in in vivo HCC
model
2.6.1 Selection of in vivo treatment
Since toxicity is an additional factor for combination treatment in vivo as
compared with in vitro studies, we chose to give weekly oxaliplatin treatment and
precede the first oxaliplatin treatment with SGI-110 in a 14-day cycle.
For oxaliplatin, weekly treatment has been widely used for in vivo studies, while
it is mostly given once every two weeks in the clinic. In in vivo studies, its activity might
vary in different animal models. When used twice a week at 5 mg/kg (i.p.), oxaliplatin
can inhibit growth of HCT-116 derived xenograft in BALB/c nude mice. (Zeng et al.,

38

2014) When it was used at 5 mg/kg for weekly injections (i.p.) in HT-29 xenograft
bearing C.B.17 SCID model, a mild delay in tumor growth was observed. (Selvakumaran
et al., 2013) However, in another athymic nude mice model with HT-29 xenograft,
weekly injection (i.v.) of oxaliplatin at 6.7 mg/kg as a single agent did not have any
significant effect on tumor size. (Gaur et al., 2014) To best evaluate clinically relevant
schedules and doses, as well as to avoid potential toxicity introduced by the combination,
we chose to give oxaliplatin at 5 mg/kg once a week.
Decitabine or SGI-110 can be toxic as a long-term repeated treatment. Based on
previous studies for the compound, QD5 treatment or bi-weekly treatment is used to
achieve therapeutic effect without systemic toxicity. In SCID mice, SGI-110 given as
daily treatment for 5 consecutive days at 3 mg/kg showed consistent and robust
hypomethylation and induction of Cancer Testis Antigen (CTA) genes in leukemia as
well as ovarian cancer xenograft without significant toxicity (Srivastava et al., 2014;
Srivastava et al., 2015), suggesting good balance between efficacy and toxicity at this
treatment schedule. In the ongoing phase II clinical trial of SGI-110 for the treatment of
advance hepatocellular carcinoma (HCC) (NCT01752933), SGI-110 is proposed to be
given daily on Days 1-5 every 28 days. In order to best recapitulate clinical settings, and
evaluate SGI-110 as a single agent as well as in combination with oxaliplatin in the same
experimental setting, we decided to give SGI-110 at 2 mg/kg on Days 1-5 in the 14-day
cycle, rather than a more frequent schedule, to achieve the best potential efficacy without
toxicity.
Considering the shorter in vivo half-life of SGI-110 (4h in vivo vs. 21 h in vitro)
(Yoo et al., 2007; Tellez et al., 2014), the first oxaliplatin dose in the 14-day cycle was

39

given 4 h after the last repetitive SGI-110 treatment on Day 5, matching the optimal
oxaliplatin treatment schedule discovered in Fig. 2-4.

2.6.2 SGI-110 as single treatment or in combination with oxaliplatin in SNU-398
xenograft model
To investigate the in vivo antitumor efficacy of SGI-110 alone and in combination
with oxaliplatin, xenograft studies were performed in athymic nude mice. Subcutaneous
human HCC xenografts from SNU-398 cells were established on the dorsal flank of the
immunodeficient mice, and treated with SGI-110, oxaliplatin, combination or vehicle
until tumor size in the group reached 2000 mm
3
(Fig. 2-6A). Although oxaliplatin (5
mg/kg) did not show efficacy, SGI-110 (2 mg/kg) treatment significantly suppressed
growth of tumors after 15 days of treatment. Combination with oxaliplatin further
delayed tumor growth, where significant difference in tumor sizes was achieved as early
as day 8. On day 19, when average tumor sizes in control and oxaliplatin treatment
groups passed the experiment endpoint of 2000 mm
3
, the average tumor size was 1010 ±
247 mm
3
(p = 0.0039) for SGI-110 treatment alone, and only 391 ± 100 mm
3

(p=0.0001) for combination with oxaliplatin. SGI-110 treatment was able to delay
endpoint from day 19 to day 26, and combination treatment further delayed the endpoint
to day 31 (Fig. 2-6B), indicating substantial survival benefits from the treatments.
SGI-110 treatment as a single agent and in combination with oxaliplatin
significantly decreased Ki67 levels in tumors, suggesting inhibition of cell proliferation
(Fig. 2-6C). Shown by the increasing slope of the tumor size curve in Fig. 6B, tumor

40

growth rates increased regardless of treatments as the study progressed. All tumor
samples were collected at the average size of 2000 mm
3
at their respective endpoints.
Rather than showing similar molecular profiles, SGI-110 and combination treated tumors,
which were collected at later time points of the study, contained a lower percentage of
proliferating cells than controls.

Figure 2-6. Single agent SGI-110 and its combination treatment with oxaliplatin inhibit
tumor growth in SNU-398 liver cancer xenograft model. A) Treatment schedule. Mice
with SNU-398 tumors were randomized into 4 treatment groups (n=5) and received
vehicle, oxaliplatin (5 mg/kg weekly), SGI-110 (2 mg/kg from day 1 to day 5 in a 2-week
B
C
Control
Oxaliplatin
SGI-110
SGI-110+Ox
0
10
20
30
40
50
p = 0.015
p = 0.018
Ki67 Index (%)
Control$(20X)$ OxaliplaBn$
SGI;110$ SGI;110$+$Ox$
0 5 10 15 20 25 30
0
1000
2000
3000
Vehicle
Oxaliplatin (5 mg/kg)
SGI-110 (2 mg/kg)
SGI-110 + Oxaliplatin
Day
Tumor Volume(mm
3)
*
*
** ***
*
**
**
Figure 6-1
Day 1-5 Day 8-12 Day 15-19 Day 22-26 Day 29-31
Control
Treatment schedules
SGI-110, 2 mg/kg (S)
Oxaliplatin, 5 mg/kg (Ox)
A
Oxaliplatin
SGI-110
S + Ox
DNMT1
Survivin
PCNA
Control SGI-110 Oxaliplatin
SGI-110 +
Oxaliplatin
SNU-398 tumor lysate
β-Tubulin
1 2 3 1 2 3 1 2 3 1 2 3
170 kDa
17 kDa
34 kDa
55 kDa

41

cycle) or combination treatment. Mice were sacrificed when tumor size reached 2000
mm
3
. B) Tumor sizes were significantly reduced in mice treated with SGI-110, and
further delayed in mice with SGI-110 and oxaliplatin combination treatment. Statistical
significance was calculated using Student’s t-test. Error bars indicate Mean ± SEM. and
*p < 0.05, ****p < 0.0001. C) Ki67 immunohistochemistry staining in tumor sections.
Ki67 index was calculated as percentage of Ki67 positive cells in total number of cells in
the field (n = 6, 3 fields of view from 2 tumors per group). Graphical data is presented as
Mean ± SD. P-values were calculated using Student’s t-test.

No systemic symptoms of toxicity such as weakness, weight loss or lethargy were
observed in any treatment group (Fig. 2-7A). H&E stained organ sections of liver,
kidney, heart, lung, spleen and pancreas did not reveal major histopathological changes,
further confirming the safety of the treatments (Fig. 2-7B).

A
B
0 5 10 15 20 25 30
10
15
20
25
30 Vehicle
Oxaliplatin (5 mg/kg)
SGI-110 (2 mg/kg)
SGI-110 + Oxaliplatin
Day
Weight (g)
Liver" Kidney" Heart" Lung" Spleen" Pancreas"
Control"
Oxalipla@n"
SGIC110"
SGIC110"+"Ox"
Figure 6-2

42

Figure 2-7. SGI-110 and its combination treatment with oxaliplatin did not exert
systemic toxicity in vivo. A) Animal weights did not change significantly during the
course of treatment. Error bars indicate Mean ± SEM. B) Representative micrographs of
hematoxylin and eosin (H&E)-stained organ sections. Images were taken with an
Olympus IX83 inverted microscope at 20X magnification. In the histopathology study,
no significant morphological changes were detected in major organs after SGI-110 or
combination treatment.

In the in vivo study, our treatments were able to delay growth of HCC tumor and
provide survival benefits, however, neither complete suppression nor tumor regression
was observed with our constant low doses of both compounds. Based on our in vitro data,
we believe that increasing doses of oxaliplatin or SGI-110 might well improve the
therapeutic effect of the combination treatment, and this can be further investigated in
future preclinical or clinical studies in pursuit of better antitumor efficacy and definition
of the therapeutic window. At the current settings, combination treatment did not induce
any toxicity in our mice. The significant delay in tumor growth with favorable safety
profile indicates great potentials of this combination treatment for HCC patients, and
warrants further clinical evaluation of the regimen.

2.7 Bru-seq identified major signaling regulations by SGI-110 and oxaliplatin
treatment
2.7.1 Major signaling regulations revealed by GSEA
To better understand the potential mechanism of SGI-110 and oxaliplatin synergy,
we performed Bru-seq to examine the global changes in transcription in HCC cells. SNU-
398 cells were treated with SGI-110 at 100 nM for 72 h, followed by oxaliplatin at 3 µM

43

for 4 h. Nascent RNA was labeled by bromouridine 30 min before sample collection, and
subjected to RNA sequencing. 22984 genes were analyzed and filtered by gene size
(>300 bp) and synthesis level (RPKM > 0.5). The gene list were then pre-ranked by fold
change of treatment over PBS-treated control and subjected to Gene Set Enrichment
Analysis (GSEA) (Table. 2-2 to 2-7). Gene sets with false discovery rate (FDR) q-value
lower than 0.25 are considered true enrichment.

44

Table 2-2. Top 30 gene sets positively associated with SGI-110 treatment

Table S1. Top 30 gene sets positively associated with SGI-110 treatment

NAME SIZE ES NES
NOM
p-val
FDR
q-val
FWER
p-val
RANK
AT MAX
JAEGER_METASTASIS_DN 34 0.609 2.195 0.000 0.054 0.051 765
MISSIAGLIA_REGULATED_BY_METHYLATION_UP 49 0.560 2.177 0.000 0.036 0.067 1415
LABBE_WNT3A_TARGETS_DN 21 0.681 2.172 0.000 0.026 0.072 220
WU_CELL_MIGRATION 50 0.564 2.145 0.000 0.028 0.106 740
YAO_TEMPORAL_RESPONSE_TO_PROGESTERONE_CLUSTER_6 28 0.596 2.033 0.000 0.079 0.320 884
GSE18791_CTRL_VS_NEWCASTLE_VIRUS_DC_2H_DN 71 0.479 1.970 0.000 0.139 0.564 932
KINSEY_TARGETS_OF_EWSR1_FLII_FUSION_DN 121 0.435 1.945 0.000 0.152 0.654 904
GAUSSMANN_MLL_AF4_FUSION_TARGETS_E_UP 21 0.615 1.929 0.006 0.159 0.716 358
VANHARANTA_UTERINE_FIBROID_DN 19 0.619 1.915 0.003 0.166 0.780 106
ODONNELL_TARGETS_OF_MYC_AND_TFRC_UP 16 0.633 1.887 0.008 0.201 0.867 1169
LEF1_UP.V1_DN 44 0.499 1.886 0.003 0.185 0.868 500
GSE360_CTRL_VS_L_DONOVANI_DC_DN 60 0.462 1.878 0.000 0.181 0.889 671
DUTERTRE_ESTRADIOL_RESPONSE_24HR_DN 217 0.377 1.867 0.000 0.186 0.916 974
MODULE_92 15 0.645 1.863 0.005 0.180 0.922 545
MEMBRANE_LIPID_BIOSYNTHETIC_PROCESS 27 0.546 1.860 0.003 0.173 0.925 621
BHAT_ESR1_TARGETS_VIA_AKT1_DN 30 0.530 1.858 0.003 0.165 0.928 1329
MULTI_ORGANISM_PROCESS 34 0.509 1.842 0.000 0.180 0.956 787
GSE18791_UNSTIM_VS_NEWCATSLE_VIRUS_DC_1H_UP 16 0.635 1.837 0.003 0.178 0.961 800
LEIN_CHOROID_PLEXUS_MARKERS 26 0.551 1.828 0.000 0.185 0.965 62
MODULE_493 36 0.497 1.810 0.003 0.207 0.983 1730
V$GATA_Q6 45 0.471 1.804 0.000 0.207 0.987 727
BOYLAN_MULTIPLE_MYELOMA_D_CLUSTER_DN 21 0.560 1.799 0.017 0.208 0.988 948
NAKAYAMA_FRA2_TARGETS 23 0.555 1.791 0.014 0.213 0.992 760
GSE25087_FETAL_VS_ADULT_TCONV_DN 94 0.406 1.789 0.000 0.209 0.992 1149
IL2_UP.V1_DN 35 0.496 1.789 0.003 0.201 0.992 469
IVANOVA_HEMATOPOIESIS_STEM_CELL_LONG_TERM 124 0.392 1.788 0.000 0.193 0.992 912
MIKKELSEN_MEF_HCP_WITH_H3_UNMETHYLATED 17 0.605 1.787 0.006 0.188 0.993 868
SEKI_INFLAMMATORY_RESPONSE_LPS_UP 15 0.613 1.780 0.013 0.192 0.994 1143
GLYCEROPHOSPHOLIPID_BIOSYNTHETIC_PROCESS 17 0.591 1.763 0.006 0.215 1.000 621
LIPOPROTEIN_BIOSYNTHETIC_PROCESS 16 0.599 1.763 0.005 0.208 1.000 566

45

Table 2-3. Top 30 gene sets negatively associated with SGI-110 treatment

46

Table 2-4. Top 30 gene sets positively associated with oxaliplatin treatment

47

Table 2-5. Top 30 gene sets negatively associated with oxaliplatin treatment

48

Table 2-6. Top 30 gene sets positively associated with combination treatment

49

Table 2-7. Top 30 gene sets negatively associated with combination treatment

50

In SGI-110 treatment, we identified enrichment of gene sets that were similarly
upregulated by decitabine treatment in pancreatic cancer cells (Fig. 2-8) (Missiaglia et
al., 2005). This finding validates the similar transcriptional regulation by SGI-110 and
decitabine, and it also suggests overlap of methylation-regulated genes in HCC and
pancreatic cancer.

Figure 2-8. Enrichment plot of the gene set MISSIAGLIA_REGULATED_BY_
METHYLATION_UP in the pre-ranked gene list from SGI-110 treatment. The gene set
comprising genes upregulated in PaCa44 and CFPAC1 cells (pancreatic cancer) after
treatment with decitabine is over-represented at the top of the pre-ranked gene list,
suggesting positive correlation between SGI-110 and decitabine treatment in the two
settings.

As a demethylating agent, SGI-110 is expected to change cellular DNA
methylation profiles to restore a more normal transcriptome leading to antitumor activity.
To elucidate the potential mechanisms of SGI-110 as a single agent and in combination
with oxaliplatin, we employed Bru-seq, which captures gene synthesis without RNA
post-transcription processing, allowing direct assessment on gene transcription status
after epigenetic modulation. Instead of studying only the select genes or targets predicted
Figure S4

51

by previous studies, Bru-seq provides an un-biased method to explore changes in nascent
RNA synthesis for the whole genome.
As revealed by GSEA, simultaneous inhibition of WNT/β-catenin, IGF, and EGF
signaling contributed to the synergistic effect of the combination treatment.

2.7.2 WNT
Using the false discovery rate (FDR, q-value) of 0.25 as the cut-off, oxaliplatin
uniquely upregulated a cluster of hypermethylated genes characterized in AML, but no
other gene sets were observed as positively associated with the treatment (Table 2-4). In
cells treated with SGI-110, the WNT3A gene set was among the most highly upregulated
sets by SGI-110 alone, and it was the only significantly enriched gene set in the
combination treatment (Fig. 2-9A, Table 2-2, Table 2-6). The heat map showing relative
synthesis levels indicates upregulated expression of listed genes by SGI-110 or its
combination with oxaliplatin (Fig. 2-9B). WNT/β-catenin signaling is a major signature
pathway of liver cancer; nearly half of HCC patients exhibit activation of the pathway
(Lachenmayer et al., 2012). The enriched gene set represents genes downregulated
following WNT3A treatment (Labbe et al., 2007). Increased synthesis of genes
negatively regulated by WNT3A suggests potential inhibition of WNT/β-catenin
signaling by SGI-110 treatment. In support of this hypothesis, we found that the
expression levels of the endogenous β-catenin inhibitor, E-cadherin, were gradually
upregulated after SGI-110 treatment. In contrast, the β-catenin target gene, survivin, was
downregulated, consistent with the findings from GSEA (Fig. 2-9C).

52

Figure 2-9. Bru-Seq reveals inhibition of Wnt signaling by SGI-110 and its combination
with oxaliplatin. A) Enrichment plots of LABBE_WNT3A_TARGETS_DN gene set
over-represented on the top of pre-ranked gene lists from both SGI-110 and combination
treatment. B) Heat map for relative transcription levels of genes in the
LABBE_WNT3A_TARGETS_DN gene set over-represented on the top of pre-ranked
gene lists from both drugs and combination treatment. C) In SNU-398 and Hep-G2 cells,
E-Cadherin and survivin were modulated by SGI-110 in a time-dependent manner. Cells
were treated with SGI-110 at 100 nM for up to 72 h. Cell culture media containing SGI-
110 were removed and cells were left to recover in fresh cell media. Samples were
collected at indicated time points from 24 h to 168 h after SGI-110 removal.

GSEA of the Bru-seq data revealed inhibition of WNT/β-catenin signaling with
SGI-110 treatment. It is noteworthy that synergistic effects of SGI-110 and oxaliplatin
were only observed in cell lines with mutation-induced WNT/β-catenin pathway activity
(Hep-G2 and SNU-398 both possess a mutation in CTNNB1, and SNU-475 has a deletion
in AXIN1(Satoh et al., 2000; Yuzugullu et al., 2009)), but not in SNU-387, which has no
such genetic aberrations. Suggested by these data, the WNT/β-catenin signaling might be
C
E-Cadherin
Survivin
Treatment duration (h)
Time after SGI-110 removal (h)
- 24 48 72 72 72 72 72 72 -
- - - - 24 48 72 96 168 -
PCNA
SNU-398
(SGI-110 100 nM)
130 kDa
17 kDa
34 kDa
130 kDa
17 kDa
34 kDa
E-Cadherin
Survivin
Treatment duration (h)
Time after SGI-110 removal (h)
- 24 48 72 72 72 72 72 72 72 72
- - - - 24 48 72 96 120 144 168
PCNA
Hep-G2
(SGI-110 100 nM)
A B
SGI-110 + Oxaliplatin SGI-110
Figure 3-1
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
PPARD
TXNIP
CLDN11
TSHZ1
TMEM132A
ORC2
CASP7
PTS
FMR1
SLC30A1
PANK1
UGCG
CHRNA7
TTC37
CHEK1
PTER
GPR19
NAB1
PGK1
GSTA4
HEBP1
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
PPARD
TXNIP
CLDN11
TSHZ1
TMEM132A
ORC2
CASP7
PTS
FMR1
SLC30A1
PANK1
UGCG
CHRNA7
TTC37
CHEK1
PTER
GPR19
NAB1
PGK1
GSTA4
HEBP1
Control
LABBE_WNT3A_TARGETS_DN
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
TXNIP
ATP1B1
CTSL1
NEU1
SQSTM1
BLZF1
BCL11A
DUSP1
PANX1
MAP1LC3B
YPEL5
MIR22HG
TIMP1
LGMN
UBE2H
LY6E
CTSB
BECN1
IFNGR1
PSEN1
PSAP
ATOX1
SDC4
ATP6V0E1
PRDX5
PON2
ARF4
DNAJB2
SCAMP2
KDM2A
RRBP1
RAB9A
SAT1
STAT1
KIAA0247
RNF103
CAMSAP2
USP9X
DUSP6
ATP6V1G1
MFGE8
TMEM59
HIST1H2BK
SOD2
ST3GAL5
PLAUR
ARL6IP5
CDC123
GADD45A
Oxaliplatin
SGI-110
SGI + Ox
LABBE_WNT3A_TARGETS_DN

53

a major targeted pathway as well as a potential patient selection marker for SGI-110
treatment.

2.7.3 Gene Sets for cancer gene neighborhood
Among the gene sets negatively associated with treatments (Fig. 3E),
GNF2_CCNB2 was selected as a representative for cancer gene neighborhood sets,
where gene synthesis was downregulated by SGI-110 alone or with the combination
treatment (Fig. 2-10A and B). Another interesting finding is the downregulation of cancer
gene neighborhood sets by SGI-110. Defined by correlated expression with certain
cancer-associated genes in the human tissue compendia, such computational gene sets
represent cancer-oriented features of the transcriptome. It is still unclear whether it is the
direct effect from SGI-110 or secondary effect from a primary modulation of upstream
targets; however, this significant enrichment of cancer-associated gene sets at the bottom
of our pre-ranked gene list implies an overall suppression of cancer-specific features by
SGI-110 treatment.

54

Figure 2-10. Inhibition of cancer related genes by SGI-110 and its combination with
oxaliplatin. A) Enrichment plots of the GNF2_CCNB2 gene set over-represented on the
bottom of pre-ranked gene lists from both SGI-110 and combination treatment. Several
computational gene sets for cancer gene neighborhood were enriched in the same
manner, and GNF2_CCNB2 was selected as a representative. B) Heat map for relative
transcription levels of genes in the GNF2_CCNB2 gene sets, which were over-
represented on the bottom of pre-ranked gene lists from both SGI-110 and combination
treatment.

2.7.4 IGF/EGF
Interestingly, the combination treatment induced a unique enrichment of the
PACHER_TARGETS_OF_IGF1_AND_IGF2_UP (Fig. 2-11A) and MAGASHIMA
_EGF_ SIGNALING_UP gene sets (Fig. 2-11C), downregulating target genes from
Insulin-like growth factor (IGF) (Fig. 2-11B) and epidermal growth factor (EGF)
signaling (Fig. 2-11D), whose activation or overexpression are often observed in HCC
(Psyrri et al., 2012).
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
TXNIP
ATP1B1
CTSL1
NEU1
SQSTM1
BLZF1
BCL11A
DUSP1
PANX1
MAP1LC3B
YPEL5
MIR22HG
TIMP1
LGMN
UBE2H
LY6E
CTSB
BECN1
IFNGR1
PSEN1
PSAP
ATOX1
SDC4
ATP6V0E1
PRDX5
PON2
ARF4
DNAJB2
SCAMP2
KDM2A
RRBP1
RAB9A
SAT1
STAT1
KIAA0247
RNF103
CAMSAP2
USP9X
DUSP6
ATP6V1G1
MFGE8
TMEM59
HIST1H2BK
SOD2
ST3GAL5
PLAUR
ARL6IP5
CDC123
GADD45A
GNF_CCNB2
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
PLK4
MCM4
ZWINT
SMC2
MCM2
PTTG1
CENPF
FEN1
CKAP2
KIF20A
HMGB2
RRM1
ASPM
RRM2
GMNN
SMC4
MKI67
UBE2S
KIF4A
ESPL1
AURKB
CDCA8
KIF2C
PLK1
CCNB2
CKS1B
TOP2A
H2AFX
RAN
BIRC5
CDC20
CENPE
HMMR
WHSC1
NUSAP1
NCAPH2
TYMS
TTK
KIF11
SPAG5
CKS2
PPRC1
AURKA
DLGAP5
RACGAP1
TPX2
CDK1
MELK
FOXM1
SHCBP1
CDCA3
RFC4
NDC80
CCNA2
PCNA
UBE2C
KIF18B
PBK
BUB1B
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
PLK4
MCM4
ZWINT
SMC2
MCM2
PTTG1
CENPF
FEN1
CKAP2
KIF20A
HMGB2
RRM1
ASPM
RRM2
GMNN
SMC4
MKI67
UBE2S
KIF4A
ESPL1
AURKB
CDCA8
KIF2C
PLK1
CCNB2
CKS1B
TOP2A
H2AFX
RAN
BIRC5
CDC20
CENPE
HMMR
WHSC1
NUSAP1
NCAPH2
TYMS
TTK
KIF11
SPAG5
CKS2
PPRC1
AURKA
DLGAP5
RACGAP1
TPX2
CDK1
MELK
FOXM1
SHCBP1
CDCA3
RFC4
NDC80
CCNA2
PCNA
UBE2C
KIF18B
PBK
BUB1B
SGI-110 + Oxaliplatin SGI-110
GNF_CCNB2
A
Figure 3-2
Control
Oxaliplatin
SGI-110
SGI + Ox
B

55

Figure 2-11. Inhibition of IGF and EGF signaling by the combination of SGI-110 with
oxaliplatin. A) Enrichment plot of IGF gene set over-represented on the bottom of pre-
ranked gene list from the combination treatment only. B) Heat maps for relative
transcription levels of genes in the IGF gene sets. C) Enrichment plot of EGF gene set
over-represented on the bottom of pre-ranked gene list from the combination treatment
only. D) Heat maps for relative transcription levels of genes in the EGF gene sets.

In the combination treatment, the EGF and IGF signaling pathways were uniquely
inhibited as revealed by Bru-seq, implying a potential mechanism for the synergistic
effect of SGI-110 and oxaliplatin. Considering the substantial molecular changes in liver
cancer, targeting multiple signature pathways with one treatment regimen is a plausible
strategy. In HCC, overexpression of EGFR proteins and amplification of the EGFR gene
were confirmed with Immunohistochemistry and FISH studies by Buckley et al. (Buckley
et al., 2008). However, single treatment targeting EGFR with gefitinib (O'Dwyer PJLD;
Kauh, 2006), cetuximab (Zhu et al., 2007) or lapatinib (Bekaii-Saab et al., 2009) has not
shown efficacy in Phase II studies. In a recent Phase III clinical trial of erlotinib in
combination with sorafenib enrolling 720 advanced HCC patients, no significant
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
PHLDA1
ANKRD57
TIPARP
KLF10
DUSP1
DNAJB1
BCL10
HES1
KBTBD2
AEN
NAB2
RYBP
CYR61
FOS
MIR22HG
TRIB1
SPRED2
MCL1
TNFRSF12A
BHLHE40
KLF6
EGR1
MYC
JUNB
DLX2
IER2
ID1
ID3
ZFP36
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
PHLDA1
ANKRD57
TIPARP
KLF10
DUSP1
DNAJB1
BCL10
HES1
KBTBD2
AEN
NAB2
RYBP
CYR61
FOS
MIR22HG
TRIB1
SPRED2
MCL1
TNFRSF12A
BHLHE40
KLF6
EGR1
MYC
JUNB
DLX2
IER2
ID1
ID3
ZFP36
PACHER_TARGETS_OF
_IGF1_AND_IGF2_UP
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
IFRD1
SLC1A4
SERTAD2
MID1
SLC7A5
SLC7A11
LARP1B
VEGFA
PSPH
ASNS
MAP1B
PHGDH
SLC3A2
DDIT3
SDF2L1
HERPUD1
BHLHE40
SESN2
GADD45A
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
IFRD1
SLC1A4
SERTAD2
MID1
SLC7A5
SLC7A11
LARP1B
VEGFA
PSPH
ASNS
MAP1B
PHGDH
SLC3A2
DDIT3
SDF2L1
HERPUD1
BHLHE40
SESN2
GADD45A
NAGASHIMA_EGF_SIGNALING_UP
relative
-3 0 3
Control
Oxaliplatin
SGI-110
SGI + Ox
Id Description
TXNIP
ATP1B1
CTSL1
NEU1
SQSTM1
BLZF1
BCL11A
DUSP1
PANX1
MAP1LC3B
YPEL5
MIR22HG
TIMP1
LGMN
UBE2H
LY6E
CTSB
BECN1
IFNGR1
PSEN1
PSAP
ATOX1
SDC4
ATP6V0E1
PRDX5
PON2
ARF4
DNAJB2
SCAMP2
KDM2A
RRBP1
RAB9A
SAT1
STAT1
KIAA0247
RNF103
CAMSAP2
USP9X
DUSP6
ATP6V1G1
MFGE8
TMEM59
HIST1H2BK
SOD2
ST3GAL5
PLAUR
ARL6IP5
CDC123
GADD45A
A
Figure 3-2
Control
Oxaliplatin
SGI-110
SGI + Ox
B
SGI-110 + Oxaliplatin SGI-110 + Oxaliplatin
IGF1_AND_IGF2_UP EGF_SIGNALING_UP
Control
Oxaliplatin
SGI-110
SGI + Ox
C
D

56

improvement of survival was observed with erlotinib plus sorafenib compared to placebo
plus sorafenib (median overall survival of 9.5 vs. 8.5 months), and disease control rate
was even significantly lower (43.9% vs. 52.5%) (Zhu et al., 2015). This negative trial
further ensures the refractory nature of HCC and suggests that EGFR inhibition is not
sufficient for treatment of HCC. For the Insulin-like growth factor (IGF) axis,
overexpression or aberrant activity of the pathway has been reported in HCC (Yang et al.,
2003; Hopfner et al., 2006; Desbois-Mouthon et al., 2009). Although preclinical studies
suggested growth inhibitory effects of the IGF-1R monoclonal antibody cixutumumab
(Tovar et al., 2010), a Phase II study with cixutumumab monotherapy did not show
clinically meaningful efficacy in unselected HCC patient populations (Abou-Alfa et al.,
2014). Interestingly, treatment with the IGF-1R antibody AVE1642 caused activation of
HER3 in an EGFR-dependent manner counteracting its growth inhibitory effects. When
AVE1642 was combined with the EGFR inhibitor gefitinib, significant reduction of HCC
cell viability was achieved (Desbois-Mouthon et al., 2009). On the other hand, activation
and nuclear translocation of IGF-1R was observed in an induced gefitinib-resistant HCC
cell line (Bodzin et al., 2012). These studies highlight the compensatory nature of the two
signature pathways, suggesting that simultaneous inhibition of EGFR and IGF-1R could
be effective in HCC treatments. In our studies, the combination of SGI-110 and
oxaliplatin suppressed activity of both pathways in addition to the inhibition of the
expression of WNT/β-catenin signaling genes. Simultaneous suppression of these three
major signature pathways supports the great potential for the combination of SGI-110 and
oxaliplatin in treating HCC.

57

2.8 Validation of biomarkers DNMT1, ephrin-B2 and survivin in HCC cell lines
and HCC tumors
DNMT1 and survivin expression levels changed after SGI-110 treatment in a
time-dependent manner as discussed above. To address their potential as PD markers, we
evaluated the expression levels of these proteins over different times of SGI-110 and/or
oxaliplatin treatments (Fig. 2-12A). While oxaliplatin treatment did not affect DNMT1
protein levels in SNU-398 cells (Fig. 2-12B), SGI-110 treatment reduced DNMT1 levels
in these cells. Following drug removal, DNMT1 protein levels started to recover from the
SGI-110-induced depletion within 72 h. Importantly, when oxaliplatin was given after
SGI-110 pretreatment, no recovery of DNMT1 protein levels occurred.
Survivin is a member of the inhibitor of apoptosis family that directly inhibits the
apoptotic process and promotes cell survival (Cheung et al., 2013). As a member of the
inhibitor of apoptosis protein (IAP) family, survivin is overexpressed in most cancers.
Inhibition of survivin promotes cell death in cancer cells (Kelly et al., 2011). In liver
cancer, survivin was identified as a target gene of Wnt/β-catenin pathway (Gedaly et al.,
2014). Survivin levels were downregulated after long-term SGI-110 treatment as
mentioned above. Interestingly, survivin was not affected within 72 h of SGI-110 or
oxaliplatin treatment, but significant decrease was detected within 48 h of the
combination treatment, suggesting that the combination treatment leads to a rapid
disruption of the survival signaling in these cells. Since cell death is involved in the anti-
cancer activity of the combination, we also investigated levels of the apoptotic markers
cleaved PARP and caspase 3 after 72 h combination treatment. While significant dose-
dependent decrease in DNMT1 and survivin levels were observed with the combination,

58

only mild increase of the two apoptotic markers were detected (Fig. 2-13), suggesting
that DNMT1 and survivin are more robust as biomarkers in SNU398 this early in the
treatment.
To further validate these potential markers for the combination treatment, we
performed the same experiment in the most sensitive cell line Hep-G2, and the least
sensitive cell lines SNU-475 and SNU-387. Similarly, the combination treatment
decreased the levels of survivin and blocked DNMT1 protein recovery following SGI-
110 removal in Hep-G2 cells (Fig. 2-12C). In order to achieve similar blockade of
DNMT1 recovery and decreased survivin levels in the oxaliplatin-resistant SNU-475
cells, higher doses were required when compared to the sensitive SNU-398 cells (Fig. 2-
12D). In the other less sensitive cell line SNU-387, where no significant synergism was
detected, DNMT1 and survivin levels showed different responses after treatment (Fig. 2-
14). There was no difference of DNMT1 and survivin levels between single and
combination treatments in this resistant cell line.

59

Figure 2-12. Oxaliplatin treatment downregulates DNMT1 and survivin levels in SGI-
110 pre-treated cells. A) Schematic for SGI-110 and oxaliplatin treatment. Cells were
treated with SGI-110 for 72 h, changed to fresh media and treated with oxaliplatin for up
to 72 h. Samples were collected every 24 h after SGI-110 treatment to study the time
course effects of SGI-110 and oxaliplatin treatments. Control, oxaliplatin, SGI-110 and
combination treated cell lysates were blotted for DNMT1 and survivin. Representative
results are shown for B) SNU-398, C) Hep-G2 and D) SNU-475 cells.
170 kDa
17 kDa
34 kDa
170 kDa
17 kDa
34 kDa
B
C
A
DNMT1
Survivin
PCNA
100 nM SGI-110 (h)
1 µM Oxaliplatin (h)
SNU-398
0 0 0 0
0 0 0 0
0 0 0 0
0 24 48 72
72 72 72 72
0 24 48 72
Time after SGI-110 removal (h) 0 24 48 72 0 24 48 72 0 24 48 72
72 72 72 72
0 24 48 72
0 0 0 0
Control Oxaliplatin SGI-110 SGI-110 + Oxaliplatin
DNMT1
Survivin
PCNA
100 nM SGI-110 (h)
0.5 µM Oxaliplatin (h)
Hep-G2

0 0 0 0
0 0 0 0
0 0 0 0
0 24 48 72
72 72 72 72
0 24 48 72
Time after SGI-110 removal (h) 0 24 48 72 0 24 48 72 0 24 48 72
72 72 72 72
0 24 48 72
0 0 0 0
Control Oxaliplatin SGI-110 SGI-110 + Oxaliplatin
Coat cells at 100,000/well
Add SGI-110 every 24h
Change media and treat
with oxaliplatin (72h)
Collect cell lysates for western
Figure 5
DNMT1
Survivin
PCNA
400 nM SGI-110 (h)
8 µM Oxaliplatin (h)
SNU-475
0 0 0 0
0 0 0 0
0 0 0 0
0 24 48 72
72 72 72 72
0 24 48 72
Time after SGI-110 removal (h) 0 24 48 72 0 24 48 72 0 24 48 72
72 72 72 72
0 24 48 72
0 0 0 0
Control Oxaliplatin SGI-110 SGI-110 + Oxaliplatin
D
170 kDa
17 kDa
34 kDa

60

Figure 2-13. Oxaliplatin treatment does not affect levels of cleaved caspase 3 or PARP
significantly in SGI-110 pre-treated SNU-398 cells, while decreased levels of DNMT1
and survivin are observed after 72 h combination treatment. Cells were treated with SGI-
110 for 72 h, changed to fresh media and treated with oxaliplatin at 0.25-2 µM for 72 h.
Samples were collected after treatment to study the dose-dependent effect of SGI-110 and
oxaliplatin treatment. Control, oxaliplatin, SGI-110 and combination treated cell lysates
were blotted for DNMT1, ephrin B2, survivin, cleaved caspase 3 and PARP.

Figure 2-14. Oxaliplatin treatment does not affect DNMT1 or survivin levels in SGI-110
pre-treated SNU-387 cells. Cells were treated with SGI-110 for 72 h, changed to fresh
media and treated with oxaliplatin for up to 72 h. Samples were collected every 24 h after
SGI-110 treatment to study the time course effect of SGI-110 and oxaliplatin treatment.
Control, oxaliplatin, SGI-110 and combination treated cell lysates were blotted for
DNMT1 and survivin.
Figure S7
DNMT1
SGI-110 pretreatment
(100 nM, 72 h)
- + - + - + - + - +
PARP
Ephrin-B2
Cleaved-
caspase 3
Survivin
SNU-398
PCNA
170 kDa
43 kDa
17 kDa
17 kDa
25 kDa
130 kDa
100 kDa
72 kDa
34 kDa
0 0.25 0.5 1 2
Oxaliplatin (µM )
DNMT1&
Survivin&
PCNA&
400&nM&SGI4110&(h)&
8&uM&Oxalipla>n&(h)&
SNU$387(
&0&&&&&&&&0&&&&&&&0&&&&&&&0&
&0&&&&&&&&0&&&&&&&0&&&&&&&0&&&
&0&&&&&&&0&&&&&&&0&&&&&&&0&
&0&&&&&&24&&&&&48&&&&&72&
72&&&&&72&&&&&72&&&&&72&
&0&&&&&&24&&&&&48&&&&&72&
Time&aCer&SGI4110&removal&(h)& &0&&&&&&&24&&&&&48&&&&&72& &0&&&&&&24&&&&&48&&&&&72& &0&&&&&&24&&&&&48&&&&&72&
72&&&&&72&&&&&72&&&&&72&
&0&&&&&&24&&&&&48&&&&&72&
&0&&&&&&&0&&&&&&&0&&&&&&&0&&&
Control& Oxalipla>n& SGI4110& SGI4110&+&Oxalipla>n&
Figure S8

61

Considering their expression profiles across the 4 HCC cell lines, the protein
levels of DNMT1 and survivin reflect the molecular responses of these cells to tested
drugs, suggesting that they could be used as novel PD markers for the effectiveness of the
combination treatment.
Evaluation of the potential PD markers in tumor tissues found that DNMT1 levels
were significantly downregulated in the combination group and mildly decreased with
SGI-110 treatment, whereas no significant changes were detected with vehicle or
oxaliplatin treatment. Similar to our results in in vitro experiments, survivin protein levels
were markedly decreased with the combination (Fig. 2-15). These findings further
validate the two PD markers for the combination treatment, suggesting their potential use
in future clinical studies.

Figure 2-15. Oxaliplatin treatment downregulates DNMT1 and survivin levels in SGI-
110 pre-treated tumors. Lysates from 3 tumors per treatment group from SNU-398
xenograft studies were blotted for DNMT1 and survivin.

In our studies, combination treatment in SNU-398, Hep-G2 and SNU-475 cells
decreased survivin levels rapidly where neither SGI-110 nor oxaliplatin single treatment
affected survivin levels. Downregulation of survivin levels was also observed following
the combination treatment in SNU-398 tumors. Interestingly, no change in survivin levels
B
C
Control
Oxaliplatin
SGI-110
SGI-110+Ox
0
10
20
30
40
50
p = 0.015
p = 0.018
Ki67 Index (%)
Control$(20X)$ OxaliplaBn$
SGI;110$ SGI;110$+$Ox$
0 5 10 15 20 25 30
0
1000
2000
3000
Vehicle
Oxaliplatin (5 mg/kg)
SGI-110 (2 mg/kg)
SGI-110 + Oxaliplatin
Day
Tumor Volume(mm
3)
*
*
** ***
*
**
**
Figure 6-1
Day 1-5 Day 8-12 Day 15-19 Day 22-26 Day 29-31
Control
Treatment schedules
SGI-110, 2 mg/kg (S)
Oxaliplatin, 5 mg/kg (Ox)
A
Oxaliplatin
SGI-110
S + Ox
DNMT1
Survivin
PCNA
Control SGI-110 Oxaliplatin
SGI-110 +
Oxaliplatin
SNU-398 tumor lysate
β-Tubulin
1 2 3 1 2 3 1 2 3 1 2 3
170 kDa
17 kDa
34 kDa
55 kDa

62

was observed following the combination treatment in SNU-387 cells, in which the
combination did not show synergistic effect. We conclude from our data that the
expression levels of survivin correlated well with the cytotoxicity of the combination
treatments, suggesting that survivin may be an important effector of the synergism. These
results imply that survivin can be used as a PD marker in response to SGI-110 and
oxaliplatin treatment.

2.9 Bru-seq identified EFNB2 as the most upregulated gene by SGI-110
treatment in SNU-398
Another important goal of this preclinical study was to establish potential PD
markers for future clinical evaluations. Using Bru-seq to investigate the rates of
transcription genome-wide we observed that the transcription of the EFNB2 gene was
significantly upregulated in SNU-398 cells by SGI-110 treatment (15.0 fold) and when
combined with oxaliplatin (13.5 fold) (Fig. 2-16A). When SNU-398 cells were treated
with SGI-110 for 72 h, protein levels of the EFNB2 gene product ephrin-B2 increased
dose-dependently, where significant induction was observed at doses as low as 100 nM
(Fig. 2-16B). Upon treatment with SGI-110 (100 nM), the expression of ephrin-B2
protein increased from 72 h post treatment and achieved maximal induction at 96 h (Fig.
2-16C), consistent with the EFNB2 upregulation after 72 h of SGI-110 treatment revealed
by Bru-seq. Ephrin-B2 is a ligand for tyrosine kinase receptor EPHB4 and EPHA4.
Ephrin-B2/EPHB4 signaling is known to suppress tumor growth in neuroblastoma (Tang
et al., 2004), breast cancer (Noren et al., 2006) and colon cancer (Liu et al., 2002). This

63

is the first study to report upregulation of ephrin-B2 levels by hypomethylating agents in
liver cancer.
In MethHC, a database of DNA methylation and gene expression in human
cancers, EFNB2 is ranked 29
th
in the most differentially methylated cells in the promoter
region between tumor samples and normal samples, with an increased average
methylation level of 0.2674 (in a 0-1 scale) in 204 hepatocellular carcinoma samples
from TCGA (Huang et al., 2015) (Fig. 2-17). This data suggests great possibility for SGI-
110-mediated regulation of EFNB2 gene transcription in HCC, thus implying EFNB2 as
a potential biomarker for the treatment.

Figure 2-16. EFNB2 expression is upregulated by SGI-110 and its combination treatment
with oxaliplatin. A) Synthesis of EFNB2 nascent RNA is upregulated by SGI-110 and its
combination treatment with oxaliplatin in SNU-398 cells as identified by Bru-Seq. The
gene map is from RefSeq Genes (UCSC genome browser, http://genome.ucsc.edu/). B)
Ephrin-B2 protein levels were upregulated dose-dependently by SGI-110 treatment. C)
Ephrin-B2 protein levels were upregulated time-dependently by SGI-110 treatment, and
the induced ephrin-B2 expression remained up to 72 h after SGI-110 removal.

EFNB2
0
2
4
6
8
20
40
Control
SGI-110 + Ox
RPKM
Synthesis UP 13.5 fold
EFNB2
107,160,000 107,170,000 107,180,000
20 kb hg19
Scale
Chr13:
Ephrin-B2
PCNA
SGI-110 treatment (nM) 0 10 30 100 300 1000
SNU-398
(72h treatment)
Ephrin-B2
PCNA
SGI-110 treatment (100nM) 0 24 48 72 72 72 72
Time after SGI-110 removal 0 0 0 0 24 48 72
SNU-398
B
C
EFNB2
0
2
4
6
8
20
40
Control
SGI-110
RPKM
Synthesis UP 15.0 fold
EFNB2
107,160,000 107,170,000 107,180,000
20 kb hg19 Scale
Chr13:
A
43 kDa
34 kDa
34 kDa
43 kDa
34 kDa
34 kDa
Figure 4

64

Figure 2-17. EFNB2 promoter is highly methylated in liver cancer. Promoter DNA
methylation data from TCGA (The Cancer Genome Atlas) in five different cancers of the
digestive system. Data was integrated by MethHC database. Each data point represents
promoter methylation level of a single sample. Mean ± SD is shown with each group.
Unpaired t test was used for statistic analysis between respective tumor and normal
samples. *, p<0.05; ***, p<0.001; ****, p<0.0001.

However, in our effort to further validate the role of EFNB2 in HCC, we found
that the SGI-110-induced ephrin-B2 expression is only observed in SNU-398 among a
panel of 12 cancer cell lines including 6 HCC cell lines (SNU-398, Hep-3B, SNU-449,
Hep-G2, SNU-475, SNU-387), 3 pancreatic cancer cell lines (BxPC-3, Panc-1, MiaPaCa-
2) and 3 other cancer cell lines (HCT-116, LnCap, U87), indicating that upregulation of
ephrin-B2 is a cell line specific effect. (Fig. 2-18) As far as with the current samples,
EFNB2 does not qualify as a robust PD marker for efficacy of the combination treatment
among unselected HCC patients. However, EFNB2 might play a role in the combination
treatment in SNU-398 cells, and its importance in HCC remains to be further
characterized.
Figure S5
Methylation level at EFNB2 promoter
0.0
0.2
0.4
0.6
0.8
1.0
Tumor samples
Normal samples
Liver
Cancer
Colon
Cancer
Rectum
Cancer
Stomach
Cancer
Pancreatic
Cancer
**** * ***
Methylation level (%)

65

Figure 2-18. Induction of ephrin-B2 protein levels is only observed in SNU-398 cells.
Six hepatocellular carcinoma cell lines, three pancreatic cancer cell lines and three other
cancer cell lines were treated with SGI-110 at 300 nM for 72 h, and subjected to Western
blotting for DNMT1 and ephrin-B2 levels. While DNMT1 levels are decreased by SGI-
110 treatment in all cell lines, ephrin B2 protein levels are only increased in SNU-398
cells as a cell line specific effect. The basal expression levels of Ephrin-B2 are highly
different among cell lines, and the different banding patterns could attribute to
glycosylation and phosphorylation of the cell surface trans-membrane ligand.

2.10 Conclusions
In this preclinical study, we have shown the significant antitumor effect of SGI-
110 alone or in combination with oxaliplatin in HCC models. Application of Bru-seq led
to the identification of Wnt/β-catenin, EGFR and IGFR signaling as key pathways
inhibited by the combination treatment. Such simultaneous inhibition of three signature
liver cancer pathways by the combination treatment supports the use of a DNA
demethylating agent in combination with a cytotoxic agent as an effective therapy for
HCC. We expect that the combination of SGI-110 and oxaliplatin at low doses will delay
disease progression and prolong overall survival without significant toxicity. Our
findings provide a strong rationale for a Phase I/II clinical trial with SGI-110 and
oxaliplatin in HCC patients.
DNMT1
SGI-110
(300 nM, 72 h)
- + - + - + - + - + - + - + - + - + - + - + - + - +
Ephrin-B2
PCNA
Hepatocellular Carcinoma Pancreatic Cancer Other Cancers
170 kDa
43 kDa
34 kDa
Figure S6

66

2.11 Materials and methods
2.11.1 Cell culture
SNU-398, SNU-449, SNU-387, SNU-475, Hep-3B, Hep-G2 hepatocellular
carcinoma cell lines were obtained from ATCC (Manassas, VA) in June 2012.
Isoenzymology and STR analyses were performed by ATCC to confirm species and cell
line identity. No further authentication was performed in-house. Cells were expanded into
10 tubes (1x10
6
/ tube) and frozen immediately. All cell lines were cultured as monolayers
and maintained in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) in a
humidified atmosphere with 5% CO
2
at 37°C. Cells were kept in culture for 20 passages
and discarded, then a new batch of cells was used in subsequent experiments. The
PlasmoTest
TM
(InvivoGen, San Diego, CA) was performed every three weeks to confirm
all cell lines were mycoplasma-free.

2.11.2 Compound preparation
For in vitro experiments, 10 mM stock solution was prepared by dissolving SGI-
110 (Astex Pharmaceuticals, Dublin, CA) in PBS. Solution was kept at -80°C for storage.
For in vivo experiments, SGI-110 was diluted in reconstitution diluent (65% propylene
glycol, 10% ethanol and 25% glycerin). Solution was stored at 4°C. Oxaliplatin was
purchased from BIOTANG Inc. (Lexington, MA) and freshly dissolved in DMSO to
prepare a 10 mM stock solution. Z-VAD-fmk (Tocris, Minneapolis, MN) and
Necrostatin-1 (Cayman, Ann Arbor, MI) were freshly dissolved in DMSO for 40 mM
stock solutions.

67

2.11.3 MTT assay
Cytotoxicity of compounds was evaluated with 3-(4,5-dimethylthiazol-2-yl)-2,5-
diphenyltetrazolium bromide (MTT) assay. Cells were placed in 96-well plate at 1000
cells/well on Day 1. After overnight attachment, SGI-110 was added to the wells at
sequential dilutions (10 nM – 1 µM for most cell lines) on Day 2. Due to the hydrolysis
of the compound, SGI-110 treatment was repeated every 24 h. After 72 h treatment (on
Day 5), SGI-110 containing media was carefully removed and fresh cell culture media
was added to the plate. For combination treatment, oxaliplatin was added on Day 5 after
changing the media, and kept in culture for 72 h treatment. On Day 8, compound-
containing media was carefully removed and fresh cell culture media was added to the
plates. On Day 12, MTT was added into the media to a final concentration of 300 µg/mL.
Cells were incubated for 3 h at 37°C, and the insoluble formazan converted by viable
cells was dissolved in 150 µL of DMSO. Absorbance at 570 nm was read in a microplate
reader (Molecular Devices, Sunnyvale, CA), and inhibition of cell proliferation was
calculated using the following formula:
Inhibition of cell proliferation (%) = (1 - OD
treatment
/OD
control
) x 100%
Synergistic effect of the combination treatment was evaluated by computing
combination index (CI) with CompuSyn using the Chou-Talalay method (Chou, 2006).
CI values lower than 1 indicates synergistic effect.

68

2.11.4 Colony formation assay
Cells were placed in a 96-well plate at 200 cells/well or in a 6-well plate at 5000
cells/well on Day 1. Treatment schedules were performed as described for the MTT
assay. After treatment, cells were kept in culture until colonies were observed in control
wells. Colonies were then fixed and stained with 0.05% crystal violet solution (2%
formaldehyde, 40% methanol in distilled water), washed with water to remove excess
stain, imaged with Odyssey Imaging Systems (LI-COR Biosciences, Lincoln, NE), and
quantified with Image J software.

2.11.5 Western blotting
Cells (4 x 10
5
) were cultured in 60 mm tissue culture dishes and treated with SGI-
110 or oxaliplatin at designated concentrations. After treatment, cells were lysed with cell
lysis buffer at 4°C for 30 min and centrifuged (12000 rpm, 10 min, 4°C). Protein
concentrations of supernatants were measured with BCA assay (Thermo Fisher
Scientific, Rockford, IL). Protein (40 µg per sample) was subjected to SDS-PAGE
analysis. Proteins were electro-transferred to methanol activated immobilon-FL PVDF
membranes (EMD Millipore, Billerica, MA). Membranes were blocked with 5% skim
milk in TBST buffer and incubated with primary antibodies (anti-E-Cadherin, anti-
survivin, anti-PARP and anti-cleaved caspase 3 from Cell Signaling (Beverly, MA), anti-
DNMT1 and anti-β-tubulin from Santa Cruz Biotechnology (Santa Cruz, CA), anti-
ephrin-B2 and anti-PCNA from Sigma-Aldrich (Saint Louis, MO)) 1:1000 dilutions
overnight at 4°C. Membranes were then washed with TBST (10 min x3), incubated with

69

Dylight 800-conjugated secondary antibodies (Thermo Fisher Scientific) 1:5000 dilutions
in 5% milk for 1 h at room temperature, and washed with TBST (10 min x2) and TBS (10
min). Fluorescent signal was then scanned by Odyssey Imaging Systems (LI-COR
Biosciences).

2.11.6 Bru-seq analysis for nascent RNA synthesis
Bru-seq analysis was performed as previously reported (Paulsen et al., 2014).
Briefly, 4 x 10
6
SNU-398 cells were placed in 10 cm dishes on Day 1. On Day 2, cells
were treated with PBS or SGI-110 at 100 nM for 72 h with fresh drug addition every 24
h. Cells were changed to fresh media on Day 5 and treated with DMSO or oxaliplatin at 3
µM for 4 h. Bromouridine, at a final concentration of 2 mM, was added into the media to
label newly synthesized nascent RNA in the last 30 min of treatment. Cells were then
collected in TRIZOL and total RNA was isolated. The bromouridine-containing RNA
population was further isolated and sequenced. Sequencing reads were mapped to the
HG19 reference genome. Pre-ranked gene lists were generated for each treatment by
ranking genes by fold changes in gene synthesis levels compared to control, and analyzed
with GSEA (Broad Institute, MA) (Mootha et al., 2003; Subramanian et al., 2005).

2.11.7 Xenograft study
SNU-398 cells (2.0 x 10
6
)

in a 100 µL suspension of RPMI 1640 were injected
subcutaneously into the dorsal flank of 6-week old athymic nude mice (The Jackson
Laboratory, Bar Harbor, ME). Tumor size was monitored twice a week by caliper

70

measurement using the following equation: V=d
2
x D/2, where d represents width and D
represents length of the tumor. Mice were randomly grouped (n = 5 per group) when
average tumor size reached 100 mm
3
. Treatment was given in 14-day cycles. On Day 1-5,
SGI-110 was administrated by subcutaneous injection in a 100 µL vehicle to SGI-110
single treatment and combination groups. On Day 5 (4 h after SGI-110 administration)
and Day 12, oxaliplatin treatment was given by intraperitoneal injection in 100 µL saline
to oxaliplatin single treatment and combination group. Control mice received vehicle
only. Study was concluded when tumor size in the group reached 2000 mm
3
. Unpaired
Student’s t-test was performed for data analysis and p < 0.05 was considered significant.

2.11.8 Histochemical analysis
On necropsy, tumors, hearts, kidneys, livers, lungs, spleens and pancreases were
collected, fixed in 10% neutral buffered formalin, embedded in paraffin, and sectioned.
Sections (5 µm) were stained with hematoxylin and eosin to facilitate histologic
examination. For Ki67 expression level, immunohistochemistry staining was performed
on sections with Ki67 antibody. Embedding, sectioning and staining of samples were
performed by the ULAM pathology core for animal research at the University of
Michigan. Representative images were taken on an Olympus IX83 microscope with 20X
magnification.

71

CHAPTER 3 Mechanistic studies of ROS modulators in pancreatic cancer
3.1 Pancreatic cancer
Pancreatic cancer is the fourth leading cause of cancer related death in both
genders in the United States, claiming 39590 lives in 2014 alone (Siegel et al., 2014).
Given its asymptomatic and metastatic nature, over 50% of pancreatic cancer cases are
diagnosed at late stages, when the tumor has metastasized and is unresectable. Therefore,
treatment of pancreatic cancer is largely dependent on systemic chemotherapy. Ever since
its approval by the FDA in 1996, gemcitabine-based regimes have been the standard of
care for pancreatic cancer (Ryan et al., 2014). However, limited by late-stage diagnosis
and inherent/acquired resistance to current chemotherapy, the overall five-year survival
rate of pancreatic cancer is only 6.7%, one of the lowest among all types of cancers.
Recently, two combination regimens with modest clinical activity have been added to the
options. The addition of nab-paclitaxel (albumin-bound paclitaxel) to gemcitabine
increased median overall survival from 6.7 to 8.5 months (Von Hoff et al., 2013). The
combination FOLFIRINOX (oxaliplatin, irinotecan, fluorouracil and leucovorin) was
approved for the treatment of metastatic pancreatic cancer by increasing median overall
survival from 6.8 months in the gemcitabine group to 11.1 months in the FOLFIRINOX
group (Conroy et al., 2011), but increased toxicity is the major concern for these new
treatment options. Therefore, novel therapeutics are urgently needed to enhance the
survival of patients with this devastating disease.

72

3.2 Modulation of redox balance as promising strategy for treatment of
pancreatic cancer
Altered redox homeostasis has been observed in various types of cancers
including PDAC. ROS levels, as the byproduct of cellular respiration, are increased as a
result of elevated energy demand in cancer cells. To adapt to changes in the
microenvironment, cancer cells hijack the intracellular antioxidant machinery to reach a
new redox state that can facilitate their proliferation. Cancer cells are dependent on such
a new altered redox state, therefore providing novel therapeutic opportunities for therapy.
A strategy to alter the redox balance for cancer treatment is to induce ROS production to
overwhelm the antioxidant machinery selectively killing cancer cells.
Altered redox homeostasis in cancer cells provides a new opportunity for tumor
intervention. Reactive oxygen species (ROS), a natural byproduct from mitochondrial
respiration, play an important role as second messengers in cell signaling (Li et al.,
2013). However, when present at high concentrations, ROS can be detrimental to cellular
processes, inducing damage to DNA, lipids and proteins by oxidation. Therefore,
excessive intracellular ROS are constantly eliminated by antioxidants regulated by the
ROS-detoxifying machinery to ensure a healthy redox state. In tumor cells, antioxidant
enzymes are often active as a result of elevated levels of intrinsic ROS (Fruehauf et al.,
2007). Oncogenic mutations like Kras
G12D
, commonly present in PDAC, activate the
master antioxidant switch Nrf2 in the basal state (DeNicola et al., 2011; Kong et al.,
2013). Altered redox homeostasis in tumors make them more susceptible to induced
oxidative stress that overwhelms their adaptive antioxidant capacity and triggers ROS-
mediated cell death (Pelicano et al., 2004; Sabharwal et al., 2014).

73

Previously, we showed that the quinazolinedione QD232 exerts ROS-dependent
cytotoxicity in pancreatic cancer models (Pathania et al., 2014; Pathania et al., 2015). In
this study, we performed a lead optimization campaign and identified QD325 as a lead
compound for in-depth preclinical and mechanistic studies. Our results show for the first
time that selective inhibition of the mitochondrial D-loop can be efficacious and be
further explored as innovative therapeutic approach to target cancers that heavily depend
on mitochondrial function.

3.3 Optimization of ROS modulators lead to discovery of QD325
To establish a robust structure-activity relationships for the QD series of
compounds, we designed and synthesized 25 new analogues of our previous lead
compound, QD232, to better elucidate their mechanism of action. We first tested the
cytotoxicity of these compounds using MTT assay in three PDAC cell lines MiaPaCa-2,
Panc-1 and BxPC-3. Nine of these novel analogues showed improved cytotoxicity in at
least two cell lines (Table 3-1). QD325 is the best analogue with IC
50
values <1 µM in all
three cell lines.
The QD analogues can be grouped into 5 major classes by chemical structures.
QD325 with phenyl group substitution on QD232 achieves more than two fold
improvement in cytotoxicity, however, further modification with alkyl, methoxy, amine
or fluorine substituted phenyl group did not further improve potency. Another major
improvement in potency was achieved by methoxy substitution on QD232.

74

Table 3-1. Structure and cytotoxicity of QD compounds in MiaPaCa-2, Panc-1 and
BxPC-3 cells by MTT assay. QD compounds are grouped by structure.

QD 232, 325-338, 340, 353-357, 359 QD 323

ID
Substitution group IC
50
(µM)
[1]

R R
1
R
2
MiaPaCa-2 Panc-1 BxPC-3
232 COCH
3
H H 2.3 ± 0.2 0.9 ± 0.2 5.2 ± 0.8
325 H Ph H 0.9 ± 0.2 0.4 ± 0.1 0.5 ± 0.1
356 H H NHCH
2
-(4-F-Ph) 1.7 ± 0.2 1.0 ± 0.1 1.4 ± 0.2
335 H 4-Et-Ph H 2.0 ± 0.1 1.2 ± 0.1 3.1 ± 0.7
336 H 4-OCH
3
-Ph H 2.1 ± 0.5 2.3 ± 0.3 3.5 ± 0.5
337 H 4-NH
2
-Ph H 2.5 ± 0.2 3.7 ± 0.1 3.5 ± 0.6
334 H 4-F-Ph H 3.5 ± 1.0 3.2 ± 0.8 4.4 ± 0.9
338 F 4-CH
3
-Ph H 4.6 ± 1.1 4.8 ± 0.1 5.0 ±0.7
326 OCH
3
OCH
3
OCH
3
1.5 ± 0.1 0.8 ± 0.1 1.6 ± 0.3
353 H H OCH
3
1.8 ± 0.3 0.6 ± 0.1 1.8 ± 0.1
354 H OCH
3
H 1.9 ± 0.2 0.8 ± 0.2 1.7 ± 0.2
355 H OCH
3
OCH
3
1.8 ± 0.1 0.9 ± 0.3 1.5 ± 0.2
357 OCH
3
H OCH
3
7.7 ± 2.0 7.2 ± 0.8 16.3 ± 1.5
327 H OCF
3
H 1.4 ± 0.2 0.9 ± 0.1 0.9 ± 0.1
324 H O-Ph H 3.7 ± 0.7 1.8 ± 0.2 3.6 ± 0.4
328 H SO
2
NH
2
H >10 >10 >10
333 B(OH)
2
H H >10 9.0 ± 1.0 >10
331 H COOCH
3
H 2.2 ± 0.4 1.1 ± 0.4 5.8 ± 0.3
329 H CH
2
OH H 3.5 ± 1.3 1.0 ± 0.2 5.7 ± 0.3
332 H COOCH
2
CH
3
H 5.5 ± 1.5 1.6 ± 0.3 5.9 ± 0.1
330 H CONH
2
H 8.0 ± 0.9 6.3 ± 0.3 >10
323 N/A N/A N/A 9.4 ± 0.9 18.0 ± 2.5 19.4 ± 1.6
339 N/A N/A N/A >30 >30 >30
358 N/A N/A N/A > 30 > 30 > 30
331 H COOCH
3
H 2.2 ± 0.4 1.1 ± 0.4 5.8 ± 0.3
340 H CONH(CH
2
)
3
(TPP)
+
Br
-[2]
H 15.3 ± 2.5 11.7 ± 1.5 21.5 ± 2.3
232 COCH
3
H H 2.3 ± 0.2 0.9 ± 0.2 5.2 ± 0.8
359 CONH(CH
2
)
3
(TPP)
+
Br
-[2]
H H 16.3 ± 3.5 14.3 ± 1.5 21.3 ± 2.5

[1]
Data is presented as Mean ± SD from three independent experiments.
[2]
TPP stands for triphenylphosphonium.
N
N
O
O
H
N
R
R
1
R
2
R
1
N
N
O
O
N
N
O
O
H
N
5
R
R
1
6a-s
R
H
2
N
i
R
2
R
2
i
ii
5 6
QD339 QD340
H
2
N Br P + HBr
. H
2
N P HBr
.
Br
H
N P
O
H
N P
Br
Br
H
2
N
O
H
N
N
N
O
O
+
N
N
O
O
+
QD323
O
OH
H
2
N
iii
i
ii
5 6
QD358 QD359
H
2
N Br P + HBr
. H
2
N P HBr
.
Br
N
H
P
O
Br
+
N
N
O
O
+
QD323
H
2
N
iii
OH
O
H
2
N
N
H
P
O
Br
H
N
N
N
O
O

75

3.4 QD325 exhibits ROS dependent cytotoxicity in pancreatic cancer models
To quantify ROS induction by our redox modulators, we developed a high
throughput ROS assay in 384-well plates using H2DCFDA as the ROS detection probe
(Fig. 3-1A). Using H
2
O
2
as the positive control, we detected time- and dose-dependent
changes with a Z factor of 0.879, demonstrating good sensitivity and reproducibility of
the assay (Fig. 3-1B).
Treatment with QD compounds elicits significant ROS accumulation in MiaPaCa-
2 cells. Among the 25 analogues, QD325, QD335 and QD326 exhibit significantly higher
ROS induction than the lead compound QD232 after 24 h treatment (Fig. 3-1C), whereas
six other analogues show similar ROS induction as the earlier lead. After 24 h the ROS
dependent DCF fluorescence plateaued for all compounds and this time point was chosen
for compound comparison. Inhibition of cell proliferation and ROS induction by QD
compounds showed linear correlation with Pearson’s correlation coefficient r of 0.66 at
3.3 µM (p = 0.00002) and 0.7080 at 10 µM (p < 0.0001), suggesting positive correlations
in both cases (Fig. 3-2). In general, the most cytotoxic compounds are also high ROS
inducers (Table 3-1).

76

Figure 3-1. QD compounds induce ROS accumulation in MiaPaCa-2 cells. A)
Illustration of cell-based ROS detection assay. Cell permeable H2DCFDA probe is
loaded into Mia PaCa-2 cells and converted into highly fluorescent DCF in the presence
of ROS. Fluorescent signal is then detected by BioTek H1 plate reader as indicator of
ROS level. B) H
2
O
2
, a form of ROS, induces conversion of H2DCFDA into DCF dose
and time dependently. C) Arranged by structural groups, new QD analogs show different
ROS induction activity at 10 µM after 24 h treatment. H
2
O
2
treatment at 300 µM for 24 h
is used as positive control representing full activation. ROS induction activity of QD
compounds is normalized to positive control. Graphical data is presented as Mean ± SD
from three independent experiments.

To validate ROS induction as the mechanism for cytotoxicity, the effect of QD
compounds were evaluated in the presence and absence of the antioxidant N-acetyl-
A
Coat cells
Load H2DCFDA
probe
Wash
Add compound
Read
fluorescence
H2DCFDA
non- fluorescent
DCF
fluorescent
oxidation
by ROS
B
C
10 100 1000
10
3
10
4
0
15
30
45
120
180
240 min
0
H
2
O
2
(µM)
fluorescence intensity
H
2
O
2
232
325
356
335
336
337
334
338
326
353
354
355
357
327
324
328
333
331
329
332
330
323
339
340
358
359
0
20
40
60
80
100
ROS induction (%)
0 15 30 45 60120180240
10
3
10
4
2000
1000
500
250
125
60
30
15 µM
time (min)
fluorescence intensity

77

cysteine (NAC). For the lead compound QD232 and the two active analogues QD325 and
QD326, we observed a time- and dose-dependent accumulation of ROS (Fig. 3-3A). A
negative control without the H2DCFDA probe was included to exclude potential
fluorescence of compounds interfering with the assay. While H
2
O
2
treatment leads to
immediate conversion of H2DCFDA to fluorescent DCF, treatment with QD compounds
leads to a gradual induction of the fluorescent signal, implying ROS accumulation. For
QD232, QD325 and QD326 treatments, ROS accumulation reaches peak levels after 4-6
h. At 10 and 3.3 µM, both QD325 and QD326 induce rapid and high ROS accumulation.

Figure 3-2. Cytotoxicity of QD compounds correlates with ROS induction. Cytotoxicity
of QD compounds is represented by inhibition of cell proliferation (%) at 3.3 or 10 µM
after 72 h treatment in MiaPaCa-2 cells. ROS induction was determined for QD
compounds at 3.3 or 10 µM after 24 h treatment in MiaPaCa-2 cells. Data points
represent the mean values of three independent experiments. Linear correlation was
analyzed by Prism.

When cells were pretreated with 5 mM NAC, ROS induction by H
2
O
2
and QD
compounds was blocked (Fig. 3-3B). In the MTT assay, NAC decreased cytotoxicity of
H
2
O
2
, QD232, QD325 and QD326 (Fig. 3-3C). These results demonstrate that ROS
accumulation is the primary mechanism for cytotoxicity of QD compounds. However,
QD compounds at 3.3 µM QD compounds at 10 µM
0 20 40 60 80 100
0
20
40
60
80
100
R
2
= 0.4388
p = 0.0002
ROS induction (%)
inhibition of cell proliferation (%)
0 20 40 60 80 100
0
20
40
60
80
100
R
2
= 0.5013
p < 0.0001
ROS induction (%)
inhibition of cell proliferation (%)
Figure S1. Cytotoxicity of QD compounds correlates with ROS induction. Cytotoxicity of QD compounds
is represented by inhibition of cell proliferation (%) at 3.3 or 10 µM after 72 h treatment in MiaPaCa-2
cells. ROS induction was determined for QD compounds at 3.3 or 10 µM after 24 h treatment in
MiaPaCa-2 cells. Data points represent the mean values of three independent experiments. Linear
correlation was analyzed by Prism.

78

NAC treatment did not completely block the cytotoxicity of QDs and H
2
O
2
suggesting
additional cellular effects responsible for the inhibition of cell proliferation.

Figure 3-3. Cytotoxicity of QD compounds is scavenged by NAC in MiaPaCa-2 cells. A)
Parental compound QD232 induces ROS accumulation dose and time dependently. New
analogs QD325 and QD326 induce stronger and more rapid ROS accumulation in
MiaPaCa-2 cells. Compounds were tested at 10, 3.3 or 1.1 µM. DMSO was used as
negative control to determine basal signal of the assay (DMSO). Cells without preloaded
H2DCFDA were treated with compounds at 10 µM at the same conditions to determine
the endogenous fluorescence of the compounds (no stain). Data points represent Mean ±
SD from duplicates. Graphs are representatives of three independent experiments. B)
ROS induction by QD232, 325 and 326 is inhibited by NAC pretreatment (5 mM for 30
min). Data points represent Mean ± SD from duplicates. Graphs are representatives of
three independent experiments. C) Presence of NAC at 5 mM decreases cytotoxicity of
0 2 4 6 8 24
10
3
10
4
10
3.3
1.1 µM
DMSO
no stain
QD232
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
10
3.3
1.1 µM
DMSO
no stain
QD326
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
10
3.3
1.1 µM
DMSO
no stain
QD325
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
300
100
33 µM
DMSO
no stain
H
2
O
2
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
3.3 µM
no stain
3.3 µM
(+NAC)
QD232
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
3.3 µM
no stain
3.3 µM
(+NAC)
QD326
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
3.3 µM
no stain
3.3 µM
(+NAC)
QD325
time (h)
fluorescence intensity
0 2 4 6 8 24
10
3
10
4
100 µM
no stain
100 µM
(+NAC)
H
2
O
2
time (h)
fluorescence intensity
A B C
1 10 100
0
20
40
60
80
100
Ctrl
NAC (5 mM)
0
H
2
O
2
concentration (µM)
inhibition of cell proliferation (%)
0.1 1 10
0
20
40
60
80
100
Ctrl
NAC (5 mM)
0
QD232
concentration (µM)
inhibition of cell proliferation (%)
0.1 1 10
0
20
40
60
80
100
Ctrl
NAC (5 mM)
0
QD325
concentration (µM)
inhibition of cell proliferation (%)
0.1 1 10
0
20
40
60
80
100
Ctrl
NAC (5 mM)
QD326
0
concentration (µM)
inhibition of cell proliferation (%)

79

QD 232, 325 and 326. Cytotoxicity was determined by MTT assay after 72 h treatment.
Data points represent Mean ± SD from three independent experiments.

3.5 QD325 exhibits anti-tumor activity in in vivo pancreatic cancer models
3.5.1 QD325 shows selectivity for PDAC cells
QD232, QD325, QD326 all show similar cytotoxicity in MiaPaCa-2 and a
gemcitabine-resistant cell line MiaPaCa-2-GR (Ali et al., 2010) (Table 3-2). In the
HPV16-E6E7 gene immortalized pancreatic cell line, HPDE (Ouyang et al., 2000),
gemcitabine produces similar IC
50
values as in MiaPaCa-2 cells, while the most potent
QD325 shows 3-fold selectivity for MiaPaCa-2 (Table S2).

Table 3-2. Cytotoxicity of QD compounds in gemcitabine resistant MiaPaCa-2 cells and
normal pancreatic cells by MTT assay

[1]
Data is presented as Mean ± SD from three independent experiments.

3.5.2 QD325 as single treatment
To investigate the in vivo antitumor efficacy of the most potent analogue, QD325,
xenograft studies were performed in NOD/SCID mice. Subcutaneous human pancreatic
cancer xenografts from MiaPaCa-2 cells were established on the dorsal flank of the
ID
IC
50
(µM)
[1]

MiaPaCa-2 MiaPaCa-2-GR HPDE
232 2.3 ± 0.2 3.6 ± 0.6 4.5 ± 0.6
325 0.9 ± 0.2 1.0 ± 0.3 2.7 ± 0.3
326 1.5 ± 0.1 2.0 ± 0.1 3.2 ± 0.4
340 15.3 ± 2.5 17.7 ± 1.8 18.3 ± 2.1
359 16.3 ± 3.5 16.2 ± 1.9 24.3 ± 1.2
Gemcitabine 0.11 ± 0.07 3.3 ± 0.6 0.14 ± 0.05

80

immune-deficient mice, and treated with QD325 or vehicle until tumor size in the control
group passed 1200 mm
3
. QD325 (5 mg/kg) treatment significantly suppressed growth of
tumors in the treatment period of 44 days. On day 44, when average tumor size in control
group was 1291 ± 168 mm
3
, it was only 308 ± 72 mm
3
(p=2.1E6) for QD325 treatment
group (Fig. 3-4A).
No symptoms of gross toxicity such as weakness, weight loss or lethargy were
observed in any treatment group (Fig. 3-4B). H&E stained organ sections of liver,
kidney, heart, lung, spleen and pancreas did not reveal major histopathological changes,
further confirming the safety of the treatments (Fig. 3-4C). Following the 44-day
treatment, two mice were kept on each group to evaluate efficacy and safety of QD325 at
higher doses. While tumors in the control group exhibited rapid growth, QD325 treatment
was able to delay growth of the tumors, and no systemic toxicity was observed at doses
as high as 20 mg/kg (Fig. 3-4D&E).
In line with the tumor growth inhibition, QD325 treatment decreased Ki67 levels
in tumor tissues, suggesting inhibition of cell proliferation (Fig. 3-5).

81

Figure 3-4. QD325 inhibits tumor growth of MiaPaCa-2 xenograft without systemic
toxicity. A) QD325 treatment at 5 mg/kg inhibits growth of MiaPaCa-2 xenograft in
NOD/SCID mice. MiaPaCa-2 engrafted mice were randomized into vehicle control (n=5)
or QD325 treatment (n=5) group when tumor size reached 65 mm
3
. QD325 was given at
5 mg/kg five times a week until day 44. B) Body weight of engrafted mice was not
affected by QD325 treatment at 5 mg/kg. Data points represent Mean ± SEM. C)
Representative micrographs of hematoxylin and eosin (H&E)-stained organ sections.
Images were taken with Olympus IX83 inverted microscope at 20X magnification. In
histopathology study, no major microscopic changes were detected in major organs after
QD325 treatment. D) Tumor volumes of study continued after data shown in panel A.
QD325 was given at 5 mg/kg five times a week until day 44. Three mice from each group
were euthanized for tissue analysis. Two mice remained in each group after day 44 and
QD325 doses were increased from 5 mg/kg to 20 mg/kg until day 67. E) Body weight of
engrafted mice was not affected by QD325 treatment from 5-20 mg/kg. Error bars
indicate Mean ± SEM.

D E
0 10 20 30 40 50 60 70
0
500
1000
1500
2000
Vehicle (n=5)
QD325 (n=5)
Vehicle (n=2)
QD325 (n=2)
From Day 57, 20 mg/kg
From Day 50, 15 mg/kg
From Day 44, 10 mg/kg
From Day 1, 5 mg/kg
day
tumor volume (mm
3
)
0 10 20 30 40 50 60 70
0
10
20
30
Vehicle (n=5)
QD325 (n=5)
Vehicle (n=2)
QD325 (n=2)
From Day 57, 20 mg/kg
From Day 50, 15 mg/kg
From Day 44, 10 mg/kg
From Day 1, 5 mg/kg
day
body weight (g)
A B
0 10 20 30 40 50
0
500
1000
1500
Vehicle
p=2.1E6
QD325
(5 mg/kg)
day
tumor volume (mm
3
)
0 10 20 30 40 50
0
10
20
30
40
Vehicle
QD325 (5 mg/kg)
day
body weight (g)
A B
Control
QD325
Liver Kidney Heart Lung Spleen Pancreas
C

82

Figure 3-5. QD325 inhibits tumor cell proliferation in MiaPaCa-2 xenograft. A)
Representative immunohistochemistry images for Ki67 staining of MiaPaCa-2 xenograft
sections. B) QD325 decreased Ki67 index (percentage of Ki67 positive cells in the field)
of treated tumors. Data represents Mean ± SD (n=9, 3 tumors from each group, 3 images
of each tumor section). P values were calculated using student’s t-test.

In our in vivo model for pancreatic cancer, QD325 showed substantial antitumor
efficacy with favorable safety profile. Our lead compound QD232 significantly delays
tumor growth at 20 mg/kg in a similar MiaPaCa-2 xenograft model as reported earlier
(Pathania et al., 2015), suppressing growth of tumor by 65% in the 31-day study. QD325,
the most active compound in this lead optimization campaign, shows well-improved
potency by achieving comparable tumor inhibitory effect at the dose of only 5 mg/kg.

3.5.3 Addition of QD325 to gemcitabine based treatment
Gemcitabine is a major component of the standard of care treatment for pancreatic
cancer patients. Unfortunately, inherent or acquired resistance to gemcitabine represents a
major challenge for treatment of the disease. With this consideration, we sought to
explore the potential of administering QD325 as a single agent or in combination with
gemcitabine.
Control
QD325
Liver Kidney Heart Lung Spleen Pancreas
Control QD325
Control
QD325
0
10
20
30
40
50
p=9.0 E8
Ki67 Index (%)
NQO1
HO-1
GRP78
CHOP
Beta-tubulin
34 kDa
34 kDa
26 kDa
72 kDa
55 kDa
1 2 1 2
A B

83

In mice studies, gemcitabine is usually given at high doses (40-160 mg/kg) twice
weekly. Considering its low tolerance in NOD/SCID mice, we compared antitumor
activity of two different gemcitabine treatment schedules in a MiaPaCa-2 xenograft
model in this mouse strain: 1) 15 mg/kg once a week for 48 days; 2) 15 mg/kg twice a
week for the first 15 days. Similar antitumor activity was achieved by both schedules
(Fig. 3-6A). In both cases, gemcitabine was well tolerated and no weight loss was
observed (Fig. 3-6B). Therefore, schedule 1 was used for comparison of efficacy with
QD325 at 5 mg/kg and the combination of gemcitabine and QD325. QD325 was given at
5 mg/kg five times a week and gemcitabine was given at 15 mg/kg once a week (Fig. 3-
6C). At the end of the 48-day treatment period, average tumor size was 1503 ± 189 mm
3

for the control group, 387 ± 74 mm
3
(p=0.0049) for gemcitabine, 248 ± 72 mm
3

(p=0.0030) for QD325, and 163 ± 83 mm
3
(p=0.0023) for the combination of
gemcitabine and QD325 (Fig. 3-6C). Single agent treatment with QD325 at 5 mg/kg
showed similar anti-tumor activity as gemcitabine. In this experiment, both gemcitabine
and QD325 greatly inhibited tumor growth as single agents and the combination
treatment did not further reduce tumor size. Importantly, the combination was well
tolerated and no weight loss was observed in any of the treatment groups, suggesting a
reasonable safety profile of the drug combination (Fig. 3-6D).

84

Figure 3-6. QD325 is well tolerated with gemcitabine treatment in NOD/SCID mice. A)
Gemcitabine treatment at 15 mg/kg inhibits growth of MiaPaCa-2 xenograft in
NOD/SCID mice. MiaPaCa-2 engrafted mice were randomized into vehicle control
(n=4), gemcitabine treatment 1 (n=3), gemcitabine treatment 2 (n=4) groups when tumor
size reached 75mm
3
. In treatment 1, gemcitabine was given at 15 mg/kg once a week for
48 days; in treatment 2, gemcitabine was given at 15 mg/kg twice a week for 15 days.
Data points represent Mean ± SEM. B) Body weight of engrafted mice is not affected by
gemcitabine treatment in either dosing frequency. C) QD325 treatment at 5 mg/kg
inhibits growth of MiaPaCa-2 xenograft in NOD/SCID mice. MiaPaCa-2 engrafted mice
were randomized into vehicle control (n=4), gemcitabine treatment (n=3), QD325
treatment (n=3) and combination treatment groups when tumor size reached 75mm
3
.
QD325 was given at 5 mg/kg five times a week and gemcitabine was given at 15 mg/kg
once a week. Data points represent Mean ± SEM. D) Body weight of engrafted mice was
not affected by gemcitabine or QD325 treatment.

Gemcitabine is the standard of care for pancreatic cancer patients. In our studies,
QD325 shows similar antitumor effect as gemcitabine, supporting the application of
QD325 as a novel therapeutic opportunity for the disease, especially for patients with
gemcitabine resistance. Furthermore, addition of QD325 to gemcitabine at their
respective active doses did not induce any toxicity in our xenograft model. The good
A B
0 10 20 30 40 50
0
10
20
30
40
Vehicle
Gem 2 Gem 1
day
body weight (g)
C
D
0 10 20 30 40 50
0
10
20
30
40
Vehicle
QD325 + Gem
Gem 1
QD325
day
body weight (g)
Figure S7. QD325 inhibits tumor growth of MiaPaCa-2 xenograft. A) QD325 treatment at 5 mg/kg inhibits
growth of MiaPaCa-2 xenograft in NOD/SCID mice. MiaPaCa-2 engrafted mice were randomized into vehicle
control (n=5) or QD325 treatment (n=5) group when tumor size reached 65mm
3
. QD325 were given at 5 mg/kg
five times a week until day 44. Three mice from each group were euthanized for tissue analysis. Two mice
remained in each group after day 44 and QD325 doses were increased from 5 mg/kg to 20 mg/kg until day 67.
B) Body weight of engrafted mice was not affected by QD325 treatment from 5 -20 mg/kg. Error bars indicate
Mean ± SEM. C) Gemcitabine treatment at 15 mg/kg inhibits growth of MiaPaCa-2 xenograft in NOD/SCID
mice. MiaPaCa-2 engrafted mice were randomized into vehicle control (n=4), gemcitabine treatment 1 (n=3),
gemcitabine treatment 2 (n=4) groups when tumor size reached 75mm
3
. In treatment 1, gemcitabine was given
at 15 mg/kg once a week for 48 days; in treatment 2, gemcitabine was given at 15 mg/kg twice a week for 15
days. Data points represent Mean ± SEM. D) Body weight of engrafted mice is not affected by gemcitabine
treatment in either dosing frequency. E) QD325 treatment at 5 mg/kg inhibits growth of MiaPaCa-2 xenograft in
NOD/SCID mice. MiaPaCa-2 engrafted mice were randomized into vehicle control (n=4), gemcitabine
treatment (n=3), QD325 treatment (n=3) and combination treatment groups when tumor size reached 75mm
3
.
QD325 was given at 5 mg/kg five times a week and gemcitabine was given at 15 mg/kg once a week. Data
points represent Mean ± SEM. F) Body weight of engrafted mice was not affected by gemcitabine or QD325
treatment.
0 10 20 30 40 50
0
500
1000
1500 Vehicle
Gem 2
Gem 1
p=0.0011
p=0.0049
day
tumor volume (mm
3
)
Gem 2
Gem 1
0 10 20 30 40 50
0
500
1000
1500 Vehicle
QD325 + Gem
Gem 1
QD325
p=0.0030
p=0.0049
p=0.0023
QD325
Gem 1
day
tumor volume (mm
3
)

85

safety profile qualifies QD325 as a promising addition to gemcitabine-base therapy for
further anticancer benefits.

3.5.4 Anti-cancer activity of QD325 is reproducible in MiaPaCa-2 xenograft
Considering reproducibility of the xenograft studies, we performed two
independent experiments under the same conditions (Fig. 3-5A and Fig. 3-6C), and
observed similar results (76% vs. 83% inhibition of tumor growth) for QD325 in both
studies, implicating that the MiaPaCa-2 xenograft model is robust and reproducible. The
exceptional in vivo efficacy of QD325 further supports our discovery in in vitro models,
making QD325 the best translational candidate in the class. With much lower effective
dose, QD325 is expected to be used at low doses in future clinical studies, which gives
greater versatility to formulation and largely reduces chance for adverse effects. Having
its active dose as low as 5 mg/kg, QD325 did not show any toxicity at repetitive doses as
high as 20 mg/kg, further suggesting a favorable therapeutic window of this compound.

3.6 Nrf2-mediated oxidative stress and unfolded protein response are identified
as major mechanisms of action by using Bru-seq
We used our recently established bromouridine labeled RNA sequencing (Bru-
seq) technique to better characterize molecular mechanisms of these novel agents. Bru-
seq is able to capture real-time synthesis of the nascent RNA, so as to provide
information on global gene transcription without interference by RNA stability or biased
gene selection (Paulsen et al., 2013; Paulsen et al., 2014). Profiling of all genes with

86

>1.5-fold change in expression upon treatment of QD232 or QD325 identified NRF2-
mediated oxidative stress response and unfolded protein response (UPR) as key pathways
implicated in drug action (Fig. 3-7). Similar transcription signatures were observed for
QD232 and QD325 through Ingenuity Pathway Analysis (IPA) or Geneset Enrichment
Analysis (GSEA) (Fig. 3-8 to Fig. 3-11), implying similar mechanisms of action for the
two compounds.

Figure 3-7. Top 15 canonical pathways regulated by QD232 or QD325 treatment as
revealed by IPA analysis of Bru-seq data. MiaPaCa-2 cells were treated by QD232 (at 1,
2 or 3 times IC
50
) or QD325 (at 1, 2 or 5 times IC
50
) for 4 h. Nascent RNA was labeled
by bromouridine in the last 30 min of treatment, isolated, and subjected to next
generation sequencing.

NRF2-mediated Oxidative Stress Response
Unfolded protein response
Protein Ubiquitination Pathway
Role of BRCA1 in DNA Damage Response
Hereditary Breast Cancer Signaling
tRNA Charging
Hypoxia Signaling in the Cardiovascular System
Adipogenesis pathway
p53 Signaling
Pentose Phosphate Pathway
Endoplasmic Reticulum Stress Pathway
Pentose Phosphate Pathway (Oxidative Branch)
Superpathway of Cholesterol Biosynthesis
ATM Signaling
DNA Double-Strand Break Repair by HR
QD232 IC
50
2 IC
50
3 IC
50
Unfolded protein response
NRF2-mediated Oxidative Stress Response
Endoplasmic Reticulum Stress Pathway
Protein Ubiquitination Pathway
tRNA Charging
Role of BRCA1 in DNA Damage Response
Hypoxia Signaling in the Cardiovascular System
Adipogenesis pathway
Hereditary Breast Cancer Signaling
Superpathway of Cholesterol Biosynthesis
Aldosterone Signaling in Epithelial Cells
Nucleotide Excision Repair Pathway
Glucocorticoid Receptor Signaling
Cell Cycle: G1/S Checkpoint Regulation
Assembly of RNA Polymerase II Complex
IC
50
2 IC
50
5 IC
50
QD325
- log ( p-value )

14.2

0

- log ( p-value )

11.6

0

87

Figure 3-8. Top 30 canonical pathways affected by QD compound treatments as shown
with IPA. List was generated by IPA based on activation z score, which indicates
activation / inhibition of specific pathway.

Canonical(pathway(
232(
IC50(
232(
2xIC50(
232(
3xIC50(
325(
IC50(
325(
2xIC50(
325(
5xIC50(
359(
5(uM(
NRF2-mediated Oxidative Stress Response
Apoptosis Signaling
April Mediated Signaling
Rac Signaling
B Cell Activating Factor Signaling
Glioma Invasiveness Signaling
Role of BRCA1 in DNA Damage Response
UVB-Induced MAPK Signaling
p53 Signaling
NGF Signaling
Renal Cell Carcinoma Signaling
Pancreatic Adenocarcinoma Signaling
IGF-1 Signaling
Dopamine-DARPP32 Feedback in cAMP Signaling
ATM Signaling
JAK/Stat Signaling
PCP pathway
Antioxidant Action of Vitamin C
Role of JAK1, JAK2 and TYK2 in Interferon Signaling
Lymphotoxin β Receptor Signaling
4-1BB Signaling in T Lymphocytes
Death Receptor Signaling
Role of CHK Proteins in Cell Cycle Checkpoint Control
Aryl Hydrocarbon Receptor Signaling
PEDF Signaling
Neuregulin Signaling
TWEAK Signaling
Induction of Apoptosis by HIV1
eNOS Signaling
Nitric Oxide Signaling in the Cardiovascular System
activation z score

3.74

-2.67

Figure S2. Top 30 canonical pathways affected by QD compound treatments as shown with IPA. List was
generated by IPA based on activation z score, which indicates activation / inhibition of specific pathway.

88

Figure 3-9. Top 30 canonical pathways affected by QD compound treatments as shown
with IPA. List was generated by IPA comparison analysis based on p value. List was
sorted by hierarchical clusters.

-log(p-value)

14.2

0

Canonical(pathway(
232(
IC50(
232(
2xIC50(
232(
3xIC50(
325(
IC50(
325(
2xIC50(
325(
5xIC50(
359(
5(uM(
NRF2-mediated Oxidative Stress Response
Unfolded protein response
Endoplasmic Reticulum Stress Pathway
Protein Ubiquitination Pathway
tRNA Charging
Role of BRCA1 in DNA Damage Response
Hereditary Breast Cancer Signaling
Hypoxia Signaling in the Cardiovascular System
Superpathway of Cholesterol Biosynthesis
Adipogenesis pathway
Mismatch Repair in Eukaryotes
Heme Biosynthesis II
Glucocorticoid Receptor Signaling
Assembly of RNA Polymerase II Complex
Nucleotide Excision Repair Pathway
Estrogen Receptor Signaling
Cell Cycle: G1/S Checkpoint Regulation
TNFR1 Signaling
Induction of Apoptosis by HIV1
ATM Signaling
DNA Double-Strand Break Repair by HR
Cholesterol Biosynthesis III (via Desmosterol)
Cholesterol Biosynthesis II (via 24,25-
dihydrolanosterol)
Cholesterol Biosynthesis I
p53 Signaling
Aldosterone Signaling in Epithelial Cells
Granzyme A Signaling
Superpathway of Serine and Glycine Biosynthesis I
Serine Biosynthesis
Pentose Phosphate Pathway (Oxidative Branch)
Figure S3. Top 30 canonical pathways affected by QD compound treatments as shown with IPA. List was
generated by IPA comparison analysis based on p value. List was sorted by hierarchical clusters.

89

Figure 3-10. Top 50 gene sets upregulated by QD compound treatments as shown with
GSEA. Top 30 gene sets (FDR q-value < 0.1) affected by each QD treatment were
selected and compiled. The compiled list across all treatments was sorted according to
sum of normalized enrichment score (NES). Top 50 gene sets are shown from the sorted
list. Cells in gray indicates blank, which means the specific gene set was not among the
top 30 gene sets affected by the indicated treatment. Heat map was generated based on
NES.
NES

3.34

1.98

Gene Set
232#
IC50#
232#
2xIC50#
232#
3xIC50#
325#
IC50#
325#
2xIC50#
325#
5xIC50#
359#
5#uM#
GARGALOVIC_RESPONSE_TO_OXIDIZED_PHOSPHOLIPIDS_BLUE_UP !! !! !! !! !! !! !!
PODAR_RESPONSE_TO_ADAPHOSTIN_UP !! !! !! !! !! !! !!
NAGASHIMA_NRG1_SIGNALING_UP !! !! !! !! !! !! !!
KRIGE_AMINO_ACID_DEPRIVATION !! !! !! !! !! !! !!
PICCALUGA_ANGIOIMMUNOBLASTIC_LYMPHOMA_DN !! !! !! !! !! !! !!
CONCANNON_APOPTOSIS_BY_EPOXOMICIN_UP !! !! !! !! !! !! !!
HELLER_SILENCED_BY_METHYLATION_DN !! !! !! !! !! !! !!
BLUM_RESPONSE_TO_SALIRASIB_UP !! !! !! !! !! !! !!
ZHAN_MULTIPLE_MYELOMA_CD1_VS_CD2_UP !! !! !! !! !! !! !!
HELLER_HDAC_TARGETS_DN !! !! !! !! !! !! !!
GERY_CEBP_TARGETS !! !! !! !! !! !! !!
DIRMEIER_LMP1_RESPONSE_EARLY !! !! !! !! !! !! !!
BURTON_ADIPOGENESIS_PEAK_AT_2HR !! !! !! !! !! !! !!
NAGASHIMA_EGF_SIGNALING_UP !! !! !! !! !! !! !!
TIEN_INTESTINE_PROBIOTICS_24HR_DN !! !! !! !! !! !! !!
MITSIADES_RESPONSE_TO_APLIDIN_UP !! !! !! !! !! !! !!
KAN_RESPONSE_TO_ARSENIC_TRIOXIDE !! !! !! !! !! !! !!
BOQUEST_STEM_CELL_CULTURED_VS_FRESH_UP !! !! !! !! !! !! !!
ENK_UV_RESPONSE_KERATINOCYTE_UP !! !! !! !! !! !! !!
HELLER_HDAC_TARGETS_SILENCED_BY_METHYLATION_DN !! !! !! !! !! !! !!
PACHER_TARGETS_OF_IGF1_AND_IGF2_UP !! !! !! !! !! !! !!
BHAT_ESR1_TARGETS_VIA_AKT1_UP !! !! !! !! !! !! !!
PEREZ_TP63_TARGETS !! !! !! !! !! !! !!
DACOSTA_UV_RESPONSE_VIA_ERCC3_UP !! !! !! !! !! !! !!
BENPORATH_ES_WITH_H3K27ME3 !! !! !! !! !! !! !!
HELLER_HDAC_TARGETS_SILENCED_BY_METHYLATION_UP !! !! !! !! !! !! !!
DACOSTA_UV_RESPONSE_VIA_ERCC3_COMMON_UP !! !! !! !! !! !! !!
MENSE_HYPOXIA_UP !! !! !! !! !! !! !!
PEREZ_TP53_AND_TP63_TARGETS !! !! !! !! !! !! !!
UZONYI_RESPONSE_TO_LEUKOTRIENE_AND_THROMBIN !! !! !! !! !! !! !!
BENPORATH_PRC2_TARGETS !! !! !! !! !! !! !!
DAZARD_RESPONSE_TO_UV_NHEK_UP !! !! !! !! !! !! !!
MISSIAGLIA_REGULATED_BY_METHYLATION_UP !! !! !! !! !! !! !!
ALTEMEIER_RESPONSE_TO_LPS_WITH_MECHANICAL_VENTILATION !! !! !! !! !! !! !!
KIM_RESPONSE_TO_TSA_AND_DECITABINE_UP !! !! !! !! !! !! !!
QI_HYPOXIA !! !! !! !! !! !! !!
ADDYA_ERYTHROID_DIFFERENTIATION_BY_HEMIN !! !! !! !! !! !! !!
REACTOME_UNFOLDED_PROTEIN_RESPONSE !! !! !! !! !! !! !!
GROSS_HYPOXIA_VIA_ELK3_DN !! !! !! !! !! !! !!
ONDER_CDH1_TARGETS_1_UP !! !! !! !! !! !! !!
REACTOME_DIABETES_PATHWAYS !! !! !! !! !! !! !!
KEGG_SYSTEMIC_LUPUS_ERYTHEMATOSUS !! !! !! !! !! !! !!
REACTOME_RNA_POL_I_PROMOTER_OPENING !! !! !! !! !! !! !!
REACTOME_AMYLOIDS !! !! !! !! !! !! !!
SCHLOSSER_MYC_TARGETS_REPRESSED_BY_SERUM !! !! !! !! !! !! !!
CUI_GLUCOSE_DEPRIVATION !! !! !! !! !! !! !!
REACTOME_CYTOSOLIC_TRNA_AMINOACYLATION !! !! !! !! !! !! !!
PENG_RAPAMYCIN_RESPONSE_DN !! !! !! !! !! !! !!
HAMAI_APOPTOSIS_VIA_TRAIL_DN !! !! !! !! !! !! !!
PELLICCIOTTA_HDAC_IN_ANTIGEN_PRESENTATION_UP !! !! !! !! !! !! !!
Figure S4 . Top 50 gene sets up regulated by QD compound treatments as shown with GSEA. Top 30 gene sets ( FDR q-value < 0.1)
affected by each QD treatment were selected and compiled. The compiled list across all treatments was sorted according to sum of
normalized enrichment score (NES). Top 50 gene sets are shown from the sorted list. Cells in gray indicates blank, which means the
specific gene set was not among the top 30 gene sets affected by the indicated treatment. Heat map was generated based on NES.

90

Figure 3-11. Top 50 gene sets downregulated by QD compound treatments as shown
with GSEA. Top 30 gene sets (FDR q-value < 0.1) affected by each QD treatment were
selected and compiled. The compiled list across all treatments was sorted according to
sum of normalized enrichment score (NES). Top 50 gene sets are shown from the sorted
list. Cells in gray indicates blank, which means the specific gene set was not among the
top 30 gene sets affected by the indicated treatment. Heat map was generated based on
NES.
Gene set
232(
IC50(
232(
2xIC50(
232(
3xIC50(
325(
IC50(
325(
2xIC50(
325(
5xIC50(
359(
5(uM(
REACTOME_MEIOTIC_RECOMBINATION !! !! !! !! !! !! !!
REACTOME_RNA_POL_I_PROMOTER_OPENING !! !! !! !! !! !! !!
REACTOME_AMYLOIDS !! !! !! !! !! !! !!
KEGG_SYSTEMIC_LUPUS_ERYTHEMATOSUS !! !! !! !! !! !! !!
REACTOME_PACKAGING_OF_TELOMERE_ENDS !! !! !! !! !! !! !!
REACTOME_MEIOTIC_SYNAPSIS !! !! !! !! !! !! !!
REACTOME_RNA_POL_I_TRANSCRIPTION !! !! !! !! !! !! !!
REACTOME_DEPOSITION_OF_NEW_CENPA_CONTAINING_NUCLEOSOMES !! !! !! !! !! !! !!
REACTOME_TELOMERE_MAINTENANCE !! !! !! !! !! !! !!
REACTOME_MEIOSIS !! !! !! !! !! !! !!
REACTOME_CHROMOSOME_MAINTENANCE !! !! !! !! !! !! !!
ZHANG_TLX_TARGETS_36HR_DN !! !! !! !! !! !! !!
DACOSTA_UV_RESPONSE_VIA_ERCC3_COMMON_DN !! !! !! !! !! !! !!
REICHERT_MITOSIS_LIN9_TARGETS !! !! !! !! !! !! !!
GARGALOVIC_RESPONSE_TO_OXIDIZED_PHOSPHOLIPIDS_TURQUOISE_DN !! !! !! !! !! !! !!
GABRIELY_MIR21_TARGETS !! !! !! !! !! !! !!
SENESE_HDAC2_TARGETS_DN !! !! !! !! !! !! !!
REACTOME_RNA_POL_I_RNA_POL_III_AND_MITOCHONDRIAL_TRANSCRIPTION !! !! !! !! !! !! !!
ACEVEDO_LIVER_CANCER_WITH_H3K9ME3_DN !! !! !! !! !! !! !!
ACEVEDO_LIVER_CANCER_WITH_H3K27ME3_DN !! !! !! !! !! !! !!
DAZARD_UV_RESPONSE_CLUSTER_G6 !! !! !! !! !! !! !!
DACOSTA_UV_RESPONSE_VIA_ERCC3_TTD_DN !! !! !! !! !! !! !!
KONG_E2F3_TARGETS !! !! !! !! !! !! !!
LEE_EARLY_T_LYMPHOCYTE_UP !! !! !! !! !! !! !!
ZHAN_MULTIPLE_MYELOMA_PR_UP !! !! !! !! !! !! !!
DACOSTA_UV_RESPONSE_VIA_ERCC3_XPCS_DN !! !! !! !! !! !! !!
WENDT_COHESIN_TARGETS_UP !! !! !! !! !! !! !!
ZHANG_TLX_TARGETS_UP !! !! !! !! !! !! !!
DING_LUNG_CANCER_EXPRESSION_BY_COPY_NUMBER !! !! !! !! !! !! !!
MARTINEZ_RESPONSE_TO_TRABECTEDIN !! !! !! !! !! !! !!
DAZARD_RESPONSE_TO_UV_NHEK_DN !! !! !! !! !! !! !!
MITSIADES_RESPONSE_TO_APLIDIN_DN !! !! !! !! !! !! !!
ROSTY_CERVICAL_CANCER_PROLIFERATION_CLUSTER !! !! !! !! !! !! !!
BENPORATH_EED_TARGETS !! !! !! !! !! !! !!
AMUNDSON_GAMMA_RADIATION_RESPONSE !! !! !! !! !! !! !!
BUYTAERT_PHOTODYNAMIC_THERAPY_STRESS_DN !! !! !! !! !! !! !!
PYEON_CANCER_HEAD_AND_NECK_VS_CERVICAL_UP !! !! !! !! !! !! !!
TOYOTA_TARGETS_OF_MIR34B_AND_MIR34C !! !! !! !! !! !! !!
ISHIDA_E2F_TARGETS !! !! !! !! !! !! !!
PUJANA_BRCA_CENTERED_NETWORK !! !! !! !! !! !! !!
WHITFIELD_CELL_CYCLE_G2 !! !! !! !! !! !! !!
GENTILE_UV_LOW_DOSE_DN !! !! !! !! !! !! !!
PUJANA_XPRSS_INT_NETWORK !! !! !! !! !! !! !!
SHEN_SMARCA2_TARGETS_UP !! !! !! !! !! !! !!
SENGUPTA_NASOPHARYNGEAL_CARCINOMA_UP !! !! !! !! !! !! !!
HAMAI_APOPTOSIS_VIA_TRAIL_UP !! !! !! !! !! !! !!
SMIRNOV_RESPONSE_TO_IR_2HR_DN !! !! !! !! !! !! !!
PID_INSULIN_PATHWAY !! !! !! !! !! !! !!
PYEON_HPV_POSITIVE_TUMORS_UP !! !! !! !! !! !! !!
SLEBOS_HEAD_AND_NECK_CANCER_WITH_HPV_UP !! !! !! !! !! !! !!
NES

-1.85

-3.56

Figure S5 . Top 50 gene sets down regulated by QD compound treatments as shown with GSEA. Top 30 gene sets ( FDR q-value < 0.1)
affected by each QD treatment were selected and compiled. The compiled list across all treatments was sorted according to sum of
normalized enrichment score (NES). Top 50 gene sets are shown from the sorted list. Cells in gray indicates blank, which means the
specific gene set was not among the top 30 gene sets affected by the indicated treatment. Heat map was generated based on NES.

91

3.7 Validation of biomarkers HO-1, NQO1, CHOP and GRP78 in pancreatic
cancer cell lines and PDAC tumors
3.7.1 NRF2 mediated oxidative stress
NRF2 (NFE2L2, nuclear factor erythroid-derived 2 like 2) is a transcription factor
from the cap’n’collar (CNC) family that plays a pivotal role in response to oxidative and
electrophilic stresses by regulating transcription of detoxifying enzymes (Jaiswal, 2004).
Activity of NRF2 is regulated by the cytosolic inhibitor protein KEAP1 (Itoh et al.,
1999). Under normal conditions, KEAP1 sequesters NRF2 in the cytoplasm through
direct interaction and targets the protein for proteasomal degradation. Upon oxidative
challenges, modification on reactive cysteine residues causes conformational changes on
KEAP1 leading to dissociation of Nrf2 from KEAP1 and translocation to the nucleus
(Dinkova-Kostova et al., 2002; Zhang et al., 2003). Heterodimers of NRF2 and MAF
protein activate transcription of antioxidant genes containing the ARE (antioxidant
response element) or the MARE (MAF recognition element) cis-acting enhancer.
A key signaling-pathway affected by treatment with QD compounds, as revealed
by Bru-seq, is the Nrf2-mediated oxidative stress response that is induced within 4 h (Fig.
3-7). The cytoprotective Nrf2 signaling pathway is oncogenic under certain conditions
(Moon et al., 2014; Na et al., 2014). Activating mutations in Nrf2 and KEAP1 are often
seen in cancers as adaptation to elevated intrinsic ROS levels (Hayes et al., 2009). In
PDAC, where such mutations are not very common, the redox state of cancer cells can be
supported through other pathways. The Kras
G12D
mutation, a major oncogenic mutation in
>90% of PDAC cases, activates the Nrf2 signaling through the MAPK pathway, thus

92

modifying the antioxidant program of PDAC cells leading to a decrease in intrinsic ROS
levels that can promote tumor progression (DeNicola et al., 2011). This subtle balance is
crucial to the proliferation of cancer cells as knockdown of Nrf2 in PDAC cell lines
results in decrease cell viability (Lister et al., 2011), showing dependence of cell
proliferation on the antioxidant pathway. Induction of ROS by QD treatment alters this
delicate balance and thus cells respond by upregulating the Nrf2 pathway in an effort to
counter the insult. However, the excess ROS overwhelms the system to the point of no
return leading to cell death.
The Nrf2 inhibitor, trig, can significantly sensitize PDAC cells to etoposide or
TRAIL-induced apoptosis, and also enhances the antitumor response to etoposide in
xenograft models (Arlt et al., 2013). However, trig was not cytotoxic as a single agent
and the sensitization of cells was dependent on the proteasomal pathway. Another
strategy to target the altered redox balance in cancer cells would be to trigger ROS
accumulation. ROS inducing-agents such as QD232 (Pathania et al., 2014),
piperlongumine (Dhillon et al., 2014), and imexon (Dorr et al., 2005) have shown single
agent anti-tumor activities in pancreatic cancer models. A phase I trial of imexon and
gemcitabine in patients with advanced pancreatic cancer demonstrated feasibility and
antitumor responses encouraging continuation to phase II trials (Cohen et al., 2010).
These results support the notion that manipulating redox homeostasis by inducing ROS is
a viable therapeutic strategy.
NQO1 and HMOX1 are two target genes in the NRF2 signaling pathway that
mediate responses to oxidative stress (Alam et al., 1999; Nioi et al., 2003). NQO1
encodes the flavoprotein NAD(P)H:quinone oxidoreductase 1 that catalyzes the two-

93

electron reduction of quinones to hydroquinones and exhibits chemo protective effects
(Dinkova-Kostova et al., 2000; Ross et al., 2000). HMOX1 encodes heme oxygenase 1
(HO-1), whose antioxidant properties arise from degradation of the pro-oxidant heme and
production of antioxidant bilirubin from biliverdin (Choi et al., 1996). As revealed by
Bru-seq, synthesis of NQO1 and HMOX1 RNAs is dose-dependently upregulated by
QD232 and QD325 treatment (Fig. 3-12).

Figure 3-12. Transcription of oxidative stress responsive genes NQO1 and HMOX1 was
upregulated by QD232 or QD325 treatment in MiaPaCa-2 cells dose dependently.

3.7.2 Unfolded protein responses
Another significant pathway upregulated upon drug treatment is the unfolded
protein response (UPR) (Fig. 3-7) as the cells try to cope with increased ROS levels.
Accumulation of ROS can affect a series of molecular signaling and cellular events. In
the ER, where an oxidizing environment is required for formation of disulfide bonds, the
0
20
40
60
80
Control
QD325 - 5×IC
50
RPKM
0
10
20
30 Control
QD325 - 2×IC
50
RPKM
0
10
20
30 Control
QD325 - IC
50
RPKM
0
20
40
60 Control
QD232 - 3×IC
50
RPKM
0
10
20
30
Control
QD232 - 2×IC
50
RPKM
0
10
20
30 Control
QD232 - IC
50
RPKM
NQO1 – NRF2-mediated Oxidative Stress Response HMOX1 – NRF2-mediated Oxidative Stress Response

0
2
4
6
QD232 - IC
50
RPKM
Control
0
2
4
6
RPKM
QD232 - 2×IC
50
Control
0
2
4
6
RPKM
Control
QD325 - IC
50
0
2
4
6
Control
RPKM
QD325 - 2×IC
50
chr22
HMOX1
Scale
Scale
chr22:
HMOX1
HMOX1
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(141)
RepeatMasker
5 kb hg19
35,785,000 35,790,000
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
35,785,000 35,790,000
5 kb hg19
chr16
NQO1
Scale
69,750,000 69,760,000
5 kb hg19
Scale
chr16:
NQO1
NQO1
NQO1
NQO1
NQO1
NQO1
NQO1
NQO1
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(141)
RepeatMasker
5 kb hg19
69,750,000 69,760,000
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
0
10
20
30
40
50
Control
QD325 - 5×IC
50
RPKM
0
10
20
30
QD232 - 3×IC
50
Control
RPKM

94

luminal redox state is constantly under close regulation. Serving as the ER surveillance
system, UPR is triggered when there is imbalance between protein-folding demands and
protein folding capacity, as well as any homeostatic perturbation that can cause protein
misfolding, such as nutrition deprivation or redox insults (Malhotra et al., 2007).
UPR comprises three different pathways regulated respectively by the ER trans-
membrane proteins inositol-requiring enzyme 1a (IRE1a), activating transcription factor
6 (ATF6), and protein kinase RNA-like endoplasmic reticulum kinase (PERK) (Shamu et
al., 1996; Haze et al., 1999; Harding et al., 2000). As the stress sensor, the ER chaperone
78-kDa glucose regulated protein (GRP78) complexes with IRE1a, ATF6 and PERK in
their inactive form under normal conditions. However, during ER stress, misfolded
proteins in the ER lumen bind to GRP78 competitively. Dissociation of GRP78 leads to
activation of IRE1a, ATF6 and PERK and downstream responses to UPR (Hetz, 2012).
Depending on the severity and duration of the ER stress, the UPR can function as a pro-
survival mechanism and restore homeostasis, or trigger apoptosis when the stress burden
is beyond the capacity of this adaptive response (Kim et al., 2006; Verfaillie et al., 2013).
DDIT3 and HSPA5 are representative genes of UPR signaling. HSPA5 encodes
GRP78, the master regulatory protein of ER stress. DDIT3 is a downstream target gene
that responds to all three arms of UPR. As a transcription factor, the DDIT3 gene product
CHOP (CCAAT-enhancer-binding protein homologous protein) promotes apoptosis
under prolong ER stress (Oyadomari et al., 2004; Nishitoh, 2012). Transcription of the
two stress responsive genes DDIT3 and HSPA5 is significantly increased by QD232 or
QD325 treatment dose dependently (Fig. 3-13).

95

Figure 3-13. Transcription of unfolded protein response target genes DDIT3 and HSPA5
was upregulated by QD232 or QD325 treatment in MiaPaCa-2 cells.

Upregulation of mRNA synthesis is further translated into increased protein levels
of these major stress responsive genes. We observed increased protein levels of CHOP
and GRP78 in MiaPaCa-2, Panc-1, and BxPC-3 cells (Fig. 3-14) confirming UPR as a
major drug mechanism. For the oxidative responsive genes, HO-1 was upregulated by
QD treatments in MiaPaCa-2 and BxPC-3, while no significant change was detected in
Panc-1. Of note, NQO1 gene is deleted in Panc-1 cells, and no expression of the gene
was observed in this cell line. In MiaPaCa-2 and BxPC-3, NQO1 showed high basal
expression levels, thus no further induction were observed. These results suggest that
oxidative stress responses are more sensitive in MiaPaCa-2 and BxPC-3 cells.
0
5
10
15 Control
QD232 - 3×IC
50
RPKM
DDIT3 – Unfolded protein response
0
2
4
6
8 Control
QD232 - IC
50
RPKM
0
2
4
6
8 Control
QD232 - 2×IC
50
RPKM
0
5
10
15 Control
QD325 - IC
50
RPKM
0
5
10
15 Control
QD325 - 2×IC
50
RPKM
0
20
40
60
80
Control
QD325 - 5×IC
50
RPKM
0
20
40
60
80 Control
QD232 - IC
50
RPKM
0
20
40
60
80 Control
QD232 - 2×IC
50
RPKM
0
100
200
300 Control
QD232 - 3×IC
50
RPKM
0
20
40
60
80 Control
RPKM
QD325 - 2×IC
50
HSPA5 – Unfolded protein response
0
20
40
60
80 Control
QD325 - IC
50
RPKM
chr9
HSPA5
Scale 2 kb hg19
Scale
chr9:
HSPA5
HSPA5
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(141)
RepeatMasker
2 kb hg19
128,000,000
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
128,000,000 chr12
DDIT3
Scale 1 kb hg19
57,912,000
Scale
chr12:
MARS
MARS
MARS
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
Mir_616
MARS
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
MIR616
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(141)
RepeatMasker
1 kb hg19
57,912,000 57,913,000 57,914,000
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
57,914,000
0
100
200
300 Control
QD325 - 5×IC
50
RPKM

96

Figure 3-14. QD compounds induce protein expression of target genes for oxidative
stress and unfolded protein response. Protein expression levels of oxidative stress
responsive genes NQO1, HO-1 and unfolded protein response target genes CHOP and
GRP78 are regulated to different extents by QD232 or QD325 treatment time
dependently in A) MiaPaCa-2, B) Panc-1 and C) BxPC-3 cells. Protein levels were
quantified by ImageJ and normalized to respective loading controls. Data on
quantification plots represent Mean ± SD from three independent experiments. P values
were calculated using student’s t-test. *, p<0.05; **, p<0.01, ***, p<0.001.

Figure 4
A
B
C
34 kDa
34 kDa
26 kDa
72 kDa
43 kDa
ACTIN
NQO1
GRP78
HO-1
CHOP
MiaPaCa-2
0 4 8 12 24
QD325
(2 µM)
0 4 8 12 24 (h)
QD232
(5 µM)
34 kDa
34 kDa
26 kDa
72 kDa
43 kDa ACTIN
NQO1
GRP78
HO-1
CHOP
Panc-1
0 4 8 12 24
QD325
(2 µM)
0 4 8 12 24 (h)
QD232
(5 µM)
34 kDa
34 kDa
26 kDa
72 kDa
43 kDa
ACTIN
NQO1
GRP78
HO-1
CHOP
BxPC-3
0 4 8 12 24
QD325
(2 µM)
0 4 8 12 24 (h)
QD232
(5 µM)
HO-1
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0.0
0.5
1.0
1.5
2.0
Relative expression
(fold change)
GRP78
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
2
4
6
QD325
(2 µM)
QD232
(5 µM)
** *
*
*
*
*
Relative expression
(fold change)
CHOP
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
5
10
QD325
(2 µM)
QD232
(5 µM)
**
*
*
*
Relative expression
(fold change)
HO-1
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
1
2
3
4
5
**
*
*
*
Relative expression
(fold change)
NQO1
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0.0
0.5
1.0
1.5
Relative expression
(fold change)
CHOP
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
1
2
3
4
*
*
*
Relative expression
(fold change)
QD325
(4 µM)
QD232
(10 µM)
GRP78
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0.0
0.5
1.0
1.5
2.0
**
*
*
*
Relative expression
(fold change)
QD325
(4 µM)
QD232
(10 µM)
NQO1
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0.0
0.5
1.0
1.5
2.0
Relative expression
(fold change)
HO-1
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
1
2
3
4
**
***
***
*
*
***
*
Relative expression
(fold change)
CHOP
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
5
10
QD325
(2 µM)
QD232
(5 µM)
**
*
*
Relative expression
(fold change)
GRP78
ctrl
4h
8h
12h
24h
ctrl
4h
8h
12h
24h
0
5
10
QD325
(2 µM)
QD232
(5 µM)
*
*
*
**
Relative expression
(fold change)

97

UPR has a critical role as a sensitive cytoprotective response to integrated stresses
including oxidative stress. The activation of Nrf2-mediated oxidative stress response and
UPR both imply induction of oxidative stress by QD232 and QD325. While Nrf2
response is more of a prosurvival system for alleviating oxidative insults, serving here as
an indicator of substantial oxidative stress; activation of UPR is speculated to be both an
adaptive response and a major executor of cytotoxicity of QD compounds. While
response to oxidative stress was triggered as a result of ROS accumulation to restore
redox homeostasis, switches in stress signaling directs cells down the apoptosis path for
elimination when the stress is beyond repair. IPA analysis suggests significant activation
of apoptosis signaling at higher concentration of QD232 (3 times IC
50
) or QD325 (5
times IC
50
) after 4 h treatment (Fig. 3-8), which suggest ER mediated apoptosis can be a
major mechanism of cytotoxicity of QD compounds.

To further evaluate the mechanisms of action of QD325 in vivo, we examined
protein levels of stress responsive markers in tumor lysates. HO-1, CHOP and GRP78
protein levels were significantly upregulated in QD325 treated tumors compared to
vehicle controls, further confirming induction of oxidative stress and unfolded protein
response as major mechanisms of action for QD325 in pancreatic cancer models (Fig. 3-
15)

98

Figure 3-15. Protein levels of NQO1, HO-1, CHOP and GRP78 in MiaPaCa-2 xenograft.

3.8 Inhibition of mitochondria gene transcription by QD325 and QD232
Mitochondria play an important role in redox homeostasis in mammalian cells.
Deregulation of mitochondrial genes can lead to interruption of the OXPHOS process and
accumulation of ROS. Mitochondrial DNA (mtDNA) encodes 13 genes that possess
important functions in the electron transport chain. The double-stranded circular DNA
comprises the guanine-rich heavy strand and the cytidine-rich light strand.
QDs preferentially inhibited large genes suggesting that they function by
interfering with transcription elongation. QD232 treatment significantly inhibited the
transcription of mitochondrial (mt)DNA from both the heavy strand promoter HSP2 (top
long arrow in Fig. 3-16) and the light strand promoter LSP (bottom arrow), thus
inhibiting the expression of mitochondrial genes that are essential for mitochondrial
oxidative phosphorylation. However, the activity of the heavy strand promoter HSP1 (top
short arrow), which regulates transcription of 12s rRNA and 16s rRNA, was not affected
by QD232 or QD325 treatment. Both compounds decreased COX III protein levels
confirming decrease in mtDNA gene products (Fig. 3-17). These results strongly suggest
disruption of mitochondrial function contributing to ROS accumulation.
NQO1
HO-1
GRP78
CHOP
ACTIN
34 kDa
34 kDa
26 kDa
72 kDa
43 kDa
1 2 1 2

99

The displacement loop (D-loop) is the only longer, non-coding control region on
the mtDNA sequence where promoters for transcription of the H and L strands (HSP and
LSP), as well as the origin of replication of H strand (O
H
) are located (Falkenberg et al.,
2007). mtDNA alterations in D-loop region have been reported as a frequent event in
lung, hepatocellular, gastric, colorectal, breast and cervical cancers (Suzuki et al., 2003;
Guleng et al., 2005; Wheelhouse et al., 2005; Zhu et al., 2005; Hung et al., 2010;
Kabekkodu et al., 2014). Cancer patients with D-loop mutations, or in particular with
heteroplasmy of the mtDNA D-loop polymorphism, have significantly poorer prognosis
(Lievre et al., 2005; Ye et al., 2014). There are two promoters for the H-strand, while
transcripts from HSP1 terminates right after the two rRNAs, transcripts from HSP2
covers the full strand subjected to subsequent processing into individual mRNA, tRNA
and rRNA molecules (Ojala et al., 1981; Montoya et al., 1983). We have observed
substantial inhibition of mtDNA synthesis from the HSP2 and LSP promoters (Fig. 3-16),
however, the inhibition on transcription from the HSP1 promoter containing synthesis of
12s and 16s rRNA was modest, suggesting selectivity in the transcription inhibition. We
proposed oxidative damage-mediated mtDNA degradation as potential mechanism for
repressed RNA synthesis, and confirmed decrease of mtDNA content with QD compound
treatment. However, the selective transcriptional inhibition implies that mechanisms
other than universal oxidative damaging contributes to such suppressed transcription, and
might serve as a pharmacologically induced model to further elucidate regulation of
mtDNA transcription.
The Bru-seq data suggests that the mechanism by which QD232 and QD325
inhibit mitochondrial function and induce ROS may be at least partially related to

100

blockade of transcription from the mitochondrial genome. To provide additional proof in
support of our findings, we performed similar studies using UV, a well-established ROS-
inducer. We found no significant effect of UV on transcription from mitochondrial
promoters. Similarly, no such effect was observed with additional 16 novel drugs
developed in our laboratory and the DNA topoisomerase I inhibitor Camptothecin (not
shown).
Lower levels of transcription could be a result of decreased DNA templates or
effects on transcription efficiency. Using mtDNA specific primers, we were able to
compare the mitochondrial DNA content among different cell treatments. A small but
significant decrease in mtDNA content was observed 6 hours after H
2
O
2
or QD
compound treatment (Fig. 3-18). Downregulation of mtDNA content is a time dependent
effect. This decrease in DNA templates could be caused by accumulation of ROS, as
suggested by H
2
O
2
treatment, leading to mtDNA damage and degradation (Shokolenko et
al., 2013).
mtDNA is more vulnerable to increased oxidative stress than nuclear DNA
(Halliwell et al., 1991; Alexeyev et al., 2013). Upon treatment with H
2
O
2
(Yakes et al.,
1997) or glucose/glucose oxidase H
2
O
2
-generating system (Salazar et al., 1997),
polymerase-blocking lesions such as strand breaks and oxidative DNA adducts can
accumulate in mtDNA, thus changing the replication and transcription profiles.
Substantial mtDNA damage can further lead to depletion of mtDNA content, which
accounts for the decreased mtDNA content and transcriptional inhibition detected by
Bru-seq after QD232 or QD325 treatment.

101

Figure 3-16. QD compounds inhibit transcription of mitochondrial genome at HSP2 and
LSP promoters. A) Nascent RNA synthesis of Mia PaCa-2 cells was inhibited by 4 h
QD232 (at 6.9 µM) or QD325 (at 5.0 µM) treatment. Top forward arrows represent
transcripts from the heavy strand. While the shorter arrow represents the shorter
transcript regulated by the H1 promoter, the longer arrow represents transcript regulated
by the H2 promoter that covers full length of the mitochondrial genome. Bottom reverse
arrow represents the light strand transcript regulated by the L promoter. Signal from
control is shown in yellow, signal from QD232 treated sample is shown in blue, and
signal from QD325 treated sample is shown in red. The full-length transcripts from both
heavy and light strands are further processed into functional tRNA, rRNA and mRNA
molecules, whose corresponding genes are shown at the bottom of the panel.

5,000 10,000 15,000 chrM
Scale 5 kb hg19
-20000
0
20000
40000
Control
QD232 - 3×IC
50
RPKM
-20000
0
20000
Control
QD325 - 5×IC
50

RPKM

102

Figure 3-17. QD compounds inhibit COXIII protein expression in MiaPaCa-2 cells.

Figure 3-18. QD compounds downregulate mitochondrial DNA content. A) Relative
levels of mtDNA are decreased by 6 h treatment of active QD compounds in MiaPaCa-2.
mtDNA content is calculated by comparing mtDNA 12S rRNA to genomic 18S rRNA,
and data is normalized to controls. Data is shown as Mean ± SD from three independent
experiments. B) Relative levels of mtDNA are decreased time dependently after QD232
or QD325 treatment in MiaPaCa-2.

In this study, we observed variation in basal levels of mtDNA transcription in
different cell lines from both the HSP2 and the LSP promoters. However, there were no
defects in initiation and elongation of the messages. Therefore, blockade of transcription
from the D-loop is unique to the mechanism of QD compounds.

3.9 Conclusions
Previously, we provided proof of concept studies with a tool compound, QD232
that showed remarkable and early onset ROS induction in PDAC cells leading to cell
72 kDa
34 kDa
43 kDa
34 kDa
GAPDH
COXIII
0 4 8 16 24 (h) 0 4 8 16 24 (h)
QD325
(2 uM)
QD232
(5 uM)
TWINKLE
0 4 8 16 24 (h)
QD325
(2 µM)
0 4 8 16 24 (h)
QD232
(5 µM)
GAPDH
B
Ctrl
H
2
O
2
(300 µM)
QD232 (5 µM)
QD325 (2 µM)
QD340 (20 µM)
QD359 (20 µM)
0.0
0.2
0.4
0.6
0.8
1.0
*
**
***
*
*
mtDNA content (12S/18S)
normalized to ctrl (%)
2h 4h 6h 2h 4h 6h
0.0
0.2
0.4
0.6
0.8
1.0
Ctrl
QD232 (5 µM)
QD325 (2 µM)
*
**
***
**
*
mtDNA content (12S/18S)
normalized to ctrl (%)
A
Figure S6. QD compounds inhibit transcription of mitochondrial genome. A) Nascent RNA synthesis of
MiaPaCa-2 cells was inhibited by 4 h QD232 (at 6.9 µM) or QD325 (at 5.0 µM) treatment. Top curves
represent reads mapped to the heavy strand transcript regulated by the HSP2 promoter, bottom curves
represent reads mapped to the light strand transcript regulated by the LSP promoter. Signal from control is
shown in yellow, signal from QD232 treated sample is shown in blue, and signal from QD325 treated sample
is shown in red. The full-length transcripts from both heavy and light strands are further processed into
functional tRNA, rRNA and mRNA molecules, whose corresponding genes are shown at the bottom of the
panel. B) Relative levels of mtDNA are decreased by 6 h treatment of active QD compounds in MiaPaCa-2.
mtDNA content is calculated by comparing mtDNA 12S rRNA to genomic 18S rRNA, and data is normalized
to controls. Data is shown as Mean ± SD from three independent experiments. C) Relative levels of mtDNA
are decreased time dependently after QD232 or QD325 treatment in MiaPaCa-2.

103

death. In order to expand upon our previous observations and to select a compound for
future clinical development we performed a lead optimization campaign and identified
QD325 with better ROS induction capability and improved cytotoxicity in PDAC cells
(Table 3-1, Figure 3-3). Our mechanistic studies presented in this study have further
elucidated the anti-tumor activity of these compounds, and have provided new
perspectives for ROS manipulation as cancer therapy.
In this study, we successfully optimized the anticancer ROS modulator QD232,
leading to the discovery of QD325, which shows significant ROS dependent anticancer
activity in PDAC models. Characterization by Bru-seq revealed cellular responses in the
nucleus, ER, and mitochondrial for the ROS-mediated cytotoxicity in pancreatic cancer
cells, providing more in-depth understanding for mechanisms of this active class.
Exceptional antitumor efficacy and safety profile of QD325 provides strong rationale for
target the mitochondria for treatment of PDAC. In the lead optimization study for
QD232, we identified QD325 as a potent analog that significantly delays tumor growth
through a unique mechanism and can be safely added to gemcitabine-based therapy for
PDAC.

3.10 Materials and methods
3.10.1 Cell culture
Mia Paca-2, Panc-1 and BxPC-3 pancreatic cancer cell lines were obtained from
the ATCC. Normal pancreatic cells HPDE and HPNE were kindly provided by Dr. Diane
Simeone (Translational Oncology Program, University of Southern California, Ann
Arbor, MI). Gemcitabine resistant cell line Mia Paca-2-GR (gemcitabine resistant) was

104

kindly provided by Dr. Sarkar (Department of Pathology, Wayne State University,
Detroit, MI) All cell lines were cultured as monolayer and maintained in RPMI1640
supplemented with 10% fetal bovine serum (FBS) in a humidified atmosphere with 5%
CO
2
at 37°C. Mia Paca-2-GR culture was supplemented with 200 nM gemcitabine.

3.10.2 MTT assay
Cytotoxicity of compounds was evaluated with 3-(4,5-dimethylthiazol-2-yl)-2,5-
diphenyltetrazolium bromide (MTT) assay. Cells were placed in 96-well plate at 3000-
8000 cells/well. After overnight attachment, compounds were added to the wells at
sequential dilutions (30 nM – 10 µM for most cell lines). After 72h treatment, MTT was
added into the media to a final concentration of 300 µg/mL. Cells were incubated for 3 h
at 37°C, and the insoluble formazan converted by viable cells were dissolved in 150 µL
of DMSO. Absorbance at 570nm was read by microplate reader (Molecular devices,
Sunnyvale, CA), and inhibition of cell proliferation was calculated using the following
formula: Inhibition of cell proliferation (%) = (1 - OD
treatment
/OD
control
) x 100%

3.10.3 ROS detection assay
Cells were detached by 0.05% trypsin-EDTA, neutralized, centrifuged (1200 rpm,
5min) and resuspended in cell culture media. Suspension were then treated with 20 µM
cell permeable H2DCFDA for 30 min at 37°C. Cells were then centrifuged (1200 rpm,
5min) and washed with cell culture media to remove excess probe. After washing, cells
were placed in black-wall 384-well plate at 20,000 cells/well, incubated for 30 min and

105

treated by compounds at designated conditions. Fluorescent signal were then read at
493nm/523nm on BioTek H1 plate reader for ROS detection.

3.10.4 Bru-seq analysis for nascent RNA synthesis
Bru-seq analysis was performed as previously reported (Paulsen et al., 2014).
Briefly, 4x10
6
Mia Paca-2 cells were placed in 10 cm dishes on Day 1. On Day 2, cells
were treated with DMSO, QD232 or QD325 for 4 h. Bromouridine was added into the
media to label newly synthesized nascent RNA in the last 30 min of treatment to a final
concentration of 2 mM. Cells were then collected in TRIZOL and total RNA was
isolated. Bromouridine containing RNA population was further isolated and sent for
sequencing. Sequencing reads were mapped to the HG19 reference genome. Pre-ranked
gene lists were generated for each treatment through ranking genes by fold changes in
gene synthesis levels compared with control, and analyzed with GSEA (Broad Institute,
MA) (Mootha et al., 2003; Subramanian et al., 2005). Gene lists with fold changes over
1.5 fold were analyzed with Ingenuity Pathway Analysis (Ingenuity Systems, Redwood
City, CA).

3.10.5 Western blotting analysis
Cells (4x10
5
) were cultured in 60 mm tissue dishes and treated with DFC
compounds at designated concentrations. After treatment, cells were lysed with cell lysis
buffer at 4°C for 30min and centrifuged (12000 rpm, 10min, 4°C). Protein concentrations
of supernatants were measured with BCA assay (Thermo Fisher Scientific). 40 µg protein

106

per sample was subjected to SDS-PAGE analysis. Proteins were then electro transferred
to methanol activated immobilon-FL PVDF membranes (EMD Millipore, Billerica, MA).
Membranes were blocked with 5% skim milk in TBST buffer and incubated with primary
antibodies (anti-NQO1, anti-HO-1, anti-CHOP, and anti-GAPDH from Cell Signaling,
anti-COXIII, anti-beta-tubulin and anti GRP78 from Santa Cruz Biotechnology, anti-
TWINKLE from Sigma) 1:1000 dilutions overnight at 4°C. Membranes were then
washed with TBST (10 min x3), incubated with Dylight 800-conjugated secondary
antibodies (Thermo Fisher Scientific, Rockford, IL) 1:5000 dilutions in 5% milk for 1 h
at room temperature, and washed with TBST (10 min x2) and TBS (10 min). Fluorescent
signal was then scanned by Odyssey Imaging Systems (LI-COR Biosciences, Lincoln,
NE).

3.10.6 Measurement of mtDNA Content by qPCR
To assess mtDNA content, genomic DNA was isolated with QIAamp® DNA
mini kit (Qiagen, Germantown, MD) from Mia PaCa-2 cells. The mtDNA content was
evaluated by co-amplifying a DNA fragment encoding mitochondrial 12S rRNA (forward
primer: 5’-TAGCCCTAAACCTCAACAGT-3’; reverse primer: 5’-
TGCGCTTACTTTGTAG CCTTCAT-3’) and a DNA fragment encoding the nuclear 18S
rRNA (forward primer: 5’- CCCTGCC CTTTGTACACACC-3’; reverse primer: 5’-
GATCCGAGGGCCTCACTA-3’). (Vadrot et al., 2012) Real-time qPCR was performed
on Viia7 cycler (Applied Biosystems). Amplifications were monitored and analyzed by
measuring the intercalation of the fluorescent dye from Fast SYBR
®
Green Master Mix

107

(Applied Biosystems). Relative mtDNA contents were calculated using 18S rRNA as
gene reference.

3.10.7 Xenograft studies
Mia Paca-2 cells (2.0 x 10
6
) in a 100 µL suspension of RPMI1640 was injected
subcutaneously into dorsal flank of 6-week NOD/SCID mice. Tumor size was monitored
twice a week through caliper measurement using the following equation: V=d
2
x D/2,
where d represents width and D represents length of the tumor.
In study 1, mice were randomly grouped (n=5 per group) when average tumor
size reached 65 mm3. Daily treatment was given at five days on two days off cycles.
QD325 was given at 5 mg/kg in 100 µL vehicle (5% DMSO, 60% Propylene glycol, 35%
Saline) by intraperitoneal injection. Study was concluded on day 44 when average tumor
size in the group reached 1200 mm
3
. Unpaired t test was performed for data analysis and
p< 0.05 was considered significant. For tolerance test, two mice remained on each group
beyond day 44 and QD325 dose was gradually increased to 20 mg/kg until day 67.
In study 2, MiaPaCa-2 engrafted mice were randomized into vehicle control
(n=4), gemcitabine treatment 1 (n=3), gemcitabine treatment 2 (n=4), QD325 treatment
(n=3) and combination treatment groups (n=3) when tumor size reached 75mm
3
. In
treatment 1, gemcitabine was given at 15 mg/kg once a week for 48 days; in treatment 2,
gemcitabine was given at 15 mg/kg twice a week for 15 days. For QD325, the compound
was given at 5 mg/kg five times a week. The combination group receives QD325 at 5
mg/kg five times a week as well as gemcitabine at 15 mg/kg once a week. Study was
concluded on day 48.

108

3.10.8 Histochemical Analysis
On necropsy, tumors, hearts, kidneys, livers, lungs, spleens and pancreases were
collected, fixed in 10% neutral buffered formalin, embedded in paraffin, and sectioned.
Sections (5 µM) were stained with hematoxylin and eosin to facilitate histologic
examination. For Ki67 expression level, immunohistochemistry staining was performed
on sections with Ki67 antibody. Embedding, sectioning and staining of samples were
performed by ULAM pathology core for animal research at the University of Michigan.
Representative images were taken on Olympus IX83 microscope with 20X
magnification.

109

CHAPTER 4 Discovery of novel N-(8-quinolinyl) nicotinamides (QNs) for the
treatment of pancreatic cancer
4.1 Phenotypic screen for cancer drug discovery
High throughput screening represents a major contributor to drug discovery in the
past decades with development in the high throughput experimental platforms (Sundberg,
2000; Mayr et al., 2009) and expanding collections of synthetic small molecules prepared
by new organic chemistry strategies (Nielsen et al., 2008; CJ et al., 2012). Advances in
target-based screening approaches are especially prominent, which has enabled rapid
discovery of potent and selective molecules against single targets. Discovery of cancer
drug candidates has largely adopted this approach and generated a variety of targeted
therapy candidates that entered clinical trials. However, only 5% of such investigational
agents are licensed after demonstrating efficacy in phase III trials (Hutchinson et al.,
2011). As a multifactorial disease, cancer displays low therapeutic susceptibility to single
targeted treatment by hijacking multiple signaling networks with functional redundancies
to fulfill its deleterious features. The high heterogeneity in target mechanisms among
patients adds to complexity of the disease, leading to high drug attrition rates in
oncology. As a result, the therapeutic demand for cancer treatment is highly unmet, and it
calls for discovery of therapeutics with novel mechanisms. To address such challenges in
cancer drug discovery, phenotypic screen with representative models resurges as a
promising strategy (Moffat et al., 2014). Without predetermined therapeutic assumptions
on certain targets, this approach uses selected phenotype as experimental readout, giving
rise to discovery of new drug candidates with novel targets and unique mechanisms of
action in the disease-relevant context. In addition, when combined with pathway profiling

110

or genomic analysis, phenotypic assays can guide rational drug combination, which
represents the current standard of care for cancer (Al-Lazikani et al., 2012; Yap et al.,
2013; Dawson et al., 2014). Taken together, phenotypic screen demonstrates great
potential as the starting point for cancer drug discovery.

4.2 Structure activity relationship of QN compounds in pancreatic cancer cell
lines
Previously, we performed a phenotypic screen of a library of 20,000 small-
molecule compounds representing 5 million chemicals. We identified QN519 as a
promising hit compound for further analysis based on its novelty, drug-like properties,
and in vitro activity profile in a panel of 10 cancer cell lines. In collaboration with Dr. Jia
Zhou we designed and synthesized a series of analogs to perform a robust SAR analysis.
Initially, we tested the 50 analogs in three pancreatic cancer cell lines using MTT
assay. Sixteen compounds produced IC
50
values < 1 µM in at least one cell line.
The 6-substituted 3-methyl-pyrazine was found to be important for cytotoxicity of
QN519. Changing the 3-methyl-pyrazine group to substituted pyridine (QN522, 524,
532) substantially decreased its cytotoxicity. Removing the methyl group at 3 position of
pyrazine resulted in a potent compound QN523, implicating that the additional methyl
group is not necessary for compound-target interaction. Changing the heterocyclic group
to pyridazine (QN566) largely decreased the activity, while changing to pyrimidine
(QN567) only slightly affected the cytotoxicity when the substitution remains on the 6
position of the ring.

111

The 8-subsituted quinoline is also necessary for the activity of QN519.
Substitutions on the 5 or 6 position (QN520, 521) of the quinoline decreased activity of
the lead compound. Additional side chain on the 2 position of the quinoline also led to
loss of activity (QN632, 634, 659, 660). Fluorine (QN651, 652 and 658) or methoxy
(QN663, QN107, QN113) substitution on the 6 position improved the activity modestly,
suggesting the position as sight for potential modification.
In an effort to investigate this potential modification site, a series of compounds
with longer side chains were designed to 1) improve solubility and 2) increase specificity
through additional binding sites. Select compound from this round of optimization was
intended for linkage with fluorescent or biotin probe for target identification in vitro
using biochemistry methods. However, the 18 compounds with different linker side
chains do not retain similar cytotoxicity, so future investigation will continue in
addressing the purposes mentioned above.
Table 4-1. IC
50
values of HJC compounds in pancreatic cancer cell lines.
Code Compound Structure M.W.
IC50 Values (µM)
[1]

MiaPaCa-2 Panc-1 BxPC-3
Gemcitabine 0.11 ± 0.07 0.20 ± 0.10 0.05 ± 0.02
*
[2]

QN519

264.28 0.50 ± 0.18 1.80 ± 0.35 9.40 ± 0.51
QN520

264.28 >10 >10 >10
CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

112

QN521

264.28 >10 >10 >10
QN522

263.29 >10 >10 >10
* QN523

250.26 0.11 ± 0.03 0.50 ± 0.07 3.30 ± 0.26
QN524

328.16 >10 >10 >10
QN529

328.16 1.83 ± 0.72 4.00 ± 0.80 >10
QN530

328.16 >10 >10 >10
QN532

267.26 >10 >10 >10
QN566

250.26 8.50 ± 1.03 >10 >10
* QN567

250.26 0.33 ± 0.06 >10 >10
QN571

343.19 2.23 ± 0.31 6.83 ± 1.73 >10
CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N
N
N
NH
O
N
N
HJC-5-19
(F8P33)
HJC-5-21
N
H
N
O
N
N
HJC-5-20
N
NH
O
N
HJC-5-22
N
NH
O
N
N
HJC-5-23
N
NH
O
N
Br
HJC-5-24
N
NH
O N
Br
HJC-5-29
N
NH
O N
HJC-5-30
Br
N
NH
O
N
F
HJC-5-32
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
13
N
3
O
Mol. Wt.: 263.2939
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
BrN
3
O
Mol. Wt.: 328.1634
C
15
H
10
FN
3
O
Mol. Wt.: 267.2578

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

113

QN572

300.31 >10 >10 >10
QN573

284.70 7.00 ± 1.30 9.30 ± 0.80 >10
QN608

379.21 >10 >10 >10
QN609

331.32 >10 >10 >10
QN610

389.40 >10 >10 >10
* QN618

329.15 0.75 ± 0.09 7.02 ± 2.92 7.51 ± 1.32
QN632

278.31 >10 >10 >10
QN634

264.28 >10 >10 >10
CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O
N N
N
NH
O N
N
N
NH
O N
N Br
N
NH
O N
N
N
NH
O N
N
Cl
N
NH
O N
N
Br
N
NH
O N
OH
OH
N
NH
O N
OMe
OMe
OMe
N
NH
O N
N Br
HJC-5-66 HJC-5-67 HJC-5-71
HJC-5-72 HJC-5-73 HJC-6-8
HJC-6-9 HJC-6-10
HJC-6-18
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
14
H
10
N
4
O
Mol. Wt.: 250.2554
C
15
H
11
BrN
4
O
Mol. Wt.: 343.1780
C
18
H
12
N
4
O
Mol. Wt.: 300.3141
C
14
H
9
ClN
4
O
Mol. Wt.: 284.7005
C
18
H
11
BrN
4
O
Mol. Wt.: 379.2101
C
19
H
13
N
3
O
3
Mol. Wt.: 331.3248
C
22
H
19
N
3
O
4
Mol. Wt.: 389.4040
C
14
H
9
BrN
4
O
Mol. Wt.: 329.1515

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

114

* QN651

268.24 0.13 ± 0.06 0.67 ± 0.25 >10
* QN652

282.27 0.31 ± 0.16 2.12 ± 0.94 >10
QN655

251.24 7.70 ± 1.53 6.72 ± 1.02 7.22 ± 1.02
* QN658

268.24 0.30 ± 0.10 5.33 ± 1.53 >10
QN659

307.35 >10 >10 >10
QN660

264.28 >10 >10 >10
* QN663

294.31 0.27 ± 0.15 0.83 ± 0.21 6.70 ± 0.60
CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

CONFIDENTIAL

Jia Zhou Lab
Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555
Phone: 409-772-9748 (O) 409-772-9749 (Lab) || Fax: 409-772-9818
N
NH
O N
N
HJC-6-32
N
NH
O N
N
HJC-6-34
N
NH
O N
N
F
HJC-6-51
N
NH
O N
N
F
HJC-6-52
N
N
NH
O N
N
HJC-6-55
N
NH
O N
N
F
HJC-6-58
N
NH
O N
N
N
HJC-6-59
N
NH
O N
N
HJC-6-60
N
NH
O N
N
MeO
HJC-6-63
C
16
H
14
N
4
O
Mol. Wt.: 278.3086
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
15
H
11
FN
4
O
Mol. Wt.: 282.2724
C
13
H
9
N
5
O
Mol. Wt.: 251.2435
C
14
H
9
FN
4
O
Mol. Wt.: 268.2459
C
17
H
17
N
5
O
Mol. Wt.: 307.3498
C
15
H
12
N
4
O
Mol. Wt.: 264.2820
C
16
H
14
N
4
O
2
Mol. Wt.: 294.3080

115

QN792

284.70 2.83 ± 0.76 2.70 ± 0.52 >10
QN793

298.73 1.50 ± 0.50 1.77 ± 0.25 5.2 ± 0.3
* QN794

284.70 0.80 ± 0.05 4.67 ± 1.53 >10
* QN107

280.28 0.28 ± 0.04 0.77 ± 0.21 5.33 ± 1.81
QN113

280.28 1.30 ± 0.35 9.00 ± 1.41 >10
QN137

379.41 1.70 ± 0.52 8.17 ± 1.61 5.17 ± 1.61
* QN138

393.43 0.73 ± 0.15 3.50 ± 0.50 2.00 ± 0.87
N
NH
N
N
O Cl
N
NH
N
N
O Cl
N
NH
N
N
O Cl
N
NH
N
N
O MeO
N
NH
N
N
O MeO
N
NH
O
N
N
O
N
O
N
NH
O
N
N
O
N
O

116

QN142

266.25 >10 >10 >10
* QN144

312.29 0.37 ± 0.15 0.53 ± 0.21 4.33 ± 0.76
QN147

337.37 >10 >10 >10
QN148

351.40 1.80 ± 0.46 4.50 ± 0.50 9.67 ± 0.58
* QN151

409.17 0.87 ± 0.15 9.50 ± 0.50 9.17 ± 0.76
H
2
N
O
O
N
N
N
N
NH
O
N
N
O
F
N
NH
O
N
N
O
N
N
NH
O
N
N
O
N
N
NH
O
N
N
O
NH
Boc

117

* QN152

423.19 0.67 ± 0.12 2.17 ± 0.76 3.67 ± 0.76
QN153

409.17 >10 >10 >10
QN154

309.12 >10 >10 >10
QN156

323.13 >10 >10 >10
QN159

477.23 3.67 ± 1.15 9.50 ± 0.50 3.33 ± 0.58
N
NH
O
N
N
O
NH
Boc
N
NH
O
N
N
O
NH
Boc
N
NH
O
N
N
O
NH
2
N
NH
O
N
N
O
NH
2
N
NH
O
N
N
O
N
Boc

118

QN160

491.25 9.00 ± 1.73 9.67 ± 0.58 9.33 ± 1.15
QN161

477.23 >10 >10 >10
QN162

309.12 >10 >10 >10
QN163

377.18 >10 >10 >10
N
NH
O
N
N
O
N
Boc
N
NH
O
N
N
O
N
Boc
N
NH
O
N
N
O
NH
2
N
NH
O
N
N
O
NH

119

QN164

391.20 6.17 ± 1.04 >10 8.33 ± 1.53
QN165

377.18 >10 >10 >10

[1]
Values are represented as Mean ± SD from three independent MTT assay experiments.
[2]
Marks for compounds with IC
50
lower than 1 µM in at least one cell line.

4.3 QN523 shows significant cytotoxicity in pancreatic cancer cell lines
4.3.1 QN523 is cytotoxic in a panel of cancer cell lines
In our lead optimization campaign, QN523 was identified as the most potent
compound in the series, with IC
50
value of 0.11 µM in MiaPaCa-2 cells, which is
comparable to gemcitabine, the current standard of care therapy for pancreatic cancer.
In order to understand its potential selectivity for different types of cancer and
choose the best model for further characterization, we tested QN523 in a panel of 12
cancer cell lines with various genetic and pathological backgrounds. QN523 showed
N
NH
O
N
N
O
NH
N
NH
O
N
N
O
NH

120

significant cytotoxicity with IC
50
values ranging from 0.1 to 5.7 µM across all 12 cell
lines.
QN523 was highly potent in the pancreatic cancer cell line MiaPaCa-2, leukemia
cell line Jurkat and colorectal cancer cell line HCT116, with IC
50
value around 0.1 µM.
Because of remarkable potency of QN523 in pancreatic cancer we performed in-depth
preclinical studies in this disease model.
Table 4-2. IC
50
values of HJC compounds in cancer cell lines.

[1]
Values are presented as Mean ± SD from three independent MTT assay experiments.

4.3.2 QN523 is cytotoxic in pancreatic cancer cell lines
We then further evaluated the cytotoxicity of QN523 in three pancreatic cancer
cell lines using colony formation assay. Using numbers and sizes of colonies as the
experimental readout, this long-term assay estimates both anti-proliferative and cytotoxic
effect. QN523 showed more potent activity in colony formation assay than in MTT assay
(Fig. 4-1A). Complete inhibition of colony formation was observed in all three cell lines
at 1 µM. At 0.1 µM QN523 treatment, we observed a complete suppression of MiaPaCa-
2 colonies, suggesting further evaluation is warranted.
Figure 1. HJC-523 is cytotoxic in cancer cell lines. A) HJC-523 is cytotoxic in a panel of 12 cancer cell lines with
IC50 values ranging from 0.1 to 5.7 µM. B) HJC-523 inhibits colony formation in PDAC cell lines. Cells were
treated with HJC-523 for 24 h and left in culture in fresh media until colonies were observed in control wells. C)
HJC-523 inhibits cell proliferation time dependently in MiaPaCa-2 cells. Cells were treated with HJC-523 for 1, 4,
8, 24, 48, 72 h and left in culture in fresh media. MTT assay was performed 72 h after initiation of treatment. Data
points represent Mean ± SD from three independent experiments
A B
IC
50
(µM)
[1]
PDAC cell lines HCC cell lines Other cancer cell lines
Mia
PaCa-2
Panc-1 BxPC-3 HepG2 Hep3B SNU398 SNU387 SNU449 SNU475 Ovcar 8 Jurkat
HCT116
p53+/+
QN523 0.11±0.03 0.50±0.07 3.30±0.26 0.50±0.10 0.21±0.09 1.90±0.60 5.73±0.46 0.40±0.15 2.67±0.84 0.30±0.12 0.10±0.04 0.10±0.03
MiaPaCa-2
Panc-1
BxPC-3
HJC-523
Control
HJC-523
Control
HJC-523
Control
1 (µM) 0.3 0.1 0.03
0.01 0.1 1 10
-20
0
20
40
60
80
100
1h
4h
8h
24h
48h
72h
HJC-523 Concentraltion (µM)
Inhibition of Cell Proliferation (%)

121

To test the durability of treatment, MiaPaCa-2 cells were treated with QN523 for
1, 4, 8, 24, 48, 72 h, washed with PBS, and assayed 72hrs later. A time dependent
cytotoxicity effect was observed for QN523 treatment (Fig. 4-1A) suggesting a delayed
onset for activity.

Figure 4-1. QN523 is cytotoxic in pancreatic cancer cell lines. A) QN523 inhibits colony
formation in PDAC cell lines. Cells were treated with QN523 for 24 h and left in culture
in fresh media until colonies were observed in control wells. B) QN523 inhibits cell
proliferation time dependently in MiaPaCa-2 cells. Cells were treated with QN523 for 1,
4, 8, 24, 48, 72 h and left in culture in fresh media. MTT assay was performed 72 h after
initiation of treatment. Data points represent Mean ± SD from three independent
experiments

4.4 QN523 exhibits anti-cancer activity in in vivo pancreatic cancer xenograft
model
To further evaluate the therapeutic potentials of QN523 in pancreatic cancer,
MiaPaCa-2 xenograft were implanted in NOD/SCID mice. When tumor size reached 100
mm
3
, mice were randomized to either vehicle control (n=5) or QN523 treatment (n=5)
Figure'4)1.'QN523&is&cytotoxic&in&pancrea3c&cancer&cell&lines.&A)&QN523&inhibits&colony&forma3on&in&PDAC&cell&lines.&Cells&
were&treated&with&QN523&for&24&h&and&leB&in&culture&in&fresh&media&un3l&colonies&were&observed&in&control&wells.&B)&QN523&
inhibits&cell&prolifera3on&3me&dependently&in&MiaPaCaG2&cells.&Cells&were&treated&with&QN523&for&1,&4,&8,&24,&48,&72&h&and&leB&
in&culture&in&fresh&media.&MTT&assay&was&performed&72&h&aBer&ini3a3on&of&treatment.&Data&points&represent&Mean&±&SD&from&
three&independent&experiments&

A B
MiaPaCa-2
Panc-1
BxPC-3
QN523
Control
QN523
Control
QN523
Control
1 (µM) 0.3 0.1 0.03
0.01 0.1 1 10
-20
0
20
40
60
80
100
1h
4h
8h
24h
48h
72h
QN523 Concentraltion (µM)
Inhibition of Cell Proliferation (%)

122

group. QN523 was initially given at 10 mg/kg with intraperitoneal administration. Since
no significant delay in tumor growth was observed in the QN523 treatment group from
day 1 to day 9, dose of QN523 was increased to 20 mg/kg from day 10 and continued
until day 44.
QN523 treatment delayed growth of the MiaPaCa-2 xenograft starting from day
17 (p < 0.01). On day 44, when mean tumor volume of the vehicle control group reached
1291 ± 72 mm
3
, mean tumor volume of the QN523 treatment group was only 259 ± 38
mm
3
(p < 0.0001) indicating 80% inhibition of tumor growth (Fig. 4-2A).
No symptoms of gross toxicity such as weakness, weight loss or lethargy were
observed in the QN523 treatment group (Fig. 4-2B). H&E stained organ sections of liver,
kidney, heart, lung, spleen and pancreas did not reveal major histopathological changes,
further confirming the safety of the treatment (Fig. 4-2C). Following the 44-day
treatment, two mice were kept on each group to evaluate efficacy and safety of QN523 at
higher doses. While tumors in the control group exhibited rapid growth, QN523 treatment
was able to delay growth of the tumors, and no systemic toxicity was observed at doses
as high as 40 mg/kg (Fig. 4-2D&E).
In line with the tumor growth inhibition, QN523 treatment decreased Ki67 levels
in tumor tissues, suggesting inhibition of cell proliferation (Fig. 4-3).
Taken together, in vivo studies in MiaPaCa-2 xenograft model suggests promising
anti-cancer activity and safety profile of QN523, supporting further characterization of
the compound as drug candidate for the treatment of pancreatic cancer.

123

Figure 4-2. QN523 inhibits tumor growth of MiaPaCa-2 xenograft without systemic
cytotoxicity. A) Tumor volume of MiaPaCa-2 xenograft of vehicle or QN523 treated
mice. MiaPaCa-2 xenograft was established in NOD/SCID mice. When tumor size
reached 100 mm
3
, mice were randomized into vehicle control group (n=5) or treatment
group (n=5). QN523 was given by i.p. injection five times a week at 10 mg/kg from day
1 to day 9, then at 20 mg/kg from day 10 to day 44. B) Body weight of vehicle of QN523
treated mice. Representative micrographs of hematoxylin and eosin (H&E)-stained organ
sections. Images were taken with Olympus IX83 inverted microscope at 20X
magnification. In histopathology study, no major microscopic changes were detected in
major organs after QN523 treatment. D) Tumor volumes of study continued after data
shown in panel A. QN523 was given at 20 mg/kg five times a week until day 44. Three
mice from each group were euthanized for tissue analysis. Two mice remained in each
A
Figure' 4)2.' QN523& inhibits& tumor& growth& of& MiaPaCaG2& xenograB& without& systemic& cytotoxicity.& A)& Tumor& volume& of&
MiaPaCaG2&xenograB&of&vehicle&or&QN523&treated&mice.&MiaPaCaG2&xenograB&was&established&in&NOD/SCID&mice.&When&
tumor&size&reached&100&mm
3
,&mice&were&randomized&into&vehicle&control&group&(n=5)&or&treatment&group&(n=5).&QN523&was&
given&by& i.p.&injec3on&five&3mes&a&week&at&10&mg/kg&from&day&1&to&day&9,&then&at&20&mg/kg&from&day&10&to&day&44.&B)&Body&
weight& of& vehicle& of& QN523& treated& mice.& Representa3ve& micrographs& of& hematoxylin& and& eosin& (H&E)Gstained& organ&
sec3ons.&Images&were&taken&with&Olympus&IX83&invertedµscope&at&20X&magnifica3on.&In&histopathology&study,&no&major&
microscopic&changes&were&detected&in&major&organs&aBer&QN523&treatment.&D)&Tumor&volumes&of&study&con3nued&aBer&data&
shown&in&panel&A.&QN523&was&given&at&20&mg/kg&five&3mes&a&week&un3l&day&44.&Three&mice&from&each&group&were&
euthanized&for&3ssue&analysis.&Two&mice&remained&in&each&group&aBer&day&44&and&QN523&dose&was&increased&to&30&mg/kg&
from&day&45,&then&to&40&mg/kg&from&day&51&to&day&60.&E)&Body&weight&of&engraBed&mice&was¬&affected&by&QN523&
treatment&at10&G&40&mg/kg.&Error&bars&indicate&Mean&±&SEM.&
B
Control
QN523
Liver Kidney Heart Lung Spleen Pancrea
s
C
D E
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
20
25
30
Vehicle (n=5)
QN523 (n=5)
Vehicle (n=2)
QN523 (n=2)
From Day 51, 40 mg/kg
From Day 44, 30 mg/kg
From Day 10, 20 mg/kg
From Day 1, 10 mg/kg
Day
Body Weight (g)
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
0
300
600
900
1200
1500
1800
2100
Vehicle (n=5)
QN523 (n=5)
Vehicle (n=2)
QN523 (n=2)
From Day 51, 40 mg/kg
From Day 44, 30 mg/kg
From Day 10, 20 mg/kg
From Day 1, 10 mg/kg
Day
Tumor Volume (mm
3
)
0 5 10 15 20 25 30 35 40 45
15
20
25
30
Vehicle (n=5)
QN523 (n=5)
10 mg/kg 20 mg/kg
Day
Body Weight (g)
0 5 10 15 20 25 30 35 40 45
300
600
900
1200
1500
Vehicle (n=5)
QN523 (n=5)
10 mg/kg 20 mg/kg
Day
Tumor Volume (mm
3
)

124

group after day 44 and QN523 dose was increased to 30 mg/kg from day 45, then to 40
mg/kg from day 51 to day 60. E) Body weight of engrafted mice was not affected by
QN523 treatment at10 - 40 mg/kg. Error bars indicate Mean ± SEM.

Figure 4-3. QN523 inhibits tumor cell proliferation in MiaPaCa-2 xenograft. A)
Representative immunohistochemistry images for Ki67 staining of MiaPaCa-2 xenograft
sections. B) QN523 decreased Ki67 index (percentage of Ki67 positive cells in the field)
of treated tumors. Quantification of Ki67 positive cells were performed with image J on
three fields of each sample, three samples were tested for each treatment group. Data
represents Mean ± SD. P values were calculated using student’s t-test, **** indicates
p<0.0001.

4.5 Bru-seq analysis identified stress signaling and autophagy as major cellular
responses to treatment with QN523
In order to understand the mechanism of action for QN523 in pancreatic cancer,
we performed Bru-seq to evaluate global changes in gene synthesis after QN523
treatment. Our preliminary observations suggest that QN series of compounds require a
minimum of 24 hours to exert significant pharmacological effects. During this time, the
cells begin initiating cascade of pharmacological events unique to these compounds.
However, cell death pathways will be initiated at later time points well beyond 24 hrs.
Therefore, the Bru-seq experiments were performed after 24 hrs drug (1 µM) exposure.
A
Figure'4)3.'QN523&inhibits&tumor&cell&prolifera3on&in&MiaPaCaG2&xenograB.&A)&Representa3ve&immunohistochemistry&images&
for&Ki67&staining&of&MiaPaCaG2&xenograB&sec3ons.&B)&QN523&decreased&Ki67&index&(percentage&of&Ki67&posi3ve&cells&in&the&
field)&of&treated&tumors.&Quan3fica3on&of&Ki67&posi3ve&cells&were&performed&with&image&J&on&three&fields&of&each&sample,&
three&samples&were&tested&for&each&treatment&group.&Data&represents&Mean&±&SD.&P&values&were&calculated&using&student’s&tG
test,&****&indicates&p<0.0001.&&
B
Control QN523
Control
QN523
0
10
20
30
40
50
****
Ki67 Index (%)

125

Using RPKM > 0.5, gene size > 300 bp as the cut off values to eliminate
background noises, there were totally 8521 expressed genes in the QN523 and DMSO
control samples out of around 22,000 genes in the reference genome. 275 genes were
significantly upregulated more than two fold with QN523, and 123 genes were
downregulated by the treatment.

4.5.1 IPA and DAVID revealed QN523-induced stress responses in MiaPaCa-2
For general understanding of cellular functions and pathways regulated by QN523
treatment, the up and downregulated gene lists were analyzed by Ingenuity Pathway
Analysis (IPA). Induction of unfolded protein response, ER stress pathway and circadian
rhythm signaling were most significant with QN523 treatment, where about 25% of
genes in these pathways were upregulated, suggesting activation of stress signaling in
MiaPaCa-2 cells.
The lists were also analyzed by Database for Annotation, Visualization and
Integrated Discovery (DAVID), which identify enriched biological themes with particular
focus on gene ontology terms, as well as functional-related gene groups (Huang da et al.,
2009b; Huang da et al., 2009a). Apoptosis, (bZIP) transcription factors, ER related genes
and stress responses were upregulated by QN523, while chromosomal proteins and cell
cycle genes were downregulated by the treatment (Fig. 4-5). These results again suggest
activation of stress responses by QN523 as shown with IPA analysis. In addition,
induction of apoptosis and disruption of cell cycle might be important cellular events
contributing to cytotoxicity of QN523.

126

Figure 4-4. QN523 induces stress responses in MiaPaCa-2 cells as revealed by Ingenuity
Pathway Analysis. Top 20 canonical pathways regulated by QN523. Histogram
represents percentage of genes regulated in the pathway; numbers on the histogram are
total numbers of genes in the specific pathway. Red color stands for upregulated genes,
and green color stands for downregulated genes. Orange line represents statistical
significance in regulation of the indicated pathway.

Figure 4-5. QN523 induces apoptosis and stress responses in MiaPaCa-2 cells as
revealed DAVID analysis. A) Top 5 biological themes in genes upregulated by QN523
treatment. B) Top 5 biological themes in genes downregulated by QN523 treatment.
Histogram represents statistical significance in regulation of the indicated theme.

0 1 2 3 4 5
Chromosomal protein
Nucleus
M phase
Cell cycle
Histone H2A
-log (p-value)
Up-regulated Down-regulated
0 2 4 6 8 10
Stress response
Apoptosis
Endoplasmic Reticulum
(bZIP) transcription factor
Regulation of apoptosis
-log (p-value)
Figure 2
A B
0 1 2 3 4 5
Chromosomal protein
Nucleus
M phase
Cell cycle
Histone H2A
-log (p-value)
Up-regulated Down-regulated
0 2 4 6 8 10
Stress response
Apoptosis
Endoplasmic Reticulum
(bZIP) transcription factor
Regulation of apoptosis
-log (p-value)
Figure 2
A B

127

4.5.2 GSEA suggested inverse correlation with estradiol-regulated transcription
We also applied Gene Set Enrichment Analysis to the pre-ranked list of all
expressed genes to discover gene sets potentially affected by QN523 treatment, and
identified a list of gene sets enriched in the top or bottom of the pre-ranked list,
suggesting correlation of these functional groups of genes with QN523 treatment. The top
20 up or downregulated gene sets are listed in Table 4-3 and 4-4.
Except for the frequently enriched large gene sets associated with adaphostin,
salirasib, tosedostat and oxidized phospholipids treatments, which show low specificity
and were often found in our analysis with other compounds, there are several highly
enriched gene sets that are of particular interests. Induction of apoptosis and inhibition on
cell cycle were observed among the enriched gene sets, which are in agreement with our
discovery with DAVID analysis. Similar transcription profiles with IKK inhibitor plus
TNF treatment, neuregulin (NRG) treatment and hypoxia were observed, suggesting
potential involvement or similarity with these treatment-related signaling profiles.
Interestingly, the transcription profile of QN523 in MiaPaCa-2 shows inverse
correlation with that of estradiol treatment in MCF7. While the role of estrogen and its
receptor is not well characterized in pancreatic cancer, it is a major promoting factor that
induces cell proliferation in breast cancer cases. Estrogen-regulated genes identified in
breast cancer models contribute to cell motility and cell cycle regulations (Dutertre et al.,
2010). Although the cellular context might be different in breast cancer and pancreatic
cancer cells, potential phenotypic simulation of estrogen inhibition by QN523 in
pancreatic cancer implies that the anti-proliferative activity of QN523 might involve

128

estrogen-regulated genes. This result also supports evaluation of QN523 in estrogen
dependent breast cancer models for validation, further characterization of the compound
and putative therapeutic effects.
Table 4-3. Top 20 gene sets upregulated with QN523 treatment

HJC523'–'Up+GSEA'
NAME SIZE NES FDR q-val
1 PODAR_RESPONSE_TO_ADAPHOSTIN_UP 104 3.0437453 <10E-6
2 TIEN_INTESTINE_PROBIOTICS_24HR_DN 183 2.8273673 <10E-6
3 BLUM_RESPONSE_TO_SALIRASIB_UP 204 2.774138 <10E-6
4 HELLER_SILENCED_BY_METHYLATION_DN 58 2.770397 <10E-6
5 ZHANG_RESPONSE_TO_IKK_INHIBITOR_AND_TNF_UP 112 2.7220533 <10E-6
6 NAGASHIMA_NRG1_SIGNALING_UP 116 2.7184274 <10E-6
7 BOQUEST_STEM_CELL_CULTURED_VS_FRESH_UP 159 2.6981082 <10E-6
8 DUTERTRE_ESTRADIOL_RESPONSE_24HR_DN 244 2.6971843 <10E-6
9 BILD_HRAS_ONCOGENIC_SIGNATURE 142 2.6907232 <10E-6
10 KRIGE_RESPONSE_TO_TOSEDOSTAT_24HR_UP 472 2.6795332 <10E-6
11 GROSS_HYPOXIA_VIA_ELK3_DN 102 2.6527967 <10E-6
12 ONDER_CDH1_TARGETS_1_UP 66 2.6429064 <10E-6
13 GARGALOVIC_RESPONSE_TO_OXIDIZED_PHOSPHOLIPIDS_BLUE_UP 107 2.640958 <10E-6
14 PICCALUGA_ANGIOIMMUNOBLASTIC_LYMPHOMA_DN 102 2.6376972 <10E-6
15 KAN_RESPONSE_TO_ARSENIC_TRIOXIDE 66 2.624026 <10E-6
16 ADDYA_ERYTHROID_DIFFERENTIATION_BY_HEMIN 45 2.602759 <10E-6
17 CONCANNON_APOPTOSIS_BY_EPOXOMICIN_UP 149 2.5632482 <10E-6
18 ELVIDGE_HYPOXIA_BY_DMOG_UP 67 2.5623298 <10E-6
19 ELVIDGE_HYPOXIA_UP 88 2.5620668 <10E-6
20 GARGALOVIC_RESPONSE_TO_OXIDIZED_PHOSPHOLIPIDS_TURQUOISE_UP 65 2.5583937 <10E-6

129

Table 4-4. Top 20 gene sets downregulated with QN523 treatment

4.5.3 Top genes regulated by QN523 treatment
While bioinformatics studies of the differentially expressed gene lists provide
information on functional regulation of the QN523 treatment, it is also important to
identify cellular responders that correlate with QN523 activity and might contribute to its
cytotoxicity. Such selected genes could serve as markers for mechanistic studies in vitro,
and as pharmacodynamics markers for future in vivo applications.
Robust and significant regulation by the treatment is required for potential
biomarkers, so we chose the top genes regulated by QN523 as candidates. The top 20
genes up or downregulated by QN523 are reported here and further studied for their
cellular function (Tables 4-5 and 4-6).

HJC523'–'Down+GSEA'
NAME SIZE NES FDR q-val
1 ROSTY_CERVICAL_CANCER_PROLIFERATION_CLUSTER 128 -2.8185263 <10E-6
2 GARGALOVIC_RESPONSE_TO_OXIDIZED_PHOSPHOLIPIDS_TURQUOISE_DN 44 -2.6491916 <10E-6
3 ZHAN_MULTIPLE_MYELOMA_PR_UP 42 -2.5831153 <10E-6
4 AMUNDSON_GAMMA_RADIATION_RESPONSE 37 -2.5632613 <10E-6
5 BURTON_ADIPOGENESIS_PEAK_AT_24HR 35 -2.5429745 <10E-6
6 LEE_EARLY_T_LYMPHOCYTE_UP 78 -2.5337672 <10E-6
7 CROONQUIST_IL6_DEPRIVATION_DN 86 -2.531633 <10E-6
8 DUTERTRE_ESTRADIOL_RESPONSE_24HR_UP 257 -2.5003061 <10E-6
9 CROONQUIST_NRAS_SIGNALING_DN 65 -2.4969666 <10E-6
10 WHITEFORD_PEDIATRIC_CANCER_MARKERS 101 -2.4715047 <10E-6
11 ISHIDA_E2F_TARGETS 50 -2.4621954 <10E-6
12 MORI_LARGE_PRE_BII_LYMPHOCYTE_UP 78 -2.457367 <10E-6
13 GRAHAM_CML_DIVIDING_VS_NORMAL_QUIESCENT_UP 140 -2.4507458 <10E-6
14 KANG_DOXORUBICIN_RESISTANCE_UP 51 -2.4243076 <10E-6
15 SOTIRIOU_BREAST_CANCER_GRADE_1_VS_3_UP 148 -2.4237332 <10E-6
16 ZHOU_CELL_CYCLE_GENES_IN_IR_RESPONSE_6HR 80 -2.3799622 <10E-6
17 BLUM_RESPONSE_TO_SALIRASIB_DN 276 -2.3799057 <10E-6
18 ZHOU_CELL_CYCLE_GENES_IN_IR_RESPONSE_24HR 113 -2.3664784 <10E-6
19 PID_PLK1_PATHWAY 43 -2.3617427 <10E-6
20 GOBERT_OLIGODENDROCYTE_DIFFERENTIATION_UP 463 -2.3504372 <10E-6

130

Table 4-5. Top 20 genes upregulated by QN523 treatment

Table 4-6. Top 20 genes upregulated by QN523 treatment

4.5.3.1 Stress responsive genes were induced by QN523
The unfold protein response genes DDIT3 and HSPA5 (see chapter 3 for detailed
discussion on these two genes) are among the top 20 upregulated genes, where DDIT3
synthesis was increased by 10.7 fold and HSPA5 by 7.9 fold (Fig. 4-6). Induction of these
Rank ID
Fold
Change
Entrez Gene Name Location Type(s)
1 GDF15 42.8 growth differentiation factor 15 Extracellular Space growth factor
2 ATF3 22.1 activating transcription factor 3 Nucleus transcription regulator
3 UPP1 18.4 uridine phosphorylase 1 Cytoplasm enzyme
4 LOC344887 16.7 NmrA-like family domain containing 1 pseudogene Other other
5 FAM129A 16.0 family with sequence similarity 129, member A Cytoplasm other
6 WIPI1 15.3 WD repeat domain, phosphoinositide interacting 1 Cytoplasm other
7 TRIB3 14.3 tribbles pseudokinase 3 Nucleus kinase
8 HMOX1 13.6 heme oxygenase (decycling) 1 Cytoplasm enzyme
9 CXCL3 11.1 chemokine (C-X-C motif) ligand 3 Extracellular Space cytokine
10 DDIT3 10.7 DNA-damage-inducible transcript 3 Nucleus transcription regulator
11 SLFN5 10.1 schlafen family member 5 Nucleus enzyme
12 HERPUD1 10.0 homocysteine-inducible, endoplasmic reticulum stress-inducible, ubiquitin-like domain member 1 Cytoplasm other
13 GABARAPL1 8.2 GABA(A) receptor-associated protein like 1 Cytoplasm other
14 OSGIN1 8.0 oxidative stress induced growth inhibitor 1 Other growth factor
15 HSPA5 7.9 heat shock 70kDa protein 5 (glucose-regulated protein, 78kDa) Cytoplasm enzyme
16 DNAJB9 7.8 DnaJ (Hsp40) homolog, subfamily B, member 9 Nucleus other
17 SAT1 7.2 spermidine/spermine N1-acetyltransferase 1 Cytoplasm enzyme
18 MAP1LC3B 6.9 microtubule-associated protein 1 light chain 3 beta Cytoplasm other
19 CCNG2 6.6 cyclin G2 Nucleus other
20 CD55 6.3 CD55 molecule, decay accelerating factor for complement (Cromer blood group) Plasma Membrane other
Rank ID
Fold
Change
Entrez Gene Name Location Type(s)
1 C17orf62 -2.5 chromosome 17 open reading frame 62 Other other
2 C9orf140 -2.5 suppressor APC domain containing 2 Nucleus other
3 THAP11 -2.5 THAP domain containing 11 Nucleus other
4 NAT14 -2.4 N-acetyltransferase 14 (GCN5-related, putative) Extracellular Space other
5 FASN -2.4 fatty acid synthase Cytoplasm enzyme
6 SEMA6B -2.4 sema domain, transmembrane domain (TM), and cytoplasmic domain, (semaphorin) 6B Plasma Membrane other
7 CCDC85C -2.4 coiled-coil domain containing 85C Plasma Membrane other
8 HPDL -2.4 4-hydroxyphenylpyruvate dioxygenase-like Other other
9 ZDHHC12 -2.4 zinc finger, DHHC-type containing 12 Other enzyme
10 TERC -2.4 telomerase RNA component Other other
11 SCARNA10 -2.4 small Cajal body-specific RNA 10 Other other
12 SLC37A4 -2.3 solute carrier family 37 (glucose-6-phosphate transporter), member 4 Cytoplasm transporter
13 RANGRF -2.3 RAN guanine nucleotide release factor Plasma Membrane transporter
14 MXD3 -2.3 MAX dimerization protein 3 Nucleus transcription regulator
15 C19orf60 -2.3 chromosome 19 open reading frame 60 Other other
16 RPL23AP32 -2.3 ribosomal protein L23a pseudogene 32 Other other
17 PRKCDBP -2.3 protein kinase C, delta binding protein Cytoplasm other
18 AURKAIP1 -2.3 aurora kinase A interacting protein 1 Nucleus enzyme
19 C19orf76 -2.3 adrenomedullin 5 (putative) Other other
20 RNASEH2C -2.3 ribonuclease H2, subunit C Other other
Rank ID
Fold
Change
Entrez Gene Name Location Type(s)
1 GDF15 42.8 growth differentiation factor 15 Extracellular Space growth factor
2 ATF3 22.1 activating transcription factor 3 Nucleus transcription regulator
3 UPP1 18.4 uridine phosphorylase 1 Cytoplasm enzyme
4 LOC344887 16.7 NmrA-like family domain containing 1 pseudogene Other other
5 FAM129A 16.0 family with sequence similarity 129, member A Cytoplasm other
6 WIPI1 15.3 WD repeat domain, phosphoinositide interacting 1 Cytoplasm other
7 TRIB3 14.3 tribbles pseudokinase 3 Nucleus kinase
8 HMOX1 13.6 heme oxygenase (decycling) 1 Cytoplasm enzyme
9 CXCL3 11.1 chemokine (C-X-C motif) ligand 3 Extracellular Space cytokine
10 DDIT3 10.7 DNA-damage-inducible transcript 3 Nucleus transcription regulator
11 SLFN5 10.1 schlafen family member 5 Nucleus enzyme
12 HERPUD1 10.0 homocysteine-inducible, endoplasmic reticulum stress-inducible, ubiquitin-like domain member 1 Cytoplasm other
13 GABARAPL1 8.2 GABA(A) receptor-associated protein like 1 Cytoplasm other
14 OSGIN1 8.0 oxidative stress induced growth inhibitor 1 Other growth factor
15 HSPA5 7.9 heat shock 70kDa protein 5 (glucose-regulated protein, 78kDa) Cytoplasm enzyme
16 DNAJB9 7.8 DnaJ (Hsp40) homolog, subfamily B, member 9 Nucleus other
17 SAT1 7.2 spermidine/spermine N1-acetyltransferase 1 Cytoplasm enzyme
18 MAP1LC3B 6.9 microtubule-associated protein 1 light chain 3 beta Cytoplasm other
19 CCNG2 6.6 cyclin G2 Nucleus other
20 CD55 6.3 CD55 molecule, decay accelerating factor for complement (Cromer blood group) Plasma Membrane other
Rank ID
Fold
Change
Entrez Gene Name Location Type(s)
1 C17orf62 -2.5 chromosome 17 open reading frame 62 Other other
2 C9orf140 -2.5 suppressor APC domain containing 2 Nucleus other
3 THAP11 -2.5 THAP domain containing 11 Nucleus other
4 NAT14 -2.4 N-acetyltransferase 14 (GCN5-related, putative) Extracellular Space other
5 FASN -2.4 fatty acid synthase Cytoplasm enzyme
6 SEMA6B -2.4 sema domain, transmembrane domain (TM), and cytoplasmic domain, (semaphorin) 6B Plasma Membrane other
7 CCDC85C -2.4 coiled-coil domain containing 85C Plasma Membrane other
8 HPDL -2.4 4-hydroxyphenylpyruvate dioxygenase-like Other other
9 ZDHHC12 -2.4 zinc finger, DHHC-type containing 12 Other enzyme
10 TERC -2.4 telomerase RNA component Other other
11 SCARNA10 -2.4 small Cajal body-specific RNA 10 Other other
12 SLC37A4 -2.3 solute carrier family 37 (glucose-6-phosphate transporter), member 4 Cytoplasm transporter
13 RANGRF -2.3 RAN guanine nucleotide release factor Plasma Membrane transporter
14 MXD3 -2.3 MAX dimerization protein 3 Nucleus transcription regulator
15 C19orf60 -2.3 chromosome 19 open reading frame 60 Other other
16 RPL23AP32 -2.3 ribosomal protein L23a pseudogene 32 Other other
17 PRKCDBP -2.3 protein kinase C, delta binding protein Cytoplasm other
18 AURKAIP1 -2.3 aurora kinase A interacting protein 1 Nucleus enzyme
19 C19orf76 -2.3 adrenomedullin 5 (putative) Other other
20 RNASEH2C -2.3 ribonuclease H2, subunit C Other other

131

two genes accounts for ER stress and activation of unfolded protein responses as revealed
by IPA and DAVID analysis, and thus would serve as robust marker for stress signaling.
GDF15 and ATF3 RNA synthesis was highly upregulated by QN523, with > 20
fold increase (Fig. 4-6). Interestingly, these two genes are also found to be stress-related
(Hai et al., 2001; Vanhara et al., 2012), which correlates with our results from
bioinformatics analysis.

Figure 4-6. QN523 induces transcription of stress responsive genes.

GDF15/NAG-1 is a TGF-beta family member that can be induced by non-
steroidal anti-inflammatory drugs (NSAID) such as sulindac sulfide. It is proposed to
inhibit inflammatory cytokine production. Transgenic mice expressing human NAG-
HSPA5
0
50
100
150
200
Control
QN523
RPKM
DDIT3
0
10
20
30
40
50
Control
QN523
RPKM
ATF3
0
2
4
6
8
Control
QN523
RPKM
GDF15
0
2
4
6
8
Control
QN523
RPKM
1 kb hg19
57,912,000 chr12
DDIT3
Scale
Scale
chr12:
MARS
MARS
MARS
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
Mir_616
MARS
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
DDIT3
MIR616
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
1 kb
hg19
57,912,000 57,913,000 57,914,000
ENST00000537638
ENST00000262027
ENST00000545888
ENST00000548944
ENST00000552914
ENST00000547665
ENST00000552499
ENST00000551172
ENST00000547303
ENST00000551116
ENST00000346473
ENST00000552740
ENST00000547526
ENST00000385293
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
Scale
chr1:
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
ATF3
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
20 kb
hg19
ENST00000366981
ENST00000366987
ENST00000341491
ENST00000465155
ENST00000366985
ENST00000492118
ENST00000366983
ENST00000464547
ENST00000336937
ENST00000578962
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
212,750,000
20 kb
chr16
ATF3
Scale
hg19
Scale
chr19:
GDF15
MIR3189
GDF15
MIR3189
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
1 kb
hg19
18,498,000 18,499,000
ENST00000597765
ENST00000595973
ENST00000252809
ENST00000578735
ENST00000594925
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
1 kb
18,498,000 chr19
GDF15
Scale hg19
hg19
128,000,000
2 kb
chr9
Scale
Scale
chr9:
HSPA5
HSPA5
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
2 kb
hg19
128,000,000
ENST00000324460
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
HSPA5
Induc&on(of(stress(responsive(genes(by(QN523(

132

1/GDF15 (NAG-1
Tg/Lox
) are leaner with lower body weight and are resistant to chemically
or genetically induced intestinal tumors (Kim et al., 2013). GDF15 is also identified as a
p53 target gene and inhibits prostate carcinoma cell growth through TGF-beta signaling
pathway (Tan et al., 2000). Induction of DNA damage and p53 overexpression triggers
growth arrest and apoptosis in breast cancer cells through GDF15 expression (Li et al.,
2000).
Importance of GDF15 for inhibition of tumor growth and metastasis has been well
explored in the context of prostate cancer; however, its cellular receptor is still unknown
(Vanhara et al., 2012). GDF15 expression is increased by TPA (10 ng/mL for 1.5-24h) in
prostate cancer cell line LnCaP, and its induced expression could be abolished by
pretreatment with PKC inhibitor (GF109230x) but not other kinase inhibitors. Forced
expression of constitutively active PKC-alpha or PKC-theta could upregulate basal
expression of GDF15 as well, suggesting PKC as a direct regulator of GDF15 expression
in LnCaP. Inhibition of GDF15 expression by siRNA partially blocks the TPA induced
apoptosis in LnCaP cells, further confirming GDF15 as an inducer of growth
arrest/apoptosis (Shim et al., 2005). In DU-145 cells, treatment with GDF15 also shows
anti-tumor effect by inhibiting cell migration and inducing apoptosis (Liu et al., 2003).
ATF3 is a member of the bZIP family transcription factor and recognized as a
tumor suppressor (Hai et al., 2001). For anti-cancer effect of the folate antimetabolite
pemetrexed in NSCLC, induction of ATF3 is necessary for NOXA-mediated apoptosis
(Yan et al., 2014). GDF15 and ATF3 are co-induced by several compounds, including
indole-3-carbinol, 5F-203 and sulindac (Bottone et al., 2003; Monks et al., 2003; Baek et
al., 2004; Lee et al., 2005). Possessing the C/EBP binding site at its promoter region,

133

GDF15 transcription is activated upon association with C/EBPβ and ATF3 in HCT-116
model (Lee et al., 2010). The study on conjugated linoleic acid (CLA, 50 µM, 24 h)
further revealed AKT/GSK3b/ATF3 dependent expression of GDF15 in colon cancer
cells (HCT-116 and HT-29) in p53-independent manner as compared with all the above
agents. Constitutively active β-catenin construct increased cyclin D1 promoter activity,
but not GDF15 transcription. In this study, ATF3 expression precedes GDF15 expression
as early as 3 hrs after treatment, and is responsible for GDF15 promoter activity
(luciferase reporter construct) as confirmed by drug induced ATF3 and plasmid mediated
overexpression of ATF3. siRNA of GDF15 can partially block CLA induced apoptosis
(Lee et al., 2006).

4.5.3.2 Autophagy related genes were induced by QN523
Surprisingly, characterization of the upregulated gene list revealed three
autophagy related genes among the top 20 (Fig. 4-7). WIPI1, GABARAPL1 and
MAP1LC3B are all reported as autophagy component proteins (Deretic et al., 2013;
Yang et al., 2013). However, this functional group of genes was not identified by
bioinformatics analysis, suggesting lack of autophagy characterization in the current
bioinformatics databases. The concurrent induction of autophagic markers strongly
suggests involvement of autophagy in QN523 cellular activity, and warrants further
studies using these markers.

134

Figure 4-7. QN523 induces transcription of autophagy related genes.

Autophagy is a cellular process for repositioning and recycling building blocks,
representing a central component of the integrated stress response (Kroemer et al., 2010).
It starts by forming double-layer-membrane vesicles from intracellular organelles like
ER. Autophagy proteins accumulate on the vesicles and the vesicles (which can be
detected by puncta formation by GFP-LC3B) are then fuse with lysosome to form
autolysosome and trigger degradation and reuse of the vesicle contents. Protein levels of
LC3B, the MAP1LC3B gene product, are often taken as a marker for autophagy
activation.
MAP1LC3B
0
10
20
30
Control
QN523
RPKM
GABARAPL1
0
1
2
3
4
5
Control
QN523
RPKM
WIPI1
0
1
2
3
4
Control
QN523
RPKM
10,370,000
5 kb
chr12
GABARAPL1
Scale
Scale
chr12:
GABARAPL1
GABARAPL1
GABARAPL1
GABARAPL1
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
5 kb
hg19
10,370,000 10,375,000
ENST00000539408
ENST00000542722
ENST00000545859
ENST00000266458
ENST00000421801
ENST00000544284
ENST00000541453
ENST00000545047
ENST00000543602
ENST00000540424
ENST00000537201
ENST00000545887
ENST00000541960
ENST00000546017
ENST00000539289
ENST00000539170
ENST00000535576
ENST00000545290
ENST00000538416
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
10 kb hg19
66,430,000 chr17
WIPI1
Scale
Scale
chr17:
WIPI1
WIPI1
MIR635
PRKAR1A
WIPI1
MIR635
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
10 kb
hg19
66,430,000 66,440,000 66,450,000
ENST00000448504
ENST00000590353
ENST00000591567
ENST00000586515
ENST00000592030
ENST00000592645
ENST00000262139
ENST00000589459
ENST00000589316
ENST00000585393
ENST00000546360
ENST00000384830
ENST00000590402
ENST00000591744
ENST00000587731
ENST00000591494
ENST00000586815
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
Scale
chr16:
MAP1LC3B
MAP1LC3B
MAP1LC3B
MAP1LC3B
Sequences
SNPs
Human mRNAs
Spliced ESTs
DNase Clusters
Txn Factor ChIP
Rhesus
Mouse
Dog
Elephant
Chicken
X_tropicalis
Zebrafish
Lamprey
Common SNPs(142)
RepeatMasker
5 kb
hg19
87,430,000 87,435,000
ENST00000268607
ENST00000570189
ENST00000565788
ENST00000564844
ENST00000564638
ENST00000534986
ENST00000569147
Layered H3K27Ac
100 -
0 _
100 Vert. Cons
4.88 -
-4.5 _
0 -
87,430,000
5 kb hg19
chr16
Scale
MAP1LC3B
Induc&on(of(autophagy(related(genes(by(QN523(
hg19

135

QN523 induced upregulation of MAP1LC3B, GABARAPL1 and WIPI1. In gene
ontology, these genes fall into two major functional groups that are crucial for autophagy.
MAP1LC3A, MAP1LC3B, MAP1LC3C, GABARAP, GABARAPL1, and
GABARAPL2 are yeast ATG8 orthologs; WIPI (WD repeat protein interacting with
phosphoinositides) family members including WIPI1, WIPI2, WDR45B and WDR45 are
yeast ATG18 orthologs.
WIPI1 was identified as a marker of autophagosome formation across a wide
range of cell lines following thapsigargin and C2-ceramide treatment (Tsuyuki et al.,
2014). Thapsigargin and tunicamycin are ER stress inducers with similar temporal
changes in expression profile of genes with unfolded protein response element (UPRE)
and ER stress element (ERSE) (Dombroski et al., 2010). In HeLa cells, 0.5 µM
thapsigargin or 2 µg/mL tunicamycin treatment for 8h induced cellular stress including
autophagy and ER stress by increasing cellular calcium ion concentration; and induced
WIPI1 mRNA transcription was associated with ER-stress related autophagy (Ogata et
al., 2006; Sakaki et al., 2008). As a sensitive marker for formation of autophagosome,
WIPI1 serves as the back up preparation for protein synthesis after autophagy, and is
eventually degraded in the autolysosome (Tsuyuki et al., 2014).
Interestingly, NSAIDs are also associated with activation of autophagy. Aspirin
inhibits mTOR signaling in colorectal cancer cells by inhibiting S6K1(p-Thr389), S6 (p-
Ser235) and 4EBP1(p-Ser65) at 5 mM 8-16 h treatment, activated AMPK, and induces
autophagy as shown with LC3B accumulation (Din et al., 2012). Sulindac sulfide induces
autophagic death in gastric epithelial cells, where pretreatment with autophagy inhibitors
3-methyladenine and chloroquine inhibits autophagy as well as cell death associated with

136

Sulindac treatment. Celecoxib (80-120 µM, 48 h) also induces both apoptosis and
autophagy in HT-29 and HCT-116. However, inhibition of autophagy increases the
celecoxib-induced apoptosis in this model (Huang et al., 2010). These previous studies
show that cellular stress induced by NSAIDs could trigger activation of the autophagy
program, however, cell fate determination might be context or condition-dependent.

4.5.3.3 Proposed markers for cytotoxicity of QN523
Considering the similarity in stressed-associated transcription profiles of QN523
and NSAIDs, we propose that activation of stress signaling program and autophagy might
be the major mechanisms for QN523 cytotoxicity. The four highly upregulated stress
responsive genes HSPA5, DDIT3, ATF3 and GDF15, and the three autophagic markers
WIPI1, GABARAPL1 and MAP1LC3B could serve as markers as well as potential
drivers for QN523 anti-cancer activity. Their roles in cytotoxicity of QN523 warrant
further investigation.
On the basis of cellular regulatory networks characterized with our preliminary
data and studies mentioned above, we propose a stress associated mechanistic model of
QN523 (Fig. 4-8), where QN523 triggers specific stress signaling pathways to activate
autophagy, growth arrest and apoptosis. We hypothesize the application of signaling
inhibitors and genetic modulators in validation process, so as to understand the role of
each cellular responder or pathway in the context of QN523 anti-cancer activity.

137

Figure 4-8. Proposed model for mechanisms of action of QN523. Reagents in purple
represent potential pharmacological or genetic tools that can be used for characterization
of QN523 activity and validation of the model.

4.6 Discovery of compounds showing similar activity with QN523
The similar transcriptional regulation on stress responsive genes and autophagy-
related genes of QN523 and NSAIDs prompted us to investigate whether there are other
compounds that triggers similar transcriptional profiles as QN523. Identification of such
compounds will not only help to understand QN523’s mechanisms of action, but might
also inspire positioning of this compound as chemical tool or treatment for diseases
besides cancer.
ATF3%
PKC$
GDF15%
c/EBP$ P$
Autophagy%
MAP1LC3B%
GABARAPL1%
WIPI1%
Stress%
HSPA5%
DDIT3%
Apoptosis%
&%
Growth%arrest%
GSK3beta$
QN523%
AKT$
PI3K$
mTOR$
BIM$II$
RO73178220$
Ro=lerin$
Bafilomycin$A1$
Chloroquine$
37MA$
Rapamycin$
Wortmannin$
LY294002$
AR7A014418$$$$$$$$ siRNA$
siRNA$
UpFregulated%genes%
DownFregulated%genes%
P$
Potential mechanisms of action
Signaling$inhibitors$

138

4.6.1 NextBio analysis
Application of the NextBio database allows for discovery of compounds
regulating the gene of interest. Here we report the top 20 compounds regulating
expression of our seven marker genes HSPA5, DDIT3, ATF3, GDF15, WIPI1,
GABARAPL1 and MAP1LC3B (Table 4-7 to 4-13).
Distinct lists were obtained for each marker. We used correlation score of 50 as
selection criteria and compared the lists for different markers. While no compound
upregulates all markers at the same time, the liver X receptor non-steroidal agonist GW
3965 could concurrently upregulate HSPA5, DDIT3, ATF3, GDF15, WIPI1 and
MAP1LC3B. The compound was reported to inhibit development of atherosclerosis in
mice (Joseph et al., 2002) and reduce angiotensin II-mediated vasopressor responses in
rats (Leik et al., 2007). On the other hand, the natural steroid lactone withaferin A
upregulates the six markers except HSPA5. Withaferin A displays anti-inflammatory and
antitumor activity by inhibiting IKKβ and NF-κB activation (Kaileh et al., 2007). It is
also a potent inhibitor of angiogenesis (Mohan et al., 2004).
Using GDF15 as a marker for stress target genes, DDIT3 for unfolded protein
response, and WIPI1 for autophagy related signaling, there are 10 compounds
concurrently activating these three pathways, but clear correlation in cellular functions of
these compounds were not observed (Fig. 4-9A). For the lists of the four stress related
genes, there are 6 compounds in common, namely niclosamide, loperamide, GW 3965,
lasalocid, doxycycline and homochlorocyclizine (Fig. 4-9B). For the three autophagy-

139

related genes, withaferin A is the only compound that shows concurrent upregulation.
(Fig. 4-9C)

Table 4-7. Top 20 compounds affecting HSPA5 expression in NextBio

Table 4-8. Top 20 compounds affecting DDIT3 expression in NextBio

compounds compounds score compounds group # Studies Effect on Query
1 mebhydroline 100 Neurotransmitter Agents 1 up-regulated
2 Protriptyline 91.82152794 Neurotransmitter Agents 1 up-regulated
3 Trifluoperazine 84.55492451 Neurotransmitter Agents 1 up-regulated
4 Arecoline 84.42716643 Neurotransmitter Agents 1 up-regulated
5 Procyclidine 80.4707731 Neurotransmitter Agents 1 up-regulated
6 Dextromethorphan 80.4707731 Neurotransmitter Agents 1 up-regulated
7 Propafenone 80.4707731 Unclassified Mechanisms of Action 1 up-regulated
8 butoconazole 78.30004229 Unclassified Mechanisms of Action 1 up-regulated
9 Tunicamycin 76.40694171 Unclassified Mechanisms of Action 12 up-regulated
10 Nefopam 76.02398954 Unclassified Mechanisms of Action 1 up-regulated
11 AICA ribonucleotide 75.07220517 Unclassified Mechanisms of Action 1 down-regulated
12 Coumarins 75.02173008 Unclassified Mechanisms of Action 1 down-regulated
13 Mycophenolic Acid 74.0440831 Enzyme Inhibitors 2 down-regulated
14 Doxycycline 69.19710395 Unclassified Mechanisms of Action 4 up-regulated
15 1-(5-Isoquinolinesulfonyl)-2-Methylpiperazine 69.14055065 Enzyme Inhibitors 1 down-regulated
16 Hexetidine 66.58030963 Unclassified Mechanisms of Action 1 up-regulated
17 GW 3965 65.54187446 Unclassified Mechanisms of Action 2 up-regulated
18 bacterial lysate 65.2288713 Unclassified Mechanisms of Action 1 up-regulated
19 versipelostatin 64.70946821 Enzyme Inhibitors 1 up-regulated
20 chlorcyclizine 63.88901118 Neurotransmitter Agents 1 up-regulated
Compounds)up*regula0ng)HSPA5)expression)
NEXTBIO
compounds compounds score compounds group # Studies Effect on Query
1
2-tert-butyl-9-fluoro-3,6-dihydro-7H-
benz(h)imidazo(4,5-f)isoquinoline-7-one
100 Enzyme Inhibitors 1 up-regulated
2 Gossypol 96.06378595 Unclassified Mechanisms of Action 1 up-regulated
3 Ethionamide 95.21587961 Antimetabolites 3 up-regulated
4 Thapsigargin 95.0460866 Enzyme Inhibitors 4 up-regulated
5 rottlerin 93.59798309 Enzyme Inhibitors 1 up-regulated
6 tyrphostin AG 1478 93.01833575 Enzyme Inhibitors 1 up-regulated
7 GW 3965 90.46613224 Unclassified Mechanisms of Action 2 up-regulated
8 Tunicamycin 90.2571037 Unclassified Mechanisms of Action 11 up-regulated
9 halofuginone 87.67104908 Enzyme Inhibitors 2 up-regulated
10 Hypericum extract LI 160 86.99050969 Unclassified Mechanisms of Action 1 up-regulated
11 Lasalocid 84.14299 Membrane Transport Modulators 1 up-regulated
12 cyclobenzaprine 84.01944662 Unclassified Mechanisms of Action 1 up-regulated
13 lactacystin 82.75915967 Enzyme Inhibitors 2 up-regulated
14 withaferin A 82.75362316 Unclassified Mechanisms of Action 2 up-regulated
15 syrosingopine 81.16394084 Unclassified Mechanisms of Action 1 up-regulated
16 Dequalinium 81.08332783 Unclassified Mechanisms of Action 1 up-regulated
17 benzyloxycarbonylleucyl-leucyl-leucine aldehyde 80.37611397 Enzyme Inhibitors 4 up-regulated
18 BW B70C 80.33532353 Enzyme Inhibitors 1 up-regulated
19 Deoxycholic Acid 79.13922935 Unclassified Mechanisms of Action 2 up-regulated
20 Monensin 78.72961707 Membrane transport modulator 1 up-regulated
Compounds)up*regula0ng)DDIT3)expression)
NEXTBIO

140

Table 4-9. Top 20 compounds affecting ATF3 expression in NextBio

Table 4-10. Top 20 compounds affecting GDF15 expression in NextBio

Table 4-11. Top 20 compounds affecting WIPI1 expression in NextBio

compounds compounds score compounds group # Studies Effect on Query
1 1,3-dichloro-2-propanol 100 Unclassified Mechanisms of Action 1 up-regulated
2 BW B70C 95.75439152 Enzyme Inhibitors 1 up-regulated
3 1-hydroxycholecalciferol 95.34075062 Unclassified Mechanisms of Action 4 up-regulated
4 ferulic acid 94.93951628 Antioxidants 2 up-regulated
5 Hexetidine 93.18846827 Unclassified Mechanisms of Action 1 up-regulated
6 geraniol 93.05269705 Unclassified Mechanisms of Action 4 up-regulated
7 Acyclovir 91.58587069 Unclassified Mechanisms of Action 4 up-regulated
8 Astemizole 91.07344824 Neurotransmitter Agents 1 up-regulated
9 Ethionamide 90.78306743 Antimetabolites 2 up-regulated
10 pyrvinium 89.52931316 Unclassified Mechanisms of Action 2 up-regulated
11 Bepridil 89.23207495 Membrane Transport Modulators 1 up-regulated
12 bromperidol 88.94271426 Unclassified Mechanisms of Action 1 up-regulated
13 cyanoginosin LR 88.56272774 Enzyme Inhibitors 3 up-regulated
14 cetraxate 88.28387843 Unclassified Mechanisms of Action 2 up-regulated
15 azacyclonol 87.77912444 Unclassified Mechanisms of Action 1 up-regulated
16 Proadifen 87.77912444 Enzyme Inhibitors 1 up-regulated
17 eperisone 87.3855813 Membrane Transport Modulators 2 up-regulated
18 Vanadates 87.24961305 Membrane Transport Modulators 2 up-regulated
19 Cinnarizine 87.16163703 Membrane Transport Modulators 3 up-regulated
20 Deoxycholic Acid 86.64258625 Unclassified Mechanisms of Action 3 up-regulated
Compounds)up*regula0ng)ATF3)expression)
NEXTBIO
compounds compounds score compounds group # Studies Effect on Query
1
4-amino-6-hydrazino-7-beta-D-ribofuranosyl-7H-
pyrrolo(2,3-d)-pyrimidine-5-carboxamide
100 Enzyme Inhibitors 1 up-regulated
2 Mitomycin 95.7056062 Alkylating Agents 5 up-regulated
3 Ethyl Methanesulfonate 94.3983552 Alkylating Agents 1 up-regulated
4 Deoxycholic Acid 86.51680945 Unclassified Mechanisms of Action 1 up-regulated
5 Potassium Dichromate 74.69238994 Unclassified Mechanisms of Action 1 up-regulated
6 Demecolcine 73.59595293 Mitosis Modulators 1 up-regulated
7 Papaverine 73.13980763 Enzyme Inhibitors 3 up-regulated
8 lactacystin 72.84717128 Enzyme Inhibitors 2 up-regulated
9 methixene 69.39413606 Unclassified Mechanisms of Action 1 up-regulated
10 2,2'-(hydroxynitrosohydrazono)bis-ethanamine 69.37492206 Nitric Oxide Donors 1 up-regulated
11 Lasalocid 68.90592316 Membrane Transport Modulators 1 up-regulated
12 pyrvinium 65.98164722 Unclassified Mechanisms of Action 2 up-regulated
13 Monensin 65.88926271 Membrane Transport Modulators 1 up-regulated
14 securinine 65.1400566 Neurotransmitter Agents 1 up-regulated
15 1,3-dichloro-2-propanol 64.2813855 Unclassified Mechanisms of Action 1 up-regulated
16 Amodiaquine 64.0301516 Unclassified Mechanisms of Action 1 up-regulated
17 GW 3965 62.91015204 Unclassified Mechanisms of Action 2 up-regulated
18 Danazol 62.90351426 Unclassified Mechanisms of Action 8 up-regulated
19 naftifine 60.86658033 Unclassified Mechanisms of Action 1 up-regulated
20 Niclosamide 60.73935532 Unclassified Mechanisms of Action 1 up-regulated
Compounds)up*regula0ng)GDF15)expression)
NEXTBIO
compounds compounds score compounds group # Studies Effect on Query
1 Desipramine 100 Enzyme Inhibitors 1 up-regulated
2 Methiothepin 96.66027784 Neurotransmitter Agents 2 up-regulated
3 monastrol 93.30234878 Unclassified Mechanisms of Action 3 up-regulated
4 Trimipramine 88.69696304 Neurotransmitter Agents 5 up-regulated
5 Flupenthixol 88.38301982 Neurotransmitter Agents 6 up-regulated
6 Lidoflazine 87.51725941 Membrane Transport Modulators 7 up-regulated
7 dimethisoquin 87.4829848 Unclassified Mechanisms of Action 8 up-regulated
8 Lasalocid 85.51468732 Membrane Transport Modulators 9 up-regulated
9 Trifluoperazine 83.4783669 Neurotransmitter Agents 10 up-regulated
10 homochlorocyclizine 83.37332255 Neurotransmitter Agents 11 up-regulated
11 Prochlorperazine 82.95354155 Neurotransmitter Agents 12 up-regulated
12 bafilomycin A 82.57299728 Enzyme Inhibitors 13 up-regulated
13 Clopenthixol 81.04285669 Neurotransmitter Agents 14 up-regulated
14 Bufexamac 80.91397366 Unclassified Mechanisms of Action 15 up-regulated
15 acetorphan 80.25373028 Enzyme Inhibitors 16 up-regulated
16 isocorydine 80.25373028 Unclassified Mechanisms of Action 18 up-regulated
17 Aclarubicin 79.9799851 Unclassified Mechanisms of Action 19 up-regulated
18 Monensin 79.38144581 Membrane Transport Modulators 20 up-regulated
19 Nicergoline 79.00912279 Neurotransmitter Agents 21 up-regulated
20 Chenodeoxycholic Acid 78.56708322 Unclassified Mechanisms of Action 22 up-regulated
Compounds)up*regula0ng)WIPI1)expression)
NEXTBIO

141

Table 4-12. Top 20 compounds affecting GABARAPL1 expression in NextBio

Table 4-13. Top 20 compounds affecting MAP1LC3B expression in NextBio

compounds compounds score compounds group # Studies Effect on Query
1 clemizole 100 Neurotransmitter Agents 1 down-regulated
2 Oxyphenbutazone 94.54340488 Unclassified Mechanisms of Action 1 up-regulated
3 Streptomycin 91.97135628 Enzyme Inhibitors 1 up-regulated
4 Xylazine 90.57377267 Neurotransmitter Agents 1 down-regulated
5 butoconazole 86.61005582 Unclassified Mechanisms of Action 1 up-regulated
6 Ampicillin 86.51099014 Unclassified Mechanisms of Action 1 up-regulated
7 PI103 77.34687372 Enzyme Inhibitors 1 up-regulated
8 Trioxsalen 74.89571659 Unclassified Mechanisms of Action 1 up-regulated
9 fenbufen 71.04088548 Enzyme Inhibitors 1 down-regulated
10 Acetohexamide 70.92962897 Unclassified Mechanisms of Action 1 up-regulated
11 acetylleucine 70.46435417 Unclassified Mechanisms of Action 1 down-regulated
12 Apazone 70.46435417 Unclassified Mechanisms of Action 1 up-regulated
13 Pantothenic Acid 70.46435417 Unclassified Mechanisms of Action 1 up-regulated
14 bortezomib 67.93989966 Enzyme Inhibitors 7 up-regulated
15 N-benzyladenine 64.24225268 Unclassified Mechanisms of Action 1 up-regulated
16 Meclofenoxate 64.08969585 Unclassified Mechanisms of Action 1 down-regulated
17 butamben 64.06215313 Unclassified Mechanisms of Action 1 down-regulated
18 Cefoperazone 63.81543286 Unclassified Mechanisms of Action 1 down-regulated
19 Pentolinium Tartrate 63.01565541 Neurotransmitter Agents 1 down-regulated
20 pimethixene 62.85661086 Neurotransmitter Agents 1 up-regulated
Compounds)up*regula0ng)GABARAPL1)expression)
NEXTBIO
compounds compounds score compounds group # Studies Effect on Query
1 Ethoxyquin 100 Antioxidants 1 up-regulated
2 trimethylcolchicinic acid 75.96088895 Unclassified Mechanisms of Action 1 up-regulated
3 Cymarine 70.55146591 Unclassified Mechanisms of Action 1 up-regulated
4 Acetylmuramyl-Alanyl-Isoglutamine 68.31431976 Unclassified Mechanisms of Action 1 down-regulated
5 carbetapentane 68.01648238 Unclassified Mechanisms of Action 1 down-regulated
6 Gossypol 66.313867 Unclassified Mechanisms of Action 1 up-regulated
7 Chorionic Gonadotropin 66.21053244 Unclassified Mechanisms of Action 4 up-regulated
8 Hydroxyzine 62.14316764 Neurotransmitter Agents 1 down-regulated
9 monobenzone 61.02261463 Unclassified Mechanisms of Action 1 up-regulated
10 GW 3965 60.68424478 Unclassified Mechanisms of Action 2 up-regulated
11 rottlerin 60.39782588 Enzyme Inhibitors 1 up-regulated
12 Clioquinol 59.01263647 Unclassified Mechanisms of Action 1 up-regulated
13 Histidinol 57.00502349 Enzyme Inhibitors 1 up-regulated
14 pioglitazone 56.88519534 Unclassified Mechanisms of Action 3 up-regulated
15 Phenoxybenzamine 56.78538959 Neurotransmitter Agents 1 up-regulated
16 Spiperone 56.71940139 Neurotransmitter Agents 1 up-regulated
17 piperlonguminine 56.58188398 Enzyme Inhibitors 1 up-regulated
18 Fendiline 56.58108686 Membrane Transport Modulators 1 up-regulated
19 Nerve Growth Factors 55.97089611 Unclassified Mechanisms of Action 1 down-regulated
20 withaferin A 55.29247785 Unclassified Mechanisms of Action 2 up-regulated
Compounds)up*regula0ng)MAP1LC3B)expression)
NEXTBIO

142

Figure 4-9. Venn diagrams for comparison of compounds regulating potential marker
genes (compound correlation score > 50). A) Comparison of GDF15, DDIT3 and WIPI1.
B) Comparison of stress responsive genes HSPA5, DDIT3, ATF3 and GDF15. C)
Comparison of autophagy related genes WIPI1, GABARAPL1 and MAP1LC3B.

4.6.2 CMAP analysis
Besides using the most highly regulated genes as key transcription signatures as
described with NextBio analysis discussed above, comparison of the overall transcription
profile serve as another approach to identify compounds with similar cellular activity.
We used the up and downregulated gene lists to query the CMAP database for
overall transcription profiles of in-house perturbagens. The top 20 perturbagens
(compounds) correlating with QN523 transcription profile is reported here (Table 4-14).
C
B
A

143

Five adrenergic or dopamine receptor antagonists were identified in the list. Although
these compounds require systemic administration for their therapeutic benefits, similarity
of transcription profiles with these compounds suggests potential correlation in
mechanisms of action. Also two Hsp90 inhibitors geldanamycin and 17-AAG are
identified as showing similar transcription profiles with QN523, suggesting the
involvement of stress responses.
Compounds identified by NextBio or CMAP do not show significant structural
similarity with QN523 (Fig. 4-10). However, correlation of these compounds hints on
potential mechanisms QN523 activity, and application of these compounds as tools for
comparison might be a plausible approach to further characterize QN523 in different
biological systems.

Table 4-14. Top 20 compounds correlating with QN523 transcription profile in
connectivity map

rank cmap name mean n enrichment p specificity
percent
non-null
note
1 phenoxybenzamine 0.795 4 0.972 0 0.0891 100 A non-selective, irreversible alpha antagonist
2 puromycin 0.754 4 0.966 0 0.0393 100 An antibiotic that inhibits translation
3 GW-8510 -0.599 4 -0.946 0 0.0687 100 An inhibitor of cyclin kinase 2 (CDK2)
4 geldanamycin 0.619 15 0.83 0 0.0054 100 A benzoquinone ansamycin antibiotic that inhibits Hsp90
5 thioridazine 0.683 20 0.755 0 0.0091 100 An antipsychotic binding D2, M1, alpha1 and 5-HT
6 15-delta prostaglandin J2 0.632 15 0.7 0 0.0447 86 Selective PPARγ agonist
7 trifluoperazine 0.594 16 0.698 0 0.0048 93 An antipsychotic binding D1, D2 and adrenergic receptors
8 prochlorperazine 0.561 16 0.619 0 0.0437 87 A dopamine (D2) receptor antagonist
9 tanespimycin 0.484 62 0.567 0 0.0259 83 17-AAG, a derivative of the antibiotic geldanamycin
10 trichostatin A 0.464 182 0.518 0 0.2654 82 Selective inhibitor for class I and II HDAC
11 anisomycin 0.683 4 0.938 0.00002 0.0412 100 Inhibits peptidyl transferase or the 80S ribosome system
12 astemizole 0.798 5 0.923 0.00002 0.019 100 A histamine H1-receptor antagonist
13 gossypol 0.639 6 0.839 0.00002 0 100 Inhibitor for several dehydrogenase enzymes
14 thapsigargin 0.8 3 0.981 0.00004 0.0573 100 A non-competitive inhibitor of the sarco/ER Ca
2+
ATPase
15 valinomycin 0.661 4 0.925 0.00004 0.0174 100 A dodecadepsipeptide antibiotic
16 LY-294002 0.278 61 0.283 0.00004 0.3893 63 Inhibitor for PI3Ks
17 fluphenazine 0.434 18 0.515 0.00006 0.0622 77 A antipsychotic binding the dopamine D2 receptors
18 terfenadine 0.762 3 0.963 0.00008 0.0197 100 An antihistamine
19 pyrvinium 0.721 6 0.815 0.00008 0.0279 100 An anthelmintic effective for pinworms
20 6-bromoindirubin-3'-oxime -0.506 7 -0.763 0.00008 0.0047 85 BIO, a potent inibitor of GSK3α/β
Top 20 CMAP perturbagens for HJC-523
Phenoxybenzamine
Geldanamycin
Thioridazine Trifluoperazine
17-AAG
Prochlorperazine

144

Figure 4-10. Compounds exhibiting similar transcription signatures with QN523. A)
Compounds identified by NextBio analysis. B) HSP90 inhibitors identified by
connectivity map (CMAP) C) Alpha or dopamine receptor antagonists identified by
CMAP.

4.7 Validation of biomarkers GDF15, ATF3, DDIT3, HSPA5, WIPI1,
GABARAPL1 and MAP1LC3B in pancreatic cancer cell lines
To validate our findings from bioinformatics analysis basing on Bru-seq
experiments, we further tested the proposed biomarkers in pancreatic cancer cell lines.
In MiaPaCa-2, we observed dose dependent upregulation in protein levels of the
stress responsive genes HSPA5, DDIT3, ATF3 and GDF15 (Fig. 4-11), suggesting that
the regulation in RNA synthesis was further translated into changes in protein levels,
which are essential for actual functional regulations in the cancer cells. The stress
responsive marker HSPA5 was also induced by Sulindac treatment, in agreement with
literature in other cancer models mentioned above, while other markers only showed mild
Phenoxybenzamine
Geldanamycin
Thioridazine Trifluoperazine
17-AAG
Prochlorperazine Fluphenazine
GW 3965 Withaferin A

145

induction. Of note, potency of Sulindac is much lower than that of QN523, with IC
50
of
300 µM in MiaPaCa-2.
We also tested three select compounds from CMAP analysis. The adrenergic
antagonist pheoxybenzamine was not cytotoxic in MiaPaCa-2 (IC
50
> 30 µM), while the
phenothiazine dopamine receptor antagonists thioridazine (IC
50
= 12 µM) and
prochlorperazine (IC
50
= 17 µM) induced inhibition of cell proliferation. When tested at
their IC
50
values (except for non-cytotoxic phenoxybenzamine), the three compounds
exhibits induction of the stress responsive markers. Among them, phrochlorperazine
showed most robust induction of GRP78 and CHOP.

Figure 4-11. QN523 induces protein expression of stress markers dose-dependently.
MiaPaCa-2 cells were treated by QN523, Sulindac, phenoxybenzamine, thioridazine or
prochlorperazine for 72 h and subjected to western blotting analysis of stress responsive
proteins GRP78, CHOP, ATF3, GDF15.

We also observed accumulation of autophagy-related markers WIPI1,
GABARAPL1 and LC3B with QN523 and Sulindac in MiaPaCa-2 upon 72 h treatment
(Fig. 4-12). Accumulation of these proteins indicated activation of autophagy in the
protein level, hence validating our discovery by analysis of nascent RNA transcriptome.
05/04/2015
Beta-tubulin
GDF15
ATF3
GRP78
CHOP
QN523 Sulindac
Stress
markers
* PO – Phenoxybenzamine, 20 µM
TH – Thioridazine, 12 µM
PR – Prochlorperazine, 17 µM
MiaPaCa-2
(72 h treatment)
PO TH PR* 0 0.25 0.5 1 2 0 75 150 300 600 (µM)

146

Among the three markers, WIPI1 showed the most significant dose-dependent induction
by QN523 treatment. Significant induction of the protein was detected at concentrations
as low as 0.25 µM.
Interestingly, the phenothiazine antipsychotic compounds thioridazine and
prochlorperazine displayed robust activation of autophagy at their IC
50
values.
Autophagy activation by the phenothiazine trifluoroperazine was identified in human
glioblastoma cell line H4 through an image-based screening by detecting LC3-GFP
accumulation on autophagosomal membrane (Zhang et al., 2007). Our findings suggest a
potential shared mechanism of autophagy induction by phenothiazines.

Figure 4-12. QN523 induces protein expression of autophagy markers dose-dependently.
MiaPaCa-2 cells were treated by QN523, Sulindac, phenoxybenzamine, thioridazine or
prochlorperazine for 72 h and subjected to western blotting analysis of autophagy related
proteins WIPI1, GABARAPL1 and LC3B.

Going back to bench, we have successfully validated the induction of stress
responses and autophagy at the protein level. Using Sulindac as a positive control for the
proposed markers, we demonstrated the highly potent cellular activity of QN523 in
MiaPaCa-2. In addition, we showed that the FDA-approved compounds identified by
05/04/2015
Beta-tubulin
QN523 Sulindac
PO TH PR* 0 0.25 0.5 1 2 0 75 150 300 600 (µM)
* PO – Phenoxybenzamine, 20 µM
TH – Thioridazine, 10 µM
PR – Prochlorperazine, 20 µM
MiaPaCa-2
(72 h treatment)
WIPI1
GABARAPL1
LC3B
Autophagy
markers

147

CMAP exhibited similar activation of cellular events with QN325. Collectively, our
bioinformatics findings have so far demonstrated robust translation in the pancreatic
cancer cell line MiaPaCa-2, supporting further evaluation of our hypothesis generated
from the Bru-seq analysis.

4.8 Conclusions
In this study, we identified QN523 as a promising compound from a list of 50
analogs of the original hit, QN519. QN523 exhibits potent cytotoxicity in pancreatic
cancer cell lines and significantly delayed growth of MiaPaCa-2 xenograft in NOD/SCID
mice model without systemic toxicity, suggesting promising therapeutic potentials of this
compound in pancreatic cancer.
Bioinformatics analysis of the Bru-seq data using IPA, DAVID and GSEA
revealed stress associated apoptosis and growth arrest by QN523 in MiaPaCa-2 cells. The
highly upregulated genes GDF15, ATF3, DDIT3 and HSPA5 suggest involvement of
stress responses in QN523 treatment. Upregulated expression of WIPI1, GABARAPL1
and MAP1LC3B implicated activation of autophagy. In addition, NextBio and CMAP
identified commercial compounds showing similar transcription signatures with QN523.
We validated the bioinformatics findings at protein expression levels by demonstrating
dose-dependent upregulation of the proposed markers by QN523. Interestingly, similar
cellular effects were observed with the phenothiazine identified by CMAP. A proposed
model was generated for mechanisms of action of QN523 basing on our current
experimental observations. Further validation will be carried out to test this hypothetic

148

model and characterize potential cellular effectors for QN523. Comparison with such
compound will also be carried out as an approach to validate the bioinformatics
discoveries as well as to further characterize potential application of QN523.

4.9 Materials and methods
4.9.1 Compounds
50 QN analogs were synthesized by Dr. Jia Zhou from The University of Texas
medical branch at Galveston. For in vitro studies, compounds were dissolved in DMSO at
10 mM as stock solutions and stored at -20°C. For in vivo studies, QN523 was dissolved
in a vehicle containing 5% DMSO, 35% propylene glycol and 60% saline, and
administered at 100 uL through intraperitoneal injections.
Pheoxybenzamine, thioridazine and prochlorperazine were from Cayman
chemical. Sulindac was obtained from LKT Labs.

4.9.2 Western blotting
Please see chapter 2 for general methods of western blotting. In addition, primary
antibodies for CHOP, GDF15 and LC3B are from Cell Signaling Technology; GRP78,
ATF-3 and beta-tubulin are from Santa Cruz; WIPI-1 is from Pierce; GABARAPL1 is
from Proteintech.

149

4.9.3 Bioinformatics analysis
Bru-seq data from QN523 treatment (1 µM, 24 h) and vehicle controls in
MiaPaCa-2 was processed by filtering for genes with RPKM > 0.5, gene size > 300 bp.
Expressed genes were then preranked according to fold change comparing to control and
subjected to analysis. Please see chapter 2 for application of GSEA and IPA on the gene
lists.
For DAVID analysis, lists of genes upregulated or downregulated by at least 2
fold with QN523 were generated, and subjected to identification of common biological
themes in each list. Top functional terms identified by DAVID are reported.
For NextBio analysis, the seven marker genes were searched independently for
pharmaco atlas. The top 20 correlated compounds are reported. Compounds with
correlation score higher than 50 in each list were subjected to comparison among lists of
the seven markers. Comparison was performed using the online Venn diagram server at
http://bioinformatics.psb.ugent.be/webtools/Venn/.
For connectivity map analysis, same lists for DAVID analysis were used as a pair
of description for QN523 treatment and queried the CMAP database. Top 20 enriched
compounds are reported.

150

CHAPTER 5 Conclusions and future directions
5.1 Workflow for mechanistic studies of novel compounds using next generation
sequencing
Application of the proposed workflow has lead to successful characterization of
the anti-cancer activities of SGI-110 and oxaliplatin combination treatment in
hepatocellular carcinoma, the optimized ROS modulator QD325 in pancreatic cancer, and
the novel N-(8-quinolinyl) nicotinamide QN523 in pancreatic cancer. We employed
various assays to fully characterize the mechanism of action of these novel agents paving
the way for their future developments as innovative therapy for cancer. (Fig. 5-1)

Figure 5-1. Methods used for preclinical evaluation of novel small molecule compounds

2"
Cell$culture$models$
Xenogra1$models$
Next$genera4on$
sequencing$
• MTT$assay$
• Colony$forma4on$assay$
• MTD$study$
• Efficacy$study$on$xenogra1$
• Histology$of$major$organs$$
• BruCseq$
Bioinforma4cs$analysis$
• GSEA$(Gene$set$enrichment$analysis)$
• IPA$(Ingenuity$Pathway$Analysis)$
• CMAP$(Connec4vity$map)$
• NextBio$/$Oncomine$
Valida4on$in$cell$
culture/tumor$samples$
• Western$bloTng$$
• qPCR$
• Gene4c$/$pharmacological$interven4on$of$poten4al$targets$
Methods$for$preclinical$evalua4on$

151

Cell culture models provide a highly scalable, reproducible and efficient platform
to evaluate cytotoxicity and molecular responses of small molecules in representative
genetic and pathologic background. Compounds showing preferable bioactivity in cell
culture models can then be taken into in vivo evaluation for their safety profile and anti-
cancer efficacy. Applications of in vivo models are indispensible in that they well
recapitulate clinical settings and provide valuable information on drug adsorption,
distribution, metabolism, excretion as well as the efficacy and toxicity profiles.
Successful anti-cancer drug candidates should be the ones that show exceptional activity
in these two systems. Lead compounds that fail in these two rounds, either with low anti-
cancer activity, poor cellular uptake, or unwanted toxicity, should be triaged or optimized
before proceeding to further characterization.
Successful candidates can then be profiled to generate transcriptional signatures.
Here in this workflow we have focused on the gene expression profiles, which are then
deciphered with bioinformatics analysis including IPA, GSEA, CMAP and NextBio. In
addition, the abundance of genomic data gives rise to immense discovery opportunities
such as regulation on long non-coding RNAs, which is not discussed in our model
currently, but can be great addition in the future. Application of bioinformatics can
generate informed hypothesis that can then be tested in the validation step.
For validation of the bioinformatics discoveries, qPCR and western blotting are
often employed to evaluate regulated expression of certain marker genes in a panel of
relevant cell lines or tumor samples, so as to confirm the observation from sequencing.
Genetic and pharmacological interventions can be employed to further characterize the
proposed target proteins or mechanisms of action.

152

Our proposed workflow for preclinical evaluation of novel compounds is proved
useful in better elucidating the mechanism of action of a drug. The information we have
gained from this systems-based approach would have not been possible using traditional
molecular pharmacology.

5.2 Limitation of the current workflow for preclinical evaluation
We propose the application of our workflow as a general guideline for preclinical
evaluation of drug candidates. However, there are some limitations of the current method
despite its success in generating informative preclinical results.
We have used adherent cell line models and xenograft models for in vitro and in
vivo studies, for their advantages in good reproducibility, high efficiency, and preferable
affordability. The xenograft model, as well as the 2-D culture model of cancer cell lines,
has limitations in their representativeness in the absence of proper microenvironment. As
a result, data obtained from these models reflects disease response and is informative for
preclinical understanding of the treatment, but might not necessarily translate into the
clinical settings.
For the analysis of transcriptional signature, we have focused on synthesis of
annotated genes with protein products. While this analysis can provide rich information
on pathway regulations and functional responses to our drug candidates, protein-coding
genes only comprise 3% of the human genome, and there are other major scopes of
transcriptional regulations that remain out of the picture. As revealed by the
Encyclopedia of DNA Elements (ENCODE) project, three quarters of the genome is

153

capable of being transcribed (Djebali et al., 2012). Such findings prompt redefinition of
the gene concept, leading to discovery of the profound significance of non-coding RNAs
in biological processes. Once considered background noise of transcription, non-coding
RNAs from all regions of the genome arise as key regulators for gene transcriptions
(Esteller, 2011; Wang et al., 2011; Ling et al., 2013). Although in synthesis levels,
microRNA(miRNA) or long non-coding RNA(lncRNA) might not interfere with gene
transcription directly, changes in synthesis of such RNA species might reflect drug-
induced regulations and exert down-stream effects on their targets. With such
consideration, expanding the bioinformatics analysis pipeline to include non-coding RNA
transcripts would be necessary to understand the global transcriptional regulations
triggered by drug candidates.
Using next generation sequencing, we were able to capture the cellular responses
to our experimental compounds globally, and identify significant biomarkers related to
the treatment. Such studies efficiently facilitate understanding of potential mechanisms of
action. However, basing on transcriptional signature, the current workflow does not
provide direct information on target identification, another major challenge in the drug
discovery process. For the multi-leveled signaling events and regulations, primary and
subsequent changes are mingled in the transcription snapshots and remain
undistinguished. Currently, there is no established method to define order of observed
responses, making it difficult to elucidate the regulatory path triggered by the treatment.
Although we can design case-dependent experiments to reveal passage of regulatory
signals, addition of general methods aiming to trace upstream regulators from
downstream observation will largely benefit the preclinical evaluation process.

154

5.3 Techniques as potential additions to the current workflow
With rapid growth in knowledge of cancer biology and development of new
technology, the field of cancer drug discovery has seen unprecedented evolution in the
past decades. It is our goal to apply the newly developed techniques or knowledge to our
studies in the pursuit of better solution for the treatment of cancer. By incorporating the
best platforms available, our current workflow will further improve as the guideline for
preclinical studies. Here I would like to discuss some potential additions to the current
workflow, which might benefit the target identification and mechanistic characterization
process.
The current screening and in vitro characterization is performed in the traditional
monolayer adherent cell culture model, which is highly convenient and consistent but
might confer non-physiological feature to the cancer cells. Introduction of three
dimensional culture technique, the hanging drop system as an example, has reframed the
application of cell culture models by allowing physiological cell-cell contacts and
diffusion-limited nutrition/compound uptakes. (Hsiao et al., 2012; LaBarbera et al.,
2012) By co-culturing different cell types, cell culture can also recapitulate tumor
environment heterogeneity and tumor-stroma interactions in a screening friendly format.
(Friedrich et al., 2009) Introducing such spheroid culture system into preclinical
evaluation might give rise to more clinically relevant results that can be better translated
into patient care.
There has also been great progress in the murine models representing different
types of cancer that can be added to our studies. The current study uses subcutaneous
xenograft model that is highly efficient and reproducible for test of anti-cancer efficacy.

155

However, lack of proper microenvironment and immune responses remain the drawbacks
of this system. Introducing new murine models could better inform the preclinical
evaluation. Taking pancreatic cancer as an example, Boj, S et al. has recently developed
an organoid model where patient-derived resected tumors and biopsies can be
orthotopically transplanted into the murine host and simulate the stage of pancreatic
ductal adenocarcinoma tumorigenesis. (Boj et al., 2015) Genetically engineered mouse
models have also been developed where the tumorigenesis process and physiological
immune responses can be assayed. (Guerra et al., 2013)
Our workflow focuses on using next generation sequencing for transcriptional
profiling. However, application of next generation sequencing in drug discovery is not
limited to RNA-seq. For compounds that might have genetic targets, DNA-oriented
techniques FAIRE-seq, ChIP-seq and CATCH-IT can reveal chromatin alterations, while
Chem-seq with biotinylated small molecule compound allows isolation of DNA
fragments directly associated with the compound of interest and facilitate identification
of direct target sites. (Anders et al., 2014) High-throughput sequencing was employed to
identify sites of action of small molecules genome-wide, for example, the G4 structures
in gene bodies was discovered by its binding with the DNA-damaging agent pyridostatin
(Rodriguez et al., 2012).
In the selection of sequencing conditions and samples for transcriptome profiling,
more flexibility could be added for different purposes. For example, taking samples at
different time points after identical treatments can capture the sensitive time dependent
changes, making it possible to elucidate drug-induced effect on active transcriptional
modulators as a time-dependent regulatory process. Profiling the same compounds with

156

different cell line models including sensitive, inherent resistant, and acquired-resistant
cell lines might help to identify mechanisms of action as well as potential resistance.
In our effort to analyze transcriptional profiles, lack of agreements in results from
different analytical tools remains an issue for drawing biologically relevant conclusions.
Such problem arises from the large difference in collections of studies from different
databases. There is limitation in every database and its corresponding analytical tool, so
we have performed analysis in a couple of bioinformatics tools so as to extract potential
information from all different aspects. However, when it comes to conclusion, there is no
way in comparing this methods and it is impossible to decide which analytical result is
more important or reliable. The newly published online database PathCards consolidates
3215 human pathways from 12 sources into a set of 1073 SuperPaths, which sought to
balance reduced redundancy and optimized pathway-related information for individual
genes (Belinky et al., 2015). This unified database should provide nearly exhaustive gene
coverage and systematic pathway interrelations without redundancy, and might serve as
next searchable analysis tool for pathway studies. Other newly developed bioinformatics
tools should also be considered to increase the power of our analysis on the
transcriptional signature.
The role of non-coding RNA in regulation of biological processes is increasingly
being recognized. Non-coding RNAs including microRNA and long non-coding RNA
have been associated with cancer and identified as potential targets for anticancer drug
development (Ling et al., 2013). Considering the importance and abundance of non-
coding RNA as a major part of the transcriptome, analysis for changes in non-coding
RNA synthesis should be included in our pipeline. Defined as transcripts longer than 200

157

nucleotides, long non-coding RNA comprises the majority of non-coding RNA
transcripts. The GENCODE consortium within the framework of the ENCODE project
provides annotation for human and mouse non-coding transcripts, serving as a valuable
resource for studies of long non-coding RNA (Derrien et al., 2012). From curated
publications, NONCODE provides another database for integrated long non-coding RNA
annotation with focus on recently identified lncRNA genes. From public RNA-seq data,
NONCODE presents expression profiles of lncRNA genes as well as functional
predictions (Xie et al., 2014). Such annotation tools and lncRNA databases can largely
facilitate lncRNA analysis of RNA-seq data, and decipher this indispensible part of the
regulatory puzzle, giving a more comprehensive understanding of cellular responses to
our candidate compounds.
In our workflow, Bru-seq and bioinformatics analysis identify cellular responses
but do not provide direct information on drug targets, or distinguish order of responses.
Bioinformatics or experimental methods could be incorporated to address these important
scopes, so as to generate a thorough understanding of drug actions in the preclinical
evaluation process. In bioinformatics analysis, the causal network analysis module in IPA
uses literature-based cause-effect association of proteins to predict the signaling stream
reflected in the transcriptional signature (Chindelevitch et al., 2012; Kramer et al., 2014).
Basing on the affected pathways, IPA predicts upstream regulators potentially involved in
the regulation. Such upstream regulators might not necessarily experience detectable
transcriptional changes, but they could act as signaling nodes or even hubs mediating
drug-triggered responses through changes in other regulatory forms. Examples are
phosphorylation modification on receptors tyrosine kinases and nuclear translocation of

158

Nrf2 or PERK in the absence of cytoplasmic sequesters. Such changes play key roles in
signaling regulation, but are not directly detectable with transcriptional analysis.
Prediction for upstream regulators allows us to go beyond the observed transcriptional
changes and trace upstream of the regulatory path, giving rise to identification of major
signaling mediators or even cellular targets. Basing on the bioinformatics-derived
hypothesis and nature of the potential mediators or targets, we can then design specific
experiments to validate roles of such potential key nodes in our drug treatment.
Intervention of the biological system with genetic or pharmacological manipulation will
serve for the mechanistic validation. Chemical labeling of the compound can also add in
great experimental values, where locating the compound in cellular compartments and
identifying direct binding partners becomes possible. The affinity-based biochemical
methods have not been used in the current study, but it will certainly serve as a powerful
addition to the discovery workflow, either as a target validation method or a parallel
characterization approach.
For discovery of novel small molecules as candidates for anti-cancer therapeutics,
every compound is unique. Specific assays need to be designed in order to address the
biological questions in different contexts. Being open to every possibility and able to
integrate available resources might be the key factors to improve the preclinical
evaluation process, and eventually lead to a translational impact for cancer patients.

159

BIBLIOGRAPHY

Abou-Alfa GK, Capanu M, O'Reilly EM, Ma J, Chou JF, Gansukh B, et al. (2014). A
phase II study of cixutumumab (IMC-A12, NSC742460) in advanced hepatocellular
carcinoma. J. Hepatol. 60(2): 319-324.

Al-Lazikani B, Banerji U, Workman P (2012). Combinatorial drug therapy for cancer in
the post-genomic era. Nature biotechnology 30(7): 679-692.

Alam J, Stewart D, Touchard C, Boinapally S, Choi AM, Cook JL (1999). Nrf2, a
Cap'n'Collar transcription factor, regulates induction of the heme oxygenase-1 gene. The
Journal of biological chemistry 274(37): 26071-26078.

Alexeyev M, Shokolenko I, Wilson G, LeDoux S (2013). The maintenance of
mitochondrial DNA integrity--critical analysis and update. Cold Spring Harbor
perspectives in biology 5(5): a012641.

Ali S, Almhanna K, Chen W, Philip PA, Sarkar FH (2010). Differentially expressed
miRNAs in the plasma may provide a molecular signature for aggressive pancreatic
cancer. American journal of translational research 3(1): 28-47.

Anders L, Guenther MG, Qi J, Fan ZP, Marineau JJ, Rahl PB, et al. (2014). Genome-
wide localization of small molecules. Nature biotechnology 32(1): 92-96.

Arlt A, Sebens S, Krebs S, Geismann C, Grossmann M, Kruse ML, et al. (2013).
Inhibition of the Nrf2 transcription factor by the alkaloid trigonelline renders pancreatic
cancer cells more susceptible to apoptosis through decreased proteasomal gene
expression and proteasome activity. Oncogene 32(40): 4825-4835.

Arrowsmith J (2011). Trial watch: Phase II failures: 2008-2010. Nature reviews. Drug
discovery 10(5): 328-329.

Baek SJ, Kim JS, Jackson FR, Eling TE, McEntee MF, Lee SH (2004). Epicatechin
gallate-induced expression of NAG-1 is associated with growth inhibition and apoptosis
in colon cancer cells. Carcinogenesis 25(12): 2425-2432.

Bekaii-Saab T, Markowitz J, Prescott N, Sadee W, Heerema N, Wei L, et al. (2009). A
multi-institutional phase II study of the efficacy and tolerability of lapatinib in patients
with advanced hepatocellular carcinomas. Clin. Cancer Res. 15(18): 5895-5901.

Belinky F, Nativ N, Stelzer G, Zimmerman S, Iny Stein T, Safran M, et al. (2015).
PathCards: multi-source consolidation of human biological pathways. Database : the
journal of biological databases and curation 2015.

160

Bodzin AS, Wei Z, Hurtt R, Gu T, Doria C (2012). Gefitinib resistance in HCC mahlavu
cells: upregulation of CD133 expression, activation of IGF-1R signaling pathway, and
enhancement of IGF-1R nuclear translocation. J. Cell. Physiol. 227(7): 2947-2952.

Boj SF, Hwang CI, Baker LA, Chio, II, Engle DD, Corbo V, et al. (2015). Organoid
models of human and mouse ductal pancreatic cancer. Cell 160(1-2): 324-338.

Bottone FG, Jr., Martinez JM, Collins JB, Afshari CA, Eling TE (2003). Gene
modulation by the cyclooxygenase inhibitor, sulindac sulfide, in human colorectal
carcinoma cells: possible link to apoptosis. The Journal of biological chemistry 278(28):
25790-25801.

Buckley AF, Burgart LJ, Sahai V, Kakar S (2008). Epidermal growth factor receptor
expression and gene copy number in conventional hepatocellular carcinoma. Am. J. Clin.
Pathol. 129(2): 245-251.

Buehler E, Khan AA, Marine S, Rajaram M, Bahl A, Burchard J, et al. (2012). siRNA
off-target effects in genome-wide screens identify signaling pathway members. Scientific
reports 2: 428.

Chang Z, Li Z, Wang X, Kang Y, Yuan Y, Niu J, et al. (2013). Deciphering the
mechanisms of tumorigenesis in human pancreatic ductal epithelial cells. Clinical cancer
research : an official journal of the American Association for Cancer Research 19(3):
549-559.

Cheung CH, Huang CC, Tsai FY, Lee JY, Cheng SM, Chang YC, et al. (2013). Survivin
- biology and potential as a therapeutic target in oncology. Onco Targets Ther. 6: 1453-
1462.

Chindelevitch L, Ziemek D, Enayetallah A, Randhawa R, Sidders B, Brockel C, et al.
(2012). Causal reasoning on biological networks: interpreting transcriptional changes.
Bioinformatics 28(8): 1114-1121.

Choi AM, Alam J (1996). Heme oxygenase-1: function, regulation, and implication of a
novel stress-inducible protein in oxidant-induced lung injury. American journal of
respiratory cell and molecular biology 15(1): 9-19.

Chou TC (2006). Theoretical basis, experimental design, and computerized simulation of
synergism and antagonism in drug combination studies. Pharmacol. Rev. 58(3): 621-681.

Chuang JC, Warner SL, Vollmer D, Vankayalapati H, Redkar S, Bearss DJ, et al. (2010).
S110, a 5-Aza-2'-deoxycytidine-containing dinucleotide, is an effective DNA
methylation inhibitor in vivo and can reduce tumor growth. Mol. Cancer Ther. 9(5):
1443-1450.

161

CJ OC, Beckmann HS, Spring DR (2012). Diversity-oriented synthesis: producing
chemical tools for dissecting biology. Chemical Society reviews 41(12): 4444-4456.

Cohen SJ, Zalupski MM, Modiano MR, Conkling P, Patt YZ, Davis P, et al. (2010). A
phase I study of imexon plus gemcitabine as first-line therapy for advanced pancreatic
cancer. Cancer chemotherapy and pharmacology 66(2): 287-294.

Conroy T, Desseigne F, Ychou M, Bouche O, Guimbaud R, Becouarn Y, et al. (2011).
FOLFIRINOX versus gemcitabine for metastatic pancreatic cancer. The New England
journal of medicine 364(19): 1817-1825.

Dawson JC, Carragher NO (2014). Quantitative phenotypic and pathway profiling guides
rational drug combination strategies. Frontiers in pharmacology 5: 118.

DeNicola GM, Karreth FA, Humpton TJ, Gopinathan A, Wei C, Frese K, et al. (2011).
Oncogene-induced Nrf2 transcription promotes ROS detoxification and tumorigenesis.
Nature 475(7354): 106-109.

Deretic V, Saitoh T, Akira S (2013). Autophagy in infection, inflammation and
immunity. Nature reviews. Immunology 13(10): 722-737.

Derrien T, Johnson R, Bussotti G, Tanzer A, Djebali S, Tilgner H, et al. (2012). The
GENCODE v7 catalog of human long noncoding RNAs: analysis of their gene structure,
evolution, and expression. Genome research 22(9): 1775-1789.

Desbois-Mouthon C, Baron A, Blivet-Van Eggelpoel MJ, Fartoux L, Venot C, Bladt F, et
al. (2009). Insulin-like growth factor-1 receptor inhibition induces a resistance
mechanism via the epidermal growth factor receptor/HER3/AKT signaling pathway:
rational basis for cotargeting insulin-like growth factor-1 receptor and epidermal growth
factor receptor in hepatocellular carcinoma. Clin. Cancer Res. 15(17): 5445-5456.

Dhillon H, Chikara S, Reindl KM (2014). Piperlongumine induces pancreatic cancer cell
death by enhancing reactive oxygen species and DNA damage. Toxicology reports 1:
309-318.

Din FV, Valanciute A, Houde VP, Zibrova D, Green KA, Sakamoto K, et al. (2012).
Aspirin inhibits mTOR signaling, activates AMP-activated protein kinase, and induces
autophagy in colorectal cancer cells. Gastroenterology 142(7): 1504-1515 e1503.

Dinkova-Kostova AT, Holtzclaw WD, Cole RN, Itoh K, Wakabayashi N, Katoh Y, et al.
(2002). Direct evidence that sulfhydryl groups of Keap1 are the sensors regulating
induction of phase 2 enzymes that protect against carcinogens and oxidants. Proceedings
of the National Academy of Sciences of the United States of America 99(18): 11908-
11913.

162

Dinkova-Kostova AT, Talalay P (2000). Persuasive evidence that quinone reductase type
1 (DT diaphorase) protects cells against the toxicity of electrophiles and reactive forms of
oxygen. Free radical biology & medicine 29(3-4): 231-240.

Djebali S, Davis CA, Merkel A, Dobin A, Lassmann T, Mortazavi A, et al. (2012).
Landscape of transcription in human cells. Nature 489(7414): 101-108.

Dombroski BA, Nayak RR, Ewens KG, Ankener W, Cheung VG, Spielman RS (2010).
Gene expression and genetic variation in response to endoplasmic reticulum stress in
human cells. American journal of human genetics 86(5): 719-729.

Dorr RT, Raymond MA, Landowski TH, Roman NO, Fukushima S (2005). Induction of
apoptosis and cell cycle arrest by imexon in human pancreatic cancer cell lines.
International journal of gastrointestinal cancer 36(1): 15-28.

Dutertre M, Gratadou L, Dardenne E, Germann S, Samaan S, Lidereau R, et al. (2010).
Estrogen regulation and physiopathologic significance of alternative promoters in breast
cancer. Cancer research 70(9): 3760-3770.

Eder J, Sedrani R, Wiesmann C (2014). The discovery of first-in-class drugs: origins and
evolution. Nature reviews. Drug discovery 13(8): 577-587.

Esteller M (2011). Non-coding RNAs in human disease. Nature reviews. Genetics
12(12): 861-874.

Falkenberg M, Larsson NG, Gustafsson CM (2007). DNA replication and transcription in
mammalian mitochondria. Annual review of biochemistry 76: 679-699.

Fan H, Lu X, Wang X, Liu Y, Guo B, Zhang Y, et al. (2014). Low-dose decitabine-based
chemoimmunotherapy for patients with refractory advanced solid tumors: a phase I/II
report. J Immunol Res 2014: 371087.

Fan H, Zhao ZJ, Cheng J, Su XW, Wu QX, Shan YF (2009). Overexpression of DNA
methyltransferase 1 and its biological significance in primary hepatocellular carcinoma.
World J. Gastroenterol. 15(16): 2020-2026.

Fang F, Munck J, Tang J, Taverna P, Wang Y, Miller DF, et al. (2014). The novel, small-
molecule DNA methylation inhibitor SGI-110 as an ovarian cancer chemosensitizer.
Clin. Cancer Res. 20(24): 6504-6516.

Ferlay J SI, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D,
Bray F. (2013). GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide:
IARC CancerBase No.11; Lyon, France: International Agency for Research on Cancer;
Available from: http://globocan.iarc.fr.

163

Forner A, Llovet JM, Bruix J (2012). Hepatocellular carcinoma. Lancet 379(9822): 1245-
1255.

Friedrich J, Seidel C, Ebner R, Kunz-Schughart LA (2009). Spheroid-based drug screen:
considerations and practical approach. Nature protocols 4(3): 309-324.

Fruehauf JP, Meyskens FL, Jr. (2007). Reactive oxygen species: a breath of life or death?
Clinical cancer research : an official journal of the American Association for Cancer
Research 13(3): 789-794.

Gaur S, Chen L, Ann V, Lin WC, Wang Y, Chang VH, et al. (2014). Dovitinib
synergizes with oxaliplatin in suppressing cell proliferation and inducing apoptosis in
colorectal cancer cells regardless of RAS-RAF mutation status. Mol. Cancer 13: 21.

Gedaly R, Galuppo R, Daily MF, Shah M, Maynard E, Chen C, et al. (2014). Targeting
the Wnt/beta-Catenin Signaling Pathway in Liver Cancer Stem Cells and Hepatocellular
Carcinoma Cell Lines with FH535. PLoS One 9(6): e99272.

Glasspool RM, Brown R, Gore ME, Rustin GJ, McNeish IA, Wilson RH, et al. (2014). A
randomised, phase II trial of the DNA-hypomethylating agent 5-aza-2'-deoxycytidine
(decitabine) in combination with carboplatin vs carboplatin alone in patients with
recurrent, partially platinum-sensitive ovarian cancer. Br. J. Cancer 110(8): 1923-1929.

Graham JS, Kaye SB, Brown R (2009). The promises and pitfalls of epigenetic therapies
in solid tumours. Eur. J. Cancer 45(7): 1129-1136.

Guerra C, Barbacid M (2013). Genetically engineered mouse models of pancreatic
adenocarcinoma. Molecular oncology 7(2): 232-247.

Guleng G, Lovig T, Meling GI, Andersen SN, Rognum TO (2005). Mitochondrial
microsatellite instability in colorectal carcinomas--frequency and association with nuclear
microsatellite instability. Cancer letters 219(1): 97-103.

Hai T, Hartman MG (2001). The molecular biology and nomenclature of the activating
transcription factor/cAMP responsive element binding family of transcription factors:
activating transcription factor proteins and homeostasis. Gene 273(1): 1-11.

Halliwell B, Aruoma OI (1991). DNA damage by oxygen-derived species. Its mechanism
and measurement in mammalian systems. FEBS letters 281(1-2): 9-19.

Harding HP, Zhang Y, Bertolotti A, Zeng H, Ron D (2000). Perk is essential for
translational regulation and cell survival during the unfolded protein response. Molecular
cell 5(5): 897-904.

164

Hasan MN, Queudeville M, Trentin L, Eckhoff SM, Bronzini I, Palmi C, et al. (2015).
Targeting of hyperactivated mTOR signaling in high-risk acute lymphoblastic leukemia
in a pre-clinical model. Oncotarget 6(3): 1382-1395.

Hayes JD, McMahon M (2009). NRF2 and KEAP1 mutations: permanent activation of an
adaptive response in cancer. Trends in biochemical sciences 34(4): 176-188.

Haze K, Yoshida H, Yanagi H, Yura T, Mori K (1999). Mammalian transcription factor
ATF6 is synthesized as a transmembrane protein and activated by proteolysis in response
to endoplasmic reticulum stress. Molecular biology of the cell 10(11): 3787-3799.

Heck S, Qian X, Velleca M (2004). Genetically engineered mouse models for drug
discovery: new chemical genetic approaches. Current drug discovery technologies 1(1):
13-26.

Hetz C (2012). The unfolded protein response: controlling cell fate decisions under ER
stress and beyond. Nature reviews. Molecular cell biology 13(2): 89-102.

Hieronymus H, Lamb J, Ross KN, Peng XP, Clement C, Rodina A, et al. (2006). Gene
expression signature-based chemical genomic prediction identifies a novel class of
HSP90 pathway modulators. Cancer cell 10(4): 321-330.

Hirota T, Lee JW, St John PC, Sawa M, Iwaisako K, Noguchi T, et al. (2012).
Identification of small molecule activators of cryptochrome. Science 337(6098): 1094-
1097.

Hopfner M, Huether A, Sutter AP, Baradari V, Schuppan D, Scherubl H (2006).
Blockade of IGF-1 receptor tyrosine kinase has antineoplastic effects in hepatocellular
carcinoma cells. Biochem. Pharmacol. 71(10): 1435-1448.

Hsiao AY, Tung YC, Kuo CH, Mosadegh B, Bedenis R, Pienta KJ, et al. (2012). Micro-
ring structures stabilize microdroplets to enable long term spheroid culture in 384
hanging drop array plates. Biomedical microdevices 14(2): 313-323.

Hsu PD, Lander ES, Zhang F (2014). Development and applications of CRISPR-Cas9 for
genome engineering. Cell 157(6): 1262-1278.

Huang da W, Sherman BT, Lempicki RA (2009a). Bioinformatics enrichment tools:
paths toward the comprehensive functional analysis of large gene lists. Nucleic acids
research 37(1): 1-13.

Huang da W, Sherman BT, Lempicki RA (2009b). Systematic and integrative analysis of
large gene lists using DAVID bioinformatics resources. Nature protocols 4(1): 44-57.

165

Huang S, Sinicrope FA (2010). Celecoxib-induced apoptosis is enhanced by ABT-737
and by inhibition of autophagy in human colorectal cancer cells. Autophagy 6(2): 256-
269.

Huang WY, Hsu SD, Huang HY, Sun YM, Chou CH, Weng SL, et al. (2015). MethHC: a
database of DNA methylation and gene expression in human cancer. Nucleic Acids Res.
43(Database issue): D856-861.

Hughes TR, Marton MJ, Jones AR, Roberts CJ, Stoughton R, Armour CD, et al. (2000).
Functional discovery via a compendium of expression profiles. Cell 102(1): 109-126.

Hung WY, Wu CW, Yin PH, Chang CJ, Li AF, Chi CW, et al. (2010). Somatic mutations
in mitochondrial genome and their potential roles in the progression of human gastric
cancer. Biochimica et biophysica acta 1800(3): 264-270.

Hutchinson L, Kirk R (2011). High drug attrition rates--where are we going wrong?
Nature reviews. Clinical oncology 8(4): 189-190.

Itoh K, Wakabayashi N, Katoh Y, Ishii T, Igarashi K, Engel JD, et al. (1999). Keap1
represses nuclear activation of antioxidant responsive elements by Nrf2 through binding
to the amino-terminal Neh2 domain. Genes & development 13(1): 76-86.

Jaiswal AK (2004). Nrf2 signaling in coordinated activation of antioxidant gene
expression. Free radical biology & medicine 36(10): 1199-1207.

Jones PA, Taylor SM (1980). Cellular differentiation, cytidine analogs and DNA
methylation. Cell 20(1): 85-93.

Joseph SB, McKilligin E, Pei L, Watson MA, Collins AR, Laffitte BA, et al. (2002).
Synthetic LXR ligand inhibits the development of atherosclerosis in mice. Proceedings of
the National Academy of Sciences of the United States of America 99(11): 7604-7609.

Kabekkodu SP, Bhat S, Mascarenhas R, Mallya S, Bhat M, Pandey D, et al. (2014).
Mitochondrial DNA variation analysis in cervical cancer. Mitochondrion 16: 73-82.

Kaida D, Motoyoshi H, Tashiro E, Nojima T, Hagiwara M, Ishigami K, et al. (2007).
Spliceostatin A targets SF3b and inhibits both splicing and nuclear retention of pre-
mRNA. Nature chemical biology 3(9): 576-583.

Kaileh M, Vanden Berghe W, Heyerick A, Horion J, Piette J, Libert C, et al. (2007).
Withaferin a strongly elicits IkappaB kinase beta hyperphosphorylation concomitant with
potent inhibition of its kinase activity. The Journal of biological chemistry 282(7): 4253-
4264.

Kauselmann G, Dopazo A, Link W (2012). Identification of disease-relevant genes for
molecularly-targeted drug discovery. Current cancer drug targets 12(1): 1-13.

166

Kelly RJ, Lopez-Chavez A, Citrin D, Janik JE, Morris JC (2011). Impacting tumor cell-
fate by targeting the inhibitor of apoptosis protein survivin. Mol. Cancer 10: 35.

Kim JM, Kosak JP, Kim JK, Kissling G, Germolec DR, Zeldin DC, et al. (2013). NAG-
1/GDF15 transgenic mouse has less white adipose tissue and a reduced inflammatory
response. Mediators of inflammation 2013: 641851.

Kim R, Emi M, Tanabe K, Murakami S (2006). Role of the unfolded protein response in
cell death. Apoptosis : an international journal on programmed cell death 11(1): 5-13.

Kong B, Qia C, Erkan M, Kleeff J, Michalski CW (2013). Overview on how oncogenic
Kras promotes pancreatic carcinogenesis by inducing low intracellular ROS levels.
Frontiers in physiology 4: 246.

Kramer A, Green J, Pollard J, Jr., Tugendreich S (2014). Causal analysis approaches in
Ingenuity Pathway Analysis. Bioinformatics 30(4): 523-530.

Kroemer G, Marino G, Levine B (2010). Autophagy and the integrated stress response.
Molecular cell 40(2): 280-293.

Kupershmidt I, Su QJ, Grewal A, Sundaresh S, Halperin I, Flynn J, et al. (2010).
Ontology-based meta-analysis of global collections of high-throughput public data. PloS
one 5(9).

LaBarbera DV, Reid BG, Yoo BH (2012). The multicellular tumor spheroid model for
high-throughput cancer drug discovery. Expert opinion on drug discovery 7(9): 819-830.

Labbe E, Lock L, Letamendia A, Gorska AE, Gryfe R, Gallinger S, et al. (2007).
Transcriptional cooperation between the transforming growth factor-beta and Wnt
pathways in mammary and intestinal tumorigenesis. Cancer Res. 67(1): 75-84.

Lachenmayer A, Alsinet C, Savic R, Cabellos L, Toffanin S, Hoshida Y, et al. (2012).
Wnt-pathway activation in two molecular classes of hepatocellular carcinoma and
experimental modulation by sorafenib. Clin. Cancer Res. 18(18): 4997-5007.

Lamb J (2007). The Connectivity Map: a new tool for biomedical research. Nature
reviews. Cancer 7(1): 54-60.

Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, et al. (2006). The
Connectivity Map: using gene-expression signatures to connect small molecules, genes,
and disease. Science 313(5795): 1929-1935.

Lee SH, Kim JS, Yamaguchi K, Eling TE, Baek SJ (2005). Indole-3-carbinol and 3,3'-
diindolylmethane induce expression of NAG-1 in a p53-independent manner.
Biochemical and biophysical research communications 328(1): 63-69.

167

Lee SH, Krisanapun C, Baek SJ (2010). NSAID-activated gene-1 as a molecular target
for capsaicin-induced apoptosis through a novel molecular mechanism involving
GSK3beta, C/EBPbeta and ATF3. Carcinogenesis 31(4): 719-728.

Lee SH, Yamaguchi K, Kim JS, Eling TE, Safe S, Park Y, et al. (2006). Conjugated
linoleic acid stimulates an anti-tumorigenic protein NAG-1 in an isomer specific manner.
Carcinogenesis 27(5): 972-981.

Lehmann S, Bykov VJ, Ali D, Andren O, Cherif H, Tidefelt U, et al. (2012). Targeting
p53 in vivo: a first-in-human study with p53-targeting compound APR-246 in refractory
hematologic malignancies and prostate cancer. Journal of clinical oncology : official
journal of the American Society of Clinical Oncology 30(29): 3633-3639.

Leik CE, Carson NL, Hennan JK, Basso MD, Liu QY, Crandall DL, et al. (2007).
GW3965, a synthetic liver X receptor (LXR) agonist, reduces angiotensin II-mediated
pressor responses in Sprague-Dawley rats. British journal of pharmacology 151(4): 450-
456.

Li PX, Wong J, Ayed A, Ngo D, Brade AM, Arrowsmith C, et al. (2000). Placental
transforming growth factor-beta is a downstream mediator of the growth arrest and
apoptotic response of tumor cells to DNA damage and p53 overexpression. The Journal
of biological chemistry 275(26): 20127-20135.

Li X, Fang P, Mai J, Choi ET, Wang H, Yang XF (2013). Targeting mitochondrial
reactive oxygen species as novel therapy for inflammatory diseases and cancers. Journal
of hematology & oncology 6: 19.

Lievre A, Chapusot C, Bouvier AM, Zinzindohoue F, Piard F, Roignot P, et al. (2005).
Clinical value of mitochondrial mutations in colorectal cancer. Journal of clinical
oncology : official journal of the American Society of Clinical Oncology 23(15): 3517-
3525.

Ling H, Fabbri M, Calin GA (2013). MicroRNAs and other non-coding RNAs as targets
for anticancer drug development. Nature reviews. Drug discovery 12(11): 847-865.

Lister A, Nedjadi T, Kitteringham NR, Campbell F, Costello E, Lloyd B, et al. (2011).
Nrf2 is overexpressed in pancreatic cancer: implications for cell proliferation and
therapy. Molecular cancer 10: 37.

Liu T, Bauskin AR, Zaunders J, Brown DA, Pankhurst S, Russell PJ, et al. (2003).
Macrophage inhibitory cytokine 1 reduces cell adhesion and induces apoptosis in prostate
cancer cells. Cancer research 63(16): 5034-5040.

168

Liu W, Ahmad SA, Jung YD, Reinmuth N, Fan F, Bucana CD, et al. (2002).
Coexpression of ephrin-Bs and their receptors in colon carcinoma. Cancer 94(4): 934-
939.

Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. (2008). Sorafenib
in advanced hepatocellular carcinoma. N. Engl. J. Med. 359(4): 378-390.

Lomenick B, Hao R, Jonai N, Chin RM, Aghajan M, Warburton S, et al. (2009). Target
identification using drug affinity responsive target stability (DARTS). Proceedings of the
National Academy of Sciences of the United States of America 106(51): 21984-21989.

Louafi S, Boige V, Ducreux M, Bonyhay L, Mansourbakht T, de Baere T, et al. (2007).
Gemcitabine plus oxaliplatin (GEMOX) in patients with advanced hepatocellular
carcinoma (HCC): results of a phase II study. Cancer 109(7): 1384-1390.

Malhotra JD, Kaufman RJ (2007). Endoplasmic reticulum stress and oxidative stress: a
vicious cycle or a double-edged sword? Antioxidants & redox signaling 9(12): 2277-
2293.

Marine S, Bahl A, Ferrer M, Buehler E (2012). Common seed analysis to identify off-
target effects in siRNA screens. Journal of biomolecular screening 17(3): 370-378.

Matei D, Fang F, Shen C, Schilder J, Arnold A, Zeng Y, et al. (2012). Epigenetic
resensitization to platinum in ovarian cancer. Cancer Res. 72(9): 2197-2205.

Mayr LM, Bojanic D (2009). Novel trends in high-throughput screening. Current opinion
in pharmacology 9(5): 580-588.

Missiaglia E, Donadelli M, Palmieri M, Crnogorac-Jurcevic T, Scarpa A, Lemoine NR
(2005). Growth delay of human pancreatic cancer cells by methylase inhibitor 5-aza-2'-
deoxycytidine treatment is associated with activation of the interferon signalling
pathway. Oncogene 24(1): 199-211.

Moffat JG, Rudolph J, Bailey D (2014). Phenotypic screening in cancer drug discovery -
past, present and future. Nature reviews. Drug discovery 13(8): 588-602.

Mohan R, Hammers HJ, Bargagna-Mohan P, Zhan XH, Herbstritt CJ, Ruiz A, et al.
(2004). Withaferin A is a potent inhibitor of angiogenesis. Angiogenesis 7(2): 115-122.

Monks A, Harris E, Hose C, Connelly J, Sausville EA (2003). Genotoxic profiling of
MCF-7 breast cancer cell line elucidates gene expression modifications underlying
toxicity of the anticancer drug 2-(4-amino-3-methylphenyl)-5-fluorobenzothiazole.
Molecular pharmacology 63(3): 766-772.

169

Montoya J, Gaines GL, Attardi G (1983). The pattern of transcription of the human
mitochondrial rRNA genes reveals two overlapping transcription units. Cell 34(1): 151-
159.

Moon EJ, Giaccia A (2014). Dual roles of NRF2 in tumor prevention and progression:
Possible implications in cancer treatment. Free radical biology & medicine.

Moore JD (2015). The impact of CRISPR-Cas9 on target identification and validation.
Drug discovery today.

Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, Lehar J, et al. (2003).
PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately
downregulated in human diabetes. Nat. Genet. 34(3): 267-273.

Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B (2008). Mapping and
quantifying mammalian transcriptomes by RNA-Seq. Nature methods 5(7): 621-628.

Na HK, Surh YJ (2014). Oncogenic potential of Nrf2 and its principal target protein
heme oxygenase-1. Free radical biology & medicine 67: 353-365.

Nakagawa H, Wardell CP, Furuta M, Taniguchi H, Fujimoto A (2015). Cancer whole-
genome sequencing: present and future. Oncogene.

Neggers JE, Vercruysse T, Jacquemyn M, Vanstreels E, Baloglu E, Shacham S, et al.
(2015). Identifying drug-target selectivity of small-molecule CRM1/XPO1 inhibitors by
CRISPR/Cas9 genome editing. Chemistry & biology 22(1): 107-116.

Neumann O, Kesselmeier M, Geffers R, Pellegrino R, Radlwimmer B, Hoffmann K, et
al. (2012). Methylome analysis and integrative profiling of human HCCs identify novel
protumorigenic factors. Hepatology 56(5): 1817-1827.

Nielsen TE, Schreiber SL (2008). Towards the optimal screening collection: a synthesis
strategy. Angewandte Chemie 47(1): 48-56.

Nioi P, McMahon M, Itoh K, Yamamoto M, Hayes JD (2003). Identification of a novel
Nrf2-regulated antioxidant response element (ARE) in the mouse NAD(P)H:quinone
oxidoreductase 1 gene: reassessment of the ARE consensus sequence. The Biochemical
journal 374(Pt 2): 337-348.

Nishida N, Kudo M, Nagasaka T, Ikai I, Goel A (2012). Characteristic patterns of altered
DNA methylation predict emergence of human hepatocellular carcinoma. Hepatology
56(3): 994-1003.

Nishitoh H (2012). CHOP is a multifunctional transcription factor in the ER stress
response. Journal of biochemistry 151(3): 217-219.

170

Noren NK, Foos G, Hauser CA, Pasquale EB (2006). The EphB4 receptor suppresses
breast cancer cell tumorigenicity through an Abl-Crk pathway. Nat. Cell Biol. 8(8): 815-
825.

O'Dwyer PJLD; Kauh JSF, D.B.; Benson, A.B. (2006). III Gefitinib in advanced
unresectable hepatocellular carcinoma: results from the Eastern Cooperative Oncology
Group's Study E1203. J. Clin. Oncol. 24 Abstract 4143.

Ogata M, Hino S, Saito A, Morikawa K, Kondo S, Kanemoto S, et al. (2006). Autophagy
is activated for cell survival after endoplasmic reticulum stress. Molecular and cellular
biology 26(24): 9220-9231.

Ojala D, Montoya J, Attardi G (1981). tRNA punctuation model of RNA processing in
human mitochondria. Nature 290(5806): 470-474.

Ouyang H, Mou L, Luk C, Liu N, Karaskova J, Squire J, et al. (2000). Immortal human
pancreatic duct epithelial cell lines with near normal genotype and phenotype. The
American journal of pathology 157(5): 1623-1631.

Oyadomari S, Mori M (2004). Roles of CHOP/GADD153 in endoplasmic reticulum
stress. Cell death and differentiation 11(4): 381-389.

Park JG, Lee JH, Kang MS, Park KJ, Jeon YM, Lee HJ, et al. (1995). Characterization of
cell lines established from human hepatocellular carcinoma. Int. J. Cancer 62(3): 276-
282.

Pathania D, Kuang Y, Sechi M, Neamati N (2015). Mechanisms underlying the
cytotoxicity of a novel quinazolinedione-based redox modulator, QD232, in pancreatic
cancer cells. British journal of pharmacology 172(1): 50-63.

Pathania D, Sechi M, Palomba M, Sanna V, Berrettini F, Sias A, et al. (2014). Design
and discovery of novel quinazolinedione-based redox modulators as therapies for
pancreatic cancer. Biochimica et biophysica acta 1840(1): 332-343.

Paul SM, Mytelka DS, Dunwiddie CT, Persinger CC, Munos BH, Lindborg SR, et al.
(2010). How to improve R&D productivity: the pharmaceutical industry's grand
challenge. Nature reviews. Drug discovery 9(3): 203-214.

Paulsen MT, Veloso A, Prasad J, Bedi K, Ljungman EA, Magnuson B, et al. (2014). Use
of Bru-Seq and BruChase-Seq for genome-wide assessment of the synthesis and stability
of RNA. Methods 67(1): 45-54.

Paulsen MT, Veloso A, Prasad J, Bedi K, Ljungman EA, Tsan YC, et al. (2013).
Coordinated regulation of synthesis and stability of RNA during the acute TNF-induced
proinflammatory response. Proc. Natl. Acad. Sci. U. S. A. 110(6): 2240-2245.

171

Pelicano H, Carney D, Huang P (2004). ROS stress in cancer cells and therapeutic
implications. Drug resistance updates : reviews and commentaries in antimicrobial and
anticancer chemotherapy 7(2): 97-110.

Prinz F, Schlange T, Asadullah K (2011). Believe it or not: how much can we rely on
published data on potential drug targets? Nature reviews. Drug discovery 10(9): 712.

Psyrri A, Arkadopoulos N, Vassilakopoulou M, Smyrniotis V, Dimitriadis G (2012).
Pathways and targets in hepatocellular carcinoma. Expert Rev. Anticancer Ther. 12(10):
1347-1357.

Qin S, Bai Y, Lim HY, Thongprasert S, Chao Y, Fan J, et al. (2013). Randomized,
multicenter, open-label study of oxaliplatin plus fluorouracil/leucovorin versus
doxorubicin as palliative chemotherapy in patients with advanced hepatocellular
carcinoma from Asia. J. Clin. Oncol. 31(28): 3501-3508.

Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, et al. (2004).
ONCOMINE: a cancer microarray database and integrated data-mining platform.
Neoplasia 6(1): 1-6.

Rodriguez R, Miller KM, Forment JV, Bradshaw CR, Nikan M, Britton S, et al. (2012).
Small-molecule-induced DNA damage identifies alternative DNA structures in human
genes. Nature chemical biology 8(3): 301-310.

Roessler S, Jia HL, Budhu A, Forgues M, Ye QH, Lee JS, et al. (2010). A unique
metastasis gene signature enables prediction of tumor relapse in early-stage
hepatocellular carcinoma patients. Cancer Res. 70(24): 10202-10212.

Ross D, Kepa JK, Winski SL, Beall HD, Anwar A, Siegel D (2000). NAD(P)H:quinone
oxidoreductase 1 (NQO1): chemoprotection, bioactivation, gene regulation and genetic
polymorphisms. Chemico-biological interactions 129(1-2): 77-97.

Ryan DP, Hong TS, Bardeesy N (2014). Pancreatic adenocarcinoma. The New England
journal of medicine 371(11): 1039-1049.

Sabharwal SS, Schumacker PT (2014). Mitochondrial ROS in cancer: initiators,
amplifiers or an Achilles' heel? Nature reviews. Cancer 14(11): 709-721.

Sachse C, Krausz E, Kronke A, Hannus M, Walsh A, Grabner A, et al. (2005). High-
throughput RNA interference strategies for target discovery and validation by using
synthetic short interfering RNAs: functional genomics investigations of biological
pathways. Methods in enzymology 392: 242-277.

Sakaki K, Wu J, Kaufman RJ (2008). Protein kinase Ctheta is required for autophagy in
response to stress in the endoplasmic reticulum. The Journal of biological chemistry
283(22): 15370-15380.

172

Salazar JJ, Van Houten B (1997). Preferential mitochondrial DNA injury caused by
glucose oxidase as a steady generator of hydrogen peroxide in human fibroblasts.
Mutation research 385(2): 139-149.

Satoh S, Daigo Y, Furukawa Y, Kato T, Miwa N, Nishiwaki T, et al. (2000). AXIN1
mutations in hepatocellular carcinomas, and growth suppression in cancer cells by virus-
mediated transfer of AXIN1. Nat. Genet. 24(3): 245-250.

Schenone M, Dancik V, Wagner BK, Clemons PA (2013). Target identification and
mechanism of action in chemical biology and drug discovery. Nature chemical biology
9(4): 232-240.

Selvakumaran M, Amaravadi RK, Vasilevskaya IA, O'Dwyer PJ (2013). Autophagy
inhibition sensitizes colon cancer cells to antiangiogenic and cytotoxic therapy. Clin.
Cancer Res. 19(11): 2995-3007.

Shamu CE, Walter P (1996). Oligomerization and phosphorylation of the Ire1p kinase
during intracellular signaling from the endoplasmic reticulum to the nucleus. The EMBO
journal 15(12): 3028-3039.

Shim M, Eling TE (2005). Protein kinase C-dependent regulation of NAG-1/placental
bone morphogenic protein/MIC-1 expression in LNCaP prostate carcinoma cells. The
Journal of biological chemistry 280(19): 18636-18642.

Shitani M, Sasaki S, Akutsu N, Takagi H, Suzuki H, Nojima M, et al. (2012). Genome-
wide analysis of DNA methylation identifies novel cancer-related genes in hepatocellular
carcinoma. Tumour Biol. 33(5): 1307-1317.

Shokolenko IN, Wilson GL, Alexeyev MF (2013). Persistent damage induces
mitochondrial DNA degradation. DNA repair 12(7): 488-499.

Siegel R, Ma J, Zou Z, Jemal A (2014). Cancer statistics, 2014. CA Cancer J. Clin. 64(1):
9-29.

Song J, Teplova M, Ishibe-Murakami S, Patel DJ (2012). Structure-based mechanistic
insights into DNMT1-mediated maintenance DNA methylation. Science 335(6069): 709-
712.

Srivastava P, Paluch BE, Matsuzaki J, James SR, Collamat-Lai G, Karbach J, et al.
(2014). Immunomodulatory action of SGI-110, a hypomethylating agent, in acute
myeloid leukemia cells and xenografts. Leuk. Res. 38(11): 1332-1341.

Srivastava P, Paluch BE, Matsuzaki J, James SR, Collamat-Lai G, Taverna P, et al.
(2015). Immunomodulatory action of the DNA methyltransferase inhibitor SGI-110 in

173

epithelial ovarian cancer cells and xenografts. Epigenetics : official journal of the DNA
Methylation Society 10(3): 237-246.

Stratton MR, Campbell PJ, Futreal PA (2009). The cancer genome. Nature 458(7239):
719-724.

Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al.
(2005). Gene set enrichment analysis: a knowledge-based approach for interpreting
genome-wide expression profiles. Proc. Natl. Acad. Sci. U. S. A. 102(43): 15545-15550.

Suh SI, Pyun HY, Cho JW, Baek WK, Park JB, Kwon T, et al. (2000). 5-Aza-2'-
deoxycytidine leads to down-regulation of aberrant p16INK4A RNA transcripts and
restores the functional retinoblastoma protein pathway in hepatocellular carcinoma cell
lines. Cancer Lett. 160(1): 81-88.

Sundberg SA (2000). High-throughput and ultra-high-throughput screening: solution- and
cell-based approaches. Current opinion in biotechnology 11(1): 47-53.

Suzuki M, Toyooka S, Miyajima K, Iizasa T, Fujisawa T, Bekele NB, et al. (2003).
Alterations in the mitochondrial displacement loop in lung cancers. Clinical cancer
research : an official journal of the American Association for Cancer Research 9(15):
5636-5641.

Swinney DC, Anthony J (2011). How were new medicines discovered? Nature reviews.
Drug discovery 10(7): 507-519.

Tan M, Wang Y, Guan K, Sun Y (2000). PTGF-beta, a type beta transforming growth
factor (TGF-beta) superfamily member, is a p53 target gene that inhibits tumor cell
growth via TGF-beta signaling pathway. Proceedings of the National Academy of
Sciences of the United States of America 97(1): 109-114.

Tang XX, Robinson ME, Riceberg JS, Kim DY, Kung B, Titus TB, et al. (2004).
Favorable neuroblastoma genes and molecular therapeutics of neuroblastoma. Clin.
Cancer Res. 10(17): 5837-5844.

Tao YF, Lu J, Du XJ, Sun LC, Zhao X, Peng L, et al. (2012). Survivin selective inhibitor
YM155 induce apoptosis in SK-NEP-1 Wilms tumor cells. BMC cancer 12: 619.

Tellez CS, Grimes MJ, Picchi MA, Liu Y, March TH, Reed MD, et al. (2014). SGI-110
and entinostat therapy reduces lung tumor burden and reprograms the epigenome. Int. J.
Cancer.

Titov DV, Liu JO (2012). Identification and validation of protein targets of bioactive
small molecules. Bioorganic & medicinal chemistry 20(6): 1902-1909.

174

Tovar V, Alsinet C, Villanueva A, Hoshida Y, Chiang DY, Sole M, et al. (2010). IGF
activation in a molecular subclass of hepatocellular carcinoma and pre-clinical efficacy of
IGF-1R blockage. J. Hepatol. 52(4): 550-559.

Tsuyuki S, Takabayashi M, Kawazu M, Kudo K, Watanabe A, Nagata Y, et al. (2014).
Detection of WIPI1 mRNA as an indicator of autophagosome formation. Autophagy
10(3): 497-513.

Vadrot N, Ghanem S, Braut F, Gavrilescu L, Pilard N, Mansouri A, et al. (2012).
Mitochondrial DNA maintenance is regulated in human hepatoma cells by glycogen
synthase kinase 3beta and p53 in response to tumor necrosis factor alpha. PloS one 7(7):
e40879.

Vanhara P, Hampl A, Kozubik A, Soucek K (2012). Growth/differentiation factor-15:
prostate cancer suppressor or promoter? Prostate cancer and prostatic diseases 15(4):
320-328.

Verfaillie T, Garg AD, Agostinis P (2013). Targeting ER stress induced apoptosis and
inflammation in cancer. Cancer letters 332(2): 249-264.

Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Jr., Kinzler KW
(2013). Cancer genome landscapes. Science 339(6127): 1546-1558.

Von Hoff DD, Ervin T, Arena FP, Chiorean EG, Infante J, Moore M, et al. (2013).
Increased survival in pancreatic cancer with nab-paclitaxel plus gemcitabine. The New
England journal of medicine 369(18): 1691-1703.

Wacker SA, Houghtaling BR, Elemento O, Kapoor TM (2012). Using transcriptome
sequencing to identify mechanisms of drug action and resistance. Nature chemical
biology 8(3): 235-237.

Wang J, Zuo Y, Man YG, Avital I, Stojadinovic A, Liu M, et al. (2015). Pathway and
network approaches for identification of cancer signature markers from omics data.
Journal of Cancer 6(1): 54-65.

Wang KC, Chang HY (2011). Molecular mechanisms of long noncoding RNAs.
Molecular cell 43(6): 904-914.

Wang Y, Cardenas H, Fang F, Condello S, Taverna P, Segar M, et al. (2014). Epigenetic
targeting of ovarian cancer stem cells. Cancer Res. 74(17): 4922-4936.

Waring JF, Jolly RA, Ciurlionis R, Lum PY, Praestgaard JT, Morfitt DC, et al. (2001).
Clustering of hepatotoxins based on mechanism of toxicity using gene expression
profiles. Toxicology and applied pharmacology 175(1): 28-42.

175

Wei G, Twomey D, Lamb J, Schlis K, Agarwal J, Stam RW, et al. (2006). Gene
expression-based chemical genomics identifies rapamycin as a modulator of MCL1 and
glucocorticoid resistance. Cancer cell 10(4): 331-342.

Wheelhouse NM, Lai PB, Wigmore SJ, Ross JA, Harrison DJ (2005). Mitochondrial D-
loop mutations and deletion profiles of cancerous and noncancerous liver tissue in
hepatitis B virus-infected liver. British journal of cancer 92(7): 1268-1272.

Wilson FH, Johannessen CM, Piccioni F, Tamayo P, Kim JW, Van Allen EM, et al.
(2015). A Functional Landscape of Resistance to ALK Inhibition in Lung Cancer. Cancer
cell 27(3): 397-408.

Wurmbach E, Chen YB, Khitrov G, Zhang W, Roayaie S, Schwartz M, et al. (2007).
Genome-wide molecular profiles of HCV-induced dysplasia and hepatocellular
carcinoma. Hepatology 45(4): 938-947.

Xie C, Yuan J, Li H, Li M, Zhao G, Bu D, et al. (2014). NONCODEv4: exploring the
world of long non-coding RNA genes. Nucleic acids research 42(Database issue): D98-
103.

Xu S, Butkevich AN, Yamada R, Zhou Y, Debnath B, Duncan R, et al. (2012).
Discovery of an orally active small-molecule irreversible inhibitor of protein disulfide
isomerase for ovarian cancer treatment. Proceedings of the National Academy of Sciences
of the United States of America 109(40): 16348-16353.

Yakes FM, Van Houten B (1997). Mitochondrial DNA damage is more extensive and
persists longer than nuclear DNA damage in human cells following oxidative stress.
Proceedings of the National Academy of Sciences of the United States of America 94(2):
514-519.

Yamada R, Cao X, Butkevich AN, Millard M, Odde S, Mordwinkin N, et al. (2011).
Discovery and preclinical evaluation of a novel class of cytotoxic propynoic acid
carbamoyl methyl amides (PACMAs). Journal of medicinal chemistry 54(8): 2902-2914.

Yan J, Zhong N, Liu G, Chen K, Liu X, Su L, et al. (2014). Usp9x- and Noxa-mediated
Mcl-1 downregulation contributes to pemetrexed-induced apoptosis in human non-small-
cell lung cancer cells. Cell death & disease 5: e1316.

Yang J, Carra S, Zhu WG, Kampinga HH (2013). The regulation of the autophagic
network and its implications for human disease. International journal of biological
sciences 9(10): 1121-1133.

Yang JM, Chen WS, Liu ZP, Luo YH, Liu WW (2003). Effects of insulin-like growth
factors-IR and -IIR antisense gene transfection on the biological behaviors of SMMC-
7721 human hepatoma cells. J. Gastroenterol. Hepatol. 18(3): 296-301.

176

Yap TA, Omlin A, de Bono JS (2013). Development of therapeutic combinations
targeting major cancer signaling pathways. Journal of clinical oncology : official journal
of the American Society of Clinical Oncology 31(12): 1592-1605.

Ye K, Lu J, Ma F, Keinan A, Gu Z (2014). Extensive pathogenicity of mitochondrial
heteroplasmy in healthy human individuals. Proceedings of the National Academy of
Sciences of the United States of America 111(29): 10654-10659.

Yoo CB, Jeong S, Egger G, Liang G, Phiasivongsa P, Tang C, et al. (2007). Delivery of
5-aza-2'-deoxycytidine to cells using oligodeoxynucleotides. Cancer Res. 67(13): 6400-
6408.

Yuzugullu H, Benhaj K, Ozturk N, Senturk S, Celik E, Toylu A, et al. (2009). Canonical
Wnt signaling is antagonized by noncanonical Wnt5a in hepatocellular carcinoma cells.
Mol. Cancer 8: 90.

Zeng ZL, Luo HY, Yang J, Wu WJ, Chen DL, Huang P, et al. (2014). Overexpression of
the circadian clock gene Bmal1 increases sensitivity to oxaliplatin in colorectal cancer.
Clin. Cancer Res. 20(4): 1042-1052.

Zhang DD, Hannink M (2003). Distinct cysteine residues in Keap1 are required for
Keap1-dependent ubiquitination of Nrf2 and for stabilization of Nrf2 by chemopreventive
agents and oxidative stress. Molecular and cellular biology 23(22): 8137-8151.

Zhang L, Yu J, Pan H, Hu P, Hao Y, Cai W, et al. (2007). Small molecule regulators of
autophagy identified by an image-based high-throughput screen. Proceedings of the
National Academy of Sciences of the United States of America 104(48): 19023-19028.

Zhang Y, Yang B, Du Z, Bai T, Gao YT, Wang YJ, et al. (2012). Aberrant methylation
of SPARC in human hepatocellular carcinoma and its clinical implication. World J.
Gastroenterol. 18(17): 2043-2052.

Zheng W, Thorne N, McKew JC (2013). Phenotypic screens as a renewed approach for
drug discovery. Drug discovery today 18(21-22): 1067-1073.

Zhu AX, Rosmorduc O, Evans TR, Ross PJ, Santoro A, Carrilho FJ, et al. (2015).
SEARCH: A Phase III, Randomized, Double-Blind, Placebo-Controlled Trial of
Sorafenib Plus Erlotinib in Patients With Advanced Hepatocellular Carcinoma. J. Clin.
Oncol. 33(6): 559-566.

Zhu AX, Stuart K, Blaszkowsky LS, Muzikansky A, Reitberg DP, Clark JW, et al.
(2007). Phase 2 study of cetuximab in patients with advanced hepatocellular carcinoma.
Cancer 110(3): 581-589.

177

Zhu W, Qin W, Bradley P, Wessel A, Puckett CL, Sauter ER (2005). Mitochondrial DNA
mutations in breast cancer tissue and in matched nipple aspirate fluid. Carcinogenesis
26(1): 145-152.

Zimmer M, Ebert BL, Neil C, Brenner K, Papaioannou I, Melas A, et al. (2008). Small-
molecule inhibitors of HIF-2a translation link its 5'UTR iron-responsive element to
oxygen sensing. Molecular cell 32(6): 838-848.

Zimmer M, Lamb J, Ebert BL, Lynch M, Neil C, Schmidt E, et al. (2010). The
connectivity map links iron regulatory protein-1-mediated inhibition of hypoxia-inducible
factor-2a translation to the anti-inflammatory 15-deoxy-delta12,14-prostaglandin J2.
Cancer research 70(8): 3071-3079.

Abstract (if available)

Abstract A major challenge for preclinical evaluation of novel therapeutics lies in target identification and elucidation of the mechanisms of action. The novel next generation sequencing technique is now becoming available for broad application in the drug discovery field to assist more in-depth and unbiased studies for traditional therapeutics and novel compounds. ❧ In this work, the bromouridine-labeled RNA sequencing (Bru-seq) technique was employed for mechanistic studies of anti-cancer therapeutics. Capturing nascent RNA transcription, Bru-seq allows for discovery of pharmacological changes posed by small molecule treatments at the gene transcription level. The global gene transcription profile represents a unique signature for the compound of interest. With the application of published databases and bioinformatics tools including IPA, GSEA, CMAP and Oncomine, bioinformatics analysis of the signature can assist discovery of major cellular responses and pathways involved in the therapeutic activity. Bru-seq can also identify potential biomarkers. Discovery of pathways and biomarkers can further be validated in in vitro and in vivo models. ❧ Following similar workflow, three anti-cancer therapeutic candidates have been evaluated in this study, and demonstrate promising translational potential. ❧ Combination treatment of the DNA demethylating agent SGI-110 and the chemotherapeutic oxaliplatin exhibits significant synergistic effect in hepatocellular carcinoma models. Through GSEA analysis of Bru-seq data, we identified inhibition of Wnt, EGF/IGF signaling as major contributors to activity of the combination. DNMT1 and survivin have been identified as biomarkers for future clinical evaluation. ❧ Structural optimization of the quinazolinedione (QD) ROS modulator QD232 resulted in discovery of the potent anti-cancer compound QD325, which shows ROS- dependent cytotoxicity in pancreatic cancer models. Global transcription analysis of QD325- and QD232-treated cells suggested rapid activation of Nrf2-mediated oxidative stress and unfolded protein response, which serve as the major mechanisms of action for these ROS modulators. HO-1, CHOP and GRP78 have been identified as biomarkers for QD compounds. ❧ The novel class of N-(8-quinolinyl) nicotinamides (QNs) also shows substantial anti-cancer activity in pancreatic cancer models, but with a slower reaction profile. Studies of transcription signature after 24 h treatment identified significant upregulation of genes GDF15, ATF3, DDIT3, HSPA5, WIPI1, GABARAPL1 and MAP1LC3B, which suggests involvement of stress signaling and autophagy in HJC anti-cancer activity. ❧ Successful application of Bru-seq for studies of these distinct chemical identities suggests the versatility and generality of this method, supporting its general use in preclinical evaluation of future therapeutics.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Discovery of novel small molecules targeting cancer cell metabolism

PDF

Discovery of novel small molecules for ovarian cancer treatment

PDF

Targeting cellular redox modulations for pancreatic cancer treatment

PDF

Using novel small molecule modulators as a tool to elucidate the role of the Myocyte Enhancer Factor 2 (MEF2) family of transcription factors in leukemia

PDF

Understanding DNA methylation and nucleosome organization in cancer cells using single molecule sequencing

PDF

Synthetic small molecules and protein secondary structure mimetics as modulators of hypoxia-inducible transcription and integrin receptors function

Asset Metadata

Creator Kuang, Yuting (author)

Core Title Mechanistic studies of novel small molecule anti-cancer agents using next generation sequencing

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Pharmaceutical Sciences

Publication Date 07/23/2016

Defense Date 05/20/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag anti-cancer agents,autophagy,Bru-seq,DNMT1,epigenetic priming,hepatocellular carcinoma,mechanistic studies,next generation sequencing,OAI-PMH Harvest,oxaliplatin,oxidative stress,pancreatic cancer,phenotypic screen,preclinical evaluation,ROS,SGI-110,stress signaling,survivin,unfolded protein responses

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Stiles, Bangyan L. (committee chair), Duncan, Roger F. (committee member), Machida, Keigo (committee member)

Creator Email yutingku@usc.edu,yutingkuang.usc@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-607453

Unique identifier UC11299918

Identifier etd-KuangYutin-3698.pdf (filename),usctheses-c3-607453 (legacy record id)

Legacy Identifier etd-KuangYutin-3698.pdf

Dmrecord 607453

Document Type Dissertation

Format application/pdf (imt)

Rights Kuang, Yuting

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