.. . alt:
. . .3.
. \.

am

.
. _
5:1!

ill.- .
t. .a: 0n-

”Emu?

zildfifi...
ink—.13.. .
3 . .1.
{up . 2:13.. 311.3
U‘?! *A--. u. -
1..Iz.n.?1... 1.. 312833....
. 5 r}...

.3. :13) 1’14
alianv 1.

. .
v3fnrll l. .
; .11! .
95.9. (.32.... . u
.1 . I , .. f
. {seat ..
ya,"3{s ,
f (‘04;
..{
1. .0V. 4.
.53 5...:
“gal

:24...

..
1.3:?
L). ... .
:3?!

.1.

13.2.1 :1; :2
53:... .33:

 

 

2&0 UBRARY *7
Michigan State I
University I

 

This is to certify that the
dissertation entitled

CELL DEATH AND INSULIN SIGNALING IN LIVER CELLS:
MOLECULAR AND SYSTEMS BIOLOGY STUDY

presented by

Xuerui Yang

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Chemical Engineering and
Biochemistry & Molecular Biology

 

 

C. y/‘l/uaC/ék, (iffy...
Major Professor’s Signature

j/?/09
Date

 

MSU is an Affinnative Action/Equal Opportunity Employer

 

 

 

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

—AU- griffin”

I ~

 

C

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Aoc&Pres/ClRC/DateDue.indd

 

 

 

 

CELL DEATH AND INSULIN SIGNALING IN LIVER CELLS:

MOLECULAR AND SYSTEMS BIOLOGY STUDY
By

Xuerui Yang

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Chemical Engineering and
Biochemistry & Molecular Biology

2009

u
I I
‘ g
I
E
f
‘. a
i“
‘1
s
g 1'
A
E
I
I“
.15
_‘3
i-
R!
79-
“I
v.
'1
a
£3
5“
$3
’5';
‘3”
‘n
:4
.1,
‘1
i?
.h
8‘
3:
x,‘
_’;.
Ei‘
9-7
A.
.;

 

ABSTRACT

CELL DEATH AND INSULIN SIGNALING IN LIVER CELLS:
MOLECULAR AND SYSTEMS BIOLOGY STUDY

By

Xuerui Yang

Liver disorders, such as hepatic insulin resistance, non-alcoholic fatty liver
disease (NAFLD), and non-alcoholic steatohepatitis (NASH) are closely related
with each other, and their initiation and development are attributed to the
dysregulations of the hepatic cellular activities including lipotoxicity and insulin
signaling. This thesis focused on the regulatory mechanisms that control these
hepatocellular activities.

Saturated free fatty acids (FFAs), e.g., palmitate, are known to induce
lipotoxicity, which is partially due to apoptosis. Using the molecular and cellular
biology techniques, novel mechanisms involved in palmitate-induced apoptosis
were identified in this thesis. In brief, double-stranded RNA-dependent protein
kinase (PKR) was found, for the first time, to exploit an anti-apoptotic role in
human hepatocellular carcinoma cells by regulating the protein level and
phosphorylation of Bcl-2. Palmitate suppresses this PKR—mediated anti-apoptotic
machinery in HepGZ cells, thereby resulting in apoptosis.

It is well recognized that most cellular functions are regulated by networks
of genes, proteins, and other small molecules, rather than single, isolated

factor(s). Given the complexity of biological systems, a system-level

understanding could provide a broader view on the cellular activities and thereby
complement the molecular biology studies. In collaboration with statisticians and
computational experts, we developed and applied two methodologies for system-
Ievel analyses of the palmitate-induced cytotoxicity. These strategies, based on
the principles of systems biology, recovered gene networks that are specifically
responsible for the phenotype of palmitate-induced cytotoxicity, thereby shedding
light into the potential mechanisms of the phenotype.

Another part of this thesis focused on the regulation of insulin signaling.
This thesis identified a novel player of insulin signaling, PKR, which differentially
regulates the two major insulin receptor substrates (IRS1 and IRSZ). For lRS1,
PKR induces the inhibitory serine phosphorylation at 312, resulting in down-
regulation of the tyrosine phosphorylation of IRS1. On the other hand, PKR
regulates "282 at the transcriptional rather than the post-translational level.

In summary, the hepatic insulin signaling and saturated FFA—induced
cytotoxicity were investigated in this thesis. The protein PKR was found, for the
first time, to be highly involved in regulating these hepatic cellular events, and
detailed mechanisms were identified. In addition to the molecular and cellular
biology studies, novel systems biology methodologies were developed and
implemented to obtain a boarder systems level view of palmitate-induced
cytotoxicity. These systems biology approaches suggested potential research
targets and novel mechanisms that were supported by the literature as well as

our experiments.

mam: wetter

Iv Y'Nr‘O‘ 'I
J,” -

i".
I
u!»
.:..
Y
a k
D
.3.
~ _
I
‘l
l.-
5'9
.
I1:
- r‘
h.-
a
E
a
v
.‘l
c O
szv’
i6.
-3’
, ‘.
I-
I
D
I:
.‘v
-‘s
.g.
I;
P
33"
L“
-A
.v“
m
.2
I

"tip-O ' INN" ',- v V J! n- O .0 he. 3| 9' 'p ' n
’1‘! a; .4 r 3 "rm! « 41-
51.“ I _ ‘ u 1

I .~
tie-

 

DEDICATION

To my parents Weihua Yang, June Zhang and my wife Xian Cao

-WI¢“ i- l'u-v In...” In "I‘_.( H .
h u I a r 1. u < u n . q...» n; ... ,4...“
"I I I‘m". 4.1. n .l 2‘ .. h ‘ L *4 ..

a... ,_
"It“
" (.I

I?"
" LIL. II

Pu
J

P'- 7

In
v

2‘.“ ., ,.

iv

' 7
03:13

 

ACKNOWLEDGEMENTS

Five and half years ago, I joined Dr. Christina Chan’s group at Michigan
State University, with the passion of becoming a research scientist but no idea of
what I was going to experience during my PhD study. Now, I am sitting here
finishing my doctoral dissertation, thinking back of the joy and the pain in the last
five and half years. I have always been grateful to myself for what I chose to do
five and half years ago. I feel more grateful to Dr. Chan, who is my research
mentor and supported me with all the patience and kindness during the
successful and frustrating times. Thanks to Dr. Chan, the dream does come true.
I want to say thank you to Dr. Chan, for being such a great advisor. Thanks for
the generous support and priceless academic independence you gave me, the
innovative and brilliant ideas that sparkle my research, the helpful advice
whenever I encountered problems during my research, and of course the
pushing when I sometimes relaxed too much. You will always be my greatest
advisor and the most important person in my professional career.

In addition to Dr. Chan, I would also like to thank Dr. Kathleen Gallo, Dr.
Dennis Miller, Dr. James Pestka, Dr. Patrick Walton, and Dr. Honggao Yan, who
served on my dissertation committee. Thank you all for your valuable time and
efforts in helping with my dissertation project. The discussions with all of you
have always been very productive and helpful. Special thanks to Dr. Walton, who
keeps a close relationship with our group and offered many suggestions and

creative ideas for my research projects.

I. .. ..- . .
Iv»-‘W‘ lu, I4 aw- _.

.. .

i it - 1

«a rfifflrwll “are. » «I, Q .
.

I I -‘
.{Tfi‘wvm‘ I {Ott'f‘J _“V‘9:‘

-r‘

"O-

_,
,I
..
_‘?
_‘I
b
I

 

Many thanks to the members in Dr. Chan and Dr. Walton’s groups. Dr.
Zheng Li and his wife, Lufang Sheng, Dr. Shireesh Srivastava and his wife, Dr.
Sheenu Mittal kindly taught me many basic techniques in statistical analysis, cell
culturing, Western blotting, RNA isolation, various biochemistry assays, and etc.
Discussions with different group members with different specifies has been an
important way of getting new knowledge and generating creative ideas. I also
want to thank Michael Opperrnan, who helped doing some of the experiments.

A part of my PhD work was done by collaborating with other people inside
and outside our group. Special thanks to Dr. Zheng Li, Dr. Shireesh Srivastava,
Dr. Xuewei Wang, Ming Wu, Dr. Rong Jin, and Yang Zhou. I enjoyed the happy
and productive collaborations with all of you. Without your great contributions, a
lot of excellent research works would have been very difficult to do.

I would also like to thank the Department of Chemical Engineering and
Materials Science, and the Department of Biochemistry & Molecular Biology.
Special thanks to the Quantitative Biology & Modeling Initiative (QBMI) at MSU,
for founding the Interdisciplinary PhD Program between Chemical Engineering
and Biochemistry & Molecular Biology. I feel very lucky being the first PhD
student enrolled in this brilliant interdisciplinary program.

Last but not the least, I would like to thank my wife Xian Cao, and my
parents Weihua Yang and June Zhang. Thanks to my wife, for your support, help,
and love. You are the greatest thing that ever happened to me. Thanks to my
parents. You have been my guide, my role model, and you will always be.

Thank you all again!

 

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. ix
LIST OF FIGURES ................................................................................................ x
KEY TO SYMBOLS AND ABBREVIATIONS ...................................................... xii
CHAPTER 1: Introduction ..................................................................................... 1
Background ................................................................................................ 1
Hepatic Lipotoxicity .................................................................................... 2
Insulin Signaling and Insulin Resistance .................................................... 9
Summary .................................................................................................. 1 1
References ............................................................................................... 1 3

CHAPTER 2: Repression of PKR Mediates Palmitate-Induced Apoptosis in

HepG2 Cells through Bcl-2 ................................................................................. 23
Abstract .................................................................................................... 23
Introduction .............................................................................................. 24
Materials and Methods ............................................................................. 27
Results ..................................................................................................... 33
Discussion ................................................................................................ 54
References ............................................................................................... 60

CHAPTER 3: Construction of Phenotype-Specific Gene Network for Palmitate-

lnduced Cytotoxicity by Synergy Analysis ........................................................... 68
Abstract .................................................................................................... 68
Introduction .............................................................................................. 69
Design of the Methodology ....................................................................... 73
Materials and Methods ............................................................................. 74
Results ..................................................................................................... 81
Discussion ................................................................................................ 90
References ............................................................................................... 92

CHAPTER 4: Reconstruct Modular Phenotype-Specific Gene Networks by

Knowledge-Driven Matrix Factorization ............................................................. 100
Abstract .................................................................................................. 100
Introduction ............................................................................................ 101
KMF Algorithm ........................................................................................ 104
Experimental Methods ............................................................................ 1 1 1
Results and Discussion .......................................................................... 112
Conclusions ............................................................................................ 130
References ............................................................................................. 132

CHAPTER 5: PKR Differentially Regulates IRS1 and IRSZ in HepG2 Cells ..... 140

vii

Abstract .................................................................................................. 140

Introduction ............................................................................................ 141
Materials and Methods ........................................................................... 144
Results ................................................................................................... 148
Discussion .............................................................................................. 165
References ............................................................................................. 1 70
CHAPTER 6: Conclusion and Discussion ......................................................... 179
References ............................................................................................. 191

viii

LIST OF TABLES

Table 3.1. The hub genes in the synergy network. ............................................. 83
Table 4.1. Predefined gene modules. ............................................................... 110
Table 4.2. Gene clusters identified by KMF. ..................................................... 115
Table 4.3. C Matrix of the clusters. ................................................................... 115
Table 4.4. C Matrix in toxic conditions ............................................................... 118
Table 4.5. C Matrix in nontoxic conditions ......................................................... 118
Table 4.6. Difference C Matrix. ......................................................................... 118
Table 4.7. Top 10 out of 33 genes in module 4121
Table 4.8. All 10 genes in module 6. ................................................................. 122

ix

LIST OF FIGURES

Figure 2.1. Effects of palmitate on the cytotoxicity and apoptosis of Hesz cells.

............................................................................................................................ 33
Figure 2.2. Effects of palmitate and oleate on the activity of PKR ....................... 34
Figure 2.3. Role of PKR in the cytotoxicity and apoptosis of HepGZ cells. ......... 36
Figure 2.4. Effects of palmitate and oleate on the protein level of Bcl-2 .............. 38
Figure 2.5. Involvement of PKR in regulating the protein level of Bcl-2 ............... 40
Figure 2.6. Role of PKR in regulating the Nuclear NF-KB p65 level. ................... 42
Figure 2.7. Effect of Palmitate on the phosphOrylation of Bcl-2 ........................... 44
Figure 2.8. Involvement of PKR in regulating the phosphorylation of Bcl-2 at

Ser70. ................................................................................................................. 46

Figure 2.9. Involvement of JNK in regulating the phosphorylation of Bcl-2. ........ 50
Figure 2.10. Effect of inhibiting JNK activity on the cytotoxicity and apoptosis of

HepG2 cells. ....................................................................................................... 52
Figure 2.11. Proposed signaling pathways from PKR to Bel-2 induced by

palmitate. ............................................................................................................ 53
Figure 3.1. Flowchart of the Proposed Methodology ........................................... 74
Figure 3.2. Four Representative Trends of the Metabolites. ............................... 79
Figure 3.3. The Synergy Network. ...................................................................... 82

Figure 3.4. The Distribution of Shortest Path Lengths in the Synergy Network...82

Figure 3.5. The Degree Distribution of the Synergy Network. ............................. 83
Figure 3.6. P4HA1 catalyzes the formation of 4-hydroxyproline. ........................ 85
Figure 3.7. INSIG2 plays important role in lipid synthesis. .................................. 87
Figure 3.8. The Protein Domains of SH3RF2 ...................................................... 89

Figure 4.1. Sorted difference regression coefficients (DRC) for all the genes...105

 

Figure 4.2. Gene module interaction network. .................................................. 116

Figure 4.3. Effects of the fatty acids on the expression level of RABGGTA and
the role of RABGGTA in cytotoxicity. ................................................................ 125

Figure 4.4. Effects of the fatty acids on the expression level of HSP105B. ....... 126

Figure 4.5. Effects of the fatty acids on the expression level of ATP6IP1 and the
role of ATP6IP1 in cytotoxicity. ......................................................................... 128

Figure 4.6. Effect of the fatty acids on the expression level of GMPS ............... 130

Figure 5.1. Effects of ceramide and PKR inhibitors on the phosphorylation of
IRS1. ................................................................................................................. 149

Figure 5.2. Involvement of PKR in regulating the phosphorylation of IRS1. ...... 150

Figure 5.3. Involvement of JNK and IKK in regulating the phosphorylation of IRS1.
.......................................................................................................................... 153

Figure 5.4. Involvement of PKR in regulating IRSZ. .......................................... 154
Figure 5.5. Involvement of FoxO1 in mediating the effect of PKR on IRS2 ....... 159
Figure 5.6. Effect of insulin on the activity of PKR. ........................................... 162

Figure 5.7. Involvement of IRS1 and IRSZ in mediating the effect of insulin on the
phosphorylation of PKR. ................................................................................... 163

Figure 5.8. Involvement of PP1 in mediating the effect of insulin on the
phosphorylation of PKR. ................................................................................... 164

Figure 5.9. Proposed signaling pathways through which PKR is involved in
insulin signaling network in HepG2 cells. .......................................................... 165

xi

ASKl
ATPGIPI
Bel-2
Bcl-X(L)
Birn

BSA

CRAB
dsRNA
ECM
eIF-Za
ER
ERK
ETC
FFA
FoxO
GATA4

GMPS

GTP

HCV

KEY TO SYMBOLS AND ABBREVIATIONS

apoptosis signal-regulating kinase 1

ATPase, H+ transporting, Iysosomal interacting protein 1
B-cell leukemia/lymphoma 2

Bcl-2-like l

Bcl-2-interacting mediator of cell death
bovine serum albumin

heart and neural crest derivatives expressed 2
CAMP response element-binding protein
double-stranded RNA

extracellular matrix

eukaryotic initiation factor 2-a1pha
endoplasmic reticulum
extracellular-signal-regulated kinase

electron transport chain

free fatty acid

Forkhead box-containing protein, class 0
GATA binding protein 4

guanine monophosphate synthetase

gene ontology

guanine triphosphate

hepatitis C virus

xii

HepGZ
HSPIOSB
Hsp70
IGF-l
IKK

IR

IRS

LDH
MAPK
Mda7
NAFLD

NASH

PCR

PKR

PP 1

PP 1 c

PP2A
RABGGTA
RAX/PACT

RNAi

hepatocellular carcinoma

heat shock lOSkD

70-kDa heat shock protein

insulin-like growth factor 1

I-kappaB kinase

insulin receptor

insulin receptor substrate

c-Jun N-terminal kinase
knowledge-driven matrix factorization
lactate dehydrogenase
mitogen-activated protein kinase
melanoma differentiation-associated gene-7
non alcoholic fatty liver disease
non-alcoholic steatohepatitis

nuclear factor kappa B

polymerase chain reaction
double-stranded RNA-dependent protein kinase
protein phosphatase 1

catalytic domain of PP1

protein phosphatase 2A

Rab geranylgeranyltransferase

PKR activator X

RNA interference

xiii

 

 

ROS

SCAP

SH3

siRNA

SREBPS

TCA

TG

TNF-a

UCP

UPR

reactive oxygen species

SREBP Cleavage-Activating Protein

Src homology 3

small interfering RNA

sterol regulatory element binding proteins
tricarboxylic acid

tn'acyl glycerol

Tumor necrosis factor -alpha

uncoupling proteins

unfolded protein response

xiv

 

tor.

.
fl
-
.
‘F
.
Q
.
I-
'l
,..
C
‘O
,..
..
.-
..
I-x.
.
r.-
r...
I. -
.'.
..
,..-
..
.
.

. .
D'.
.-
O
>-.'
i
I}
40
4"—
’3.
W
~.
".
.
“F
l.

rIn'nJ

"' "Y'TTJT 7"!
Tim g‘tiign‘

:17“.- Dig {A} ~
.= miter-mime.

,‘ A.“
all?”

 

_. r. 0"
no I
l

CHAPTER 1: Introduction

Background

As one of the vital organs in vertebrates, the liver plays a major role in
metabolism (i.e. carbohydrates, lipids, amino acids, and proteins), thereby functioning in
glycogen storage, plasma protein synthesis, hormone production, and detoxification (1).
Primary hepatocytes make up 70-80% of the cytoplasmic mass of the liver, and perform
most of the functions of the liver (1). Dysregulation of the hepatocyte function can lead to
liver diseases, e. g., non-alcoholic fatty liver disease (NAFLD), which consists of different
stages: steatosis and non—alcoholic steatohepatitis (NASH), and, at its most severe stages,
may progress to cirrhosis and liver failure (2).

NAFLD is the most common liver disease in the United States. It is believed that
over next 20 years, NAFLD and NASH will surpass hepatitis C as the leading cause for
liver transplantation in the United States (3). The exact cause of NAFLD is still unknown.
However, it has been recognized that excess circulating free fatty acids (F FAs), in
particular long-chain saturated fatty acids, e. g. palmitate, play significant roles in the
development and progression of NAFLD by inducing hepatic lipotoxicity (4, 5).

Hepatic lipid metabolism is highly regulated by insulin, through the insulin
signaling pathway (6). Disruption of insulin signaling, i.e., insulin resistance, is common
in patients with NAFLD (7), and is believed to play an important role in the pathogenesis
of NAFLD by impairing mitochondrial F FA B-oxidation and promoting fat accumulation

in the liver (8).

 

Insulin signaling is a central signaling pathway that regulates many cellular
activities, such as glucose and lipid metabolism, protein synthesis and degradation, cell
growth and differentiation (6). Insulin resistance leads to dysregulation of glucose
homeostasis and lipid metabolism in hepatocytes (9), thereby playing a major role in the
development of type 2 diabetes as well as NAFLD. Type 2 diabetes is characterized by
high blood glucose in response to insulin resistance as well as insulin deficiency (10).
According to the NIH, in 2007 up to 23.6 million people (7.8% of the population) have
diabetes, 90-95% of which are type 2 diabetes (11).

In this thesis, my research was focused on the mechanisms of hepatic lipotoxicity

and the regulation of insulin signaling in HepGZ cells, a study model of the hepatocytes.

Hepatic Lipotoxicity

Chapters 2 to 4 discussed the investigations on hepatic lipotoxicity. Chronic
treatments of elevated FFAs induce cytotoxicity in a variety of cells, including skeletal
muscle myotubes (12), cardiomyocytes (13), pancreatic cells (14, 15), and hepatocytes
(I6, 17). Previous studies have identified that saturated FFAs, e. g. palmitate, are much
more toxic to hepatocytes and hepatoma cells than unsaturated FFAs, e. g. oleate (l 7, 18).
Over the past decades, researchers have put much effort in uncovering the mechanisms of
saturated FFAs induce cytotoxicity, which is believed to largely relate to apoptosis (18-
20). Apoptosis, programmed cell death, is one of the major cellular activities involved in
metazoan development, inflammatory responses, tumor growth control, and many other
processes (21, 22). Saturated FFAs, e. g. palmitate, induce apoptosis in many cell types,

such as cardiac cells, pancreatic beta cells, breast cancer cells, and hepatocytes. The

\IF; '

a
‘l
'f.‘
02-.
H.
1),.
3:
v...
7o.
3‘!
-;.4‘
'7}
8.-
1-
.1‘

“I

r I win“! fulll'l‘ 9 II
“I 1.9".‘1‘I
"r t“. 'I,‘

are . ,
We

.0
A «a

'ttr-x r:
Wklhflfkixtli

IO.’ ' . 'l'l'v
WW
. 0'

 

mechanism by which palmitate induces apoptosis in different types of cells has been
under intense investigation. Studies of palmitate-induced apoptosis in liver cells have
focused predominantly on Iysosomal permeabilization (23), ROS production (24), and
intracellular metabolic pathways such as beta oxidation (25), TC accumulation (26, 27),
and ceramide production (1 8, 28).

More recent studies revealed the involvement of certain Bel-2 family proteins in
mediating saturated FFA-induced apoptosis of liver cells (19, 20, 23). The Bel-2 family
proteins play critical roles in the intrinsic and extrinsic apoptosis pathways by regulating
the release of apoptogenic molecules from the mitochondria to the cytosol or the
activities of caspases (29, 30). It has been proposed that palmitate-induced apoptosis in
liver cells is related to some of the Bcl-2 family proteins, such as Bax and the Bax
antagonist Bcl-X(L) (19, 23), and Bcl-2-interacting mediator of cell death (Bim), a pro-
apoptotic Bel-2 family protein (19, 20). In addition to these pro-apoptotic Bel-2 family
proteins, an anti-apoptotic member, Bel-2, has been identified as an important factor in
regulating the apoptosis of HepGZ cells (31-33). As one of the most studied and
important anti-apoptotic members of the Bcl-2 family, Bel-2 was found to protect cells
against intrinsic (mitochondria) apoptosis by maintaining the integrity of the
mitochondrial membrane (29). In Chapter 2 of this thesis, the potential effects of FFAs,
palmitate and oleate, on the Bcl-2 protein in HepG2 cells were investigated. The findings
suggest a negative effect of palmitate on the protein level of Bcl-2 in HepGZ cells, and
further show that palmitate also down-regulates the phosphorylation of Bel-2 at Ser70,
which also plays a role in determining the anti-apoptotic function of Bel-2 (34).

Furthermore, 2 signal transduction pathways were identified, through which palmitate

exploits its regulatory effect on Bel-2. The double-stranded RNA-dependent protein
kinase (PKR) was found to be central in mediating these 2 pathways.

Initially identified as an anti-viral protein, PKR is best known for triggering cell
defense responses and initiating innate immune responses by arresting general protein
synthesis and inducing apoptosis during virus infection (35). Foreign double-stranded
RNA (dsRNA), a by—product of viral RNA polymerases during virus replication, binds
PKR, facilitating the homo-dimerization and auto-phosphorylation at Thr451 and Thr446
and thereby the activation of PKR (36, 37). Independent of dsRNA, PKR can also be
activated by other cellular proteins such as RAX/PACT (38, 39) and Mda7 (40), thereby
mediating stresses signals induced by serum starvation, H202, or ceramide (38, 39, 41).
The activated PKR, known as a eukaryotic initiation factor 2-alpha (eIF-2a) kinase,
induces the phosphorylation of eIF-2a at SerSl (3 7), which inhibits the initiation of RNA
translation by the tRNA 4OS ribosomal subunit. This function of PKR results in the
inhibition of general protein synthesis and the induction of cellular apoptosis in many
types of eukaryotic cells (37, 42). However, evidence is emerging, albeit controversial‘at
this point, suggesting that PKR also has an anti-apoptotic role in mouse embryo
fibroblasts and certain tumor cells (43-47). In this thesis, an anti-apoptotic role of PKR in
HepG2 cells was identified, mediated through Bcl-2, an anti-apoptotic protein. The
findings further showed that palmitate suppresses the phosphorylation of PKR, which
serves as a mechanism of palmitate-induced apoptosis.

Thus, in Chapter 2 using research tools from molecular and cellular biology, I
successfully uncovered a mechanism by which palmitate induces apoptosis and

cytotoxicity in liver cells. A number of mechanisms have already been proposed, in

which many important genes were suggested to play roles in saturated F FA-induced cell
death. They include Iysosomal permeabilization (23), ROS production (24), and altered
intracellular metabolic pathways (18, 25-28). Considering these multi-faceted effects, it is
unlikely that hepatic lipotoxicity can be attributed to a single or a few genes functioning
on an isolated cellular activity. For example, cytotoxicity, as well as apoptosis, were
shown to be related to ROS production, but they were not completely prevented upon
treatment with mitochondrial complex inhibitors or fi‘ee radical scavengers, suggesting
that mechanisms other than ROS production contributed to the toxicity of palnritate (24).
It is now well-recognized that most cellular fimctiorrs are regulated by complex
networks of genes, proteins, and small molecules, including metabolites and signaling
molecules (48). Focusing on a limited number of proteins/ genes, molecular biology has
made great strides in elucidating detailed mechanistic information of cellular activities.
Given the complexity of biological systems, a system-level understanding could
complement and help unravel the complex interactions in biological systems and thereby
provide a broader view on the cellular activities (48). Recent development in
biotechnology has made it possible to systematically measure profiles of genes, proteins
and metabolites of cells, enabling system-level analyses of cellular activities facilitated
by statistical or mathematical tools and grounded in molecular-level understanding (49).
With the technique of cDNA microarray, our group has obtained comprehensive
gene-expression profiles of HepGZ cells in response to different FFA exposures (50). To
systematically study the overall effect of FFAs on liver cells, in Chapters 3 and 4, I,
together with my colleague-mathematicians and computational experts, evaluated

methodologies that could extract mechanistic information from gene-expression profiles.

Assuming that saturated FFA-induced cytotoxicity is coordinately regulated by a set of
genes that interact in a complex network, we developed methods to 1) select a pool of the
genes that are potentially relevant to the phenotype of interest, i.e. saturated F FA-induced
lipotoxicity, and 2) reconstruct gene networks, with the selected genes, that give rise to
the observed phenotype. These gene networks should only recover the interactions that
are related to the phenotype of interest, herein denoted as phenotype-specific gene
networks. By reconstructing phenotype-specific gene networks we hypothesize that a
broader view of cellular activities could provide novel insights into the mechanisms
involved, and ultimately, in the mechanisms at play in diseases (51, 52).

Selection of the Genes——Selection of the genes relevant to a phenotype is usually
viewed as a feature selection problem (53, 54) and have been based predominantly on
statistical tests (55) or correlation measures (56), which were used to assess the statistical
differences among the gene profiles and the correlations between the genes and the
phenotype, respectively. These methods, purely based on microarray data, are susceptible
to the noise level in the data and usually computationally expensive. In addition, they are
likely to miss some important genes that are not differentially expressed. One strategy to
ameliorate these issues is to incorporate prior knowledge, such as domain knowledge and
functional (e. g. gene ontology (60)) information of the genes, in the gene selection
process (57). In Chapter 4 of this thesis, a Bayesian mixture regression model was
developed to quantitatively incorporates the prior knowledge of the gene functions,
upfront, in the process of gene-selection (58). This method is more efficient in
incorporating the prior knowledge, as compared to the typical knowledge-based methods,

in which the prior information is incorporated in the post-processing of the genes that are

selected by the data-driven approaches.

In addition to the Bayesian mixture regression model, an alternative strategy of
gene-selection was also evaluated in Chapter 3. This strategy integrated multiple levels of
information, i.e., gene expression and metabolite profiles, in the process of gene selection.
It is known that some alterations in metabolism are involved in the induction of
cytotoxicity by saturated FFAs (26, 59, 60). Therefore, this approach better reflected the
“multi-level” characteristic of cellular activities and took advantage of the additional
accessible biological information in the selection of genes that are involved in the
phenotype.

Reconstruction of the Gene Network—Various methods, such as correlation (61),
mutual information (62, 63), and Bayesian network analysis (64), have been used to
assess gene interactions and reconstruct gene networks. These methods do not directly
incorporate the phenotype information in identifying the gene interactions. Instead, gene
networks were built for each of the conditions, and compared across the conditions to
identify the gene interactions that are specific to a condition or a phenotype.
Consequently, these methods are computationally expensive and sensitive to the quality
of data To address this short-coming, Chapter 3 introduced an alternative method, which
identifies the synergistic gene pairs that cooperatively function on a phenotype, thereby
avoiding comparison of the networks under different conditions. This method is based on
the concept of synergy. Synergy is defined as the “additional” contribution provided by
the ‘fivhole” as compared to the sum of the contributions of the individual “parts” (65, 66).
Biological activities are regulated by multiple factors, many of which function

cooperatively, i.e., synergistically. The basic idea is that the whole (i.e., the regulatory

system) is greater than the sum of the individual parts (i.e., regulators) of a system (67).
We assess the synergistic effects between gene pairs on the phenotype, namely palmitate-
induced cytotoxicity, and with the identified synergistic gene pairs, built a synergy
network that was phenotype-specific. The topological analyses revealed the structural
characteristics of the network while the hub genes provided insights into potential
mechanism(s) involved in the induction of the phenotype.

Given that biological networks, e.g., protein-protein interaction network (68),
metabolic network (69) and transcriptional regulation network (70), are modular in
structure, identifying the gene modules and their interplay in a modular network should
provide insights into the differential mechanisms involved in different cellular states, i.e.,
normal vs. disorder. Several clustering methods, such as self-organizing mapping (71, 72),
hierarchical clustering (61) and K-means (73), have been commonly used to identify gene
modules. However, most of these approaches cannot uncover the interactions among the
modules or clusters. To address this limitation, several studies integrated clustering
methods with structure learning algorithms, such as Graphical Gaussian Modeling and
Bayesian network learning (74-76). These approaches are predominantly data-driven and
thus susceptible to noise in the expression data, and suffer from the sparse data problem
associated with limited number of experimental conditions (77, 78). Prior knowledge has
been shown to be helpful in reconstructing modular networks with sparse and noisy
expression data (70, 79-84). To exploit the prior knowledge and reconstruct modular gene
networks, we developed a framework based on knowledge-driven matrix factorization,
termed KMF, in Chapter 4. This KMF fiamework efficiently incorporates the prior

knowledge of co-regulation relationships into the network reconstruction using a

regularization scheme and successfully derives both gene modules and their interaction
simultaneously, while the other matrix factorization methods can only identify gene
clusters. Supported by the literature, these modules and their interactions potentially may
be important in saturated-FFA induced cytotoxicity. In addition, evaluating the
contribution of the genes within the highly relevant modules revealed potential genes that
may be involved in palmitate-induced cytotoxicity. Further experiments in Chapter 4
confirmed the involvement of these genes in conferring a cytotoxic phenotype and
suggested novel research targets for addressing the cytotoxicity.

In summary, in this thesis saturated FFA-induced hepatic cytotoxicity was
investigated using 2 different types of approaches, experimental and computational. In
Chapter 2, with the help of molecular and cellular biology techniques, 2 pathways
through which PKR mediates the regulation of the protein level and the phosphorylation
status of Bcl-2 were identified, providing a novel mechanism by which palmitate induced
apoptosis in HepG2 cells. On the other hand, fundamental ideas in systems biology were
applied in Chapters 3 and 4, and 2 strategies were developed, to reconstruct phenotype-
specific gene networks, one based on synergy analysis between gene pairs (Chapter 3)
and the other one based on gene clustering (Chapter 4). Both of these systematic
strategies were able to provide a systematic view and capture certain features of the

phenotype, further shedding light into potential mechanisms of lipotoxicity.

Insulin Signaling and Insulin Resistance
Chapter 5 discussed my investigations on insulin signaling. As a central signaling

pathway that regulates many cellular activities including glucose and lipid metabolism,

 

. .. . ..... . r .. ‘ I -
If." v r‘ ,n s I.-' ' . ,.. - ,.. .r
"fl-Itch“ ‘.'£’l‘i . I' (I I we" IV... .1

rig 13‘1;15;6;‘ if»; ’v'
s . :- kiln-s;- .

I

.‘ '1" ”I’J'iy'
. mlilllz‘u

protein synthesis and degradation, cell growth and differentiation (6), insulin signaling is
initiated upon binding of insulin to the insulin receptor (IR), a receptor tyrosine kinase
(85), and transmitted intracellularly by the insulin receptor substrates (IRS) (6, 85). At
least 4 of the 1R substrates belong to the IRS group, with R81 and IRSZ being
predominant and expressed in most tissues, including the liver (86, 87). Upon
phosphorylation of the tyrosine residues catalyzed by IR, the IRS proteins initiate,
through different binding mechanisms (87), various downstream signal transduction
cascades, including mitogen-activated protein kinase (MAPK) pathways (c-Jun N-
terminal kinase (JNK), extracellular signal-related kinase (ERK), and p38) (88, 89) and
phosphoinositide 3-kinase (PI3K) (90), which in turn activates Akt/protein kinase B
(Akt/PKB) (91), and atypical protein kinase C (aPKC) (92).

Insulin signaling is sophisticatedly tuned by a large number of regulators.
Pathologically, dysregulation of insulin signaling has been linked to multiple diseases and
disorders. In particular, disruption of insulin signaling leads to insulin resistance, a
pathological state in which the target cells fail to respond to physiological levels of
insulin (93). In combination with the failure of B cells to compensate, insulin resistance
directly contributes to the development of type 2 diabetes (94, 95). Insulin resistance is
also related to multiple other metabolic, endocrine and cardiovascular disorders (96, 97).
At the molecular level, dysregulation of insulin signaling or insulin resistance could
occur at several possible stages, e.g., degradation or mutation of IR (98, 99), inhibitory
phosphorylation or degradation of IRS (100, 101), or suppression of down-stream
signaling molecules, such as PI-3 kinase or Alct/PKB (reviewed in ref. (102)). However,

most of the insulin signaling disruption and insulin resistance have been attributed to

10

 

 

dysregulation of the phosphorylation of IR and, in particular, IRS (100, 101). IRS
proteins mediate intracellular insulin signaling, through the tyrosine residues, which
facilitate recruitment of IRS substrates and promote insulin signaling (87), and the serine
residues, which generally suppress the activities of IRS by blocking the interaction
between IRS and IR (103), inhibiting the tyrosine phosphorylation of IRS (104), or
inducing the degradation of IRS (105). A number of serine residues have been identified
to negatively regulate the activity of IRS], in particular, Ser307 (equivalent to Ser312 in
human IRS 1 ). Ser307 has been extensively investigated and characterized as a key
indicator of inhibitory phosphorylation of R81 and insulin resistance, and confirmed in
insulin-resistant rodent models (106).

In Chapter 5, the potential involvement of PKR in regulating the phosphorylation
of IRSl was investigated. PKR was found to up-regulate the inhibitory phosphorylation
of IRSl at Ser312, which is believed to suppress insulin signaling and induce insulin
resistance. In addition, further investigation shoWed that, IRSZ, another predominant IRS
family protein in the liver, is also affected by PKR, through a different mechanism. In
addition, we as well as others showed that PKR activity is down-regulated by insulin
through the IRS proteins (107). Therefore this feedback loop between PKR and IRS
proteins may have potential implications in the regulation of insulin signaling and the

development of insulin resistance.

Summary

Liver disorders such as NAFLD and hepatic insulin resistance are correlated with

each other and involve major cellular activities such as lipotoxicity and insulin signaling

11

in liver cells. In this thesis, my research focused on the regulatory mechanisms that
control these hepatic cellular activities.

Using molecular and cellular biology techniques, mechanisms involved in
palmitate-induced apoptosis and the regulation of insulin signaling through IRS proteins
were identified. In addition, in collaboration with other colleagues inside and outside our
research group, we applied novel systems biology methodologies to provide potential
insights into lipotoxicity. These methods could be easily adapted to other complex
biological systems, thereby serving as powerful research tools in extracting mechanistic

information and identifying potential research and drug targets.

12

 

10.

11.

References

Davey, B., Halliday, T., and Hirst, M. (2001) Human biology and health : an
evolutionary approach, Open University, Buckingham [UK] ; Phildelphia

Clark, J. M., and Diehl, A. M. (2003) Nonalcoholic fatty liver disease: an
underrecognized cause of cryptogenic cirrhosis. JAMA 289, 3000-3004

Contos, M. J ., and Sanyal, A. J. (2002) The clirricopathologic spectrum and
management of nonalcoholic fatty liver disease. Adv Anat Pathol 9, 37-51

Feldstein, A. B., Wemeburg, N. W., Canbay, A., Guicciardi, M. B., Bronk, S. F.,
Rydzewski, R., Burgart, L. J., and Gores, G. J. (2004) Free fatty acids promote
hepatic lipotoxicity by stimulating TNF-alpha expression via a lysosomal pathway.
Hepatology 40, 185-194

Wang, D., Wei, Y., and Pagliassotti, M. J. (2006) Saturated fatty acids promote
endoplasmic reticulum stress and liver injury in rats with hepatic steatosis.
Endocrinology 147, 943-951

Saltiel, A. R., and Kahn, C. R. (2001) Insulin signalling and the regulation of
glucose and lipid metabolism. Nature 414, 799-806

Hamaguchi, M., Kojima, T., Takeda, N., Nakagawa, T., Taniguchi, H., Fujii, K.,
Omatsu, T., Nakajima, T., Sarui, H., Shimazaki, M., Kato, T., Okuda, J ., and Ida, K.
(2005) The metabolic syndrome as a predictor of nonalcoholic fatty liver disease.
Ann Intern Med 143, 722-728

Salt, W. B., 2nd (2004) Nonalcoholic fatty liver disease (NAFLD): a
comprehensive review. J Insur Med 36, 27-41

Klover, P. J ., and Mooney, R. A. (2004) Hepatocytes: critical for glucose
homeostasis. Int J Biochem Cell Biol 36, 753-758

Kumar, V., Abbas, A. K., Fausto, N., Robbins, S. L., and Cotran, R. S. (2005)
Robbins and Cotran pathologic basis of disease, Elsevier Saunders, Philadelphia

Rena, 6., Woods, Y. L., Prescott, A. R., Peggie, M., Unterman, T. G., Williams, M.
R., and Cohen, P. (2002) Two novel phosphorylation sites on FKHR that are
critical for its nuclear exclusion. EMBO J 21, 2263-2271

13

12.

l3.

14.

15.

16.

17.

l8.

19.

20.

21.

22.

23.

Turpin, S. M., Lancaster, G. I., Darby, I., Febbraio, M. A., and Watt, M. J. (2006)
Ap0ptosis in skeletal muscle myotubes is induced by ceramides and is positively
related to insulin resistance. Am J Physiol Endocrinol Metab 291, E1341-1350

Hickson-Bick, D. L., Sparagna, G. C., Buja, L. M., and McMillin, J. B. (2002)
Palmitate-induced apoptosis in neonatal cardiomyocytes is not dependent on the
generation of ROS. Am J Physiol Heart Circ Physiol 282, H656-664

Landau, Z., Forti, B., Alcaly, M., and Birk, R. Z. (2006) Palmitate induced
lipoapoptosis of exocrine pancreas AR42J cells. Apoptosis 11, 717-724

Shimabukuro, M., Zhou, Y. T., Levi, M., and Unger, R. H. (1998) Fatty acid-
induced beta cell apoptosis: a link between obesity and diabetes. Proc Natl Acad
Sci U S A 95, 2498-2502

Ji, J., Zhang, L., Wang, P., Mu, Y. M., Zhu, X. Y., Wu, Y. Y., Yu, H., Zhang, B.,
Chen, S. M., and Sun, X. Z. (2005) Saturated free fatty acid, palmitic acid, induces
apoptosis in fetal hepatocytes in culture. Exp T oxicol Pathol 56, 369-376

Srivastava, S., and Chan, C. (2007) Hydrogen peroxide and hydroxyl radicals
mediate palmitate-induced cytotoxicity to hepatoma cells: relation to mitochondrial
permeability transition. Free Radic Res 41, 38-49

Wei, Y., Wang, D., Topczewski, F., and Pagliassotti, M. J. (2006) Saturated fatty
acids induce endoplasmic reticulum stress and apoptosis independently of ceramide
in liver cells. Am J Physiol Endocrinol Metab 291, E275-281

Barreyro, F. J., Kobayashi, S., Bronk, S. F., Wemeburg, N. W., Malhi, H., and
Gores, G. J. (2007) Transcriptional Regulation of Birn by FoxO3A Mediates
Hepatocyte Lipoapoptosis. J Biol Chem 282, 27141-27154

Malhi, H., Bronk, S. F., Wemeburg, N. W., and Gores, G. J. (2006) Free fatty acids
induce JNK-dependent hepatocyte lipoapoptosis. J Biol Chem 281, 12093-12101

Lawen, A. (2003) Apoptosis-an introduction. Bioessays 25, 888-896

Elmore, S. (2007) Apoptosis: a review of programmed cell death. T oxicol Pathol 35,
495-5 1 6

Feldstein, A. B., Wemeburg, N. W., Li, Z., Bronk, S. F., and Gores, G. J. (2006)
Bax inhibition protects against free fatty acid-induced Iysosomal permeabilization.
Am J Physiol Gastrointest Liver Physiol 290, G1 339-1 346

14

 

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

Srivastava, S., and Chan, C. (2006) Hydrogen peroxide and hydroxyl radicals
mediate palrrritate-induced cytotoxicity to hepatoma cells: Relation to
mitochondrial permeability transition. Free Radic Res 41, 38-49

Giordano, A., Calvani, M., Petillo, O., Grippo, P., Tuccillo, F., Melone, M. A.,
Bonelli, P., Calarco, A., and Peluso, G. (2005) tBid induces alterations of
mitochondrial fatty acid oxidation flux by malonyl-CoA-independent inhibition of
carnitine palmitoyltransferase-l. Cell Death Dijfir 12, 603-613

Listenberger, L. L., Han, X., Lewis, S. B., Cases, S., Farese, R. V., Jr., Ory, D. S.,
and Schaffer, J. E. (2003) Triglyceride accumulation protects against fatty acid-
induced lipotoxicity. Proc Natl Acad Sci U S A 100, 3077-3082

Li, Z., Srivastava, S., Yang, X., Mittal, 8., Norton, P., Resau, J ., Haab, B., and Chan,

C. (2007) A hierarchical approach employing metabolic and gene expression
profiles to identify the pathways that confer cytotoxicity in HepG2 cells. BMC Syst
Biol 1, 21

Assimacopoulos-Jeannet, F. (2004) Fat storage in pancreas and in insulin-sensitive
tissues in pathogenesis of type 2 diabetes. Int J Obes Relat Metab Disord 28 Suppl
4, $53-57

Breckenridge, D. G., and Xue, D. (2004) Regulation of mitochondrial membrane
permeabilization by BCL-2 family proteins and caspases. Curr Opin Cell Biol 16,
647-652

Li, H., Zhu, H., Xu, C. J., and Yuan, J. (1998) Cleavage of BID by caspase 8
mediates the mitochondrial damage in the Fas pathway of apoptosis. Cell 94, 491-
501

Feng, L. F., Zhong, M., Lei, X. Y., Zhu, B. Y., Tang, S. S., and Liao, D. F. (2006)
Bcl-2 siRN A induced apoptosis and increased sensitivity to 5-fluorouracil and
HCPT in HepG2 cells. J Drug Target 14, 21-26

Handayani, T., Sakinah, S., Nallappan, M., and Pihie, A. H. (2007) Regulation of
p53-, Bcl-2- and caspase-dependent signaling pathway in xanthorrhizol-induced
apoptosis of HepG2 hepatoma cells. Anticancer Res 27, 965-971

Huang, D. S., Shen, K. 2., Wei, J. F., Liang, T. B., Zheng, S. S., and Xie, H. Y.
(2005) Specific COX-2 inhibitor NS398 induces apoptosis in human liver cancer
cell line HepG2 through BCL-2. World J Gastroenterol 11, 204-207

15

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

Blagosklonny, M. V. (2001) Unwinding the loop of Bel-2 phosphorylation.
Leukemia 15, 869-874

Barber, G. N. (2005) The dsRNA-dependent protein kinase, PKR and cell death.
Cell Death Difler 12, 563-570

Zhang, R, Romano, P. R., Nagamura-Inoue, T., Tian, B., Dever, T. B., Mathews, M.

B., Ozato, K., and Hinnebusch, A. G. (2001) Binding of double-stranded RNA to
protein kinase PKR is required for dimerization and promotes critical
autophosphorylation events in the activation loop. J Biol Chem 276, 24946-24958

Taylor, S. S., Haste, N. M., and Ghosh, G. (2005) PKR and eIF2alpha: integration
of kinase dimerization, activation, and substrate docking. Cell 122, 823-825

Patel, C. V., Handy, 1., Goldsmith, T., and Patel, R. C. (2000) PACT, a stress-
modulated cellular activator of interferon-induced double-stranded RNA-activated
protein kinase, PKR. J Biol Chem 275, 37993-37998

Ito, T., Yang, M., and May, W. S. (1999) RAX, a cellular activator for double-
stranded RNA-dependent protein kinase during stress signaling. J Biol Chem 274,
15427-15432

Pataer, A., Vorburger, S. A., Chada, S., Balachandran, 8., Barber, G. N., Roth, J. A.,
Hunt, K. K., and Swisher, S. G. (2005) Melanoma difi‘erentiation-associated gene-7
protein physically associates with the double-stranded RNA-activated protein
kinase PKR. MoI Ther 11, 717-723

Ruvolo, P. P., Gao, F., Blalock, W. L., Deng, X., and May, W. S'. (2001) Cerarnide
regulates protein synthesis by a novel mechanism involving the cellular PKR
activator RAX. J Biol Chem 276, 11754-11758

Stark, G. R., Kerr, 1. M., Williams, B. R., Silverman, R. H., and Schreiber, R. D.
(1998) How cells respond to interferons. Annu Rev Biochem 67, 227-264

Donze, O., Deng, J ., Curran, J ., Sladek, R., Picard, D., and Sonenberg, N. (2004)
The protein kinase PKR: a molecular clock that sequentially activates survival and
death programs. EMBO J 23, 564-571

Kim, S. H., Forrnan, A. P., Mathews, M. B., and Gunnery, S. (2000) Human breast
cancer cells contain elevated levels and activity of the protein kinase, PKR.
Oncogene 19, 3086-3094

16

 

 

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

Kim, S. H., Gunnery, S., Choc, J. K., and Mathews, M. B. (2002) Neoplastic
progression in melanoma and colon cancer is associated with increased expression
and activity of the interferon-inducible protein kinase, PKR. Oncogene 21, 8741-
8748

Hiasa, Y., Kamegaya, Y., Nuriya, H., Onji, M., Kohara, M., Schmidt, E. V., and
Chung, R. T. (2003) Protein kinase R is increased and is functional in hepatitis C
virus-related hepatocellular carcinoma. Am J Gastroenterol 98, 2528-2534

Delgado Andre, N., and De Lucca, F. L. (2007) Knockdown of PKR expression by
RNAi reduces pulmonary metastatic potential of B16-F10 melanoma cells in mice:
Possible role of NF-kappaB. Cancer Lett 258, 118-125

Kitano, H. (2002) Computational systems biology. Nature 420, 206-210
Kitano, H. (2002) Systems biology: a brief overview. Science 295, 1662-1664

Li, Z., Srivastava, S., Mittal, S., Yang, X., Sheng, L., and Chan, C. (2007) A Three
Stage Integrative Pathway Search (TIPS) framework to identify toxicity relevant
genes and pathways. BMC Bioinformatics 8, 202

Said, M. R., Begley, T. J ., Oppenheim, A. V., Lauffenburger, D. A., and Samson, L.
D. (2004) Global network analysis of phenotypic effects: protein networks and

toxicity modulation in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 101,
18006-1801 1

Srivastava, S., Li, Z., Yang, X., Yedwabnick, M., Shaw, 8., and Chan, C. (2007)
Identification of genes that regulate multiple cellular processes/responses in the
context of lipotoxicity to hepatoma cells. BMC Genomics 8, 364

Ressom, H. W., Varghese, R. S., Zhang, Z., Xuan, J., and Clarke, R. (2008)
Classification algorithms for phenotype prediction in genomics and proteorrrics.
Front Biosci 13, 691-708

Saeys, Y., Inza, 1., and Larranaga, P. (2007) A review of feature selection
techniques in bioinforrnatics. Bioinformatics 23, 2507-2517

Tusher, V. G., Tibshirani, R., and Chu, G. (2001) Significance analysis of
microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98,
51 16-5121

Li, Z., and Chan, C. (2004) Integrating gene expression and metabolic profiles. J
Biol Chem 279, 27124-27137

17

57.

58.

59.

60.

61.

62.

63.

65.

66.

67.

68.

Le Phillip, P., Bah], A., and Ungar, L. H. (2004) Using prior knowledge to improve
genetic network reconstruction from rrricroarray data. In Silico Biol 4, 335-353

Srivastava, S., Zhang, L., Jin, R., and Chan, C. (2008) A novel method
incorporating gene ontology information for unsupervised clustering and feature
selection. PLoS ONE 3, e3 860

Lam, T. K., Carpentier, A., Lewis, G. F., van de Werve, G., Fantus, I. G., and
Giacca, A. (2003) Mechanisms of the free fatty acid-induced increase in hepatic
glucose production. Am J Physiol Endocrinol Metab 284, E863-873

Srivastava, S., and Chan, C. (2008) Application of metabolic flux analysis to
identify the mechanisms of free fatty acid toxicity to human hepatoma cell line.
Biotechnol Bioeng 99, 399-410

Eisen, M. B., Spellman, P. T., Brown, P. O., and Botsteirr, D. (1998) Cluster
analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A
95, 14863-14868

Liang, K. C., and Wang, X. (2008) Gene regulatory network reconstruction using
conditional mutual information. E URASIP journal on bioinformatics & systems
biology, 253894

Basso, K., Margolin, A. A., Stolovitzky, G., Klein, U., Dalla-Favera, R., and
Califano, A. (2005) Reverse engineering of regulatory networks in human B cells.
Nature genetics 37, 382-390

Pe'er, D., Regev, A., Elidan, G., and Friedman, N. (2001) Inferring subnetworks
from perturbed expression profiles. Bioinformatics 17 Suppl 1, 8215-224

Varadan, V., and Anastassiou, D. (2006) Inference of disease-related molecular
logic from systerns-based microarray analysis. PLoS Comput Biol 2, e68

Anastassiou, D. (2007) Computational analysis of the synergy among multiple
interacting genes. Mol Syst Biol 3, 83

Varadan, V., Miller, D. M., 3rd, and Anastassiou, D. (2006) Computational
inference of the molecular logic for synaptic connectivity in C. elegans.
Bioinformatics 22, 6497-506

Han, J. D. (2008) Understanding biological functions through molecular networks.
Cell Res 18, 224-237

18

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

Ravasz, B., Somera, A. L., Mongru, D. A., Oltvai, Z. N., and Barabasi, A. L. (2002)
Hierarchical organization of modularity in metabolic networks. Science 297, 1551-
1555

Ihmels, J ., Friedlander, G., Bergrnann, S., Sarig, O., Ziv, Y., and Barkai, N. (2002)
Revealing modular organization in the yeast transcriptional network. Nat Genet 31,
370-377

Yin, L., Huang, C. H., and Ni, J. (2006) Clustering of gene expression data:
performance and similarity analysis. BMC Bioinformatics 7 Suppl 4, S19

Toronen, P., Kolehmainen, M., Wong, G., and Castren, E. (1999) Analysis of gene
expression data using self-orgarrizing maps. FEBS Lett 451, 142-146

Ma, S. F., Grigoryev, D. N., Taylor, A. D., Nonas, S., Sammani, S., Ye, S. Q., and
Garcia, J. G. (2005) Bioinfonnatic identification of novel early stress response
genes in rodent models of lung injury. Am J Physiol Lung Cell Mol Physiol 289,
L468-477

Toh, H., and Horimoto, K. (2002) Inference of a genetic network by a combined

approach of cluster analysis and graphical Gaussian modeling. Bioinformatics 18,
287-297

Segal, B., Shapira, M., Regev, A., Pe'er, D., Botstein, D., Koller, D., and Friedman,
N. (2003) Module networks: identifying regulatory modules and their condition-
specific regulators fiom gene expression data. Nat Genet 34, 166-176

Li, Z., and Chan, 'C. (2004) Inferring pathways and networks with a Bayesian
framework. FASEB J 18, 746-748

Husmeier, D. (2003) Sensitivity and specificity of inferring genetic regulatory
interactions fiom microarray experiments with dynamic Bayesian networks.
Bioinformatics 19, 2271-2282

Yu, J., Smith, V. A., Wang, P. P., Hartemink, A. J., and Jarvis, E. D. (2004)
Advances to Bayesian network inference for generating causal networks from
observational biological data. Bioinformatics 20, 3594-3603

Bar-Joseph, Z., Gerber, G. K., Lee, T. 1., Rinaldi, N. J., Yoo, J. Y., Robert, R,
Gordon, D. B., Fraenkel, B., Jaakkola, T. 8., Young, R. A., and Gifford, D. K.
(2003) Computational discovery of gene modules and regulatory networks. Nat
Biotechnol 21, 1337-1342

19

 

__'L

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

Berman, B. P., Nibu, Y., Pfeiffer, B. D., T omancak, P., Celniker, S. E., Levine, M.,
Rubin, G. M., and Eisen, M. B. (2002) Exploiting transcription factor binding site
clustering to identify cis-regulatory modules involved in pattern formation in the
Drosophila genome. Proc Natl Acad Sci U S A 99, 757-762

Hartemink, A. J., Gifford, D. K., Jaakkola, T. S., and Young, R. A. (2002)
Combining location and expression data for principled discovery of genetic
regulatory network models. Pac Symp Biocomput, 43 7-449

Ideker, T., Thorsson, V., Ranish, J. A., Christmas, R., Buhler, J., Eng, J. K.,
Bumgarner, R., Goodlett, D. R., Aebersold, R., and Hood, L. (2001) Integrated

genomic and proteomic analyses of a systematically perturbed metabolic network.
Science 292, 929-934

Pilpel, Y., Sudarsanam, P., and Church, G. M. (2001) Identifying regulatory
networks by combinatorial analysis of promoter elements. Nat Genet 29, 153-159

Li, F ., and Yang, Y. (2004) Recovering genetic regulatory networks fiom micro-
array data and location analysis data. Genome Inform 15, 131-140

Patti, M. B., and Kahn, C. R. (1998) The insulin receptor--a critical link in glucose
homeostasis and insulin action. J Basic Clin Physiol Pharmacol 9, 89-109

Thirone, A. C., Huang, C., and Klip, A. (2006) Tissue-specific roles of IRS proteins
in insulin signaling and glucose transport. Trends Endocrinol Metab 17, 72-78

White, M. F. (1998) The IRS-signalling system: a network of docking proteins that
mediate insulin action. Mol Cell Biochem 182, 3-11

Lowenstein, E. J., Daly, R. J., Batzer, A. G., Li, W., Margolis, B., Lammers, R.,
Ullrich, A., Skolnik, E. Y., Bar-Sagi, D., and Schlessinger, J. (1992) The SH2 and
SH3 domain-containing protein GRB2 links receptor tyrosine kinases to ras
signaling. Cell 70, 431-442

Skolnik, E. Y., Lee, C. H., Batzer, A., Vicentini, L. M., Zhou, M., Daly, R., Myers,
M. J., Jr., Backer, J. M., Ullrich, A., White, M. F., and et a1. (1993) The SH2/SH3
domain-containing protein GRB2 interacts with tyrosine-phosphorylated IRS1 and
She: implications for insulin control of ras signalling. EMBO J 12, 1929-1936

Backer, J. M., Schroeder, G. G., Kahn, C. R., Myers, M. G., Jr., Wilden, P. A.,
Cahill, D. A., and White, M. F. (1992) Insulin stimulation of phosphatidylinositol
3-kinase activity maps to insulin receptor regions required for endogenous substrate
phosphorylation. J Biol Chem 267, 1367-1374

20

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

Alessi, D. R., James, S. R., Downes, C. P., Holmes, A. B., Gaffney, P. R., Reese, C.
B., and Cohen, P. (1997) Characterization of a 3-phosphoinositide-dependent
protein kinase which phosphorylates and activates protein kinase Balpha. Curr Biol
7, 261-269

Standaert, M. L., Galloway, L., Karnarn, P., Bandyopadhyay, G., Moscat, J., and
Farese, R. V. (1997) Protein kinase C-zeta as a downstream effector of
phosphatidylinositol 3-kinase during insulin stimulation in rat adipocytes. Potential
role in glucose transport. J Biol Chem 272, 30075-30082

Le Roith, D., and Zick, Y. (2001) Recent advances in our understanding of insulin
action and insulin resistance. Diabetes Care 24, 588-597

Kahn, B. B. (1998) Type 2 diabetes: when insulin secretion fails to compensate for
insulin resistance. Cell 92, 593-596

Kahn, S. E., Hull, R. L., and Utzschneider, K. M. (2006) Mechanisms linking
obesity to insulin resistance and type 2 diabetes. Nature 444, 840-846

Reaven, G. M. (1988) Banting lecture 1988. Role of insulin resistance in human
disease. Diabetes 37, 1595-1607

Sone, H., Suzuki, H., Takahashi, A., and Yarnada, N. (2001) Disease model:
hyperinsulinernia and insulin resistance. Part A-targeted disruption of insulin
signaling or glucose transport. Trends Mol Med 7, 320-322

Irnamura, T., Takata, Y., Sasaoka, T., Takada, Y., Morioka, H., Haruta, T., Sawa,
T., Iwanishi, M., Hu, Y. G., Suzuki, Y., and et a1. (1994) Two naturally occurring
mutations in the kinase domain of insulin receptor accelerate degradation of the
insulin receptor and impair the kinase activity. J Biol Chem 269, 31019-31027

McElduff, A., Hedo, J. A., Taylor, S. I., Roth, J., and Gorden, P. (1984) Insulin
receptor degradation is accelerated in cultured lymphocytes fi'om patients with
genetic syndromes of extreme insulin resistance. J Clin Invest 74, 1366-1374

Zick, Y. (2005) Ser/Thr phosphorylation of IRS proteins: a molecular basis for
insulin resistance. Sci SYKE 2005, pe4

Herschkovitz, A., Liu, Y. F., Ilan, E., Ronen, D., Boura-Halfon, S., and Zick, Y.
(2007) Common inhibitory serine sites phosphorylated by IRS-1 kinases, triggered
by insulin and inducers of insulin resistance. J Biol Chem 282, 18018-18027

21

102.

103.

104.

105.

106.

107.

 

 

Taylor, S. I., and Arioglu, E. (1998) Syndromes associated with insulin resistance
and acanthosis nigricans. J Basic Clin Physiol Pharmacol 9, 419-439

Paz, K., Henri, R., LeRoith, D., Karasik, A., Elhanany, E., Kanety, H., and Zick, Y.
(1997) A molecular basis for insulin resistance. Elevated serine/threonine
phosphorylation of IRS-1 and IRS-2 inhibits their binding to the juxtamembrane
region of the insulin receptor and impairs their ability to undergo insulin-induced
tyrosine phosphorylation. J Biol Chem 272, 29911-29918

Hotarrrisligil, G. S., Peraldi, P., Budavari, A., Ellis, R., White, M. F., and
Spiegehnan, B. M. (1996) IRS-l-mediated inhibition of insulin receptor tyrosine
kinase activity in TNF-alpha- and obesity-induced insulin resistance. Science 271,
665-668

Pederson, T. M., Kramer, D. L., and Rondinone, C. M. (2001) Serine/threonine
phosphorylation of IRS-1 triggers its degradation: possible regulation by tyrosine
phosphorylation. Diabetes 50, 24-31

Hirosumi, J., Tuncman, G., Chang, L., Gorgun, C. 2., Uysal, K. T., Maeda, K.,
Karin, M., and Hotamisligil, G. S. (2002) A central role for JNK in obesity and
insulin resistance. Nature 420, 333-336

Eley, H. L., Russell, S. T., and Tisdale, M. J. (2008) Role of the dsRNA-dependent
protein kinase (PKR) in the attenuation of protein loss from muscle by insulin and
insulin-like growth factor-1(IGF-I). Mol Cell Biochem

22

CHAPTER 2: Repression of PKR Mediates Palmitate-Induced
Apoptosis in HepGZ Cells through BcI-2

The work in this chapter has been published in Cell Research:
Yang, X. and Chan, C. (2009) Repression of PKR mediates palmitate-
induced apoptosis in HepGZ cells through regulation of Bel-2. Cell Research 19,

469-486

Abstract

The study in this chapter shows that the double-stranded RNA-dependent protein
kinase (PKR) regulates the protein expression level and phosphorylation of Bcl-2 and
exploits an anti-apoptotic role in human hepatocellular carcinoma cells (HepG2). In
various types of cells, saturated fiee fatty acids (FF As), such as palrrritate, have been
shown to induce cellular apoptosis by several mechanisms. Palmitate down-regulates the
activity of PKR and thereby decreases the level of Bcl-2 protein, mediated in part by the
NF-KB transcription factor. In addition to the level of Bel-2 protein, the phosphorylation
of Bcl-2 at different amino acid residues, such as Ser70 and Ser87, is also important in
regulating cellular apoptosis. The decrease in the phosphorylation of Bcl-2 at Ser70 upon
exposure to palmitate is mediated by PKR and possibly JN K, whereas the
phosphorylation of Bel-2 at Ser87 is unaffected by palmitate or PKR. In summary, PKR
mediates the regulation of the protein level and the phosphorylation status of Bel-2,

providing a novel mechanism of palmitate-induced apoptosis in HepG2 cells.

23

Introduction

The double-stranded RNA-dependent protein kinase (PKR), activated by foreign
double-stranded RNA (dsRNA) (l , 2), is best known for triggering cell defense responses
by inhibiting general protein synthesis and inducing cellular apoptosis in many types of
eukaryotic cells during viral infection (2, 3). However, evidence is emerging, albeit
controversial at this point, suggesting that PKR also has an anti-apoptotic role in mouse
embryo fibroblasts and certain tumor cells (4-8). As a Ser/Thr protein kinase, PKR is also
known for its role in mediating signaling pathways (9) by interacting with proteins such
as NF-KB, MAPKs, and PP2A (10-13). PKR phosphorylates and thereby releases I-KB
from NF-ch, thus activating NF-ch and promoting the translocation of NF-KB into the
nucleus (10, 11). Upon activation of NF-KB by PKR, the transcription of a number of
apoptosis-regulating genes, such as FasL (14), p53 (15), and cIAPs (4), has been shown
to be up-regulated. PKR plays a crucial role in the phosphorylation of the 3 mitogen-
activated protein kinases (MAPK) (JNK, ERK, and p3 8) upon ribotoxic stress (12).
However, the mechanism by which PKR interacts with the major MAPKs, as well as
whether its role is apoptotic or anti-apoptotic, is unclear. PKR can also phosphorylate the
PP2A regulatory subunit B56a and thus activate the catalytic subunit of PP2A,
potentially leading to the dephosphorylation of eIF—4E and the arrest of translation (13).

In addition to dsRNA, PKR has been shown to be responsive to many other
factors (e.g., endoplasmic reticulum (ER) stress (16, 17), cytokines such as tumor
necrosis factor (TNF)-or (18) and interleukin (IL)-1 (19), deoxynivalenol (DON, or
vomitoxin) (12) and lipopolysaccharide (19)). The present study shows for the first time

that palmitate down-regulates the activity of PKR in HepG2 cells.

24

Saturated FFAs (e.g., palmitate) induce apoptosis in many cell types, such as
cardiac cells, pancreatic beta cells, breast cancer cells, and hepatocytes. They are
associated with the development of a variety of diseases, including diabetes, heart disease,
and non alcoholic fatty liver disease (NAFLD) (20, 21). The mechanism by which
palmitate induces apoptosis depends on the cell type. For example, palmitate induces
apoptosis by generating intracellular reactive oxygen species (ROS) in microvascular
endothelial cells (EC) and retinal pericytes (22), but not in neonatal rat cardiomyocytes
(23). Studies of palmitate-induced apoptosis in liver cells have focused predominantly on
Iysosomal permeabilization (24), intracellular metabolic pathways such as beta oxidation
(25), TG accumulation (26, 27), and ceramide production (28, 29). In addition, increased
hydrogen peroxide (H202) and hydroxyl (*OH) radicals mediated the palmitate-induced
lipotoxicity in the human hepatocellular carcinoma (HepG2/C3A) cell line (30). However,
cytotoxicity, as well as apoptosis, was not completely prevented upon treatment with
mitochondrial complex inhibitors or free radical scavengers, suggesting that mechanisms
other than ROS production in mitochondria contribute to the toxicity of palmitate.

More recent investigations implicate certain Bel-2 family proteins in mediating
the saturated FFA-induced apoptosis of liver cells. For example, palmitate-induced
apoptosis in liver cells is related to the activation (24, 31) of Bax and a decrease in the
Bax antagonist Bcl-X(L) (24). In addition, elevated Bcl-2-interacting mediator of cell
death (Bim), a pro-apoptotic Bel-2 family protein, plays a role in stearic and palmitic
acid-induced apoptosis of several liver cell lines including HepG2 (31, 32). This process
depends on the transcription factor F oxO3a (31). In addition to these pro-apoptotic Bel-2

family proteins, an anti-apoptotic member, Bel-2, has been identified as an important

25

factor in regulating the apoptosis of HepG2 cells (33-35). In pancreatic cells, the
induction of apoptosis by palmitate is associated with reduced anti-apoptotic Bcl-2 levels
(36). We previously showed that palmitate also decreased the protein level of Bel-2 in
HepG2 cells (3 7). The present study confirms a similar effect of palmitate on the levels of
Bcl-2 and provides a potential mechanism; furthermore, it shows that palnritate also
down-regulates the phosphorylation of Bel-2 at Ser70, but not at Ser87, in HepG2 cells.

As one of the most important anti-apoptotic members in the Bel-2 family, Bcl-2
protects cells against intrinsic apoptosis by maintaining the integrity of the mitochondrial
membrane (3 8). The expression of the Bcl-2 gene is regulated by different transcription
factors depending on the cell type (39-41). In liver tumor cell lines such as U937 and
HepG2 cells, NF-ch has been identified as the central regulator of the transcription of the
Bcl-2 gene (42, 43). Bel-2 is expressed in the progenitor cells of several self-renewmg
tissues and some tumor cells, including human hepatocellular carcinoma (HepG2) cells
(44-47). Irnmunohistochemical studies indicate that Bel-2 is not expressed in primary
hepatocytes (44), although a more recent study with “hi gh-power’ ’ staining shows Bel-2
expression in primary hepatocytes (48). In the present study, I confirm by both RT-PCR
and western blotting that HepG2 cells express Bel-2, and I propose a pathway through
which palmitate regulates the Bcl-2 protein in HepG2 cells.

Post-translational modification (e.g., phosphorylation) of Bcl-2 also plays a role
in determining the anti-apoptotic role of Bcl-2 (49). Phosphorylation of Bel-2 at the anti-
apoptotic site, Ser70, sustains the anti-apoptotic role of Bcl-2 (50). On the other hand,
phosphorylation of Bcl-2 at SerS7 is believed to reduce the anti-apoptotic fiinction of

Bel-2, possibly by inhibiting the phosphorylation of Bel-2 at Ser70 or destabilizing the

26

Bcl-2 protein (51, 52). MAPKs have been proposed to mediate the phosphorylation of
Bel-2 (53-57) because the sequences surrounding both Ser70 and Ser87 residues of the
Bel-2 protein represent the consensus motif, X-X-S-P, recognized by MAPKs (57). PP2A,
a Ser/Thr-specific protein phosphatase, has been shown to dephosphorylate Bcl-2 at both
Ser70 (58) and Ser87 (52) residues.

The data in the present study support the action of PKR as an anti-apoptotic factor
and demonstrate that PKR is involved in regulating the protein level and phosphorylation
of Bcl-2 in HepG2 cells. There has been no evidence in the literature to date indicating
whether palrrritate has an effect on the activity of PKR. I show that palmitate down-
regulates the activity of PKR, which, however, does not alter the phosphorylation level of
eIF-Za, and I propose that the repression of PKR mediates apoptosis through the

regulation of Bel-2.

Materials and Methods

Cell Culture and Reagents—Human hepatocellular carcinoma cells (HepG2) were
cultured in Dulbecco's Modified Eagle Medium (DMEM) (Invitrogen, Carlsbad, CA)
with 10% fetal bovine serum (FBS) (Biomeda Corp, Foster City, CA) and penicillin-
streptomycin (penicillin: 10,000 U/ml, streptomycin: 10,000 rig/ml) (Invitrogen, Carlsbad,
CA). Freshly trypsinized HepG2 cells were suspended at 5 X 105 cells/ml in standard
HepGZ culture medium and seeded at a density of 106 cells per well in standard six-well
tissue culture plates. After being seeded, the cells were incubated at 37°C in a 90%
air/10% CO2 atmosphere, and two milliliters of fresh medium was supplied every other

day to the cultures after removal of the supernatant. The HepG2 cells were cultured in

27

 

standard medium for 5-6 days to achieve 90% confluence before treatment with FF As or
other additives. HepGZ cell number was assessed by trypan blue dye exclusion using a
hematocytometer. Phosphate-buffered saline (PBS) was purchased from Invitrogen, poly-
(I:C) and trypan blue from Sigrna-Aldrich (St. Louis, MO), and NF-KB SN50 and its
inactive control, NF-KB SNSOM, PKR inhibitor, JNK inhibitor (SP600125) and their
analogues, used as negative controls, from EMD Biosciences (San Diego, CA).

Fatty Acid Salt Treatment—Sodium salts of palmitate (P9767) and oleate (07501)
were purchased fi'om Sigma-Aldrich. Palrrritate and oleate were complexed to 0.7 mM
bovine serum albumin (BSA, fatty acid free) dissolved in the medium, which mirrrics the
physiological concentration of albumin in human blood (3.5-5%, (59)). Fatty acid free
BSA was purchased from MP Biomedicals (Chillicothe, OH). Dose responses of
palnritate at 0, 0.2, 0.4, and 0.7 mM were performed in most of the experiments. If only
one concentration was reported, the concentration of palmitate or oleate was 0.7 mM. In
all experiments, FFAs were given for 24 hours, and the vehicle (0.7 mM BSA) was used
as the control.

RNA Interference for PKR and Reverse T ransfection—Silencer® Validated siRNA
targeting human PKR mRN A was purchased fi'om Ambion (Austin, TX). The
synthesized oligonucleotides for siRNA are 5'-GGUGAAGGUAGAUCAAAGATT-3‘
and 5'-UCUUUGAUCUACCUUCACCTI‘-3'. Reverse transfection of siRNA was
performed. In general, the scrambled non-targeting siRNA as a negative control or the
siRNA targeting PKR was diluted in serum and antibiotic free Opti-MEM (Invitrogen)
and then mixed with the transfection reagent, Lipofectamine RNAiMAX (Invitrogen).

The mixture of siRNA and Lipofectamine RNAiMAX in Opti-MEM was then added into

28

6-well plates and incubated at room temperature for 20 minutes. HepG2 cells suspended
in antibiotic fi'ee medium were counted and plated into the 6-well plates at the same cell-
number per well. The cells were then incubated at 37°C for 24 hours. After transfection,
the cells were incubated in regular medium for another 24 hours and then collected. The
mRNA and protein levels of PKR were measured by RT-PCR and western blot analysis,
respectively. Titration of the siRNA and the transfection reagent was performed (not
shown), and the lowest working amounts of the siRN A and the transfection reagent were
applied in the loss-of-function (LOF) experiments in the present study.

Over-expression of PKR and Forward T ransfection—The PKR plasmid, pCMV6-
XLS-hPKR, and the empty vector, pCMV6-XL5, were purchased from Origene
(Rockville, MD). Transient transfection was performed according to the Lipofectamine
2000 (Invitrogen) method. In general, regular HepG2 cells (Fig. 2.5C and 2.8B), or the
cells in which the PKR geneiwas silenced by the siRNA of PKR (Fig. 2.36), were
washed twice with phosphate buffered saline, and the medium was replaced with 2 ml of
Opti-MEM with 1% FBS. Two micrograms per well of pCMV6-XL5-hPKR or the empty
vector pCMV6-XL5 was then mixed with 10 ul/well of Lipofectamine 2000 in Opti-
MEM and 20 minutes later the mixture was added to the wells. After 6 hours of
transfection, the cells were then cultured in regular medium for 48 hours and
subsequently harvested (Fig. 2.3C) or treated with palmitate (Fig. 2.5C and 2.8B).

Cytotoxicity Measurement—HepG2 cells were cultured in different media for 24
hours and the supematants were collected. Cells were washed with PBS and kept in 1%
triton-X-lOO in PBS for 24 hours at 37°C. The cell lysate was then collected, vortexed for

15 seconds, and centrifuged at 7000 rpm for 5 minutes. A cytotoxicity detection kit

29

(Roche Applied Science, Indianapolis, IN) was used to measure the LDH levels in the
supematants and in the cell lysates. The fraction of LDH released into the medium was
normalized to the total LDH (LDH released into the medium + LDH remaining in the cell
lysates) (30).

DNA Fragmentation—Treated HepG2 cells were lysed, and the DNA was
extracted using the DNA purification kit from Promega (Madison, WI). Two rrricrograms
of DNA samples was analyzed by electrophoresis on 1.5% agarose gels and visualized by
SYBR gold staining for 4 hours.

Caspase Analysis—For the caspase-3 substrate cleavage assay, the cells were
washed with PBS, lysed, and assayed in a 96-well plate using the Caspase-3 cellular
assay kit (Biomol, Plymouth Meeting, PA). Fluorescence was measured at emission and
excitation settings of 360 and 460 nm, respectively, with a Microplate
Spectrofluorometer from Molecular Device (Sunnyvale, CA). The caspase-3 activities
were normalized by relative beta-actin levels.

Nuclear Extraction and Detection of Nuclear NF -ch levels—Nuclear extracts
from HepG2 cells were prepared using the Nuclear/Cytosol Fraction Kit from BioVision
(Mountain View, CA). The extracted nuclear and cytoplasmic protein fractions were
subjected to western blot analysis with anti-NF-kB p65 and anti-TBP, as a loading
control for the nuclear extracts, and anti-beta actin for the cytoplasmic fractions.

Western Blot Analysis and Immunoprecipitation—The HepG2 cells were washed
twice with cold PBS and lysed in 300 uUwell of CelLytic M cell lysis buffer (Sigrna-
Aldrich) supplemented with protease inhibitor cocktail (Roche Applied Science,

Indianapolis, IN) and Ser/Thr phosphatase inhibitor cocktail (Sigrna-Aldrich). The cell

30

lysate was clarified by centrifugation at 10,000 rpm for 15 minutes, and the supernatant
was collected. Total protein levels were quantified by BCA assay kit from Pierce Inc
(Rockford, IL). A total of 20-40 ug of total protein was resolved by SDS-PAGE gels
fi'om Bio-Rad, transferred to nitrocellulose membranes, and probed with primary and
secondary antibodies. Biotinylated protein ladders (Cell Signaling, Beverly, MA) were
loaded to one well of each SDS-PAGE gel, and anti-biotin antibody was used to detect
the protein ladders on the western blots. Antibody detection was performed using the
enhanced chemiluminescence kit from Pierce Biotechnology and imaged on the
Molecular Imager ChemiDoc XRS System from Bio-Rad. For irnmunoprecipitation, the
cell lysates were incubated with appropriate primary antibodies at 4°C for 1-2 hours, and
the immunocomplexes were precipitated in a mixture with protein A affinity gel (Sigrna-
Aldrich) by overnight incubation. The irnmunoprecipitates were washed three times with
the cell lysis buffer and boiled in SDS-PAGE sample buffer, and the immune complexes
were analyzed by western blot analysis. Phospho site-specific anti-eIF2a (SerSl) and
JNK (T183/Y 185), anti-eIFZa, anti-PKR, anti-NF-kB p65, and anti-JNK rabbit
polyclonal antibodies were purchased from Cell Signaling (Beverly, MA), phospho site-
specific anti-PPZA/C (T yr307) from Abcarn (Cambridge, MA), phospho site-specific
anti-Bcl-Z (Ser70) from Upstate (Charlottesville, VA), phospho site-specific anti-phospho
PKR (Thr451) and Bcl-2 (Ser87) polyclonal antibodies from EMD Biosciences, and anti-
Bcl-2, anti-Birn, anti-TBP, and anti-beta actin antibodies from Sigma-Aldrich. Anti-
PP2A-B56a and anti- goat secondary antibodies were purchased fi'om Santa Cruz
Biotechnology (Santa Cruz, CA). Secondary anti-rabbit and anti-mouse antibodies were

purchased from Pierce Biotechnology Inc.

31

Real-time Quantitative RT -PCR Analysis—Total RNA was extracted from cells
with the RNeasy mini kit (Qiagen, Valencia, CA) and depleted of contaminating DNA
with RNase-fiee DNase (Qiagen). Equal amounts of total RNA (1 pg) were reverse-
transcribed using an iScript cDNA synthesis kit (Bio-RAD). The first-strand cDNA was
used as a template. The primers used for quantitative RT-PCR analyses of human PKR
(5'-CCT GTCCT CT GG’I'I‘CTTIT GCT-3' and 5'-GATGATI‘CAGAAGCGAGTGTGC-3')
(60), human Bcl-2 (5'-ACATCGCCCTGTGGATGACT-3' and 5'-
TCAC'I'I‘GTGGCCCAGATAGG—3'), and human GAPDH (5'-
AACT'I‘TGGTATCGTGGAAGGA-3' and 5'-CAGTAGAGGCAGGGATGATGT-3')
were synthesized by Operon Biotechnologies, Inc. (Huntsville, AL). RT-PCR was
performed in 25-ul reactions using 1/ 10 of the cDNA produced by reverse transcription,
0.2 uM of each primer, 1 X SYBR green superrnix from Bio-RAD, and an annealing
temperature of 60 °C for 40 cycles. Each sample was assayed in three independent RT
reactions and triplicate reactions were performed and normalized to the GAPDH
expression levels. Negative controls included the absence of enzyme in the RT reaction
and the absence of template during PCR. The cycle threshold (CT) values, corresponding
to the PCR cycle number at which the fluorescence emission in real time reaches a
threshold above the baseline emission, were determined using MyIQ1M Real-Time PCR
Detection System (Bio-RAD).

Statistical Analysis—All experiments were performed at least three times, and
representative results are shown. All data, unless specified, are shown as the mean :t SD.

for the indicated number of experiments. One-way AN OVA with Tukey’s post hoc

32

method was used to evaluate statistical significances between different treatment groups.

Statistical significance was set at p<0.01.

 

 

 

 

A B.
30% * 10
g 25% a} 3 8
8 20% '3 w
a “ 5 6
'3 15% E '5
Z a I 2 4
3 10A: * a 2‘3
-' 5% .3 2
0% 0
0 0.2 0.4 0.7 Oleate 0 0.2 0.4 0.7 Oleate
Palmitate Concentration (mM) Palmitate Concentration (mM)

Figure 2.1. Effects of palmitate on the cytotoxicity and apoptosis of HepG2 cells.
HepG2 cells were exposed to different levels of palmitate or 0.7 mM oleate for 24 hours.
The vehicle for the FFAs (0.7 mM BSA) was used as the control (i.e., regular medium
with BSA) (A, B). LDH release (A) and caspase-3 activity (B) were measured after
treatment with palmitate. Data are expressed as the averages of nine samples i SD from
three independent experiments. One-way ANOVA with Tukey’s post hoc method was
used to analyze the differences between the treatment groups. *, significantly higher
than control (i.e., regular medium with BSA), p<0.01.
Results

Palmitate Induces Cytotoxicity and Apoptosis of HepGZ Cells—Previous work
in our lab showed that palmitate induced cytotoxicity in HepG2 cells (30, 37), whereas
unsaturated FFAs (e.g., oleate and linoleate) were not cytotoxic (30, 37). In a separate
study, we found that upon exposure to palmitate, the HepG2 cells stained for Annexin-V
(data not shown), which indicates phosphatidylserine extemalizatiorr, a sign of early stage
apoptosis. In the present study, I further found that palmitate increased LDH release and
caspase-3 activity of HepG2 cells in a dose-dependent manner, whereas oleate did not

have a significant effect on LDH release and caspase-3 activity (Fig. 2.1 A and B),

supporting the idea that palmitate induces cytotoxicity and apoptosis in HepG2 cells. It

33

 

has also been confirmed in the literature that palmitate induces apoptosis of HepG2 cells

 

 

 

(31,32).
A.
we: E
1. , n. 1
p-PKRThr451—V-z p. an... L” 32"»
. .. . = g
'3 o
PKR hunch—”w 220.5
E Q
beta Actin w—va— n.-
a.

 

 

 

 

Palmitate (mM) 0 0.2 0.4 0.7 0 0,2 0.4 0-7

Palmitate Concentration (mM)

 

     

B.
we: g 1-2
”"1 I ‘L 1
'5 A
p"°KRT'"4~"1- § § 0.8
T. a
,5 ,3; 0.4
betaActin y.» n: 2;
E 0.2
Control Palm. Ole. a 0

 

Control Palm. Ole.

Figure 2.2. Effects of palmitate and oleate on the activity of PKR. HepG2 cells were
exposed to different levels of palmitate (A) or 0.7 mM palmitate or oleate (B) for 24
hours. The vehicle for the FFAs (0.7 mM BSA) was used as the control (i.e., regular
medium with BSA), in which the concentration of FFAs was 0 (A, B). Afier treatment,
the cells were harvested and western blot analysis was performed to detect the
phosphorylated level of PKR. The level of p-PKR Thr451 was quantified by normalizing
to the levels of total PKR and is expressed as the average of three samples 3: SD from
three independent experiments. One-way ANOVA with Tukey’s post hoc method was
used to analyze the differences between the treatment groups. *, significantly lower than
control (i.e., regular medium with BSA), p<0.01.

Palmitate Decreases the Activity of PKR in HepG2 Cells—Palmitate decreased
the phosphorylation of PKR at Thr451, which indicates the activity of PKR, in a dose-
dependent manner (Fig. 2.2A), whereas the unsaturated FFA oleate did not significantly
affect the phosphorylation of PKR (Fig. 2.28) in HepG2 cells. Considering the apoptotic

effects of palrrritate on HepG2 cells (Fig. 2.1B), I hypothesize that in HepG2 cells PKR is

34

involved in mediating the apoptosis induced by palmitate. To uncover the role of PKR in
regulating apoptosis, gene silencing and gene over-expression studies were performed.
PKR is Anti-apoptotic in HepGZ Cells—The siRNA targeting PKR that was
employed in the present study markedly inhibited the gene and protein expression of
PKR and thereby reduced the level of phosphorylated PKR (Fig. 2.3A). Silencing PKR
with this siRNA increased the activity of caspase-3 significantly (Fig. 2.33) but did not
increase the release of LDH (Fig. 2.3B), and it also induced the fragmentation of
chromatin DNA (Fig. 2.33), suggesting that PKR has an anti-apoptotic role in HepG2
cells. To confirm the role of PKR in apoptosis, I over-expressed and rescued the PKR
expression level in PKR-silenced cells and found that the caspase-3 activity was reduced
to levels close to that of the control (Fig. 2.3 C). Taken together, these results suggest that
PKR plays an anti-apoptotic role in HepG2 cells. To further confirm a catalytic role of
PKR in regulating apoptosis, I inhibited the activity of PKR with a pharmaceutical
inhibitor of PKR (61-63) and found that, similar to the siRNA of PKR, the PKR inhibitor
also induced apoptosis in HepG2 cells, as evidenced by caspase-3 activity and DNA
fragmentation (not shown). Considering the negative effect of palmitate on the activity of

PKR, I therefore proposed that palmitate induces apoptosis, in part by repressing PKR.

35

Figure 2.3. Role of PKR in the cytotoxicity and apoptosis of HepG2 cells. Reverse
transfection of suspended HepG2 cells was performed with scrambled siRNA (Control)
or siRNA of PKR for 24 hours, and the transfected cells were cultured in regular medium
for another 24 hours (A, B). Cells were then harvested, and RT-PCR and western blot
analysis were performed to detect the gene, protein and phosphorylation levels of PKR to
confirm that the PKR gene was silenced and the activity of PKR was suppressed (A).
LDH release, Caspase—3 activity, and DNA fragmentation were assayed (B). In (C),
reverse transfection of scramble siRNA (I, control) or siRNA of PKR (E, siPKR) was
performed followed by forward transfection of empty vector pCMV6-XL5 (pCMV) or
the plasmid containing PKR cDNA sequence (hPKR). Cells were then harvested and
caspase-3 activity was assayed (C). Data are expressed as the average of three (A) or nine
(B, C) samples i SD from three independent experiments. One-way ANOVA with
Tukey’s post hoc method was used to analyze the differences between the treatment
groups. * , significantly higher (B, C) or lower (A, C), than control, p<0. 01. #,
significantly lower than siPKR-CMV in Fig. C, p<0. 01.

36

 

 

 

2

1

._ a. a. ._._
0 o o o
.358 28.

:otafinxm 0:00 ”in.

   

PKR

beta Actin

p-PKR Thr451 —>

*
— A
2 0

siPKR

Ctrl

siPKR

Ctrl

siPKR

Clrl

 

 

 

 

. _

1 5

5-

1. o.
$955 22.
33:3. @8380

.
0

 

 

 

15%

10% ~
%

Se 883. In...

siPKR

Ctrl

siPKR

Chi

 

 

 

IControl

 

1 5.
0

3955 En:

22.2 983.8

1.5

pCMVG hPKR

hPKR

pCMV6

37

 

1.2

0.8

BCI-2 0.6 *

 

0.4

Palmitate(mM) o 0.2 0.4 0.7 0,

801-2 Protein Level
(fold change)

0 0.2 0.4 0.7
Palmitate Concentration (mM)

 

0.8
0.6
0.4
0.2

WB:
beta Actin \_ ~

Ctrl Palm. Ole.

Bel-2 Protein Level
(fold change)

 

Control Palm. Ole.

Figure 2.4. Effects of palmitate and oleate on the protein level of Bcl-2. HepG2 cells
were exposed to different levels of palmitate (A) or 0.7 mM palmitate or 0.7 mM oleate
(B) for 24 hours. The vehicle for the FFAs (0.7 mM BSA) was used as the control (i.e.,
regular medium with BSA), in which the concentration of FFAs was 0 (A, B). After
treatment, the cells were harvested and western blot analysis was performed to detect the
protein level of Bel-2. Bel-2 protein levels were quantified by normalizing to the beta
actin levels and are expressed as the average of three samples i SD from three
independent experiments. One-way ANOVA with Tukey’s post hoc method was used to
analyze the differences between the treatment groups. *, significantly lower than control
(i.e., regular medium with BSA), p<0.01.

Palmitate Down-regulates the Protein Expression Level of Bcl-2—Palmitate
decreased the protein level of Bcl-2 (Fig. 2.4A). Oleate did not have a significant effect
on the protein level of Bel-2 (Fig. 2.43). It was unclear from the literature how palmitate
would regulate the Bel-2 protein. However, as illustrated in Figs. 2.2 and 2.4, palmitate

concomitantly decreased the phosphorylation of PKR and the protein level of Bel-2,

suggesting a potential association between PKR and Bel-2. To confirm this association

38

and test the involvement of PKR in mediating the effect of palmitate on Bel-2, gene
silencing and over-expression of PKR were performed.

PKR is Involved in Mediating the Effects of Palmitate on the Protein Level of
Bel-2 in HepGZ Cells—Silencing the expression of the PKR gene with siRNA of PKR
(Fig. 2.3A) down-regulated the mRNA and protein levels of Bel-2 (Fig. 2.5 A and B),
suggesting that PKR positively regulates the mRNA expression (Fig. 2.5A) and, in turn,
the protein level (Fig. 2.53) of Bel-2. Similarly, inhibiting the activity of PKR with the
PKR inhibitor also suppressed the mRN A and protein levels of Bel-2 (not shown).
Considering the repressive effect of palmitate on PKR activity (Fig. 2.2), I hypothesize
that the suppression of PKR activity mediates the negative effects of palmitate on Bel-2
levels. To confirm my hypothesis that PKR is involved in the decrease of Bel-2 by
palnritate, I over-expressed the PKR gene in HepG2 cells (Fig. 2.56) and exposed the
cells to palmitate. I indeed found that the Bel-2 level was rescued by over-expression of
PKR (Fig. 2.5C). This suggests that the suppression of PKR activity indeed mediates the

decrease of Bcl-2 induced by palmitate.

39

Figure 2.5. Involvement of PKR in regulating the protein level of Bcl-2. Reverse
transfection of suspended HepG2 cells was performed with scrambled siRNA (Control)
or siRNA of PKR for 24 hours and the transfected cells were cultured in regular medium
for another 24 hours (A, B). Cells were harvested, and RT-PCR (A) and western blot
analysis (B) were performed to detect the gene (A) and protein (B) expression levels of
PKR (I, also shown in Fig. 2.3C) and Bcl-2 (a). In HepG2 cells, the forward transfection
of empty vector pCMV6-XL5 (pCMV6) or the plasmid containing PKR cDNA sequence
(pCMV6-hPKR) was performed, and the cells were then treated with different
concentrations of palmitate for 24 hours (C). The vehicle for palmitate (0.7 mM BSA)
was used as the control (i.e., regular medium with BSA), in which the concentration of
palmitate was 0 (C). Cells were then harvested and western blot analysis was performed
to detect the phosphorylation and protein levels of PKR and the protein levels of Bel-2
(C). Gene expression data are expressed as the average of nine samples :1: SD from three
independent experiments. The protein levels of Bel-2 were quantified by normalizing to
beta actin and are expressed as the average of three samples :l: SD from three independent
experiments. One-way ANOVA with Tukey’s post hoc method was used for analyzing
the differences between the treatment groups. * and #, PKR and Bel-2 levels,
respectively, significantly lower than control, p<0.01.

40

>

 

,6 1.2 3 12 ..
g 1 3 IPKR IPKR
.: 1 UBcl-2 1
o . ,, , .: DBcl—Z.
E 0-8 3 0.8
‘c’ 0 . # beta Actin WM 3% ,
g -6 - a, 0.6 #
s
3 0.4 Control siPKR g 04
a. E _|
" ' r: ‘k
l: 0.2 . * g 0.2
5 ; e
c o n. 0
cm siPKR cm siPKR

.0

WB:

mpam_~aaq

betaActin [wwmeW ,.. I

 

 

 

 

 

p-PKR Thr451

 

 

 

PKR

 

Palmitate (mM) 0 0.2 0.4 0.7 0 0.2 0.4 0.7
pCMV6 pCMV6-hPKR

 

 

3.5
3

N
r001

(fold change)
2;

Bel-2 protein level

—A

0.5

 

 

0
Palmitate(mM) 0 0.2 0.4 0.7 0 0.2 0.4 0.7
pCMV6 pCMV6-hPKR

41

A. B.

Nuclear extract

NF-KB p65 w w ' ‘ 4 Bcl-2 «OI—v” “M

beta Actin ~ w m

 

 

 

 

 

 

 

 

 

TBP h u _ '“' “-

 

l

 

 

Ctrl SN50M SN50

 

Cytoplasmic fraction
Bel-2 “WW ”4

Nuclear extract

 

 

 

NF-KB p65 ~ H "I

TBP P" hi-
beta ACti" W Ctrl Ole. Palm.

pCMV6 hPKR pCMV6 hPKR
Control siPKR

 

 

 

NF-ch p65 W w v u.-

 

 

 

 

 

 

 

Figure 2.6. Role of PKR in regulating the Nuclear NF-KB p65 level. Reverse
transfection of scramble siRNA (control, the first two lanes) or siRNA of PKR (siPKR,
the third and fourth lanes) was performed, followed by the forward transfection of empty
vector pCMV6-XL5 (pCMV) or the plasmid containing PKR cDNA sequence (hPKR)
(A). Cells were then harvested, the nuclear extract was separated from the cytoplasmic
fiaction, and western blot analysis was performed to detect the level of Bel-2 in the
cytoplasmic fiaction and the levels of NF-kB p65 in both the nuclear extract and the
cytoplasmic fiaction. TBP and beta actin were also measured as loading controls for the
nuclear extract and the cytoplasmic fi'action, respectively (A). HepG2 cells were exposed
to a cell-permeable inhibitor of NF-kB, SN50 (18 uM), or its negative control, SNSOM
(18 uM), in regular medium for 24 hours (B). After treatment, the cells were harvested,
and western blot analysis was performed to detect the protein levels of Bcl-2 in the whole
cell lysates (B). HepG2 cells were exposed to 0.7 mM palmitate or oleate for 24 hours
(C). The vehicle for the FFAs (0.7 mM BSA) was used as the control (i.e., regular
medium with BSA). After treatment, the cells were harvested, and western blot analysis
was performed to detect the protein level of NF-kB p65 in the nuclear extracts (C).

Thus far, I showed that palmitate down-regulated the activity of PKR, which
played an anti-apoptotic role in HepG2 cells. By suppressing PKR activity, palnritate
down-regulated the expression levels of Bel-2. Although the mechanism is not fully
understood, the positive effect of PKR on the anti-apoptotic protein Bel-2 could serve as

a potential pathway by which PKR protects HepG2 cells from apoptosis. PKR

42

phosphorylates I-KB, which then releases and activates NF-KB (10, 11), the key
transcription factor that up-regulates the transcription of Bel-2 in HepG2 cells (42, 43).
Indeed, silencing the gene expression of PKR decreased the level of NF-KB in the
nucleus (Fig. 2.6A, comparing lanes 1 and 3), whereas over-expressing PKR increased it
(Fig. 2.6A, comparing lanes 1 and 2). Moreover, the decreased level of NF-KB in the
nucleus was restored by rescuing the PKR expression in PKR-silenced cells (Fig. 2.6A,
comparing lanes 3 and 4). Furthermore, the PKR inhibitor also decreased the level of NF-
KB p65 in the nuclear extract of HepG2 cells (not shown). These results suggest that PKR
regulates the activity of NF-ch in HepG2 cells. NF-KB plays a key role in facilitating the
transcription of the Bel-2 gene in liver tumor cell lines, such as U937 and HepG2 cells,
and the inhibition of NF-KB results in the down-regulation of Bcl-2 gene expression (42,
43). Therefore, it is expected that the protein level of Bel-2 changes in correspondence
with the nuclear level of NF-KB, as shown in Fig. 2.6A, in which the gene expression
level of PKR is modulated. I further confirmed the role of NF-KB in regulating the
expression level of Bcl-2 with an inhibitor of NF-KB, NF-KB SN50 (Fig. 2.6B). Thus,
from my results and the literature data, I propose that the transcription factor, NF-KB,
mediates the PKR regulation of Bel-2 expression in HepG2 cells. Indeed, I also observed
that palmitate decreased the level of NF-KB in the nucleus (Fig. 2.66).

In summary, palmitate induces apoptosis of HepG2 cells, in part by reducing the
Bel-2 level, which is mediated by the repression of PKR and NF-KB activities (Fig. 2.6).
However, in addition to the protein level of Bcl-2, the anti-apoptotic role of Bcl-2 is also
regulated by the post-translational modification of Bel-2. The phosphorylation of Bcl-2 at

Ser70 sustains the anti-apoptotic role of Bel-2, whereas phosphorylation of Bcl-2 at

43

Ser87 attenuates the anti-apoptotic role of Bel-2 (50-52). Therefore, I further investigated

the involvement of palmitate and PKR in regulating the phosphorylation of Bel-2.

 

A. 1.2
IP: Bcl-2. IB: p-BcI-2 Ser70

 

I--- - -~-l

 

lP: Bcl—2, IB: Bel-2

 

p Bel-2 Ser70
(fold change)

lmuw

 

 

Palmitate (mM) 0 0.2 0.4 0.7

0 0.2 0.4 0.7
Palmitate Concentration

B.
IP: Bel-2, IB: p-Bcl-2 Ser87

IP: Bcl-Z. IB: Bcl-2

Control Palmitate

Figure 2.7. Effect of Palmitate on the phosphorylation of Bel-2. HepG2 cells were
exposed to regular medium with different levels of palmitate (A, B) for 24 hours. The
vehicle for palmitate (0.7 mM BSA) was used as the control (i.e., regular medium with
BSA), in which the concentration of palmitate was 0 (A, B). After treatment, the cells
were harvested and immunoprecipitated with anti-Bcl—2 and detected for the
phosphorylation of Bel-2 at Ser 70 (A) and Ser87 (B). The phosphorylation levels of Bcl-
2 were quantified by normalizing to total Bel-2 levels and are expressed as the average of
three samples i SD from three independent experiments. One-way ANOVA with
Tukey’s post hoc method was used for analyzing the differences between the treatment
groups. *, significantly lower than control (i.e., regular medium with BSA), p<0.01.

Palmitate Decreases the Phosphorylation of Bel-2 at Ser70, whereas the
Phosphorylation of Bcl-2 at Ser87 is Not Affected—Palmitate decreased the
phosphorylation of Bel-2 at Ser 70 in HepG2 cells (Fig. 2.7A). 0n the other hand, the
phosphorylation of Bel-2 at the pro-apoptotic residue, Ser87, was not affected by

palmitate (Fig. 2.73). Considering the concomitant repression by palmitate of the

phosphorylation of Bel-2 and the PKR activity (Fig. 2.2), I propose that PKR may be
involved in mediating the effects of palmitate on the phosphorylation of Bel-2.

PKR is Involved in Mediating the Effecm of Palmitate on the Phosphorylation
of Bel-2 at Ser70—Silencing the PKR gene using siRNA of PKR decreased the
phosphorylation of Bel-2 at Ser70 (Fig. 2.8A), indicating that PKR has a positive effect
on (i.e., enhances) the phosphorylation of Bel-2 at Ser70, whereas the phosphorylation of
Bcl-2 at Ser87 was not affected by silencing PKR (Fig. 2.8A). The effects of silencing or
inhibiting PKR on the phosphorylation of Bel-2 at Ser70 and Ser87 are consistent with
those of palmitate, and considering the negative effect of palmitate on PKR activity (Fig.
2.2), I propose that the suppression of the PKR activity mediates the negative effects of
palmitate on the phosphorylation of Bel-2 at Ser70, as well as the protein level of Bcl-2.
To confirm this hypothesis, I over-expressed the PKR gene in HepG2 cells and exposed
the cells to palmitate (Fig. 2.5C). The phosphorylation of Bcl-2 at Ser70 was restored by
over-expressing PKR in the palmitate treatment (Fig. 2.83), supporting the involvement
of PKR in mediating the effect of palmitate on the phosphorylation of Bel-2 at Ser70.
However, a co-irnmunoprecipitation study showed that PKR did not directly interact with
the Bcl-2 protein (Fig. 2.80; therefore, other intermediate signaling molecules are
involved in mediating the effect of PKR on the phosphorylation of Bel-2. PKR has been
reported to positively signal MAPKs (12), and sequence analysis of both Ser70 and Ser87
residues of Bcl-2 suggests that JNK could regulate the phosphorylation of Bel-2 (57).
Taken together, the evidence suggests that JNK may act as one of the intermediates

between PKR and Bel-2.

45

Figure 2.8. Involvement of PKR in regulating the phosphorylation of Bel-2 at Ser70.
Reverse transfection of suspended HepG2 cells was performed with scrambled siRNA
(Control) or siRNA of PKR for 24 hours and the transfected cells were cultured in regular
medium for another 24 hours (A). Cells were harvested and immunoprecipitated with
anti-Bcl-2 and detected for the phosphorylation of Bel-2 at Ser 70 and Ser87 (A). In
HepG2 cells, the forward transfection of empty vector pCMV6-XL5 (pCMV 6) or the
plasmid containing PKR cDNA sequence (pCMV6-hPKR) was performed and the cells
were then treated with different concentrations of palmitate for 24 hours (B). The vehicle
for palmitate (0.7 mM BSA) was used as the control (i.e., regular medium with BSA), in
which the concentration of palnritate was 0 (B). After the transfections and palmitate
treatment, the cells were harvested, and cell lysates were immunoprecipitated with anti-
Bcl-2 and detected for phosphorylation of Bel-2 at Ser 70 (B) (Please see Fig. 2.5C for
the over-expressed PKR levels). Confluent HepG2 cells were harvested and
immunoprecipitated with anti-Bcl-2 or anti-PKR, and western blot analysis was
performed to detect the coirnmunoprecipitation of Bel-2 and PKR (C). The
phosphorylation levels of Bel-2 were quantified by normalizing to total Bcl-2 levels and
are expressed as the average of three samples :t SD from three independent experiments.
One-way AN OVA with Tukey’s post hoc method was used for analyzing the differences
between the treatment groups. *, significantly different fiom control, p<0.01.

46

 

IP: Bel-2, IB: p-Bcl-2 Ser87

 

 

3 1
8
- 3
IP: Bel-2, IB: p-BCl—Z Ser70 g g 0'8 *
2 .9 0.4
IP: Bcl-2, lB: Bel-2 73 -
. l=E' 0.2
H' ”with 0-
Ctrl siPKR ° .
Ctrl SIPKR
B.
IP: Bel-2, lB: p-Bcl-2 Ser70
Lw ¢ ~— - “nu-«o ~ _ n —l

 

IP: Bel-2, IB: Bcl-2

 

It-N- what—“w m w—‘MH‘

 

Palmitate(mM) 0 0.2 0.4 0.7 O 0.2 0.4 0.7
pCMV6 pCMV6-hPKR

 

 

1.6
1.4
1.2

1
0.8
0.6
0.4

p-Bcl2 Ser70/total Bcl2
(fold change)

0.2

 

 

O
Palmitate(mM) o 0.2 0.4 0.7 o 0.2 0.4 0.7
pCMV6 pCMV6-hPKR

IB: Bcl- 2

IB: PKR
IP: Ctrl Bcl-2 PKR

47

 

PKR Positively Regulates JNK, and JNK Regulates the Phosphorylation of Bcl-
2 at Ser 70, but not at Ser8 7—A co-immunoprecipitation study indicated that JNK
directly interacted with both PKR and Bel-2 (Fig. 2.9A), suggesting that JNK may be a
potential intermediate protein kinase that mediates the positive effect of PKR on the
phosphorylation of Bel-2 at Ser70. Indeed, silencing the PKR gene repressed the activity
of JNK (Fig. 2.93, comparing lanes 1 and 3), whereas over-expressing PKR enhanced it
(Fig. 2.9B, comparing lanes 1 and 2). Moreover, the suppressed JNK activity was
restored by rescuing the PKR expression in PKR-silenced cells (Fig. 2.93, comparing
lanes 3 and 4), confirming the positive connection between PKR and JNK. Considering
the negative effect of palmitate on PKR activity (Fig. 2.2), I then propose that palmitate
decreases the activity of JNK by repressing PKR. Indeed, I found that treatment of
palmitate for 24 hours decreased the phosphorylation of JNK (Fig. 2.9C), and over-
expressing PKR in the palmitate treatment restored the phosphorylation of JN K (Fig.
2.9D), supporting the involvement of PKR in mediating the effect of palmitate on the
phosphorylation of JNK. In addition, to firrther assess the connection between JNK and
Bel-2, I performed a JNK inhibition study and found that the JNK inhibitor, SP600125,
suppressed the phosphorylation of Bel-2 at Ser70, but not the phosphorylation at Ser87
(Fig. 2.9 E and F), which is consistent with the distinct effects of palnritate as well as
PKR on the phosphorylation of Bel-2 at the two amino acid residues (Figs. 2.7, 2.8).
Therefore, the results in Figure 2.9 suggest JNK is an intermediate protein that mediates
the effects of pahnitate and PKR on the phosphorylation of Bcl-2 at Ser70. I showed that
PKR had an anti-apoptotic role in HepG2 cells (Fig. 2.3). To firrther confirm the positive

correlation between PKR and JNK, I investigated the role of JNK in regulating

48

cytotoxicity and apoptosis in HepG2 cells. The inhibition of JNK using SP600125
significantly increased LDH release (Fig. 2.104) and caspase-3 activity (Fig. 2.103),
suggesting that, similar to PKR, JNK has an anti-apoptotic role in HepG2 cells.

In summary, palmitate inhibits the phosphorylation of Bel-2 at the anti-apoptotic
residue, Ser70, without affecting the pro-apoptotic residue, Ser87. My data suggest that
this effect of palmitate on the phosphorylation of Bcl-2 is mediated by the signaling of
PKR and JNK. I therefore propose another anti-apoptotic pathway, which is suppressed
by palmitate, consisting of PKR, JNK, and phosphorylation of Bcl-2.

My investigation of the association between PKR and Bel-2 revealed two
different but complementary anti-apoptotic pathways that connect PKR and Bel-2. First,
PKR up-regulates the transcription of the Bel-2 gene, possibly through the transcription
factor, NF-KB. Second, PKR up-regulates the phosphorylation of Bel-2 at the anti-
apoptotic residue, Ser70, mediated by JNK (Fig. 2.113). These two pathways were down-
regulated in HepG2 cells upon exposure to palmitate, and they may constitute one of the

mechanisms by which palmitate induces apoptosis in HepG2 cells.

49

Figure 2.9. Involvement of JN K in regulating the phosphorylation of Bel-2.
Confluent HepG2 cells were harvested and immunoprecipitated with anti-Bcl-2 or anti-
PKR, and western blot analysis was performed to detect the protein level of JN K (A).
Reverse transfection of scramble siRNA (control, the first two lanes) or siRNA of PKR
(siPKR, the third and fourth lanes) was performed, followed by the forward transfection
of empty vector pCMV6-XL5 (pCMV) or the plasmid containing PKR cDNA sequence
(hPKR) (B). HepGZ cells were exposed to 0.7 mM palmitate or oleate for 24 hours (C).
The vehicle for the FFAs, 0.7 mM BSA, was used as the control (i.e., regular medium
with BSA) (C). In HepG2 cells, the forward transfection of empty vector pCMV6-XL5
(pCMV6) or plasmid containing PKR cDNA sequence (pCMV6-hPKR) was performed
and the cells were then treated with 0.7 mM palmitate for 24 hours (D). The vehicle for
palmitate, 0.7 mM BSA, was used as the control (i.e., regular medium with BSA) (D).
After treatments, cells were then harvested and western blot analysis was performed to
detect the phosphorylation and protein levels of JNKl/2 after treatment (B, C, D). HepG2
cells were exposed in regular medium for 24 hours to the pharmaceutical inhibitor of
JNK, SP600125 (25 uM) (E, F), or its analogue, SP600125 (25 uM) (E), as a negative
control. The control in F is the vehicle of palnritate, BSA. After treatment, the cells were
harvested and western blot analysis was performed to detect the protein level of Bel-2
and the phosphorylation of Bcl-2 at Ser70 (E), or the cell lysate was immunoprecipitated
with anti-Bcl-2 and western blot analysis was performed to detect the phosphorylation
level of Bcl-2 at Ser87 (F).

50

 

 

 

 

 

 

 

    

 

 

  
 

 

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A. B.
leJNK
p-JNK —+
q... - T1831Y185 _’
. —9 j
IP. Ctrl Bel-2 PKR JNK .
pCMV6 hPKR pCMV6 hPKR
Control siPKR
C. D.
p-JNK —> f j i " ' ' P-JNK -’ ~.3--- ’
T183IY185—D .. . T183IY185 -* g
JNK" . "T
r ,..} - _> , . v , . ..7
Control Palm. Ole. Ct rl Palm. Ct ” Palm.
pCMV6 pCMV6-hPKR
E. F
IP: Bel-2, IB: p-BcI-2 Ser87
IP: Bel-2, IB: Bcl-2
Bel-2 W «null-b
W W, .
Control 3P600125 Control Palm. SP600125

51

.>
5”

 

 

 

 

 

 

   

1.60% 2.5
'k
g 1.20% - :EA 2
v .5 o
g 2 g 1.5
a 0.80% 3 5
. §£
0.40%. o 0.5
0.00% 0
cm SP600125 Ctrl SP600125

Figure 2.10. Effect of inhibiting JNK activity on the cytotoxicity and apoptosis of
HepG2 cells. HepG2 cells were exposed for 24 hours to the pharmaceutical inhibitor of
JNK, SP600125 (25 pM), or its analogue, SP600125 (ZSuM), as a negative control (A, B).
LDH release (A) and caspase-3 activity (B) were measured afier treatment with the JNK
inhibitor. Data are expressed as the average of nine samples :t SD from three independent
experiments. One-way ANOVA with Tukey’s post hoc method was used for analyzing
the differences between the treatment groups. *, significantly higher than control, p<0.01.

52

 

Bim-EL—> m w
Bim-L —> ~

Ctrl siPKR

Palmitate

 

 

 

cu

 

‘.........

FoxO3a other
factors?
V
Bim
Ser70

 

genes including Bel-2

Figure 2.11. Proposed signaling pathways from PKR to Bel-2 induced by palmitate.
A. Reverse transfection of suspended HepG2 cells was performed with scrambled siRNA
(Control) or siRNA of PKR for 24 hours and the transfected cells were cultured in regular
medium for another 24 hours. Cells were harvested, and western blot analysis was
performed to detect the protein level of Bim. B. A summary of the signaling pathways
identified in this study. First, PKR up-regulates the transcription of the Bel-2 gene
through a transcription factor, likely NF-KB. Second, PKR up-regulates the
phosphorylation of Bel-2 at the anti-apoptotic residue Ser70, mediated by JNK. By
suppressing PKR, and in turn the two downstream pathways, palmitate regulates the
protein level and phosphorylation of Bcl—2 at Ser70 and therefore induces apoptosis in
HepGZ cells. Another Bel-2 family protein, Bim, which also mediates palmitate-induced
apoptosis (31, 32) (dashed lines), and the potential effect of PKR on Bim (line with the
question mark) are included.

Apoptosis regulating 1

53

DiscuSsion

In this chapter, PKR is found to be anti-apoptotic in human hepatoma cells. PKR
is best known for its pro-apoptotic role by phosphorylating eIF-Za and thereby inhibiting
general protein synthesis (2, 3). Thus, it has previously been suggested that PKR acts as a
tumor suppressor by inhibiting cell growth and inducing apoptosis (64, 65). In contrast,
more recent studies suggest that PKR has an anti-apoptotic role in regulating tumor
development and tumor cell apoptosis (5-8). Elevated PKR protein levels and activity
were observed in certain tumor cells (e.g., human breast cancer cells (5), melanoma cells
(6, 8), and hepatitis C virus (HCV)-related hepatocellular carcinoma (7)). The over-
expression and elevated activity of PKR have been attributed to the development of
tumors and the proliferation of tumor cells (5-8), but the mechanism is still unclear.
Research suggests that PKR may suppress apoptosis by activating the NF-KB pathway
before phosphorylating eIF-Za, therefore inducing cell survival initially and,
subsequently, cell death in NIH3T3 cells expressing PKR (4). The present study supports
an anti-apoptotic role of PKR in human hepatoma cells (HepG2) and suggests potential
pathways by which PKR mediates anti-apoptosis through Bel-2, thereby further
providing a potential mechanism of palmitate-induced apoptosis in HepG2 cells.
Interestingly, although the phosphorylation of PKR was significantly suppressed in
HepG2 cells, the phosphorylation level of eIF-Za was not altered by palmitate (not
shown), suggesting that the protein synthesis arrest machinery and related apoptosis
activity were not affected. Similarly, another substrate of PKR, BS6a-PP2A, was not

affected by palmitate (not shown). This suggests that other substrates of PKR, which I

54

propose may be NF-KB and JN K, are mediating the effects of altered PKR activity on the
regulation of the Bel-2 protein and its phosphorylation levels in HepGZ cells.

PKR activates several transcription factors, such as IRF-l, p53, and NF-ch (66,
67). In the present study, I show that PKR is involved in controlling the transcription of
Bel-2 in HepGZ cells, mediated by the transcription factor NF-KB. This result suggests a
novel mechanism by which palmitate down-regulates the protein level of Bel-2 (Fig. 2.4)
in HepGZ cells. By modulating the transcription of a number of anti-apoptotic genes,
such as Bel-2 (68), Bcl-xL (69), cIAPs (70), and p53 (71), NF-KB is commonly
considered an anti-apoptotic transcription factor (72, 73), which is consistent with the
anti-apoptotic role of PKR in HepG2 cells. NF -KB is the key regulator of transcription of
the Bel-2 gene in liver tumor cell lines, including HepG2 (42, 43). However, other
transcription factors that may regulate the transcription of the Bel-2 gene, for example, c-
Myb (39, 40) and STAT5 (41), have not been evaluated in the present study. It is not
known in the literature whether these transcription factors are altered by palmitate or
regulated by PKR. Therefore, it remains to be determined whether other transcription
factors, in addition to NF-kB, mediate the effect of PKR on the gene expression and
protein levels of Bel-2.

Furthermore, other studies have revealed that elevated Bim, a pro-apoptotic Bel-2
family protein, also plays an important role in the stearic and palmitic acid-induced
apoptosis of several liver cell lines, including HepG2 (31, 32), and this process has been
determined to be dependent on the transcription factor FoxO3a (31). In the present study,
I propose that the suppression of PKR mediates palmitate-induced apoptosis through Bcl-

2; therefore, to test the potential connection between my results and the previous findings

55

on Bim, I evaluated the effect of PKR on the expression of Bim. Interestingly, silencing
PKR significantly increased the protein level of Bim (Fig. 2.11A), suggesting that PKR
represses Bim expression. As a BH3-only Bel-2 protein, Bim promotes apoptosis by
binding and inactivating anti-apoptotic proteins (74); therefore, the negative effect of
PKR on Bim supports the role of PKR as an anti-apoptotic protein and could serve as
another potential mechanism by which PKR inhibits apoptosis in HepGZ cells. In
addition, I show in the present study that palmitate decreased PKR activity (Fig. 2.2), and
thus it is possible that the up-regulation of Bim by palmitate (31) may be due, in part, to
the repression of PKR activity. However, the mechanism by which PKR down-regulates
the expression of Bim is not known at this point. Future experiments will determine
whether PKR has any effect on the key transcription factor of Bim, Fox03a (31).
Previously, it was shown that PP2A, which was activated by palmitate, mediated the
effect of palmitate on the activity of Fox03a and thereby the expression of Bim (31).
However, in my study palmitate did not affect the activity of PP2A (data not shown), and
this contradictory result may be due to the different treatment time (6 vs. 24 hours) and/or
the different types of palmitate used (palmitic acid dissolved in isopropyl alcohol vs.
sodium palmitate complexed with BSA).

In the present study, I uncovered a second anti-apoptotic pathway, namely, that
PKR up-regulates the phosphorylation of Bel-2 at Ser70 (Fig. 2.8), mediated by JNK (Fig.
2.9). Although sequence analysis suggests that both Ser70 and Ser87 residues of Bel-2
can be recognized by JN K (57), I found that only phosphorylation at Ser70 is affected by
JNK (Fig. 2.9). Concomitantly, palmitate (Fig. 2.7) and PKR silencing (Fig. 2.8) both

decreased only the phosphorylation of Bel-2 at Ser70, which lends support to the

56

hypothesis that palmitate down-regulates the phosphorylation of Bel-2 at Ser70 through
PKR and JNK. In fact, all 3 MAPKs are activated by PKR upon ribotoxic stress, and the
investigators proposed that MAPKs respond to PKR in the rank order of JNK>p3 8>ERK
(12). However, I did not test the effects of PKR on the other two less responsive MAPK
proteins, ERK and p38, which were also proposed to phosphorylate Bel-2 at Ser70 (54-
56). Thus, JNK is one of the intermediates involved in mediating the signaling pathway
from PKR to Bel-2 phosphorylation. Further investigations of the 3 MAPKs are required
to fully understand the interaction between PKR and Bel-2.

Currently, it remains unclear how PKR interacts with MAPKs (e. g., JNK). An
association between PKR and apoptosis signal-regulating kinase 1 (ASKI), one of the
MAPK signaling proteins, has been previously established (75). ASK], a MAPK kinase
kinase (MAPKKK), phosphorylates SEKI/MKK4 and MKK3/MKK6, which, in turn,
activate IN K and p38 MAPK, respectively (76). My coirnmunoprecipitation study
suggests an interaction between JNK and PKR (Fig. 2.9). Nevertheless, it must be noted
that the coirnmunoprecipitation of PKR and IN K does not necessarily indicate a “direct”
binding of these two proteins. It is believed that the activation of the MAPK cascade
requires a scaffold protein that assembles the MAPKKK, MAPKK, and MAPK proteins
together into a certain signaling module (77). For instance, J IPl organizes upstream
kinase HPK-l, MAPKKK MLKI, MKK7, and JNKl/2, and it facilitates the activation of
JNK (7 8). Therefore, the coirnmunoprecipitation study suggests an interaction, but not
necessarily a direct one, between PKR and JNK. Instead, the coirnmunoprecipitation of
PKR and JNK may recruit a third protein, possibly a scaffold protein that assembles JNK

and facilitates the activation of JNK by PKR. In other words, the pathway that I identified

57

consisting of PKR and JNK does not preclude the possibility of other intermediates
between PKR and JN K. To understand how PKR interacts with JNK, the scaffold protein
and the assembly of the signaling proteins with their upstream activators must be
investigated.

JNK is involved in many signaling pathways that control diverse cellular
activities such as proliferation, differentiation, and apoptosis. Through different pathways
and substrates (79), JNK has been shown to have either pro- or anti-apoptotic functions,
depending on the cell type, stimulus, duration of its activation and activity of other
signaling pathways, and it is therefore considered a “double-edged sword” (80). In liver
tumor cell lines such as HepG2, conflicting views on the role of JNK in regulating
apoptosis have been reported in the literature. For example, JNK was shown to mediate
anti-apoptotic signals in transforming growth factor-beta l- (81) and tumor necrosis
factor-induced apoptosis (82) in HepGZ cells, whereas other researchers have proposed
pro-apoptotic roles for JNK in HepG2 cells (32, 83). In these studies on the role of JNK
in regulating apoptosis, the JNK inhibitor SP600125 was used, and the conflicting results
may be due to the different conditions applied by the investigators, such as the
concentrations and treatment times of SP600125 and the accompanying additives. In my
study, I inhibited JNK in regular medium for 24 hours and found that the IN K inhibitor
(SP600125, at 25 uM) was apoptotic, as evidenced by caspase-3 activity (Fig. 2.103).
Zhang et al. (81), treated liver tumor cell lines with 20 uM SP600125 for 50 hours,
resulting in a significant increase of apoptosis. On the other hand, Chen et al. (83) treated
the HepGZ cells with 1 uM SP600125 for 48 hours and showed that it blocked the

norcantharidin-induced apoptosis. Interestingly, in their controls without norcantharidin,

58

SP600125 or ERK inhibitor (U0126 or PD98059) slightly increased the apoptosis of the
HepGZ cells, which supports my finding with SP600125 in the control medium. Malhi et
al. (32) used SP600125 to inhibit JNK in the presence of palmitic acid for 24 hours, but
the concentration of the inhibitor used was not specified. My study suggests that JNK is
involved in phosphorylating Bel-2, an anti-apoptotic member of the Bel-2 family, at its
anti-apoptotic residue, Ser70 (Fig. 2.9C), suggesting that JNK may act as an anti-
apoptotie factor. The anti-apoptotic role exhibited by JNK is consistent with that of PKR
and therefore lends support to the hypothesis that JNK may mediate, in part, the signaling
between PKR and Bel-2. However, inhibiting JNK with SP600125 also increased the
LDH release significantly (Fig. 2.10A), whereas silencing PKR did not have a significant
effect on the LDH release (Fig. 2.33), suggesting that JNK may be involved in other
cellular activities related to cytotoxicity, in addition to the pathway fiom PKR to Bel-2.
On the other hand, it is also possible that the inhibitor SP600125 itself has other non-
specific effects that lead to cytotoxicity, despite its widespread use as an inhibitor of JNK.

In summary, I identified an anti-apoptotic role of PKR and found that it is
involved in regulating the protein level and phosphorylation status of Bel-2 in HepGZ
cells. The transcription factor NF-KB and the MAP kinase JNK appear to be involved in
mediating the effects of PKR on the protein level and the phosphorylation of Bel-2,
respectively. I propose that palmitate suppresses these two pathways (Fig. 2.4 and 2.7) by
inhibiting PKR (Fig. 2.2) and thereby attenuates the anti-apoptotic machinery (Fig.

2.113).

59

10.

References

Zhang, R, Romano, P. R, Nagamura-Inoue, T., Tian, B., Dever, T. B., Mathews, M.

B., Ozato, K., and Hinnebusch, A. G. (2001) Binding of double-stranded RNA to
protein kinase PKR is required for dimerization and promotes critical
autophosphorylation events in the activation loop. J Biol Chem 276, 24946-24958

Taylor, S. 8., Haste, N. M., and Ghosh, G. (2005) PKR and eIF2alpha: integration
of kinase dimerization, activation, and substrate docking. Cell 122, 823-825

Stark, G. R., Kerr, 1. M., Williams, B. R., Silverman, R. H., and Schreiber, R. D.
(1998) How cells respond to interferons. Annu Rev Biochem 67, 227-264

Donze, 0., Deng, J ., Curran, J., Sladek, R., Picard, D., and Sonenberg, N. (2004)
The protein kinase PKR: a molecular clock that sequentially activates survival and
death programs. EMBOJ 23, 564-571

Kim, S. H., Forrnan, A. P., Mathews, M. B., and Gunnery, S. (2000) Human breast
cancer cells contain elevated levels and activity of the protein kinase, PKR.
Oncogene 19, 3086-3094

Kim, S. H., Gunnery, S., Choe, J. K., and Mathews, M. B. (2002) Neoplastic
progression in melanoma and colon cancer is associated with increased expression
and activity of the interferon-inducible protein kinase, PKR. Oncogene 21, 8741-
8748

Hiasa, Y., Kamegaya, Y., Nuriya, H., Onji, M., Kohara, M., Schmidt, E. V., and
Chung, R. T. (2003) Protein kinase R is increased and is functional in hepatitis C
virus-related hepatocellular carcinoma. Am J Gastroenterol 98, 2528-2534

Delgado Andre, N., and De Lucca, F. L. (2007) Knockdown of PKR expression by
RNAi reduces pulmonary metastatic potential of Bl6-F10 melanoma cells in mice:
Possible role of NF-kappaB. Cancer Lett 258, 1 18-125

Williams, B. R. (2001) Signal integration via PKR. Sci STKE 2001, RE2

Kumar, A., Haque, J., Lacoste, J., Hiscott, J ., and Williams, B. R. (1994) Double-
stranded RN A-dependent protein kinase activates transcription factor NF-kappa B
by phosphorylating I kappa B. Proc Natl Acad Sci U S A 91, 6288-6292

60

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

Gil, J., Alcami, J., and Esteban, M. (2000) Activation of NF-kappa B by the
dsRNA-dependent protein kinase, PKR involves the I kappa B kinase complex.
Oncogene 19, 1369-1378

Zhou, H. R., Lau, A. S., and Pestka, J. J. (2003) Role of double-stranded RNA-
activated protein kinase R (PKR) in deoxynivalenol-induced ribotoxic stress
response. Toxicol Sci 74, 335-344

Xu, 2., and Williams, B. R. (2000) The B56alpha regulatory subunit of protein
phosphatase 2A is a target for regulation by double-stranded RNA-dependent
protein kinase PKR. Mol Cell Biol 20, 5285-5299

Watanabe, M. A., Rodrigues Souza, L., Murad, J. M., and De Lucca, F. L. (2003)
Antitumor activity induced by regulatory RNA: possible role of RNA—dependent
protein kinase and nuclear factor-kappaB. Eur J Pharmacol 465, 205-210

Yeung, M. C., and Lau, A. S. (1998) Tumor suppressor p53 as a component of the
tumor necrosis factor-induced, protein kinase PKR-mediated apoptotic pathway in
human promonocytic U937 cells. J Biol Chem 273, 25198-25202

Onuki, R., Bando, Y., Suyama, B., Katayama, T., Kawasaki, H., Baba, T., Tohyama,
M., and Taira, K (2004) An RNA-dependent protein kinase is involved in
tunicamycin-induced apoptosis and Alzheimer’s disease. EMBO J23, 959-968

He, B. (2006) Viruses, endoplasmic reticulum stress, and interferon responses. Cell
Death Differ 13, 393-403

Meusel, T. R., Kehoe, K. E., and Imani, F. (2002) Protein kinase R regulates
double-stranded RNA induction of TNF-alpha but not IL-1 beta mRNA in human
epithelial cells. J Immunol 168, 6429-6435

Goh, K. C., deVeer, M. J ., and Williams, B. R. (2000) The protein kinase PKR is

- required for p38 MAPK activation and the innate immune response to bacterial

endotoxin. EMBO J 19, 4292-4297

Feldstein, A. B., Werneburg, N. W., Canbay, A., Guicciardi, M. B., Bronk, S. F.,
Rydzewski, R., Burgart, L. J ., and Gores, G. J. (2004) Free fatty acids promote
hepatic lipotoxicity by stimulating TNF-alpha expression via a lysosomal pathway.
Hepatology 40, 185-194

Lu, Z. H., Mu, Y. M., Wang, B. A., Li, X. L., Lu, J. M., Li, J. Y., Pan, C. Y.,
Yanase, T., and Nawata, H. (2003) Saturated free fatty acids, pahnitic acid and

61

22.

23.

24.

25.

26.

27.

28.

29.

30.

stearic acid, induce apoptosis by stimulation of ceramide generation in rat testicular
Leydig cell. Biochem Biophys Res Commun 303, 1002-1007

Yamagishi, S., Okamoto, T., Amano, S., Inagaki, Y., Koga, K., Koga, M., Choei,
H., Sasaki, N., Kikuchi, S., Takeuchi, M., and Makita, Z. (2002) Palmitate-induced
apoptosis of microvascular endothelial cells and pericytes. Mol Med 8, 179-184

Hickson-Bick, D. L., Sparagna, G. C., Buja, L. M., and McMillin, J. B. (2002)
Palmitate-induced apoptosis in neonatal cardiomyocytes is not dependent on the
generation of ROS. Am J Physiol Heart Circ Physiol 282, H656-664

Feldstein, A. B., Werneburg, N. W., Li, Z., Bronk, S. F., and Gores, G. J. (2006)
Bax inhibition protects against free fatty acid-induced Iysosomal permeabilization.
Am J Physiol Gastrointest Liver Physiol 290, G1339-1346

Giordano, A., Calvani, M., Petillo, 0., Grippo, P., Tuccillo, F., Melone, M. A.,
Bonelli, P., Calarco, A., and Peluso, G. (2005) tBid induces alterations of
mitochondrial fatty acid oxidation flux by malonyl-CoA-independent inhibition of
carnitine palmitoyltransferase-l. Cell Death Difler 12, 603-613

Listenberger, L. L., Han, X., Lewis, S. B., Cases, S., Farese, R. V., Jr., Ory, D. S.,
and Schaffer, J. E. (2003) Triglyceride accumulation protects against fatty acid-
induced lipotoxicity. Proc Natl Acad Sci U S A 100, 3077-3082

Li, Z., Srivastava, S., Yang, X., Mittal, S., Norton, P., Resau, J ., Haab, B., and Chan,

C. (2007) A hierarchical approach employing metabolic and gene expression
profiles to identify the pathways that confer cytotoxicity in HepG2 cells. BMC Syst
Biol 1, 21

Wei,.Y., Wang, D., Topczewski, F., and Pagliassotti, M. J. (2006) Saturated fatty
acids induce endoplasmic reticulum stress and apoptosis independently of ceramide
in liver cells. Am J Physiol Endocrinol Metab 291, 13275-281

Assimacopoulos-Jeannet, F. (2004) Fat storage in pancreas and in insulin-sensitive
tissues in pathogenesis of type 2 diabetes. Int J Obes Relat Metab Disord 28 Suppl
4, 853-57

Srivastava, S., and Chan, C. (2006) Hydrogen peroxide and hydroxyl radicals
mediate palmitate-induced cytotoxicity to hepatoma cells: Relation to
mitochondrial permeability transition. Free Radic Res 41, 38-49

62

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

Barreyro, F. J., Kobayashi, S., Bronk, S. P., Werneburg, N. W., Malhi, H., and
Gores, G. J. (2007) Transcriptional Regulation of Bim by FoxO3A Mediates
Hepatocyte Lipoapoptosis. J Biol Chem 282, 27141-27154

Malhi, H., Bronk, S. P., Werneburg, N. W., and Gores, G. J. (2006) Free fatty acids
induce JNK-dependent hepatocyte lipoapoptosis. J Biol Chem 281, 12093-12101

Feng, L. F., Zhong, M., Lei, X. Y., Zhu, B. Y., Tang, S. S., and Liao, D. F. (2006)
Bcl-2 siRN A induced apoptosis and increased sensitivity to 5-fluorouracil and
HCPT in HepG2 cells. J Drug Target 14, 21-26

Handayani, T., Sakinah, S., Nallappan, M., and Pihie, A. H. (2007) Regulation of
p53-, Bcl-2- and caspase-dependent signaling pathway in xanthorrhizol-induced
apoptosis of HepG2 hepatoma cells. Anticancer Res 27, 965-971

Huang, D. S., Shen, K. Z., Wei, J. F., Liang, T. B., Zheng, S. S., and Xie, H. Y.
(2005) Specific COX-2 inhibitor NS398 induces apoptosis in human liver cancer
cell line HepG2 through BCL—2. World J Gastroenterol 11, 204-207

Landau, Z., Forti, B., Alcaly, M., and Birk, R. Z. (2006) Palmitate induced
lipoapoptosis of exocrine pancreas AR42J cells. Apoptosis 11, 717-724

Srivastava, S., Li, Z., Yang, X., Yedwabnick, M., Shaw, S., and Chan, C. (2007)
Identification of genes that regulate multiple cellular processes/responses in the
context of lipotoxicity to hepatoma cells. BMC Genomics 8, 364

Breckenridge, D. G., and Xue, D. (2004) Regulation of mitochondrial membrane
permeabilization by BCL-2 family proteins and caspases. Curr Opin Cell Biol 16,
647-652

Salomoni, P., Perrotti, D., Martinez, R., Franceschi, C., and Calabretta, B. (1997)
Resistance to apoptosis in CTLL—2 cells constitutively expressing c-Myb is
associated with induction of BCL-2 expression and Myb-dependent regulation of
bcl-2 promoter activity. Proc Natl Acad Sci U S A 94, 3296-3301

Taylor, D., Badiani, P., and Weston, K. (1996) A dominant interfering Myb mutant
causes apoptosis in T cells. Genes Dev 10, 2732-2744

Weber-Nordt, R. M., Egen, C., Wehinger, J ., Ludwig, W., Gouilleux-Gruart, V.,
Mertelsmann, R., and Finke, J. (1996) Constitutive activation of STAT proteins in

primary lymphoid and myeloid leukemia cells and in Epstein-Barr virus (EBV)-
related lymphoma cell lines. Blood 88, 809-816

63

42.

43.

45.

46.

47.

48.

49.

50.

51.

D'Alessio, M., Cerella, C., Amici, C., Pesce, C., Coppola, S., Fanelli, C., De Nicola,
M., Cristofanon, S., Clavarino, G., Bergamaschi, A., Magrini, A., Gualandi, G., and
Ghibelli, L. (2004) Glutathione depletion up-regulates Bcl-2 in BSO-resistant cells.
FASEB J 18, 1609-1611

Nishikawa, T., Nakajima, T., Moriguchi, M., Jo, M., Sekoguchi, S., Ishii, M.,
Takashima, H., Katagishi, T., Kimura, H., Minami, M., Itoh, Y., Kagawa, K., and
Okanoue, T. (2006) A green tea polyphenol, epigalocatechin-3 -gallate, induces
apoptosis of human hepatocellular carcinoma, possibly through inhibition of Bcl-2
family proteins. J Hepatol 44, 1074-1082

Charlotte, F., L'Hermine, A., Martin, N., Geleyn, Y., Nollet, M., Gaulard, P., and
Zafrani, E. S. (1994) Immunohistochemical detection of bcl-2 protein in normal and
pathological human liver. Am J Pathol 144, 460-465

Terada, T., and Nakanuma, Y. (1996) Expression of apoptosis, proliferating cell
nuclear antigen, and apoptosis-related antigens (bcl-2, c-myc, Fas, Lewis(y) and
p53) in human cholangiocarcinomas and hepatocellular carcinomas. Pathol Int 46,
764-770

Zeppa, P., Benincasa, G., Fulciniti, F., Cochand-Priollet, B., Troncone, G., Vetrani,
A., Celentano, B., Nardone, G., and Palombini, L. (1996) Apoptosis and cytologic
differentiation in hepatocellular carcinoma on fine needle aspiration samples. Acta
Cytol 40, 861-866

Zhao, M., Zhang, N. X., Economou, M., Blaha, I., Laissue, J. A., and Zimmermann,
A. (1994) Immunohistochemical detection of bcl-2 protein in liver lesions: bcl-2
protein is expressed in hepatocellular carcinomas but not in liver cell dysplasia.
Histopathology 25, 237-245

Pianko, S., Patella, S., and Sievert, W. (2000) Alcohol consumption induces
hepatocyte apoptosis in patients with chronic hepatitis C infection. J Gastroenterol
Hepatol 15, 798-805

Blagosklonny, M. V. (2001) Unwinding the loop of Bcl-2 phosphorylation.
Leukemia 15, 869-874

Ito, T., Deng, X., Carr, B., and May, W. S. (1997) Bcl-2 phosphorylation required
for anti-apoptosis function. J Biol Chem 272, 11671-11673

De Chiara, G., Marcocci, M. B., Torcia, M., Lucibello, M., Rosini, P., Bonini, P.,
Higashimoto, Y., Damonte, G., Armirotti, A., Amodei, S., Palamara, A. T., Russo,
T., Garaci, B., and Cozzolino, F. (2006) Bc1-2 Phosphorylation by p38 MAPK:

64

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

identification of target sites and biologic consequences. J Biol Chem 281, 21353-
21361

Simizu, S., Tamura, Y., and Osada, H. (2004) Dephosphorylation of Bcl-2 by
protein phosphatase 2A results in apoptosis resistance. Cancer Sci 95, 266-270

Srivastava, R. K., Mi, Q. S., Hardwick, J. M., and Longo, D. L. (1999) Deletion of
the loop region of Bcl-2 completely blocks paclitaxel-induced apoptosis. Proc Natl
Acad Sci USA 96, 3775-3780

Deng, X., Xiao, L., Lang, W., Gao, F., Ruvolo, P., and May, W. S., Jr. (2001)
Novel role for JNK as a stress-activated Bc12 kinase. J Biol Chem 276, 23681-
23688

Deng, X., Ruvolo, P., Carr, B., and May, W. 8., Jr. (2000) Survival function of
ERK1/2 as IL-3 -activated, staurosporine-resistant Bc12 kinases. Proc Natl Acad Sci
USA 97, 1578-1583

Mai, H., May, W. S., Gao, F., Jin, Z., and Deng, X. (2003) A fimctional role for
nicotine in Bc12 phosphorylation and suppression of apoptosis. J Biol Chem 278,
1886-1891

Basu, A., You, S. A., and Haldar, S. (2000) Regulation of Bc12 phosphorylation by
stress response kinase pathway. Int J Oncol 16, 497-500

.Ruvolo, P. P., Deng, X., Ito, T., Carr, B. K., and May, W. S. (1999) Ceramide

induces Bc12 dephosphorylation via a mechanism involving mitochondrial PP2A. J
Biol Chem 274, 20296-20300

Peters, T. (1996) All about albumin : biochemistry, genetics, and medical
applications, Academic Press, San Diego

Pataer, A., Vorburger, S. A., Chada, S., Balachandran, 8., Barber, G. N., Roth, J. A.,
Hunt, K. K, and Swisher, S. G. (2005) Melanoma differentiation-associated gene-7

protein physically associates with the double-stranded RNA-activated protein
kinase PKR. Mol Thar 11, 717-723

Jammi, N. V., Whitby, L. R., and Beal, P. A. (2003) Small molecule inhibitors of
the RNA-dependent protein kinase. Biochem Biophys Res Commun 308, 50-57

Eley, H. L., and Tisdale, M. J. (2007) Skeletal muscle atrophy, a link between
depression of protein synthesis and increase in degradation. J Biol Chem 282, 7087-
7097

65

63.

65.

66.

67.

68.

69.

70.

71.

72.

Gilfoy, F. D., and Mason, P. W. (2007) West Nile virus-induced interferon
production is mediated by the double-stranded RN A-dependent protein kinase PKR.
J Virol 81, 11148-11158

Koromilas, A. R., Roy, S., Barber, G. N., Katze, M. G., and Sonenberg, N. (1992)
Malignant transformation by a mutant of the IFN-inducible dsRNA-dependent
protein kinase. Science (New York, NY 257, 1685-1689

Meurs, E. F., Galabru, J., Barber, G. N., Katze, M. G., and Hovanessian, A. G.
(1993) Tumor suppressor function of the interferon-induced double-stranded RNA-
activated protein kinase. Proceedings of the National Academy of Sciences of the
United States of America 90, 232-236

Kumar, A., Yang, Y. L., Flati, V., Der, S., Kadereit, S., Deb, A., Haque, J ., Reis, L.,
Weissmann, C., and Williams, B. R. (1997) Deficient cytokine signaling in mouse
embryo fibroblasts with a targeted deletion in the PKR gene: role of [RF-l and NF-
kappaB. Embo J 16, 406-416

Cuddihy, A. R., Li, S., Tam, N. W., Wong, A. H., Taya, Y., Abraham, N., Bell, J.
C., and Koromilas, A. E. (1999) Double-stranded—RNA-activated protein kinase
PKR enhances transcriptional activation by tumor suppressor p53. Mol Cell Biol 19,
2475-2484

Tamatani, M., Che, Y. H., Matsuzaki, H., Ogawa, S., Okado, H., Miyake, S.,
Mizuno, T., and Tohyama, M. (1999) Tumor necrosis factor induces Bcl-2 and Bcl-

x expression through NFkappaB activation in primary hippocampal neurons. J Biol
Chem 274, 8531-8538

Chen, C., Edelstein, L. C., and Gelinas, C. (2000) The Rel/NF-kappaB family
directly activates expression of the apoptosis inhibitor Bcl-x(L). Mol Cell Biol 20,
2687-2695

Wang, C. Y., Mayo, M. W., Komeluk, R. G., Goeddel, D. V., and Baldwin, A. S.,
Jr. (1998) NF-kappaB antiapoptosis: induction of TRAFl and TRAF2 and c-IAPl
and c-IAP2 to suppress caspase-8 activation. Science 281, 1680-1683

Wu, H., and Lozano, G. (1994) NF-kappa B activation of p53. A potential
mechanism for suppressing cell growth in response to stress. J Biol Chem 269,
20067-20074

Kucharczak, J ., Simmons, M. J ., Fan, Y., and Gelinas, C. (2003) To be, or not to be:
NF-kappaB is the answer--role of Rel/NF-kappaB in the regulation of apoptosis.
Oncogene 22, 8961-8982

66

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

Li, M., Shillinglaw, W., Henzel, W. J., and Beg, A. A. (2001) The Rela(p65)
subunit of NF -kappaB is essential for inhibiting double-stranded RNA-induced
cytotoxicity. J Biol Chem 276, 1185-1194

Willis, S. N., and Adams, J. M. (2005) Life in the balance: how BH3-only proteins
induce apoptosis. Curr Opin Cell Biol 17, 617-625

Takizawa, T., Tatematsu, C., and Nakanishi, Y. (2002) Double-stranded RNA-
activated protein kinase interacts with apoptosis signal-regulating kinase 1.
Implications for apoptosis signaling pathways. Eur J Biochem 269, 6126-6132

Ichijo, H., Nishida, B., Irie, K., ten Dijke, P., Saitoh, M., Moriguchi, T., Takagi, M.,
Matsurnoto, K., Miyazono, K., and Gotoh, Y. (1997) Induction of apoptosis by
ASKl, a mammalian MAPKKK that activates SAPK/JNK and p38 signaling
pathways. Science 275, 90-94

Chang, L., and Karin, M. (2001) Mammalian MAP kinase signalling cascades.
Nature 410, 37-40

Whitmarsh, A. J ., Cavanagh, J ., Toumier, C., Yasuda, J ., and Davis, R. J. (1998) A
mammalian scaffold complex that selectively mediates MAP kinase activation.
Science 281, 1671-1674

Davis, R. J. (2000) Signal transduction by the JNK group of MAP kinases. Cell 103,
239-252

Liu, J ., and Lin, A. (2005) Role of JNK activation in apoptosis: a double-edged
sword. Cell Res 15, 36-42

Zhang, H., Ozaki, I., Mizuta, T., Yoshimura, T., Matsuhashi, S., Eguchi, Y.,
Yasutake, T., Hisatomi, A., Sakai, T ., and Yamamoto, K. (2004) Transforming
growth factor-beta l-induced apoptosis is blocked by beta l-integrin-mediated
mitogen-activated protein kinase activation in human hepatoma cells. Cancer Sci
95, 87 8-886

Liedtke, C., Plumpe, J ., Kubicka, S., Bradham, C. A., Manns, M. P., Brenner, D. A.,
and Trautwein, C. (2002) Jun kinase modulates tumor necrosis factor-dependent
apoptosis in liver cells. Hepatology 36, 315-325

Chen, Y. N., Cheng, C. C., Chen, J. C., Tsauer, W., and Hsu, S. L. (2003)
Norcantharidin-induced apoptosis is via the extracellular signal-regulated kinase

and c-Jun-NH2-terminal kinase signaling pathways in human hepatoma HepG2
cells. Br J Pharmacol 140, 461-470

67

CHAPTER 3: Construction of Phenotype-Specific Gene Network for
Palmitate-Induced Cytotoxicity by Synergy Analysis

The work in this chapter has been accepted to be published in the book Systems
Analysis of Biological Networks:

Yang, X., Wang, X., Dalkic, E., Wu, M., and Chan, 0., Construction of
Phenotype-Specific Gene Network by Synergy Analysis. Systems Analysis of

Biological Networks, Chapter 5, 2009, in press

Abstract

Complex cellular activities, such as the saturated free fatty acid (FFA)-induced
cytotoxicity, are believed to be coordinately regulated by genes that function in a network.
Reconstructing these gene networks can provide insights into the molecular mechanisms
of cell physiology and thus represents a fundamental challenge in systems biology.
Elucidating the phenotype— gene interaction through the reconstruction of context-specific
gene networks remains elusive. In this chapter, I present a methodology that integrates
multi-level biological data to infer a cooperative gene network with respect to a specific
phenotype, palmitate-induced cytotoxicity. Our method introduces the concept of synergy
and builds a network that consists of gene pairs with significant synergistic relations,
which implies cooperation in producing cytotoxicity in response to palmitate. Scale-free
characteristics and multiple-hub genes are found in the network, revealing many
important cooperative candidate genes in regulating the FF A-induced cytotoxicity. These

candidates are supported by the literature.

68

Introduction

The development of most diseases can be traced to abnormal activities of the cells
in specific tissues or parts of the body. Hepatic disorders, such as non alcoholic fatty liver
disease (NAFLD) and non-alcoholic steatohepatitis (NASH), are associated with
saturated free fatty acid (F FA)-induced cytotoxicity of liver cells (1-3). Cellular activities
are tuned by regulatory machineries involving genes, proteins, and metabolites. For
instance, saturated FFA-induced cytotoxicity is coordinately regulated by a set of genes
that interact in a complex network (4). Therefore, reconstructing the gene networks that
give rise to the different phenotypes may provide insights into the cellular mechanisms
involved, ultimately, in the deve10pment of diseases and disorders (5, 6).

This chapter describes a methodology to reconstruct phenotypic and context-
specific gene networks based on the assumption that only a subset of the genes is relevant
to the target phenotype. The phenotype addressed in this chapter is saturated FFA-
induced cytotoxicity. Methods of gene selection relevant to a phenotype have been based
predominantly on fold changes in the genes across different conditions or correlations
between the genes and the phenotype, using statistical tests (7) or correlation measures
(8), respectively. Statistical tests typically yield too many genes for analysis, while
correlation measures only select genes that are statistically correlated with the phenotype,
thereby missing potentially relevant genes that are not highly correlated to the phenotype.
Incorporating prior information, such as gene set enrichment analysis (4) or trend profiles,
into the gene selection methods can help to alleviate these problems, however, the source

and quality of the prior knowledge will affect the results.

69

FF As modulate intracellular metabolic pathways involved in glucose (9),
triglyceride (TG) (10), and amino acid (1 l), metabolism. Tuned by the gene network (4,
8), some of these alterations are involved in the induction of cytotoxicity by saturated
FFAs (9-11). Therefore, integrating multiple levels of information, i.e., gene expression
and metabolite profiles, would better reflect the “multi-level” characteristic of cellular
activities, such as saturated FFA-induced cytotoxicity, and thereby aid in the selection of
genes involved in the observed phenotype, and in the reconstruction of a phenotype-
specific gene network.

Various methods, such as correlation (12), mutual information (l3, l4), and
Bayesian network analysis (15), have been used to construct gene networks. These
methods do not directly incorporate the phenotype in identifying the gene interactions.
Instead, they typically build a gene network for each of the conditions, and compare the
networks to identify the gene interactions that are specific to a condition or phenotype.
Consequently, these methods are computationally expensive and sensitive to the quality
of data. Since the size (i.e., the number of genes included in the network reconstruction)
and the noise level of the samples can affect the networks reconstructed for each of the
conditions, it is difficult to determine whether the differences in the networks across the
conditions are real changes in the mechanisms or simply an artifact due to the size or
noise levels. Alternatively, methods have been developed to select sets of gene pairs
relevant to a phenotype based on classification models, such as support vector machine
(16, 17), decision tree (18), and probabilistic model (19). Intuitively, if a phenotype-
prediction based on a pair of genes performs better than that based on either one of the

genes, then the pair of genes is suggested to have cooperative effects on the phenotype.

7O

However, these classification methods fail to differentiate the cooperative effects of the
genes pairs from the independent contributions of the individual genes (20). To address
this short-coming, in collaboration with Xuewei Wang, in this chapter I present a method
that distinguishes the difference in the cooperative vs. individual effects of the genes on
the phenotype.

Biological activities are regulated by multiple factors, many of which function
cooperatively, i.e., synergistically. The basic idea is that the whole (i.e., the regulatory
system) is greater than the sum of the individual parts (i.e., regulators) of a system.
Synergy is defined as the “additional” contribution provided by the “whole” as compared
to the sum of the contributions of the individual “parts”. An example of synergy can been
seen with transcription factors, such as GAT A4 and dHAND, which together
cooperatively and dramatically up-regulate cardiac (target) gene expression levels more
significantly than the sum of the effects fi'om either of the transcription factors alone (21).

The concept of synergy will be used in this chapter to assess the cooperative
effect of two genes on a phenotype. In a multivariate system, the synergistic effect of two
factors on a phenotype is the gain in the “mutual information” over the sum of the
information provided by each factor on a phenotype. A positive synergy denotes that two
factors regulate a phenotype, either cooperatively (e.g., co-activating) or antagonistically
(e.g., competitive inhibiting). Thus, one can predict the phenotype with a certain
confidence from either of the two factors; however, knowing both factors brings
additional information, which enhances the confidence of the prediction. Negative
synergy denotes redundancy, thus knowing both factors brings redundant information to

the prediction of the phenotype. Zero synergy denotes that at least one of the two factors

71

has no effect on the phenotype, and therefore brings neither additional nor redundant
information to the prediction of the phenotype.

Systematic assessment of synergy was first applied in neuroscience, where the
goal was to understand the neuron code by evaluating the strength of correlations
between the neurons upon activation by a stimulus (22, 23). More recently the concept of
synergy has been applied to the field of systems biology (24-26). Investigators developed
an information theoretic measure of synergy fi'om discretized gene expression data, and
applied this measure to identify cooperative gene interactions associated with neural
interconnectivity (24) and prostate cancer development (25). More recently, the concept
of synergy and the information theoretic measure of synergy have been applied directly
to continuous gene expression data (20).

In this chapter we introduce an integrative methodology to reconstruct phenotype-
specific gene networks based on synergy analysis. First, we select the phenotype-specific
genes by integrating the gene expression and metabolite profiles in the context of
saturated FFA-induced cytotoxicity. Next, we assess the synergistic effects between the
gene pairs. Unlike other computational methods used to identify gene interactions, the
fimdamental concept of synergy is to identify the cooperative gene interactions
responsible for the phenotype, and these cooperative gene interactions may or may not
be direct interactions. Finally, with the identified synergistic gene pairs, we build a
synergy network. Topological analyses reveal the structural characteristics of the network
while the hub genes provide insights into potential mechanism(s) involved in the

induction of the phenotype, i.e., saturated F FA-induced cytotoxicity.

72

Design of the Methodology

The Framework of the proposed experiments and methodology, described in a
flowchart (Fig. 3.1), for reconstructing the phenotype-specific gene network with synergy
analysis are as follows.

(1) Identify the phenotype of interest: cytotoxicity levels induced by saturated

FFA;

(2) Obtain metabolite data;

(3) Obtain gene-expression data with cDNA microarray;

(4) Select the metabolites that are associated with the phenotype, i.e. cytotoxicity,

(5) Select the genes by matching the trend of the metabolite and gene profiles;

(6) Obtain synergistic gene pairs by calculating their synergy scores;

(7) Construct a synergy network with the gene pairs that are sigrificantly

synergistic.

In the proposed methodology, the phenotype (cytotoxicity), metabolite and gene
expression profiles were collected as “inputs” and integrated as described above. The
anticipated results include the representative trends of the metabolites relevant to the
phenotype, genes that match the representative trends of the metabolites, and gene pairs
with significant synergy scores.

Gene pairs that have statistically sigiificant synergy scores indicate possible
membership of the gene pairs in a shared pathway or potential cross-talk between
different pathways. Graphical representation of these synergistic gene pairs yields a
network of gene-gene interactions that are associated with the phenotype. Topology

analysis of the synergy network reveals the characteristics of the network, such as degree

73

distribution, modularity and centrality. The hub genes, which have the highest number of

edges, suggest that they may be central regulators in the induction of the phenotype.

FFA treatment

11

HepGZ cells

(‘1/ (2); \(f)

Cytotoxicity Metabolites Gene expression
(4) |

 

Phenotype-related
metabolites

| (5)
l

Phenotype-related
genes
1(6)
Synergistic
gene pairs
in)
Synergy Network

 

 

Figure 3.1. Flowchart of the Proposed Methodology.

Materials and Methods

Cell Culture and Reagents, Fatty Acid Salt Treatment, and Cytotoxicity
Measurement—HepG2 cells were cultured and treated by sodium salts of palmitate and
oleate as previously described in Chapter 2. Cytotoxicity of the FFAs were measured
using the cytotoxicity detection kit (Roche Applied Science, Indianapolis, IN) as
previously described in Chapter 2 (27).

RNA Isolation for Gene Expression Profiling—Cells were cultured in 10 cm tissue

culture plates until confluence and then exposed to different treatments. RNA was

74

isolated with Trizol reagent. In detail, after the treatments, cells were washed twice with
cold phosphate buffered saline (PBS). 10 ml trizol reagent was added and kept for 5 - 10
minutes. The cell lysate was then transferred to 15 ml conical tubes and vortexed.
Meanwhile, phase-lock gel (heavy) in 15 ml tubes was centrifuged at 1000 g for 10 min.
Trizol cell lysate was added to tubes containing phase-lock gel. 3 ml of chloroform was
added to the tubes and mixed. The tubes were then centrifuged at 4000 g for 30 minutes
to separate the aqueous and organic phases, with the phase-lock gel forming a central
layer. The aqueous phase was poured off into fi'esh 15 ml conical tubes. 5 ml of isopropyl
alcohol was then added to the tubes containing the aqueous phase and tubes were mixed
gently by inversion. Tubes were then centrifuged at 4000 g for 20 minutes. RNA forms a
pellet at the bottom. The supernatant liquid is decanted and the pellets were washed thrice
with ice-cold 75% ethanol. Ethanol was then carefully removed and the pellet suspended
in 1 ml of water and the solution was transferred to 1.5 ml microcenuifuge tubes. For
LiCl precipitation, 0.5 ml of 7.5 M LiCl was added to the tubes and vortexed. The tubes
were then kept overnight at -20C. The next morning, tubes were centrifuged at 13000 g
for 45 min. The pellet was washed thrice with ice-cold 75% ethanol. Finally, all the
ethanol was removed and the pellet suspended in 100 ul RNAase-free water.

Micro-array Analysis—The gene expression profiles were obtained with cDNA
microarray. Analyses were done at the Van Andel Institute, Grand Rapids, MI. The
procedure of the microarray analysis was described previously (6). The detailed
procedure is available on line (28). In brief, labeled cDNAs were generated with Reverse
I Transcriptase (RT) reaction using Cy3 or Cy5 -labeled dCTP and low-dCTP dNTP mix.

After generating labeled cDNA, the template was degraded using Rnase and the cDNA

75

further purified using QIAquick columns (Qiagen, Valencia, CA). Hybridization
reactions were performed in a 50C water bath for 16 h, following which the microarrays
were washed and read.

There were two biological replicates for each condition and each replicate was
measured with the Cy3 and Cy5 dyes, i.e. there were two technical replicates/color swaps
for each biological replicate. Color swaps indicate the arrays in which the cDNA from the
treated sample was labeled with Cy3 dye and the cDNA from the control sample was
labeled with the Cy5 dye.

Metabolites Measurements—The concentrations of the various metabolites were
measured according to (l 1, 29-31). The metabolites include glucose, lactate, glycerol,
beta-hydroxybutyrate, triglycerides, acetoacetate, Cystine, and all the amino acids. All
the measured fluxes were normalized to total protein in the cell extract, measured with
the bicinchoninic acid (BCA) method (Pierce Chemicals, Rockford, IL).

Gene Selection Based on Trends of Metabolites—The statistical significance of
the changes in the metabolite levels across the conditions, i.e. BSA (control), Palmitate
and Oleate, were assessed using two-sample t-test for each metabolite. The metabolites
altered by saturated FFAs were previously identified (1 1). Eleven metabolites differed
sigrificantly across the three conditions, and four representative trends were extracted
from these metabolites (Fig. 3.2). 7394 genes remained after removing the
EST/hypothetical proteins and ORF of unknown functions fiom the list of ~20,000 genes.
Genes with expression patterns that matched the four representative metabolite trends

were selected. Two-sample t-test was applied to each gene to assess the sigiificance in

76

their fold change across the different conditions. Finally, 610 genes were selected from
the hill list of 7394 genes. The p-value cutoff was set at 0.05.

Calculation of the Synergy Scores of Gene Pairs—An information theory based
score was calculated to quantify the synergy between the genes (26). Given two genes,
G1 and G2, and a phenotype P, the synergy score between G1 and G2 with respect to the
phenotype P is defined as:

Syn(Gl,G2;P) = I (G1,G2;P) — [I (G1;P) + I (GZ;P)]
where I(G1;P) is the mutual information between G1 and P, I(G2;P) is the mutual
information between G2 and P, and I(G1,G2;P) is the mutual information between
(G1,G2) and P. This equation reflects the definition of synergy, the additional
contribution provided by the “whole” as compared to the sum of the contributions of the
individual “parts”. Mutual information (I) was calculated using a clustering-based
method from continuous data (20).

The synergy scores range from [-1 1]. A positive synergy score indicated that two
genes jointly provided additional information on the phenotype, a negative synergy score
indicated that the two genes provided redundant information about the phenotype, and a
zero score indicated that the two genes provided no additional information about the
phenotype. The 610 genes that were selected based on the metabolite trends generated
185745 gene pairs, of which 436 pairs had sigrificant synergy scores.

Permutation Test to Evaluate the Significance of the Synergy—A permutation test
was performed to assess the statistical sigrificance of the synergy of the gene pairs. The
phenotypes, i.e., toxic and nontoxic, were randomly permutated to be un-correlated with

the gene expression profiles. The synergy scores of the genes were then re-calculated

77

based on the permutated phenotype. This process was repeated 100 times to calculate the
p-values of the synergy score for each gene pair. Finally, Benjamin-Hochberg false
discovery rate procedure (32) was performed to adjust the p-values for all the gene pairs
and thereby control the expected false discoveries. The p-value cutoff was set at 0.05.
Characterization of the Network Topology—A synergy network was built with
gene pairs that have statistically significant synergy scores. The network composed of
nodes that represented the genes, and edges that represented the synergy of the gene pairs.
Graph theoretical (topological) analysis of the reconstructed gene networks was used to
assess the generated network and how it compared with other biological networks (33).
We characterized the topology of the synergy network by its degree distribution and
shortest path length. Degree distribution provides a distribution of the number of edges
associated with the nodes. Shortest path length is the lowest number of edges that

connect two nodes, and is measured using a bread-first search algorithm (34).

78

Figure 3.2. Four Representative Trends of the Metabolites. The metabolites altered by
saturated FFAs were previously identified (11). Eleven metabolites differed significantly
across the three conditions, and four representative trends were extracted fi'om these
metabolites. Trend 1: BSA < Palm and Palm > Ole; Trend II: BSA > Palm and Palm <
Ole; Trend 111: BSA < Palm < Ole; Trend IV: BSA > Palm > Ole.

79

ow

 

>_ .82»
20
8.5.".
.= 2:5
Sin.
0.0
__ 32p

_ .82»
20

Ein—

 

Results

Based upon the concept of synergy, we reconstructed a synergy network
specifically for the phenotype of saturated FFA-induced cytotoxicity. We first selected
the phenotype-relevant genes by integrating the metabolites altered by saturated FFAs
(11) with the global gene expression profile and extracted the genes that followed the
trends of the metabolites. From the selected genes, the synergy analysis revealed
synergistic gene pairs, which were used to build a synergy network. The reconstructed
network suggested potential gene targets that may play important roles in the induction of
the phenotype.

Topological characteristics of the synergy network—The synergy network,
shown in Figure 3.3, is composed of 292 genes with 436 connection edges. The synergy
network is characterized by relatively short path lengths, ranging fi'om 2 to 10 (Fig. 3.4),
while the characteristic path length, or average diameter, of the network is 4.872. The
network demonstrates small world characteristics of real networks (35), suggesting that
the propagation of communication between the genes is relatively fast.

The degree distribution, P(k), provides the probability that a randomly selected
vertex has k links to its neighbors. A power law distribution suggests that P(k) ~k'7, where
k is the degree and y is the degree exponent, and in most biological, scale-free networks y
ranges around 2 and 3 (35). The degree distribution of our synergy network is shown in
Figure 3.5, and 7 ~ 2, suggesting the synergy network, similar to other biological
networks, is scale-free. Therefore, most of the genes are sparsely connected, while few of
the genes (hubs) are connected to many genes and play important roles in sustaining the

integity of the network, which suggest their importance in the biological function or the

81

phenotype. In summary, the topology of the synergy network differs from the bell-like

Poisson distribution characteristic of a random and statistically homogeneous network

and suggests the existence of hub genes.

 

tau-00».

Figure 3.3. The Synergy Network. The synergy network is composed of 292 genes with
436 connection edges. The size of the nodes indicates its degree.

20000"

 
 
   
 
 
 
 

16000-

12000 -

8000 -

Frequency

4000‘

 

7 8 9 10 11 12 13

1 2 3 4 5 6
Shortest path length

Figure 3.4. The Distribution of Shortest Path Lengths in the Synergy Network.

82

N
or

   
    

 

 

 

2
if
g 1.5 - Log(Ndegree) = -1.9Log(degree) + 2.3005
8 e
E ‘l l . O

0.5 -
o I I I r
0 0.2 0.4 0.6 0.8 1 1 .2 1 .4 1 .6

Log(degree)

Figure 3.5. The Degree Distribution of the Synergy Network. For a degee value k,
Ndegree is the number of genes with degree k. in the network.

Hub genes in the network—The genes in the synergy network are listed and
ranked by their degree. Table 3.1 lists the genes with the highest degree, which are
therefore “hub genes” in the synergy network. These genes include P4HA1, AHDC1,

MACF1, INSIG2, and SH3RF2.

Table 3.1. The hub genes in the synergy network.

 

 

Gene Symbol Degree Full Name

P4HA1 22 procollagen-proline, 2-oxoglutarate 4-dioxygenase
(proline 4-hydroxylase), alpha polypeptide

AHDC1 20 AT hook, DNA binding motif, containing 1

MACF1 19 microtubuIe-actin crosslinking factor 1

lNSlGZ 18 insulin induced gene 2

SH3RF2 13 SH3 domain containing ring finger 2

 

P4HA1, proline 4-hydroxylase, alpha polypeptide, is the alpha subunit of the

protein proline 4-hydroxylase (P4H). P4H is a key enzyme involved in the biosynthesis

83

of collagens (36, 37). Collagens, the most abundant protein in the extracellular matrix
(ECM) and the main protein of connective tissues, play important roles in regulating
cellular activities and are linked to multiple diseases, including cardiovascular diseases
and cancers (3 8-40). In liver, alteration in the synthesis of collagen is related to liver cell
apoptosis, hepatic fibrosis and cirrhosis (41 -43). Physiologically, the synthesis,
processing, secretion, and deg'adation of collagens are tightly modulated by regulatory
factors, including; P4HA1. As a critical functional subunit of P4H, P4HA1 is involved in
the post-translational modification of procollagen (36, 44). As shown in Figure 3.6,
P4HA1 is located within the endoplasmic reticulum and catalyzes the post-translational
formation of 4-hydroxyproline in the -Xaa-Pro-Gly- sequences in procollagens, which is
essential for proper folding of the procollagen polypeptide (44). Inhibiting P4H generates
unstable intracellular collagens which cannot be secreted (45). On the other hand, over-
expressing P4HA1 causes excess synthesis of collagen (46). The deregulation of P4H or
P4HA1 in collagen synthesis has been attributed to cytotoxicity and apoptosis in various
types of cells (47-50). P4HA1 regulates the ECM components by controlling the
synthesis and secretion of procollagens, which modulates fibrosis, cell proliferation, and
apoptosis (47-50). The role of P4HA1 in palmitate-induced lipotoxicity is unclear.
However in our data, the level of P4HA1 is sigrificantly down-regulated in palmitate as
compared to the control (p= 0.0014) and oleate (p=0.018) samples. Therefore, it is
plausible that palmitate-induced cytotoxicity is mediated, in part, by P4HA1 through
altered synthesis or improper folding of collagen peptides, although experimental

validations are needed to confirm this hypothesis.

84

o o
2-ketoglutarate procollagen-L-proline
proline 4-hydroxy1ase
0 '1" Fit
N
H0 0

C 02 + NW +

O HO

succinate procollagen trans 4-hydroxy-L-proline

P4HA1: prolyl 4-hydroxylase is a key enzyme in collagen synthesis

Figure 3.6. P4HA1 catalyzes the formation of 4-hydroxyproline. Graph adapted from
MetaCyc database (http://www.metacyc.org/) (51).

In addition, a number of proteins have been identified or predicted to interact with
P4HA1, fi'om publicly available protein-protein interaction databases, such as KEGG (52)
and STRING (51, 52). An interaction network (not shown) obtained fi'om STRING
shows some of the proteins that potentially interact with P4HA1. P4HA1 is centrally
positioned and integates different gene clusters in this network, supporting our result that
P4HA1 is a hub gene (with the highest degree, 22) in the synergy network.

AHDC1, AT hook, DNA binding motif, containing 1, contains 2 AT hook DNA
binding domains, and can be phosphorylated upon DNA damage, probably by ATM or
ATR (55). Although the in vivo function or phenotype of this gene has not been identified,

it is known that AHDC1 encodes 7 different isoforrns, some of which contain HMG-l and

85

HMG-Y, DNA-binding domains (56). HMG proteins are involved in nucleosome phasing,
3' end processing of mRNA transcripts, and transcription of genes close to AT rich
regions, and are thereby related to the pathogenesis of inflammatory and autoimmune
diseases (57-59). Although the fimction of AHDC1 is currently unknown, it may be
involved in DNA damage and inflammatory responses. The significant alteration of this
gene by palmitate (p=0.015 for palmitate vs. BSA and 0.014 for palmitate vs. oleate) and
the identification of this gene in the synergy network hints at the possibility that palmitate
may affect DNA damage and inflammatory responses through AHDC1.

MACF1, microtubule-actin crosslinking factor 1, also called ACF7 (actin cross-
linking factor 7), is a member of the spectrale family of cytoskeletal cross—linking
proteins that possess actin- and microtubule-binding domains (60, 61). It may be involved
in microtubule dynamics to facilitate actin-microtubule interactions at the cell periphery
and in coupling the microtubule network to cellular junctions (56). Cell-cell contact and
cell-surface interactions through the cytoskeleton and ECM are involved in the control
and regulation of cell motility, tissue remodeling, gene expression, differentiation, and
proliferation (62). In the literature, a large number of cytoskeletal and ECM genes were
found to be down-regulated by palmitate treatment (63, 64). Therefore, it is possible that
our method has identified two central genes, i.e. P4HA1 and MACF1, in mediating the
effect of palmitate on the ECM and cytoskeletal structure.

INSIG2, insulin-induced gene 2, encodes a protein that blocks the processing of
sterol regulatory element binding proteins (SREBPs), which regulate human lipogenic
and adipocyte metabolism (65). As shown in Figure 3.8, in endoplasmic reticulum (ER),

SREBP Cleavage-Activating Protein (SCAP) can bind to the regulation domain of

86

SREBP and transfer SREBP into Golgi Apparatus, where SCAP activates protease SlP to
cleave the regulation domain and activate the transcription activation/DNA binding
domain of SREBP. Activated SREBP then can be transported to the nucleus to bind to the
cis-element SRE to promote the expression of a series of enzymes that are involved in
lipid synthesis (66). INSIG2 can bind to SCAP and inhibit its function, thereby blocking
lipid synthesis (67). Indeed, reduced INSIG2 levels in adipocytes resulted in SREBP
activation, which increased the expression of genes involved in adipogenesis (68). In our
system of HepG2 cells, microarray analysis found that the gene expression levels of
IN SIG2 were down-regulated by both palrrritate (p=0.011) and oleate (p=0.14). Therefore,
the suppression of INSIG2 expression likely contributes to the increased TG synthesis

observed in the FFA cultures (10).

 

Figure 3.7. INSIG2 plays important role in lipid synthesis. (66, 67) Figure adapted
from Ref (66). By binding to SCAP, INSIG2 blocks the processing of SREBPs and
therefore suppresses lipid synthesis. '

87

SH3RF2, SH3 domain containing ring finger 2, is a putative protein whose
function is unknown, but from sequence analysis SH3RF2 contains 3 Src homology 3
(SH3) domains and a RING-type zinc finger domain (Fig. 3.9, from the InterPro
database). RING-type zinc finger domain is found in many E3 ubiquitin—protein ligases
(69). E3 ubiquitin-protein li gases determine the substrate specificity for ubiquitination
and are therefore involved in targeting proteins for degradation by the Ubiquitin-
Proteasome System (70). During this process, RING fingers, by interacting with E2
ubiquitin-conjugating enzymes, promotes ubiquitination (71 , 72). Although the exact
function of SH3RF2 is not known, the RING-type zinc finger domain suggests SH3RF2
as a putative E3 ubiquitin-protein ligase, which may be involved in the protein
degradation pathway through the Ubiquitin-Proteasome System. In addition, SH3RF2
contains another type of domain, SH3 domains, which are present in many proteins
involved in intracellular sigral transduction pathways (73, 74). SH3 domains recognize
and bind to the proline-rich motifs (-X-P-P-X-P-) on the associated proteins. Therefore,
SH3 domains are recruited by the sigraling proteins to direct protein-protein interactions
and thereby specify distinct regulatory pathways mediated by different protein binding
domains (75, 76). Taken together, the SH3 domains of SH3RF2 could potentially serve as
a targeting domain that determines the substrate specificity of SH3RF2 as a putative E3
ubiquitin—protein ligase, therefore facilitating SH3RF2 to subject certain types of proteins
to degradation via the ubiquitin-proteasome system. Interestingly, this gene is
significantly up-regulated in the palmitate culture (p=0.019 for palmitate vs. BSA and
0.025 for palmitate vs. oleate). This result is consistent with the literature suggesting that

palmitate induces protein degradation.

88

 

 

——an RING SH3

 

 

 

 

 

 

 

m: —

Figure 3.8. The Protein Domains of SH3RF2. Figure modified fiom the InterPro
database. SH3RF2 contains 3 Src homology 3 (SH3) domains and a RING-type zinc
finger domain.

 

 

 

Chronic treatment of palmitate increases the levels of unfolded or misfolded
proteins, which induces Endoplasmic reticulum (ER) stress (77, 78). This lends support
to our finding that P4HA1 was down-regulated in palmitate, suggesting the possibility
that procollagen may be misfolded in the palmitate culture as compared to the oleate and
control cultures. The accumulation of unfolded or misfolded proteins activates the
ubiquitin-proteasome system. Indeed, recent studies found palmitate strongly enhances
ubiquitination by activating E3 ubiquitin ligases, resulting in enhanced protein
degradation. Therefore, SH3 RF2, putative ubiquitin-protein ligase, may be potentially
involved in palmitate-induced cytotoxicity by triggering ubiquitination. In addition,
SH3RF2, composed of 3 SH3 and a zinc finger domain which are all protein-protein
interaction domains, could recruit a diverse set of proteins for degradation, supporting its
central position in our synergy network. Therefore, the synergy network identified a
novel protein, SH3RF 2, which potentially plays a central role in palmitate-induced
cytotoxicity. The domain knowledge of this protein suggests SH3RF2 may be involved in
palmitate-induced cytotoxicity by recruiting unfolded proteins and triggering
ubiquitination.

In summary, we built a synergy network specifically for palmitate-induced
cytotoxicity. This network is scale-free and has multiple hub genes. The hub genes are
related to cellular activities, such as cellular contact, cytotoxicity, metabolic pathways,

protein degradation, which may play important roles in palmitate-induced cytotoxicity.

89

Therefore, these hub genes suggest potential mechanisms involved in palnnitate-induced

cytotoxicity.

Discussion

Phenotype-specific gene network reconstruction is a useful approach to extract
gene interaction information from nnicro-array data and to help provide insight into
disease mechanisms. Multiple methods, e. g., meta-analysis (79) and pair-wise relevance
(80), have been developed to reconstruct gene networks associated with diseases. The
information theoretic measure of synergy provides a convenient method to identify
cooperative gene pairs with respect to a phenotype. Therefore in this chapter, we present
an alternative strategy to reconstruct phenotype-specific networks based on the concept
of synergy.

A major concern that often arises in network reconstruction using microarray data
is the high computational cost. To alleviate this limitation, we pre-select a subset of genes
by matching the trend of their expression profiles, across the different conditions, to the
profiles of the phenotype-relevant metabolites. This step reduces the nmnber of genes to
be analyzed. Concomitantly, the trend-based analysis allows the incorporation of prior
knowledge of the gene expression patterns. In other words, it permits the inclusion of
genes of particular interest that are known to be related to the phenotype, whether or not
these gene profiles are statistically correlated with the phenotype.

Gene pairs with statistically significant synergy scores suggest potential
combinatorial effects of those genes on the phenotype. However, the scores cannot

distinguish between the types of combinatorial effects, such as additive or antagonistic.

9O

Nevertheless, this limitation can be addressed by integating physical interaction data (i.e.
protein-protein and protein-DNA interaction) into the analysis.

In addition, the proposed analysis pipeline consists of several steps (see Fig. 3.1),
including selection of genes relevant to the phenotype, calculation of the synergy scores
for each gene pair, evaluation of the statistical significance of the synergy score, and
analysis of the network t0pology to identify biologically relevant genes. The methods for
these steps are not limited to those proposed in this chapter; alternative methods for each
of the steps in the framework can be used. For example, pattern recognition methods can
be used to select the genes, discretization-based entropy estimation can be used to
calculate the synergy score, Bayesian FDR control procedure can be used to evaluate the
statistical significance of the synergy score, etc. A comparison of the different methods
for each step could be performed to determine the optimal procedure for each step.

In summary, in this chapter we have achieved the following goals: 1). Integrated
gene and the metabolite profiles to identify a select group of genes that may be involved
in palmitate-induced cytotoxicity; 2). Reconstructed a phenotype-specific synergy
network. Topology analysis of the synergy network revealed scale-flee characteristics
and multiple hub genes, which are typical characteristics shared by many biological
networks. These hub genes suggest potential mecharnisms and may be targets for

modulating palmitate-induced cytotoxicity.

91

References

Feldstein, A. B., Werneburg, N. W., Canbay, A., Guicciardi, M. B., Bronk, S. F.,
Rydzewski, R., Burgart, L. J ., and Gores, G. J. (2004) Free fatty acids promote
hepatic lipotoxicity by stimulating TN F-alpha expression via a lysosomal pathway.
Hepatology 40, 185-194

Lu, Z. H., Mu, Y. M., Wang, B. A., Li, X. L., Lu, J. M., Li, J. Y., Pan, C. Y.,
Yanase, T., and Nawata, H. (2003) Saturated free fatty acids, palmitic acid and
stearic acid, induce apoptosis by stimulation of ceramide generation in rat testicular
Leydig cell. Biochem Biophys Res Commun 303, 1002-1007

Farrell, G. C., and Latter, C. Z. (2006) Nonalcoholic fatty liver disease: from
steatosis to cirrhosis. Hepatology 43, 899-5112

Li, Z., Srivastava, S., Yang, X., Mittal, S., Norton, P., Resau, J., Haab, B., and Charn,

C. (2007) A hierarchical approach employing metabolic and gene expression
profiles to identify the pathways that confer cytotoxicity in HepG2 cells. BMC Syst
Biol 1, 21

Said, M. R., Begley, T. J ., Oppenheim, A. V., Lauffenburger, D. A., and Samson, L.
D. (2004) Global network analysis of phenotypic effects: protein networks and

toxicity modulation in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 101,
18006-18011

Srivastava, S., Li, Z., Yang, X., Yedwabnick, M., Shaw, S., and Chan, C. (2007)
Identification of genes that regulate multiple cellular processes/responses in the
context of lipotoxicity to hepatoma cells. BMC Genomics 8, 364

Tusher, V. G., Tibshirani, R., and Chu, G. (2001) Significance analysis of
microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98,
5116-5121

Li, Z., and Chan, C. (2004) Integrating gene expression and metabolic profiles. J
Biol Chem 279, 27124-27137

Lam, T. K., Carpentier, A., Lewis, G. F., van de Werve, G., Fantus, I. G., and
Giacca, A. (2003) Mecharnisms of the free fatty acid-induced increase in hepatic
glucose production. Am J Physiol Endocrinol Metab 284, E863-873

92

10.

ll.

12.

l3.

14.

15.

l6.

17.

18.

19.

20.

21.

Listenberger, L. L., Han, X., Lewis, S. E., Cases, 8., Farese, R. V., Jr., Ory, D. S.,
and Schaffer, J. E. (2003) Triglyceride accumulation protects against fatty acid-
induced lipotoxicity. Proc Natl Acad Sci U S A 100, 3077-3082

Srivastava, S., and Chan, C. (2008) Application of metabolic flux analysis to
identify the mechanisms of fi'ee fatty acid toxicity to human hepatoma cell line.
Biotechnol Bioeng 99, 399-410

Eisen, M. B., Spellman, P. T., Brown, P. O., and Botstein, D. (1998) Cluster
analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A
95, 14863-14868

Liang, K. C., and Wang, X. (2008) Gene regulatory network reconstruction using

conditional mutual information. E URASIP journal on bioinformatics & systems
biology, 253 894

Basso, K., Margolirn, A. A., Stolovitzky, G., Klein, U., Dalla-Favera, R., and
Califano, A. (2005) Reverse engineering of regulatory networks in human B cells.
Nature genetics 37, 382-390

Pe'er, D., Regev, A., Elidan, G., and Friedman, N. (2001) Inferring subnetworks
from perturbed expression profiles. Bioinformatics 17 Suppl 1, 8215-224

Furey, T. S., Cristianini, N., Duffy, N., Bednarski, D. W., Schurnmer, M., and
Haussler, D. (2000) Support vector machine classification and validation of cancer
tissue samples using microarray expression data. Bioinformatics 16, 906-914

Tang, E. K., Sugantlnan, P. N., and Yao, X. (2006) Gene selection algorithms for
microarray data based on least squares support vector machine. BM C
Bioinformatics 7, 95

Diaz-Uriarte, R., and Alvarez de Andres, S. (2006) Gene selection and
classification of microarray data using random forest. BMC Bioinformatics 7, 3

Paul, T. K., and Iba, H. (2005) Gene selection for classification of cancers using
probabilistic model building genetic algorithm. Bio Systems 82, 208-225

Watkinson, J., Wang, X., Zheng, T., and Anastassiou, D. (2008) Identification of
gene interactions associated with disease from gene expression data using synergy
networks. BMC Syst Biol 2, 10

Dai, Y. S., Cserjesi, P., Markham, B. E., and Molkentin, J. D. (2002) The
transcription factors GATA4 and dHAND physically interact to synergistically

93

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

activate cardiac gene expression through a p3 OO-dependent mechanism. J Biol
Chem 277, 24390-24398

Schneidrnan, B., Bialek, W., and Berry, M. J., 2nd (2003) Synergy, redundancy,
and independence in population codes. J Neurosci 23, 11539-11553

Brenner, N., Strong, S. P., Koberle, R., Bialek, W., and de Ruyter van Steveninck,
R. R. (2000) Synergy in a neural code. Neural Comput 12, 1531-1552

Varadan, V., Miller, D. M., 3rd, and Anastassiou, D. (2006) Computational
inference of the molecular logic for synaptic connectivity in C. elegans.
Bioinformatics 22, e497-506

Varadan, V., and Anastassiou, D. (2006) Inference of disease-related molecular
logic fi'om systems-based microarray analysis. PLoS Comput Biol 2, e68

Anastassiou, D. (2007) Computational analysis of the synergy among multiple
interacting genes. Mol Syst Biol 3, 83

Srivastava, S., and Chan, C. (2006) Hydrogen peroxide and hydroxyl radicals
mediate palmitate-induced cytotoxicity to hepatoma cells: Relation to
mitochondrial permeability transition. Free Radic Res 41, 38-49

Institute, V. A. R. Standard Operating Procedures

Chan, C., Berthiaume, F., Lee, K., and Yarmush, M. L. (2003) Metabolic flux
analysis of hepatocyte function in horrnone- and amino acid-supplemented plasma.
Metab Eng 5, 1-15

Olsen, C. (1971) An enzymatic fluorimetric nnicromethod for the determination of
acetoacetate, -hydroxybutyrate, pyruvate and lactate. Clin Chim Acta 33, 293-300

Fiskerstrand, T., Refsum, H., Kvalheinn, G., and Ueland, P. M. (1993)
Homocysteine and other tlniols in plasma and urine: automated determination and
sample stability. Clin Chem 39, 263-271 ’

Benjarnini, Y., and Hochberg, Y. (1995) Controlling the False Discovery Rate - a
Practical and Powerful Approach to Multiple Testing. J Roy Stat Soc B Met 57,
289-300

Christensen, C., Gupta, A., Maranas, C. D., and Albert, R. (2007) Large-scale
inference and gaph-theoretical analysis of gene-regulatory networks in B-Subtilis.
Physica A 373, 796-810

94

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

Cormern, T. H., Leiserson, C. E., and Rivest, R. L. (2001) Introduction to
Algorithms, MIT press

Barabasi, A. L., and Oltvai, Z. N. (2004) Network biology: Understanding the cell's
functional organization. Nat Rev Genet 5, 101-U1 15

Chen, L., Shen, Y. H., Wang, X., Wang, J., Gan, Y., Chen, N., LeMaire, S. A.,
Coselli, J. S., and Wang, X. L. (2006) Human prolyl-4-hydroxylase alpha(l)
transcription is mediated by upstream stimulatory factors. J Biol Chem 281, 10849-
10855

Annunen, P., Autio-Harmainen, H., and Kivirikko, K. I. (1998) The novel type II
prolyl 4-hydroxylase is the main enzyme fornn in chondrocytes and capillary

endotlnelial cells, whereas the type I enzyme predominates in most cells. J Biol
Chem 273, 5989-5992

Bedossa, P., and Paradis, V. (2003) Liver extracellular matrix in health and disease.
J Pathol 200, 504-515

Rodriguez-Feo, J. A., Sluijter, J. P., de Kleijn, D. P., and Pasterkarnp, G. (2005)
Modulation of collagen turnover in cardiovascular disease. Curr Pharm Des 11,
2501-2514

Stone, P. J. (2000) Potential use of collagen and elastin degradation markers for
monitoring liver fibrosis in schistosonniasis. Acta Trop 77, 97-99

Bickel, M., Baringhaus, K. H., Gerl, M., Gunzler, V., Kanta, J ., Schmidts, L., Stapf,
M., Tschank, G., Weidmann, K., and Werner, U. (1998) Selective inhibition of

hepatic collagen accumulation in experimental liver fibrosis in rats by a new prolyl
4-hydroxylase inhibitor. Hepatology 28, 404-411

Clement, B., Chesne, C., Satie, A. P., and Guillouzo, A. (1991) Effects of the prolyl
4-hydroxylase proinhibitor HOE 077 on human and rat hepatocytes in primary
culture. J Hepatol 13 Suppl 3, 841-47

Faouzi, 8., Le Bail, B., Neaud, V., Boussarie, L., Saric, J., Bioulac-Sage, P.,
Balabaud, C., and Rosenbaum, J. (1999) Myofibroblasts are responsible for
collagen synthesis in the stroma of human hepatocellular carcinoma: an in vivo and
in vitro study. J Hepatol 30, 275-284

Myllyharju, J. (2003) Prolyl 4-hydroxylases, the key enzymes of collagen
biosyntlnesis. Matrix Biol 22, 15-24

95

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

Rocnik, E. F., Chan, B. M., and Pickering, J. G. (1998) Evidence for a role of
collagen synthesis in arterial smooth muscle cell migration. J Clin Invest 101, 1889-
1898

John, D. C., Watson, R., Kind, A. J., Scott, A. R., Kadler, K. E., and Bulleid, N. J.
(1999) Expression of an engineered form of recombinant procollagen in mouse
milk. Nat Biotechnol 17, 385-389

Xia, S. H., Wang, J., and Kang, J. X. (2005) Decreased n-6/n-3 fatty acid ratio
reduces the invasive potential of human lung cancer cells by downregulation of cell
adhesion/invasion-related genes. Carcinogenesis 26, 779-784

Huet, C., Monget, P., Pisselet, C., and Monniaux, D. (1997) Changes in
extracellular matrix components and steroidogenic enzymes during growth and
atresia of antral ovarian follicles in the sheep. Biol Reprod 56, 1025-1034

Dong, M. S., Jung, S. H., Kim, H. J., Kim, J. R., Zhao, L. X., Lee, E. S., Lee, E. J.,
Yi, J. B., Lee, N., Cho, Y. B., Kwak, W. J., and Park, Y. I. (2004) Structure-related
cytotoxicity and anti-hepatofibric effect of asiatic acid derivatives in rat hepatic
stellate cell-line, HSC-T6. Arch Pharm Res 27, 512-517

Ju, H., Hao, J ., Zhao, S., and Dixon, 1. M. (1998) Antiproliferative and antifibrotic
effects of mimosine on adult cardiac fibroblasts. Biochim Biophys Acta 1448, 51-60

Caspi, R., Foerster, H., Fulcher, C. A., Kaipa, P., Krurnmenacker, M., Latendresse,
M., Paley, S., Rhee, S. Y., Shearer, A. G., Tissier, C., Walk, T. C., Zhang, P., and
Karp, P. D. (2008) The MetaCyc Database of metabolic pathways and enzymes and
the BioCyc collection of Pathway/Genome Databases. Nucleic Acids Res 36, D623-
631

Aoki, K. F., and Kanehisa, M. (2005) Using the KEGG database resource. Curr
Protoc Bioinformatics Chapter 1, Unit 1 12

von Mering, C., Jensen, L. J., Snel, B., Hooper, S. D., Krupp, M., Foglierini, M.,
Jouffre, N., Huynen, M. A., and Bork, P. (2005) STRING: known and predicted

protein-protein associations, integrated and transferred across organisms. Nucleic
Acids Res 33, D43 3-437

von Mering, C., Jensen, L. J ., Kuhn, M., Chaffron, S., Doerks, T., Kruger, B., Snel,
B., and Bork, P. (2007) STRING 7--recent developments in the integration and
prediction of protein interactions. Nucleic Acids Res 35, D358-362

96

55.

56.

57.

58.

59.

60.

61.

'62.

63.

65.

Matsuoka, S., Ballif, B. A., Smogorzewska, A., McDonald, E. R., 3rd, Hurov, K. B.,
Luo, J ., Bakalarski, C. B., Zhao, Z., Solimirni, N., Lerenthal, Y., Shiloh, Y., Gygi, S.
P., and Elledge, S. J. (2007) ATM and ATR substrate analysis reveals extensive
protein networks responsive to DNA damage. Science 316, 1160-1166

Thierry-Mieg, D., and Thierry-Mieg, J. (2006) AceView: a comprehensive cDNA-
supported gene and transcripts annotation. Genome Biol 7 Suppl 1, 312 11-14

Voll, R. E., Urbonaviciute, V., Herrmann, M., and Kalden, J. R. (2008) High
mobility group box 1 in the pathogenesis of inflammatory and autoimmune diseases.
Isr Med Assoc J 10, 26-28

Tesniere, A., Panaretakis, T., Kepp, 0., Apetoh, L., Ghiringhelli, F., Zitvogel, L.,
and Kroemer, G. (2008) Molecular characteristics of immunogenic cancer cell
death. Cell Death Differ 15, 3-12

Jiang, W., and Pisetsky, D. S. (2007) Mechanisms of Disease: the role of high-
mobility group protein 1 in the pathogenesis of inflammatory arthritis. Nat Clin
Pract Rheumatol 3, 52-58

Kodama, A., Karakesisoglou, I., Wong, B., Vaezi, A., and Fuchs, E. (2003) ACF7:
an essential integrator of microtubule dynamics. Cell 115, 343-354

Gong, T. W., Besirli, C. G., and Lomax, M. I. (2001) MACF1 gene structure: a
hybrid of plectin and dystrophin. Mamm Genome 12, 852-861

Zarnir, E., and Geiger, B. (2001) Molecular complexity and dynamics of cell-matrix
adhesions. J Cell Sci 114, 3583-3590

Draghici, S., Khatri, P., Tarca, A. L., Amin, K., Done, A., Voichita, C., Georgescu,
C., and Romero, R. (2007) A systems biology approach for pathway level analysis.
Genome Res 17, 1537-1545

Swagell, C. D., Henly, D. C., and Morris, C. P. (2005) Expression analysis of a
human hepatic cell line in response to palrrnitate. Biochem Biophys Res Commun
328, 432-441

Krapivner, S., Popov, S., Chemogubova, B., Hellenius, M. L., Fisher, R. M.,
Hamsten, A., and van't Hooft, F. M. (2008) Insulin-induced gene 2 involvement in
human adipocyte metabolism and body weight regulation. J Clin Endocrinol Metab
93, 1995-2001

97

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

Horton, J. B., Goldstein, J. L., and Brown, M. S. (2002) SREBPs: activators of the
complete mom of cholesterol and fatty acid synthesis in the liver. J Clin Invest
109,1125-1131

Yabe, D., Brown, M. S., and Goldsteirn, J. L. (2002) Insig-2, a second endoplasmic
reticulum protein that binds SCAP and blocks export of sterol regulatory elernent-
binding proteins. Proc Natl Acad Sci U S A 99, 12753-12758

Rosen, E. D., Walkey, C. J., Puigserver, P., and Spiegelman, B. M. (2000)
Transcriptional regulation of adipogenesis. Genes Dev 14, 1293-1307

Freemont, P. S. (1993) The RING finger. A novel protein sequence motif related to
the zinc finger. Ann N Y Acad Sci 684, 174—192

Hershko, A., Heller, H., Eytan, E., and Reiss, Y. (1986) The protein substrate
binding site of the ubiquitin-protein ligase system. J Biol Chem 261, 11992-11999

Freemont, P. S. (2000) RING for destruction? Curr Biol 10, R84-87

Barinaga, M. (1999) A new finger on the protein destruction button. Science 286,
223, 225

Cohern, G. B., Ren, R., and Baltimore, D. (1995) Modular binding domains in signal
transduction proteins. Cell 80, 237-248

Pawson, T. (1995) Protein modules and signalling networks. Nature 373, 573-580

Ren, R., Mayer, B. J ., Cicchetti, P., and Baltimore, D. (1993) Identification of a ten-
arnino acid proline-rich SH3 binding site. Science 259, 1157-1161

Morton, C. J ., and Campbell, I. D. (1994) SH3 domains. Molecular 'Velcro'. Curr
Biol 4, 615-617

Guo, W., Wong, S., Xie, W., Lei, T., and Luo, Z. (2007) Palmitate modulates
intracellular signaling, induces endoplasmic reticulum stress, and causes apoptosis

in mouse 3T3-Ll and rat primary preadipocytes. Am J Physiol Endocrinol Metab
293, E576-586

Karaskov, B., Scott, C., Zhang, L., Teodoro, T., Ravazzola, M., and Volchuk, A.
(2006) Chronic palnnitate but not oleate exposure induces endoplasmic reticulum

stress, which may contribute to INS-1 pancreatic beta-cell apoptosis.
Endocrinology 147, 3398-3407

98

79.

80.

Rasche, A., Al-Hasani, H., and Herwig, R. (2008) Meta-analysis approach
identifies candidate genes and associated molecular networks for type-2 diabetes
mellitus. BMC Genomics 9, 310

Jiang, W., Li, X., Rao, S., Wang, L., Du, L., Li, C., Wu, C., Wang, H., Wang, Y.,
and Yang, B. (2008) Constructing disease-specific gene networks using pair-wise
relevance metric: application to colon cancer identifies interleukin 8, desmin and
enolase 1 as the central elements. BMC Syst Biol 2, 72

99

CHAPTER 4: Reconstruct Modular Phenotype-Specific Gene
Networks by Knowledge-Driven Matrix Factorization

The work in this chapter has been published in Bioinformatics:
Yang, X., Zhou, Y., Jin, R. and Chan, C. (2009) Reconstruct Modular
Phenotype-Specific Gene Networks by Knowledge-Driven Matrix Factorization,

Bioinformatics.

Abstract

Reconstructing gene networks from rrnicroarray data is believed to be able to
provide mechanistic information on cellular processes. A popular structure learning
method, Bayesian network inference, has been used to determine network topology
despite its shortcomings, i.e., the high computational cost when analyzing a large number
of genes and the inefficiency in exploiting prior knowledge, such as the coregulation
information of the genes. To address these limitations, in collaboration with computer
scientists, we introduced an alternative method, krnowledge-driven matrix factorization
(KMF) fiamework, to reconstruct phenotype-specific modular gene networks.
Considering the reconstruction of gene network as a matrix factorization problem, we
first use the gene expression data to estimate a correlation matrix, and then factorize the
correlation matrix to recover the gene modules and the interactions between them. Prior
knowledge from Gene Ontology is integrated into the matrix factorization. We applied
this KMF algorithm to hepatocellular carcinoma (HepG2) cells treated with free fatty

acids (FFAs). By comparing the module networks for the different conditions, we

lOO

identified the specific modules that are involved in conferring the cytotoxic phenotype
induced by palnnitate. Furtlner analysis of the gene modules of the different conditions
suggested individual genes that play important roles in palnnitate-induced cytotoxicity. In
summary, KMF can efficiently integrate gene expression data with prior knowledge,
thereby providing a powerful method of reconstructing phenotype-specific gene networks

and valuable insights into the mechanisms that govern the phenotype.

Introduction

Cellular activities are believed to be coordinately regulated by genes and proteins
that function in complex networks. Reconstructing the gene networks that give rise to the
different phenotypes may provide insights into the cellular mechanisms involved (1, 2).
Biological networks of protein-protein interaction (3), metabolic pathways (4) and
transcriptional regulation (5) are modular in structure, enabling mutations to be isolated
to specific modules without affecting the overall viability of the system (6-8). Since
organized modularity is ubiquitous in biological systems, identifying the gene modules
and their interplay in a modular network should provide insights into the differential
mechanisms involved in normal vs. disease states.

Previously we identified that saturated FFA, e. g. palrrnitate, induced cytotoxicity
in liver cells, while unsaturated FFAs, e.g. oleate and linoleate, were not significantly
cytotoxic (9-12). Palnnitate-induced cytotoxicity of liver cells has been implicated in the
pathogenesis of many obesity-related metabolic disorders, such as non-alcoholic fatty
liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) (13, 14). Tumor

necrosis factor (TNF)-o., a proinflarnmatory cytokine often is involved, along with

101

elevated FFA, in these diseases (15), and further potentiated the cytotoxicity induced by
palnnitate (2, 9, 10). Multiple mechanisms have been proposed for palnnitate-induced
cytotoxicity (16, 17). To study the multi-faceted effects of palmitate and provide insights
into potential mecharnism of saturated F FA-induced alterations, we obtained gene
expression profiles of hepatocellular carcinoma (HepG2) cells upon exposure to different
FFAs and TNF-a, and applied a module-based gene network reconstruction method that
integates prior knowledge and phenotypic information. The proposed methodology
consists of two phases. The first phase, the “gene selection phase”, selects a subset of
genes that are relevant to the phenotype, palnnitate-induced cytotoxicity, using a mixture
regession model. The second phase, the “network reconstruction phase”, clusters the
selected genes into modules, and reconstructs a module network based upon the
interactions between the modules.

Selecting the genes that are potentially relevant to the desired
metabolic/phenotypic response of the cells can be viewed as a feature selection problem
(1 8, 19), which is extensively studied in machine learrning (20, 21). Most feature selection
methods, such as the Wilcoxon's rank sum test (22) and Fisher‘s Discriminant Analysis
(FDA) (23), are data driven, and thus susceptible to the noise level of the nnicroarray data.
One strategy to ameliorate this problem is to incorporate domain knowledge and
functional information of the genes (24). Typically these knowledge-based methods
qualitatively incorporate the prior krnowledge in post-processing the genes that are
selected by the data-driven approaches. In the present work, we address this limitation
with a Bayesian mixture regession model that quantitatively incorporates the prior

knowledge of the gene firnctions, upfront, in the gene-selection phase.

102

Clustering methods, such as Self-organizing mapping (25, 26) hierarchical
clustering (27) and K-means (28), commonly used to identify gene modules, cannot
uncover the interactions among the modules or clusters. To address this limitation,
several studies integated clustering methods with structure learrning algorithms, such as
Graphical Gaussian Modeling and Bayesian network learning (29-31). These approaches
are predominantly data-driven and thus susceptible to noise in the expression data, and
suffer from the sparse data problem associated with limited number of experimental
conditions (32, 33).

Previous studies recognized the importance of exploiting prior knowledge in
reconstructing networks with sparse and noisy expression data (5, 34-39). Similarly, we
developed a fi'amework based on knowledge-driven matrix factorization, termed KMF, to
exploit the domain knowledge and reconstruct modular gene networks. This fiamework
views the gene network reconstruction as a matrix factorization problem. In brief, the
pair-wise correlation coefficients between any two genes are computed from their
expression data and used to construct a correlation matrix. This correlation matrix is
decomposed into a product of three matrices, from which the gene clusters and module
interaction information are extracted. During this process, the Gene Ontology (GO)
information is introduced as regularization in matrix factorization, which affects the
decomposed matrices and eventually the derived gene modules and interaction among
modules. Compared to the existing approaches for gene network reconstruction, the key
features of the proposed KMF fiamework are: (1) derives both the gene modules and
their interactions from a combination of expression data and GO information, (2)

incorporates the prior knowledge of co-regulation relationships into the network

103

reconstruction using a regularization scheme, and (3) presents an efficient learning
algorithm based on non-negative matrix factorization and semi-definite pro gamming.
The difference between our work and previous works on matrix factorization is that the
proposed framework is able to derive both gene modules and their interaction
simultaneously while the other matrix factorization methods can only identify gene

clusters.

KMF Algorithm

Gene Selection by Knowledge-driven Bayesian Mixture Regression Model—
Regession models are often used for gene selection, in which genes with large assigned
weights are deemed to be important for given cellular responses. However, these methods
usually suffer from the sparse data problem and are sensitive to noise in the expression
data To overcome these limitations, we incorporated the prior knowledge from GO into
the regession model with a Bayesian prior. As described previously (40), the main idea
is to integate the ontology information of the genes with their expression data (X) to
perform unsupervised classification and to identify the genes that regulate the cellular
responses (Y). A central assumption behind this method is that the genes within a G0
category should have similar function or effect on a cellular process. Thus, genes
belonging to the same GO category were constrained to have similar regession weights.
The second problem using the linear regession model for gene selection is that the genes
with the largest regession weights may not be specific to the target cellular process since
they may also be important to a number of biological processes other than the target

process. We addressed this problem by extending the linear regession model into a

104

mixture regression model. The main idea behind the mixture models is to cluster the
experimental conditions into two subgoups: the subgroup of conditions that produce the
target phenotype, i.e. cytotoxicity, and the subgroup of conditions that do not. A different
regession model was built for each subgroup of conditions, and the genes with the
largest difference in their regession weights between these two subgoups of conditions
were deemed to be specific to the target phenotype. Figure 4.1 shows the absolute
difference in their regession coefficients (denoted by DRC) of the genes between the two
regession models. The DRC values decrease exponentially for the top ranked genes
followed by a slow linear reduction. To identify a subset of genes with large DRC values,
we extracted only the genes that are related to the exponential component of the DRC
curve. This was done by identifying the point of the DRC curve where the second order
derivative starts to approach zero or a negative value. Thus, 250 genes are selected by

this process.

0.2 l
0.18 ’
0.16-
0.14 ,
0.12
0.1
0.08
0.06
0.04
0.02

0

 

abe(DRC)

 

 

 

0 100 200 300 400 500 600 700
Gene Number

Figure 4.1. Sorted difference regression coefficients (DRC) for all the genes.

Knowledge-driven Matrix Factorization (KMF) Algorithm—KMF is a technique

based on matrix factorization. It first computes pair wise correlation between two genes

105

based on their expression levels across different experimental conditions. The matrix of

pair wise gene correlation, denoted by W, is approximated by the product of three

matrices, M X C X M. A gene modular network, including gene modules and their

interaction, is derived from the decomposed matrices M and C.

We denote the gene expression data by X = (x1, x2, , xn) where n is the number
of genes, and each xi = (xi, 1, xi, 2, , xi,m) E Rm is the expression levels of the ith gene
measured under m conditions. We can compute the pairwise correlation between any two
genes using statistical correlation metrics such as Pearson correlation, mutual information
and chi-square statistics. In our experiment, we use RBF kernel function. This

computation results in a symmetric matrix W = [wi,fln X n where wi,j measures the

correlation between gene xi and xj. This estimated correlation matrix Wprovides valuable
information about the structure of the gene network since a high correlation wi, j between
two genes xi and xj could suggest that: 1) genes xi and xj belong to the same module, or 2)
gene xi regulates the expression levels of gene xj or vice versa. To derive these two types
of interactions simultaneously, we factorize W as
W e M x C x MT

where M is a matrix of size n X r and C is a matrix of size r X r, and r is the number of
modules that can be determined empirically. Matrix M = [miJJn xr represents the
memberships of the n genes in r modules where mij Z 0 indicates the confidence of
assigning the ith gene to the jth module. Matrix C = [cierxr represents the relationships
among r modules where ci,j Z 0 indicates the confidence of the two gene modules to
interact (regulate) with each other. Note that in this study, we focus on the undirected

network since the gene module regulation matrix C is symmetric.

106

The appropriate factorization of matrix W, the solution of M using the prior
knowledge from GO information, and the solution for C by the regularizer lc(C) =||C| |F
were performed according to (41).

Solving the Optimization Problem—Optimize M by fixing C: The related
optimization problem is:

arg min Fm(M) = ”W — MGM-r”? + atr(MTL(S)M)
MeRan

8.13. M,,j20,i,j=1,2,...,n
To find an optimal solution for M, we use the following bound optimization
algorithm (42). The key idea is to upper bound the objective firnction using the properties
of convex functions, and iteratively update the solution based on the solution of previous
iteration. Let M represent the solution of the previous iteration, and our goal is to find a

solution of M for the current iteration. Taking the derivative of the upper bound of Fm(M)

with respect to Mi, k, we have the optimal solution for M as follows:

NIH

 

 

where
am = [MCMTMClak’bak = aMi,sz'7
Cue = a[SM]z‘,k + [WMCk’k

107

Optimize C by fixing M: The related optimization problem is:

arg min 77 + fig — 2tr(MTWMC)
CERrXr

$.13. Ci,i=1,i=1,2,...,7“,
QJZQiJZIJVHJ,Ct0

B = MTMC, 77 2 ET: Bi,ij.ia 52 Zr: 029'

i,j = 1 i,j = 1
The above problem can be solved using semi-definite progarnming technique (43).

The KMF algorithm was applied to toxic conditions and non-toxic conditions,
separately. It was also applied to the combination of botln conditions. We denoted by Ct
and Cu the interaction matrices of toxic and non-toxic conditions, respectively, and by
Call the interaction matrix derived from all the conditions. In order to ensure that matrices
Ct and Cu are comparable, we align Ct and Cu with Call.

Parameter Tuning—There are three key parameters that can significantly affect
the outcome of the proposed algorithm: a, which is used to weight the knowledge from
the GO database, [3, which is used to control the sparseness of the interaction matrix C,
and the number of modules.

Tuning the parameters a and B: In the KMF model, parameter a weights the
prior knowledge from GO, and [3 controls the sparseness of the interaction matrix C. To
tune these two parameter for optimal result, we apply the supervised learrning technique.
First, from the 250 genes, we pick some genes that obviously should belong to the same
goup based on their biological functions, and thereby predefined 7 virtual modules
composed of 32 genes (see Table 4.1). Second, we gadually change the parameter within

a predefined range (i.e., [0.1 . . . 10] in our experiment), and choose the parameters a and

108

B that make the best prediction of the module membership compared to the predefined
modules.

Determining the number of modules: We determine the optimal number of
modules using stability analysis. The basic assumption of stability analysis is that if the
estimated number of modules is close to the true number of modules in the data, we
would expect the KMF model to yield similar results for matrix C and M with different
random initialization. In addition, the stability analysis is applied in a hierarchical fashion.
More specifically, in our experiment, the first application of stability analysis results in
two major modules. The application of the stability analysis to the two major module

yields four and five sub-modules, respectively.

109

Table 4.1. Predefined gene modules. 7 virtual modules composed of 32 genes fiom the
250 genes selected in the gene selection phase.

TCA cycle
(90) isocitrate dehydrogenase 3 (NAD+) gamma (IDH3G), mRNA.

(90) succinate-00A ligase, GDP-forming, beta subunit

(gC) succinate-00A ligase, GDP-forming, alpha subunit (SUCLG1), mRNA.

(90) citrate synthase (CS), nuclear gene encoding mitochondrial protein, mRNA.
Glycolfiis

(gC) aldolase C, fructose-bisphosphate (ALDOC), mRNA.

(gM) phosphoglycerate kinase 1 (PGK1), mRNA.

(90) lactate dehydrogenase C (LDHC), transcript variant 2, mRNA.

(gM) glucose phosphate isomerase (GPl), mRNA.

(gM) ketohexokinase (fructokinase) (KHK), transcript variant b, mRNA.
(90) hexokinase 1 (HK1), transcript variant 5, nuclear gene encoding mitochondrial protein,
mRNA.

Post-translational: Ubiguitin-Proteasome Pathway
(gC) ubiquitin D (UBD), mRNA.
(gC) ubiquitin-conjugating enzyme E23 (RAD6 homolog) (UBEZB), mRNA.

(9) proteasome (prosome, macropain) subunit, alpha type, 6 (PSMA6), mRNA.
(90) proteasome (prosome, macropain) subunit, alpha type, 1 (PSMA1), transcript variant 1,
mRNA.

(9C) 268 proteasome-associated pad1 homolog
Phosphatases
(gM) dual specificity phosphatase 3 (vaccinia virus phosphatase VH1-related)
(90) protein phosphatase 2, regulatory subunit B (B56), beta isoform (PPP2R5E), mRNA.

(90) protein tyrosine phosphatase, non-receptor type substrate 1 (PTPNSt ), mRNA.
(90) protein phosphatase 3 (formerly 28), regulatory subunit B, alpha isoform (calcineurin B,

type I)
(90) protein phosphatase 2 (formerly 2A), catalytic subunit, alpha isoform (PPPZCA), mRNA.
(gC) protein phosphatase 2, regulatory subunit B (B56), epsilon isoform (PPP2R5E), mRNA.
(gM) protein phosphatase 2, regulatory subunit B (856), gamma isoform
(90) protein tyrosine phosphatase, non-receptor type 9 (PTPNQ), mRNA.
Translation initiation
(gC) eukaryotic translation initiation factor 48 (EIF4B), mRNA.
(gC) eukaryotic translation initiation factor 4E binding protein 2
(gN) eukaryotic translation initiation factor 3, subunit 4 delta, 44kDa (EIF3S4), mRNA.
(90) eukaryotic translation initiation factor 1A, Y chromosome (ElF1AY), mRNA.
Urea cycle
(90) argininosuccinate lyase (ASL), mRNA.
(90) argininosuccinate synthetase (ASS), transcript variant 2, mRNA.
Protein folding, transmrtatlon
(gC) heat shock 105kD (HSP1OSB), mRNA.
(QC) heat shock 70kDa protein 8
_(gC) heat shock 10kDa protein 1 (chaperonin 10) (HSPE1), mRNA.

 

110

Experimental Methods

Cell Culture and Reagents, Fatty Acid Salt Treatment, and Cytotoxicity
Measurement—HepG2 cells were cultured and treated by sodium salts of palrrnitate and
oleate as previously described in Chapter 2. Cytotoxicity of the FFAs were measured as
previously described in Chapter 2 (44).

Gene Expression Profiling—As previously described in Chapter 3 (2), cells were
cultured and exposed to different treatments. RNA was isolated with Trizol reagent, and
the gene expression profiles were obtained with cDNA microarray at the Van Andel
Institute, Grand Rapids, MI.

RNA Interference and Reverse T ransfection—Silencer® select predesigned
siRNAs targeting human ATP6IP1 and RABGGTA mRN As were purchased from
Ambion (Austin, TX). The synthesized oligonucleotides are 5'-
AGUCCGAAGAUGUCCCAUATT-3' and 5'-UAUGGGACAUCUUCGGACUTG-3' for
the siRNA of ATP6IP1, and 5'-AAAGAAUGCGUGCUUUUAATT-3' and 5'-
UUAAAAGCACGCAUUCUUUCT-B' for the siRNA of RABGGTA. Reverse
transfection of siRNA was performed as described in Chapter 2 (12), with the
transfection reagent, Lipofectamine RNAiMAX (Invitrogen).

Western Blot Analysis—HepG2 cell lysate was obtained and Western Blot
analysis was performed as previously described in Chapter 2 (12). Anti-RABGGTA, anti-
ATP6AP1, and anti-GMPS antibodies were purchased from Abcam Inc. (Cambridge,
MA). Anti-beta actin antibody was purchased from Si gna-Aldrich. Secondary anti-rabbit

and anti-mouse antibodies were purchased fi'om Pierce Biotechnology Inc.

Real-time Quantitative RT -PCR Analysis—Total RNA was extracted from cells as

111

previously described in Chapter 2 (12). Equal amounts of total RNA (1 pg) were reverse-
transcribed using an iScript cDNA synthesis kit (Bio-RAD). The primers used for
quantitative RT-PCR analyses of human RABGGTA (5'-
G'ITTCCTGGAGGTGGATGAG—3' and 5'-GCCAGGAAGAGTAGTTGGAGAA-3'),
human HSPIOSB (5'-GGAGTTCCATATCCAGAAGCA-3' and 5'-
TCCACCATAGATGCCGTAGA-3'), human GMPS (5'-
CAAAGCCTGCACAACAGAAG—3' and 5'—ACTGGAGATTCCACACACGTAA-3'),
human ATP6IP1 (5'-CCTGCTCTGCTGCTCATTC-3' and 5'-
CGCTGTGTATGGGACATCTT-3 '), and human GAPDH (5'-
AACTI'I‘GGTATCGTGGAAGGA-3' and 5'-CAGTAGAGGCAGGGATGATGT-3')
were synthesized by Operon Biotechnologies, Inc. (Huntsville, AL). RT-PCR was
performed as previously described in Chapter 2 (12).

Statistical Analysis—All experiments were perfornned at least tlnree times, and
representative results are shown. All data are shown as the mean :t SD. for indicated
number of experiments. Student’s t-test was used to evaluate statistical significances

between different treatment goups. Statistical significance was set at p<0.01.

Results and Discussion

We applied the proposed KMF framework to HepG2 cells cultured in different
FFAs, with or without TNF-u for 24 hours. 250 genes selected in the gene selection phase
(see Methods section) were used to reconstruct the network. Prior knowledge can be
incorporated to help reconstruct networks with sparse and noisy expression data (5, 34-

39). Typically, the prior knowledge of the gene interaction is encoded in a Bayesian prior,

112

in which a high probability is given for each gene regulatory relationship derived from
prior krnowledge. By incorporating a Bayesian prior, Bayesian Network analysis
penalizes any gene relationship (i.e., gives a low score) when it violates the prior
krnowledge of the gene relationships, thus improving both the accuracy and efficiency of
BN analysis. In this study, the prior knowledge of the genes is taken fiom the GO
database. Although GO information does not directly reveal the gene relationships,
nevertheless it does provide coregulation relationships and functional information of the
genes, both of which are still potentially useful for reconstructing gene networks. Unlike
existing methods that apply the GO information to generate pre-defined sets of genes
based on supervised feature selection (45), our KMF algorithm applies an alternative
unsupervised feature selection, which allows us to identify the feature genes when the
classification of the experimental conditions is unknown. In addition, KMF tunes the
impact of the GO information on the model selection to obtain optimal results (see
Methods section). This is in contrast to the other methods where the GO information
takes precedence over the subsequent analysis (40).

The KMF algorithm yields two matrices, M and C. M is the module matrix. Each
element Mij in matrix M represents the confidence of assigning the ith gene to the jth
module. From the module matrix M, we can derive the member genes for each module by
only including the genes whose confidences of being assigned to the module exceeds a
certain threshold. These member genes will furthermore allow us to infer the overall
biological functions of each module. C is the network structure matrix that indicates the
connectivity between gene modules. In particular, each element Cij in matrix C represents

the strength of the interaction between modules i and j. The interaction information

113

revealed by the C matrix may shed light onto how biological information is processed
and passed between different cellular activities. Furthermore, comparing the C matrices
for the different conditions suggests structural changes in the module network in response
to the toxic conditions, and these changes may confer the cytotoxic phenotype.

Application of KMF to identify gene modules and the interactions between the
modules—Nine modules (Table 4.2) were identified by KMF. These gene modules are
highly enriched with genes involved in specific cellular firnctions or activities. A full list
of the genes in each module is available online at
http://www.chems.msu.edu/goups/chan/GO_KMFjgenecluster.xls. Next, KMF
identified the interactions between the modules, namely the connections between
different cellular fimctions, in the form of the C matrix (Table 4.3), and thereby
recovered a module network (Fig. 4.2). The bottom row (“sum”) of Table 4.3 sums the
correlation coefficients (Cg) between a module with the other 8 modules, tlnereby
capturing an overall snapshot of the module connections. A higher “sum” value indicates
that the module is more highly correlated with the other modules and thereby takes a
more central position in the overall gene module network. A map of the module network
is provided in Fig. 4.2, where the strengths of the interactions between the gene modules
are indicated by darkness and thickness of the edges.

From the C matrix (Table 4.3) and the module interaction network (Fig. 4.2),
module 6 (ATP and GTP metabolism) has the highest “sum” value among the 9 modules,
and the largest node in the module interaction map. Indeed, as the molecular currency of
intracellular energy transfer, ATP (as well as GTP) is either produced or consumed by

most of cellular activities, e.g., metabolism (catabolism and anabolism) and signaling

114

pathways. Module 6 has the highest interaction values witln modules 3, 5 and 8 in the C
matrix, reflecting that glucose metabolism (module 3) and TCA cycle (module 5) are the
major metabolic pathways that produce ATP, the electron transport chain (ETC) (module
5) produces the proton gadient across the mitochondria membrane to provide the driving
force for ATP production (46), and amino acid metabolism (module 8) is highly
dependent on the ATP levels. Therefore, KMF recovered a high connectivity between
ATP (and GTP) synthesis and the major cellular activities that are known to be related to

energy production and consumption.

Table 4.2. Gene clusters identified by KMF.

 

Module
1 Lipid metabolism and lipid processing

2 Signaling proteins, intracellular and membrane protein-mediated: G protein-coupled
receptor signaling, chemokine/TNF-a receptor signaling, ion channel-related signaling

Glucose metabolism: glycolysis and pentose phosphate pathway

Post-translational modification: ubiquitin-proteasome pathway, protein folding,
transportation, phosphorylation/dephosphorylation

ROS homeostasis, redox system regulation and the TCA cycle
Energy: ATP and GTP metabolism

Protein synthesis regulation: translation initiation and transcription
Amino acid metabolism and urea cycle

Apoptosis: executors and regulators

Function

 

uh“

@QNQUI

 

Table 4.3. C Matrix of the clusters. Rows 1-9 represent the interaction between two
clusters. Bottom row (sum) is the summation of each column.

 

Clusters 1 2 3 4 5 6 7 8 9

 

@Q‘IOOI-wa-l

 

0.152
0.234
0.195
0.191
0.275
0.101
0.236
0.176

1.56

0.152

0.177
0.155
0.152
0.214
0.092
0.183

0.14
1.265

0.234
0.177

0.236
0.215
0.305
0.107
0.284
0.209
1.767

0.195
0.155
0.236

0.204
0.295

0.12
0.249
0.188
1.642

115

0.191
0.152
0.215
0.204

0.302
0.122
0.253
0.186
1.625

0.275
0.214
0.305
0.295
0.302

0.17
0.36
0.267
2.188

0.101
0.092
0.107
0.12
0.122
0.17

0.138
0.108
0.958

0.236
0.183
0.284
0.249
0.253

0.36
0.138

0.227
1 .93

0.176

0.14
0.209
0.188
0.186
0.267
0.108
0.227

1.501

 

 

Figure 4.2. Gene module interaction network. Interactions among the nine gene
modules are visualized according to the C matrix. The nodes represent modules and the
edges indicating the strength of the interaction between modules. A higher Cij value in
the C matrix, suggesting stronger interaction, is indicated by a thicker and darker edge
line, whereas a higher “sum” value in the C matrix, suggesting more relevant module, is
indicated by a larger and darker node.

From the C matrix (Table 4.3) and the module interaction network (Fig. 4.2),
module 6 (ATP and GTP metabolism) has the highest “sum” value among the 9 modules,
and the largest node in the module interaction map. Indeed, as the molecular currency of
intracellular energy transfer, ATP (as well as GTP) is either produced or consumed by
most of cellular activities, e.g., metabolism (catabolism and anabolism) and signaling
pathways. Module 6 has the highest interaction values with modules 3, 5 and 8 in the C
matrix, reflecting that glucose metabolism (module 3) and TCA cycle (module 5) are the
major metabolic pathways that produce ATP, the electron transport chain (ETC) (module
5) produces the proton gadient across the mitochondria membrane to provide the driving
force for ATP production (46), and arrnino acid metabolism (module 8) is highly

dependent on the ATP levels. Therefore, KMF recovered a high connectivity between

116

ATP (and GTP) synthesis and the major cellular activities that are krnown to be related to
energy production and consumption.

Application of KMF to identify the interactions involved in palmitate-induced
cytotoxicity—KMF, if applied to the different conditions separately, yields different C
matrices specifically for the toxic (saturated FFAs and TNF-a) (Table 4.4) and nontoxic
(control, unsaturated FFAs and TNF-a) (Table 4.5) conditions. This is in contrast to the
average C matrix obtained using all the conditions discussed above (Table 4.3). Sinnilarly,
these condition-specific C matrices indicate module networks composed of interactions
between cellular activities for their corresponding condition. The C matrix in the toxic
conditions differs significantly fi'om the nontoxic conditions, suggesting that the
interactions between the gene modules in the toxic (saturated FFAs and TNF-or) case are
altered significantly, and these changes potentially may help explain the phenotype,
palmitate-induce cytotoxicity. To quantitatively assess tlnese changes, we subtracted the C
matrix for the nontoxic conditions from the C matrix for the toxic conditions, and
obtained a matrix we denoted as the “difference C matrix” (Table 4.6). This matrix
indicates the differences in the interactions between the gene modules for the toxic vs. the
nontoxic conditions. Positive values indicate stronger interactions between the modules
under the toxic than the nontoxic conditions, and vice versa. The surmnation of each
column in the difference C Matrix indicates the overall difference in the module
interactions of a module with the other modules for the toxic vs. the nontoxic conditions.
As shown in the difference C matrix (Table 4.6), modules 2, 3, 4 and 5 are more highly
connected to the other modules in the toxic than in the nontoxic conditions, while clusters

6 and 9 are less connected in the toxic than in the non-toxic conditions. Since modules 4

117

and 6 have the largest positive and negative “sum” values, 0.144 and -0.294, respectively,
we focused on these two modules in the supplemented discussion of their potential
involvement in palmitate-induced cytotoxicity.

Table 4.4. C Matrix in toxic conditions. Rows 1-9 represent the interaction between
two clusters under toxic conditions. Bottom row is the summation of each column.

 

 

 

Clusters 1 2 3 4 5 6 7 8 9
1 0.158 0.252 0.218 0.199 0.229 0.088 0.242 0.17
2 0.158 0.212 0.184 0.168 0.197 0.087 0.203 0.137
3 0.252 0.212 0.272 0.252 0.255 0.106 0.291 0.218
4 0.218 0.184 0.272 0.236 0.271 0.128 0.274 0.194
5 0.199 0.168 0.252 0.236 0.281 0.134 0.266 0.181
6 0.229 0.197 0.255 0.271 0.281 0.162 0.3 0.213
7 0.088 0.087 0.106 0.128 0.134 0.162 0.142 0.083
8 0.242 0.203 0.291 0.274 0.266 0.3 0.142 0.216
9 0.17 0.137 0.218 0.194 0.181 0.213 0.083 0.216

sum 1.556 1.346 1.858 1.777 1.717 1.908 0.93 1.934 1.412

 

 

Table 4.5. C Matrix in nontoxic conditions. Rows 1-9 represent the interaction between
two clusters under nontoxic conditions. Bottom row is the summation of each column.

 

Clusters

@ONOUIhNN-h

 

1

0.153
0.236
0.195
0.191
0.279
0.104
0.236
0.177
1.571

2
0.153

0.174
0.153

0.15
0.214
0.095

0.18
0.139
1.258

3

0.236
0.174

0.234
0.213
0.307
0.108
0.282
0.207
1.761

4
0.195
0.153
0.234

0.201
0.295
0.121
0.246
0.186
1.631

5
0.191
0.15
0.213
0.201

0.302
0.123
0.251
0.185
1.616

6
0.279
0.214
0.307
0.295
0.302

0.172
0.362

0.27
2.201

7

0.104
0.095
0.108
0.121
0.123
0.172

0.139
0.11

0.972

8
0.236
0.18
0.282
0.246
0.251
0.362
0.139

0.225
1 .921

9
0.177
0.139
0.207
0.186
0.185

0.27
0.11
0.225

1 .499

Table 4.6. Difference C Matrix. Subtract the C Matrix of the nontoxic conditions from
the toxic conditions.

 

Clusters 1 2 3 4 5 6 7 8 9

 

CONGOI‘WN-h

 

0.006
0.016
0.023
0.008
-0.05
-0.016
0.006
-0.007
-0.014

0.006

0.037
0.031
0.017
-0.017
—0.007
0.023
-0.002
0.088

0.016
0.037

0.038
0.039
-0.053
-0.001
0.01
0.01 1
0.097

0.023
0.031
0.038

0.034
-0.024
0.007
0.027
0.008
0.144

0.008
0.01 7
0.039
0.034

-0.021
0.011
0.015

-0.005
0.098

-0.05
-0.017
-0.053
-0.024
-0.021

-0.01
-0.062
-0.057
-0.294

-0.016
-0.007
-0.001
0.007
0.011
-0.01

0.003
-0.027
-0.04

0.006
0.023
0.01
0.027
0.015
-0.062
0.003

-0.01
0.012

-0.007
-0.002
0.01 1
0.008
-0.005
-0.057
-0.027
-0.01

-0.089

 

118

Module 4: Palnnitate and cytokines, such as TNF-or, have been shown to increase
the levels of unfolded or rrnisfolded proteins, which induces endoplasnnic reticulum (ER)
stress (47, 48) and triggers the unfolded protein response (UPR) (48, 49) and ubiquitin-
proteasome system (50, 51). Indeed, it has been suggested that ER stress together with
UPR and ubiquitin-proteasome system serve as self-healing processes for cells upon toxic
or stressed conditions (52-54). In addition, palrrnitate-induced ER stress is also correlated
with the redistribution of the ER chaperon proteins (48), which are highly involved in
post-translational modifications, such as protein transportation and folding, and the
chaperon proteins have been suggested to counteract the ER stress induced by palnnitate
(55). Therefore, although the mechanism is not fully understood, the ubiquitin-
proteasome pathway and the post-translational modifications (folding/unfolding,
transportation, degadation) of proteins have been suggested to play important roles in
cellular responses to saturated FFAs and cytokines, such as TNF-or. Indeed, module 4
(post-translational modification of proteins: ubiquitin-proteasome pathway, protein
folding, transportation, phosphorylation/dephosphorylation) has the highest positive
‘sum’ value in the difference C matrix, reflecting the fact that post-translational
modifications of proteins (folding, transportation, and recycling) are highly activated,
playing significant roles, and therefore more closely connected with other cellular
activities in the toxic conditions.

Module 6: Module 6 (ATP and GTP metabolism) has the largest negative ‘sum’
value in the difference C matrix, indicating that ATP metabolism is less correlated with
the other cellular processes in the toxic than the non-toxic conditions. Long-term

exposure of saturated FFAs, e.g. palrrnitate, leads to activation of the uncoupling proteins

119

(UCP) (56-58). Located in the mitochondrial inner membrane, UCPs act as proton
channels to dissipate the proton gadient before it is used by ATP synthases to produce
ATP. Thus, UCPs uncouple mitochondrial oxidative phosphorylation and produce heat
instead of ATP (58, 59). Therefore in the toxic conditions, the highly activated UCPs
may serve as anotlner key regulator that controls the level of ATP, in addition to the other
cellular processes that produce proton gradients or consume ATP, such as glucose and
amino acid metabolism and TCA cycle. As a result, with this additional regulation
through UCPs, the level of ATP is less connected with the cellular activities identified by
KMF, but more related to regulators such as UCPs in the toxic than the nontoxic
conditions. This could provide an explanation for the values in column 6 (ATP and GTP
metabolism) being negative in the difference C matrix (Table 4.6).

In summary, the proposed KMF algorithm identified the gene modules and their
interactions, as well as how they change in the toxic vs. non-toxic conditions. The results
suggested that post-translational modification and uncoupling proteins (U CP) may play
important roles in mediating the palmitate/TNF-a induced cellular responses, thereby
shedding light on potential mechanisms involved in palnnitate-induced cytotoxicity. Thus
far, this methodology has focused on the module network. To further uncover the specific
genes that may be responsible for the palrrnitate-induced cytotoxicity, we performed
further analysis to assess the contribution of each gene in the two gene modules that were
deemed important.

Identifying potential genes responsible for palmitate-induced cytotoxicity—As
described above, the values in the M matrix (My) indicate the strength or contribution of

gene i to module j. The rank of the genes in a module by their My- values provides a

120

relative index of the importance of a gene to the cellular function that corresponds to that
module. Under different conditions, the modules remained relatively stable with respect
to their size and gene members, however, the rank of certain genes changed significantly
in some of the modules. This suggests that the importance or the weights of these genes
in their corresponding modules varied across the different conditions; therefore these
genes may play important roles in conferring a phenotype.

Table 4.7. Top 10 out of 33 genes in module 4. Top 10 out of 33 genes in module 4
ranked according to their contributions to the module under toxic conditions. The ranking
difference was calculated by subtracting the ranking number of the specific gene under

toxic conditions from non-toxic conditions. Positive ranking difference numbers indicate
bigger ranking numbers and less contribution in non-toxic conditions.

 

Rank in Rank In Rank

 

toxic non-toxic difference Gene

1 4 3 leucine carboxyl methyltransferase (LCMT)

2 5 3 mitogen-activated protein kinase kinase kinase 12
(MAP3K12)

3 3 0 protease, serine, 2 (trypsin 2) (PRSSZ)

4 2 -2 suppression of tumorigenicity 13 (colon carcinoma) (ST13)

5 25 20 heat shock 105kD (HSP1058)

6 6 0 apolipoprotein C-l (APOC1)

7 30 23 Rab geranylgeranyltransferase, alpha subunit
(RABGGTA)

8 14 6 UV radiation resistance associated gene (UVRAG)

9 7 _2 dolichyl-phosphate mannosyltransferase polypeptide 2,
regulatory subunit (DPM2)

10 23 13 mitogen-activated protein kinase-activated protein kinase 3
(MAPKAPKS)

 

Given the importance of modules 4 (Post-translational modification of proteins)
and 6 (ATP and GTP metabolism), we ranked the genes in these two modules according
to their Mi,- values in the toxic conditions. The top 10 out of 33 genes in module 4 and all

the genes in module 6 are listed in Tables 4.7 and 4.8, respectively. The ranking numbers

121

of these genes in the toxic and nontoxic conditions are listed, as is the difference in the
rankings of the genes between these conditions (Tables 4.7 and 4.8).

Table 4.8. All 10 genes in module 6. The genes in module 6 ranked according to their
contributions to the module under toxic conditions. The ranking difference was calculated
by subtracting the ranking number of the specific gene under toxic conditions from non-
toxic conditions. Positive ranking difference numbers indicate bigger ranking numbers
and less contribution in non-toxic conditions.

 

Rank in Rank in Rank

 

toxic non-toxic difference Gene
1 5 4 ATP citrate lyase (ACLY)
1 7 6 guanine monphosphate synthetase (GMPS)
1 8 7 ATPase, H+ transporting, Iysosomal interacting protein 1
(ATP6IP1)
1 1 0 ATP synthase, H+ transporting, mitochondrial F0 complex,

subunit c (subunit 9), isoform 2 (ATPSG2)
_3 ATP synthase, H+ transporting, mitochondrial F0 complex,

5 2 subunit F6 (ATPSJ)

6 4 -2 adenosine kinase (ADK), transcript variant ADK-long

7 3 -4 adenylosuccinate synthase

8 9 1 adenosine monophosphate deaminase (isoform E) (AMPD3)
9 6 -3 Iysosomal apyrase-like 1

10 10 0 ATPase, H+ transporting, Iysosomal 13kDa, V1 subunit G
isoform 1 (ATP6V1G1)

 

In module 4, the positions of two genes, Rab geranylgeranyltransferase
(RABGGTA) and heat shock 105kD (HSPlOSB) changed significantly in the toxic
conditions, as indicated by high positive differences in rankings, 23 and 20 respectively
(Table 4.7), suggesting that tlnese two genes may be more involved in the toxic than non-
toxic conditions and tlnereby play a role in conferring the toxic phenotype. RABGGT
catalyzes the transfer of a geranyl-geranyl moiety from geranyl-geranyl pyrophosphate to
Rab proteins (GTPases) such as RABlA, RAB3A and RABSA (60). As a member of the
Ras superfarnily of monomeric G proteins, Rab proteins regulate membrane traffic,

which facilitates the trafficking of cell membrane proteins from the Golgi to the plasma

122

membrane and the recycling of the membrane proteins (61 , 62). RABGGT, by facilitating
the prenylation of Rab proteins (60), ensures that the Rab proteins are insoluble and
correctly anchored in the membrane. The response of RABGGT to saturated FF As and its
potential role, if any, in the saturated F F A-induced cytotoxicity has never been studied.
The mRNA level of RABGGT is not affected by oleate and increased by palmitate albeit
insignificantly (Fig. 4.3A). However, furtlner analysis by silencing the gene expression
level of RABGGT revealed a very interesting feature of RABGGT in regulating
cytotoxicity (Fig. 4.33). In the non-toxic conditions, i.e. BSA (vehicle of the FFAs) or
oleate, the LDH release was increased by the siRN A of RABGGT (Fig. 4.33), suggesting
that RABGGT may help to maintain normal healthy cellular activities under
physiological and non-toxic conditions. Indeed, membrane traffic patlnways, regulated by
RABGGT through Rab GTPases, are important in maintaining normal vesicle formation
and movement and membrane protein trafficking and recycling. By contrast, in the toxic
condition, i.e. palrrnitate, the LDH release was decreased by the siRNA of RABGGT (Fig.
4.38), suggesting that RABGGT may be involved in mediating the cytotoxic effect of
palmitate. The potential mechanism of the distinct roles of RABGGT under the different
conditions is unclear at this point. Given that RABGGT catalyzes the prenylation and
therefore the activation of Rab GTPases, we hypothesize that the toxic conditions (i.e.,
palnnitate) induce disordered trafficking and recycling of the membrane proteins and
disrupt the membrane integity, through RABGGT and Rab proteins, thereby enhancing
the cytotoxicity.

Containing a HSP70-like motif, HSPIOSB belongs to the Hsp105/ 110 family, a

diverged subgoup of the 70-kDa heat shock protein (Hsp70) family (63). HSP105B has

123

been found to inhibit the aggegation of denatured proteins under severe stress conditions
(64). In addition, HSP105B is usually associated with Hsp70 or Hsc70 (a constitutive
form of Hsp70) in mammalian cells and suppresses the chaperone activity of
Hsp70/Hsc70 (65). The role of HSP105B in cytotoxicity and apoptosis has been
suggested to be cell type-dependent (66—69). HSP105 suppressed staurosporine-induced
apoptosis in HeLa cells (66) and stress-induced apoptosis in neuronal PC12 cells (67).
However, HSP105 was also found to enhance the apoptosis induced by oxidative stress in
mouse embryonal F9 cells (68) and promote ER stress-induced caspase-3 activation, an
indicator of apoptosis (69). As an important chaperon protein involved in processing
denatured proteins under stress conditions (64, 65), HSP105B may also be involved in r
the cellular responses induced by the toxic conditions, potentially by regulating the post-
translational modifications, such as denaturation, folding/unfolding, transportation and
degadation. Indeed, we found that both the mRNA (Fig. 4.4A) and protein (Fig. 4.4B)
expression levels of HSP105B are significantly increased by palmitate but not by oleate,
suggesting that this gene potentially plays a role in the cytotoxicity induced by saturated
FFA. More detailed investigation is needed to clarify the exact role of HSP105 in

palnnitate-induced cytotoxicity.

124

P

 

 

 

 

   

1.6
<
'g 1.4-
m I; 1.2
a .
E 1: 1
'- U
3 g 0.8 1
e
E E 0.6 —
< "' 0.4 -
E 0.2 -
E 0
BSA Palm Ole
B.
16%
M DNeg. Ctrl.
312% . IsiRABGGTA
a **4:
0 80/ .
2 ° *
E
.1 4%
0%

 

 

 

 

BSA ' Palm ‘ Ole

Figure 4.3. Effects of the fatty acids on the expression level of RABGGTA and the
role of RABGGTA in cytotoxicity. HepG2 cells were exposed to 0.7 mM palmitate or
oleate for 24 hours (A). After treatment, the cells were harvested, and RT-PCR analysis
was performed to detect the mRNA levels of RABGGTA (A). Reverse transfection of
suspended HepG2 cells were performed with scrambled siRNA ([1, negative control) or
siRNA of RABGGTA (I, siRABGGTA) for 24 hours and the transfected cells were then
cultured in 0.7 mM pahnitate or oleate for another 24 hours (B). Cells were then
harvested, and the LDH release was assayed (B). In all the experiments with the FFAs,
the vehicle (0.7 mM BSA) was used as the control (i.e., regular media with BSA). Data
expressed as average of nine samples i SD from three independent experiments. One-
way ANOVA with Tukey’s post hoc method was used to analyze the differences between
treatment groups. *, **, ***, significantly higher (* and ***) or lower (**) than negative
control, i.e., scrambled siRNA, p<0.0l.

125

P

 

 

 

     

12
a:
g 10 - *
D. A
g 81 8—
.. 5
0 s: 6 _
a 0
In; E 4 _
i8
“é "
0 _
BSA Palm Ole
B.
HSP105B
beta Actin

 

BSA Palm. Ole.

Figure 4.4. Effects of the fatty acids on the expression level of HSP105B. HepG2 cells
were exposed to 0.7 mM palmitate or oleate for 24 hours. The vehicle for the FFAs (0.7
mM BSA) was used as the control (i.e., regular media with BSA). After treatment, the
cells were harvested, and RT-PCR (A) and western blot analysis (B) were performed to
detect the mRNA (A) and the protein (B) expression levels of HSP105B. Gene
expression data expressed as average of nine samples :t SD from three independent
experiments. Student’s t-test was used to analyze the differences between treatment
goups. *, significantly higher than control, i.e., regular media with BSA, p<0.01.

In module 6, the positions of two genes, ATPase, H+ transporting, Iysosomal interacting
protein 1 (ATP6IPl) and guarnine monophosphate synthetase (GMPS) are changed
significantly by the toxic conditions, indicated by their highest positive difference
rankings, 7 and 6 respectively (see Table 4.8), suggesting that these two genes may be

involved in conferring the toxic phenotype. As an essential component of most

eukaryotic cells, ATP6IP1 is located on the vacuole membrane, responsible for acidifying

126

vacuoles by transporting H+ into the vacuoles at the expense of ATP (70, 71). Vacuoles
are involved in removing and recycling unwanted or harmful substances, such as
misfolded proteins and foreign invaders such as bacteria (72, 73). Together with
lysosomes, vacuoles play major roles in autophagy and maintaining the balance between
biogenesis and degadation of many cellular products. It has been shown that palmitate
induces Iysosomal pernneabilization, which contributes to the cytotoxicity induced by
palmitate (16, 74), however the effect of palmitate on the vacuole membrane has not been
reported. In fact, vacuoles and lysosomes share similarities in their structures, internal pH
and major functions, and these two organelles sometimes fuse together to exchange their
internal substances. Considering the similarity between vacuoles and lysosomes, and
more importantly based on our result that suggests this vacuole membrane protein
ATP6IP1 plays an important role in the toxic (palmitate) conditions, we hypothesize that
I palrrnitate may similarly perturb the vacuole membrane, and thereby interrupt the internal
pH, and induce vacuole permeabilization, which may lead to apoptosis and cytotoxicity.
Since ATP6IP1 is the major membrane protein that produces the pH difference across the
vacuole membrane, we further propose that palmitate perturbs the vacuole membrane by
interacting with ATP6IP1. The mRNA (Fig. 4.5A) and protein (Fig. 4.5B) expression
levels of ATP6IP1 were not significantly affected by either palnritate or oleate. However,
silencing the expression of ATP6IP1 with the siRNA of ATP6IPl decreased the
cytotoxicity induced by palmitate, as evidenced by the LDH release (Fig. 4.56). This
result suggests that down-regulating vacuole membrane protein reduces palnnitate-
induced cytotoxicity, suggesting that palnrnitate may alter the vacuole membrane and

induce cytotoxicity through the vacuole membrane protein ATP6IP1.

127

 

 

 

 

E 1.6 —
g A 1.4 —
'2 g, 1.2 —
- 5 1 -
3 5 08
O . r
>
2 33 0.6 —
< a;
z 0.4 “
“E5 0.2 —
0 _
BSA Palm Ole
C.
16%
** EINeg. Ctrl.
B. .
A ElsnATP6AP1
WBI °\o
2
' I
3
BSA Palm. Ole.

 

 

 

BSA Palm Ole

Figure 4.5. Effects of the fatty acids on the expression level of ATP6IP1 and the role
of ATP6IP1 in cytotoxicity. HepG2 cells were exposed to 0.7 mM palmitate or oleate
for 24 hours (A, B). After treatment, the cells were harvested, and RT-PCR (A) and
western blot analysis (B) were performed to detect the mRNA (A) and the protein (B)
expression levels of ATP6IPl. Reverse transfection of suspended HepG2 cells were
performed with scrambled siRNA (D, negative control) or siRNA of ATP6IP1 (I,
siATP6IPl) for 24 hours and the transfected cells were then cultured in 0.7 mM palmitate
or oleate for another 24 hours (C). Cells were then harvested, and the LDH release was
assayed (C). In all the experiments with the FFAs, the vehicle (0.7 mM BSA) was used as
the control (i.e., regular media with BSA). Data expressed as average of nine samples i
SD from three independent experiments. One-way ANOVA with Tukey’s post hoc
method was used to analyze the differences between treatment goups. *, significantly
lower than negative control, i.e., scrambled siRNA, p<0.01.

GMPS is the key enzyme involved in the de novo synthesis of guanine

nucleotides by catalyzing the amination of xantlnosine monophosphate (XMP) to

128

guanosine monophosphate (GMP) (75). Guanine nucleotides are essential for DNA and
RNA synthesis, and GMP is also the substrate for GTP production, which is a source of
energy, similar to ATP, especially in protein synthesis. Another major role of GTP is
mediating signal transduction pathways, particularly G-protein signaling pathways (76).
Therefore, in addition to RNA and DNA replication, GMPS is also potentially involved
in other cellular processes by regulating the GTP level. The regulation of guanine
nucleotide synthesis by GMPS in liver cell cytotoxicity has never been studied.
Interestingly, we found that the mRNA (Fig. 4.6A) and protein (Fig. 4.68) expression
levels of GMPS were both significantly enhanced by the saturated FFA, palnnitate, but
not by unsaturated FFA, oleate. Currently it is unclear what role this enzyme plays in the
toxicity induced by saturated FFA. However, the changes at the gene and protein
expressions of GMPS in response to pahnitate suggested that GMPS may play a role in

saturated FFA induced cellular activities.

129

P

 

 

 

u
(D *
n. 3“
5325
For
0 g 2.
'6:
531a
3e1—
% 0e

0.

BSA Palm Ole
B.
GMPS

 

BSA Palm. Ole.

Figure 4.6. Effect of the fatty acids on the expression level of GMPS. HepG2 cells
were exposed to 0.7 mM palmitate or oleate for 24 hours. The vehicle for the FFAs (0.7
mM BSA) was used as the control (i.e., regular media with BSA). After treatment, the
cells were harvested, and RT-PCR (A) and western blot analysis (B) were performed to
detect the mRNA (A) and the protein (B) expression levels of GMPS. Gene expression
data expressed as average of nine samples i SD fiom three independent experiments.
Student’s t-test was used to analyze the differences between treatment groups. * ,
significantly higher than control, i.e., regular media with BSA, p<0.01.
Conclusions

KMF was used to reconstruct a gene module network, composed of fimctional
gene modules and their interactions. Comparing the gene module networks for the
different conditions revealed changes in module interactions across the conditions. Our
results showed that modules 2-6 and 9 played important roles in palnnitate-induced

cytotoxicity. These modules covered most of the popular areas of the research on the

cytotoxicity induced by saturated F FAs and TNF-or in liver cells, including the regulation

130

of apoptosis pathways (module 9) (12, 16, 77-79) and redox system (module 5) (10, 80).
In a separate study, we evaluated how saturated FFAs and TNF-or affected some of the
genes in module 9, such as PKR and Bcl-2 family proteins (11, 12), and found that the
gene expression level of Bcl-2 was suppressed by palmitate and TNF-a through PKR (l 1,
12), providing a potential mechanism by which palmitate- and T'NF-or-induced
cytotoxicity. We also investigated some genes in module 5, such as NADH
dehydrogenases, which we found are also highly involved in palmitate-induced
cytotoxicity by inducing ROS production (80).

In addition, assessing the contribution of genes within some of the highly relevant
modules (4 and 6) revealed potential genes that may be involved in palmitate-induced
cytotoxicity. Furtlner experiments confirmed the involvement of these genes in conferring
the phenotype, palmitate-induced cytotoxicity, and suggested novel research targets for
addressing the palmitate-induced cytotoxicity. In summary, by quantitatively integating
gene expression profile with prior knowledge extracted from GO database, KMF
provides a powerful tool to reconstruct modular and phenotype-specific gene networks

that elucidate possible mechanisms involved in producing a phenotype.

131

10.

References

Said, M. R., Begley, T. J ., Oppenheim, A. V., Lauffenburger, D. A., and Samson, L.
D. (2004) Global network analysis of phenotypic effects: protein networks and

toxicity modulation in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 101,
18006-1801 1

Srivastava, S., Li, Z., Yang, X., Yedwabnick, M., Shaw, S., and Chan, C. (2007)
Identification of genes that regulate multiple cellular processes/responses in the
context of lipotoxicity to hepatoma cells. BMC Genomics 8, 364

Han, J. D. (2008) Understanding biological functions through molecular networks.
Cell Res 18, 224-237

Ravasz, E., Somera, A. L., Mongu, D. A., Oltvai, Z. N., and Barabasi, A. L. (2002)
Hierarchical orgarnization of modularity in metabolic networks. Science 297, 1551-
1555

Ilnmels, J ., Friedlander, G., Bergnann, S., Sarig, 0., Ziv, Y., and Barkai, N. (2002)
Revealing modular organization in the yeast transcriptional network. Nat Genet 31,
370-377

Thieffry, D., and Romero, D. (1999) The modularity of biological regulatory
networks. Biosystems 50, 49-59

Yook, S. H., Oltvai, Z. N., and Barabasi, A. L. (2004) Functional and topological
characterization of protein interaction networks. Proteomics 4, 928-942

Jeong, H., Tombor, B., Albert, R., Oltvai, Z. N., and Barabasi, A. L. (2000) The
large-scale orgarnization of metabolic networks. Nature 407, 651-654

Li, Z., Srivastava, S., Mittal, S., Yang, X., Sheng, L., and Chan, C. (2007) A Three
Stage Integative Pathway Search (TIPS) framework to identify toxicity relevant
genes and pathways. BMC Bioinformatics 8, 202

Srivastava, S., and Chan, C. (2007) Hydrogen peroxide and hydroxyl radicals
mediate palnnitate-induced cytotoxicity to hepatoma cells: relation to mitochondrial
permeability transition. Free Radic Res 41, 38-49

132

 

ll.

12.

l3.

14.

15.

l6.

17.

18.

19.

20.

21.

22.

Srivastava, S., Li, Z., Yang, X., Yedwabnick, M., Shaw, S., and Chan, C. (2007)
Identification of genes that regulate multiple cellular processes/ responses in the
context of lipotoxicity to hepatoma cells. BMC Genomics 8, 364-379

Yang, X., and Charn, C. (2009) Repression of PKR mediates palrrnitate-induced
apoptosis in HepG2 cells through regulation of Bcl-2. Cell Res

Scheern, A. J ., and Luyckx, F. H. (2002) Obesity and liver disease. Best Pract Res
Clin Endocrinol Metab 16, 703-716

Farrell, G. C., and Latter, C. Z. (2006) Nonalcoholic fatty liver disease: from
steatosis to cirrhosis. Hepatology 43, S99-S112

Bruce, C. R., and Dyck, D. J. (2004) Cytokine regulation of skeletal muscle fatty
acid metabolism: effect of interleukin-6 and tumor necrosis factor-alpha. Am J
Physiol Endocrinol Metab 287, E616-621

Feldstein, A. B., Werneburg, N. W., Canbay, A., Guicciardi, M. E., Bronk, S. P.,
Rydzewski, R., Burgart, L. J ., and Gores, G. J. (2004) Free fatty acids promote
hepatic lipotoxicity by stimulating TNF-alpha expression via a lysosomal pathway.
Hepatology 40, 185-194

Wang, T., and Weinman, S. A. (2006) Causes and consequences of mitochondrial
reactive oxygen species generation in hepatitis C. J Gastroenterol Hepatol 21
Suppl 3, S34-37

Ressom, H. W., Varghese, R. S., Zhang, Z., Xuarn, J., and Clarke, R. (2008)
Classification algorithms for phenotype prediction in genomics and proteonrics.
Front Biosci 13, 691-708

Saeys, Y., Inza, 1., and Larranaga, P. (2007) A review of feature selection
techniques in bioinformatics. Bioinformatics 23, 2507-2517

Bhaskar, H., Hoyle, D. C., and Singh, S. (2006) Machine learning in bioinformatics:
a brief survey and recommendations for practitioners. Comput Biol Med 36, 1104-
1 125

Inza, 1., Larranaga, P., Blanco, R., and Cerrolaza, A. J. (2004) Filter versus wrapper
gene selection approaches in DNA microarray domains. ArtifIntell Med 31, 91-103

Troyanskaya, O. G., Garber, M. E., Brown, P. O., Botstein, D., and Altman, R. B.
(2002) Nonparametric methods for identifying differentially expressed genes in
microarray data. Bioinformatics 18, 1454-1461

133

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

Chan, C., Hwang, D., Stephanopoulos, G. N., Yarrnush, M. L., and Stephanopoulos,
G. (2003) Application of multivariate analysis to optimize function of cultured
hepatocytes. Biotechnol Prog 19, 580-598

Le Phillip, P., Bah], A., and Ungar, L. H. (2004) Using prior knowledge to improve
genetic network reconstruction from rrnicroarray data. In Silico Biol 4, 335-353

Yirn, L., Huang, C. H., and Ni, J. (2006) Clustering of gene expression data:
performance and similarity analysis. BMC Bioinformatics 7 Suppl 4, S19

Toronern, P., Kolehmainen, M., Wong, G., and Castren, E. (1999) Analysis of gene
expression data using self-organizing maps. FEBS Lett 451, 142-146

Eisen, M. B., Spellman, P. T., Brown, P. 0., and Botstein, D. (1998) Cluster
analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A
95, 14863-14868

Ma, S. F., Grigoryev, D. N., Taylor, A. D., Nonas, S., Sarnmarni, S., Ye, S. Q., and
Garcia, J. G. (2005) Bioinforrnatic identification of novel early stress response
genes in rodent models of lung injury. Am J Physiol Lung Cell Mol Physiol 289,
L468-477

Toh, H., and Horimoto, K. (2002) Inference of a genetic network by a combined

approach of cluster analysis and gaphical Gaussian modeling. Bioinformatics 18,
287-297

Segal, E., Shapira, M., Regev, A., Pe'er, D., Botsteirn, D., Koller, D., and Friedmarn,
N. (2003) Module networks: identifying regulatory modules and their condition-
specific regulators from gene expression data. Nat Genet 34, 166-176

Li, Z., and Chan, C. (2004) Inferring patlnways and networks with a Bayesian
framework. FASEB J 18, 746-748

Husmeier, D. (2003) Sensitivity and specificity of inferring genetic regulatory
interactions from microarray experiments with dynamic Bayesian networks.
Bioinformatics 19, 2271-2282

Yu, J., Snnith, V. A., Wang, P. P., Hartemink, A. J., and Jarvis, E. D. (2004)
Advances to Bayesian network inference for generating causal networks from
observational biological data. Bioinformatics 20, 3594-3603

Bar-Joseph, Z., Gerber, G. K., Lee, T. 1., Rinaldi, N. J., Yoo, J. Y., Robert, R,
Gordon, D. B., Fraenkel, E., Jaakkola, T. S., Young, R. A., and Gifford, D. K.

134

35.

36.

37.

38.

39.

40.

41.

42.

43.

(2003) Computational discovery of gene modules and regulatory networks. Nat
Biotechnol 21, 1337-1342

Berman, B. P., Nibu, Y., Pfeiffer, B. D., Tomancak, P., Celrniker, S. E., Levine, M.,
Rubin, G. M., and Eisern, M. B. (2002) Exploiting transcription factor binding site
clustering to identify cis-regulatory modules involved in pattern formation in the
Drosophila genome. Proc Natl Acad Sci U S A 99, 757-762

Hartemink, A. J., Gifford, D. K., Jaakkola, T. S., and Young, R. A. (2002)
Combining location and expression data for principled discovery of genetic
regulatory network models. Pac Symp Biocomput, 43 7—449

Ideker, T., Thorsson, V., Ranish, J. A., Christmas, R., Buhler, J., Eng, J. K.,
Bumgarner, R., Goodlett, D. R., Aebersold, R., and Hood, L. (2001) Integrated

genomic and proteomic analyses of a systematically perturbed metabolic network.
Science 292, 929-934

Pilpel, Y., Sudarsanarn, P., and Church, G. M. (2001) Identifying regulatory
networks by combinatorial analysis of promoter elements. Nat Genet 29, 153-159

Li, R, and Yang, Y. (2004) Recovering genetic regulatory networks fi'om rrnicro-
array data and location analysis data. Genome Inform 15, 131-140

Srivastava, S., Zhang, L., Jin, R., and Chan, C. (2008) A novel method
incorporating gene ontology information for unsupervised clustering and feature
selection. PLoS ONE 3, e3860

Yang, X., Zhou, Y., Jin, R., and Chan, C. (2009) Reconstruct Modular Phenotype-
Specific Gene Networks by Knowledge-Driven Matrix Factorization.
Bioinformatics

Zhou, Y., Li, Z., Yang, X., Zhang, L., Srivastava, S., Jin, R., and Chan, C. (2008)
Using knowledge driven matrix factorization to reconstruct modular gene
regulatory network. In AAA] Conference on Artificial Intelligence pp. 811-816, The
AAAI Press, Menlo Park, CA., Chicago, IL

Boyd, S. P., and Vandenberghe, L. (2004) Convex optimization, Cambridge
University Press, Cambridge, UK ; New York

Srivastava, S., and Chan, C. (2006) Hydrogen peroxide and hydroxyl radicals
mediate palrrnitate-induced cytotoxicity to hepatoma cells: Relation to
mitochondrial permeability transition. Free Radic Res 41, 38-49

135

 

 

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

Subramaniarn, A., Tarnayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette,
M. A., Paulovich, A., Pomeroy, S. L., Golub, T. R., Lander, E. S., and Mesirov, J. P.
(2005) Gene set enrichment analysis: a knowledge-based approach for interpreting
genome-wide expression profiles. Proc Natl Acad Sci U S A 102, 15545-15550

Lehninger, A. L., Nelson, D. L., and Cox, M. M. (2005) Lehninger principles of
biochemistry, W.H. Freeman, New York

Guo, W., Wong, S., Xie, W., Lei, T., and Luo, Z. (2007) Palmitate modulates
intracellular signaling, induces endoplasmic reticulum stress, and causes apoptosis

in mouse 3T3-L1 and rat primary preadipocytes. Am J Physiol Endocrinol Metab
293, E576-586

Karaskov, E., Scott, C., Zhang, L., Teodoro, T., Ravazzola, M., and Volchuk, A.
(2006) Chronic palnnitate but not oleate exposure induces endoplasmic reticulrun

stress, which may contribute to INS-1 pancreatic beta-cell apoptosis.
Endocrinology 147, 3398-3407

Zhang, Z., Lee, Y. C., Kim, S. J., Choi, M. S., Tsai, P. C., Xu, Y., Xiao, Y. J.,
Zhang, P., Heifer, A., and Mukherjee, A. B. (2006) Palrrnitoyl-protein tlnioesterase-l
deficiency mediates the activation of the unfolded protein response and neuronal
apoptosis in INCL. Hum Mol Genet 15, 337-346

Zhou, Q., Du, J., Hu, Z., Walsh, K., and Wang, X. H. (2007) Evidence for adipose-
muscle cross talk: opposing regulation of muscle proteolysis by adiponectin and
Fatty acids. Endocrinology 148, 5696-5705

Ding, W. X., Ni, H. M., Gao, W., Yoshimori, T., Stolz, D. B., Ron, D., and Yin, X.
M. (2007) Linking of autophagy to ubiquitin-proteasome system is important for

the regulation of endoplasmic reticulum stress and cell viability. Am J Pathol 171,
513-524

Zhao, L., and Ackennan, S. L. (2006) Endoplasmic reticulum stress in health and
disease. Curr Opin Cell Biol 18, 444-452

Shen, X., Zhang, K., and Kaufinan, R. J. (2004) The unfolded protein response--a
stress signaling patlnway of the endoplasmic reticulum. J Chem Neuroanat 28, 79-
92

Ron, D. (2002) Translational control in the endoplasmic reticulum stress response.
JClin Invest 110, 1383-1388

136

 

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

Lai, B., Bikopoulos, G., Wheeler, M. B., Rozakis-Adcock, M., and Volchuk, A.
(2008) Differential activation of ER stress and apoptosis in response to chronically
elevated free fatty acids in pancreatic beta-cells. Am J Physiol Endocrinol Metab
294, E540-550

Gonzalez-Barroso, M. M., Fleury, C., Arechaga, 1., Zaragoza, P., Levi-Meyrueis, C.,

Raimbault, S., Ricquier, D., Bouillaud, F., and Rial, E. (1996) Activation of the
uncoupling protein by fatty acids is modulated by mutations in the C-terrninal
region of the protein. Eur J Biochem 239, 445—450

Larneloise, N., Muzzin, P., Prentki, M., and Assimacopoulos-Jeannet, F. (2001)
Uncoupling protein 2: a possible link between fatty acid excess and impaired
glucose-induced insulin secretion? Diabetes 50, 803-809

Nedergaard, J ., Ricquier, D., and Kozak, L. P. (2005) Uncoupling proteins: current
status and therapeutic prospects. EMBO Rep 6, 917-921

Breen, E. P., Gouin, S. G., Murphy, A. F., Haines, L. R., Jackson, A. M., Pearson,
T. W., Murphy, P. V., and Porter, R. K. (2006) On the mechanism of mitochondrial
uncoupling protein 1 function. J Biol Chem 281, 2114-2119

Leung, K. F., Baron, R., and Seabra, M. C. (2006) Thematic review series: lipid
posttranslational modifications. geranylgeranylation of Rab GTPases. J Lipid Res
47, 467-475

Seabra, M. C., Mules, E. H., and Hume, A. N. (2002) Rab GTPases, intracellular
traffic and disease. Trends Mol Med 8, 23-30

Stenmark, H., and Olkkonen, V. M. (2001) The Rab GTPase family. Genome Biol 2,
REVIEWS3007

Ishihara, K., Yasuda, K., and Hatayama, T. (1999) Molecular cloning, expression
and localization of human 105 kDa heat shock protein, hsp105. Biochim Biophys
Acta 1444, 138-142

Yamagishi, N., Ishihara, K., Saito, Y., and Hatayama, T. (2003) Hsp105 but not
Hsp70 family proteins suppress the aggegation of heat-denatured protein in the
presence of ADP. FEBS Lett 555, 390-396

Yamagishi, N., Nishihori, H., Ishihara, K., Ohtsuka, K., and Hatayama, T. (2000)
Modulation of the chaperone activities of Hsc70/Hsp40 by Hsp105alpha and
Hsp105beta. Biochem Biophys Res Commun 272, 850-855

137

 

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

Yamagishi, N., Ishihara, K., Saito, Y., and Hatayama, T. (2006) Hsp105 family
proteins suppress staurosporine-induced apoptosis by inhibiting the translocation of
Bax to mitochondria in HeLa cells. Exp Cell Res 312, 3215-3223

Hatayama, T., Yamagishi, N., Minobe, E., and Sakai, K. (2001) Role of hsp105 in
protection against stress-induced apoptosis in neuronal PC12 cells. Biochem
Biophys Res Commun 288, 528-534

Yamagishi, N., Saito, Y., Ishihara, K., and Hatayama, T. (2002) Enhancement of
oxidative stress-induced apoptosis by Hsp105alpha in mouse embryonal F9 cells.
Eur J Biochem 269, 4143-4151

Meares, G. P., Zmijewska, A. A., and Jope, R. S. (2008) HSP105 interacts with
GRP78 and GSK3 and promotes ER stress-induced caspase-3 activation. Cell
Signal 20, 347-358

Nelson, N. (1987) The vacuolar proton-ATPase of eukaryotic cells. Bioessays 7,
251-254

Nelson, N. (1992) Evolution of organellar proton-ATPases. Biochim Biophys Acta
1100, 109-124

Ganong, W. F. (2003) Review of medical physiology, McGraw-Hill, New York
London

Alberts, B. (2002) Molecular biology of the cell, Garland Science, New York

Acosta, D., and Wenzel, D. G. (1974) Injury produced by fiee fatty acids to
lysosomes and nnitochondria in cultured heart muscle and endothelial cells.
Atherosclerosis 20, 417-426

Nakarnura, J ., and Lou, L. (1995) Biochemical characterization of human GMP
synthetase. J Biol Chem 270, 7347-7353

Berg, J. M., Tymoczko, J. L., and Stryer, L. (2002) Biochemistry, W.H. Freeman,
New York

Lu, Z. H., Mu, Y. M., Wang, B. A., Li, X. L., Lu, J. M., Li, J. Y., Pan, C. Y.,
Yanase, T., and Nawata, H. (2003) Saturated free fatty acids, palnnitic acid and
stearic acid, induce apoptosis by stimulation of ceramide generation in rat testicular
Leydig cell. Biochem Biophys Res Commun 303, 1002-1007

138

 

78.

79.

80.

Landau, Z., Forti, E., Alcaly, M., and Birk, R. Z. (2006) Palnnitate induced
lipoapoptosis of exocrine pancreas AR42J cells. Apoptosis 11, 717-724

Barreyro, F. J., Kobayashi, S., Bronk, S. F., Werneburg, N. W., Malhi, H., and
Gores, G. J. (2007) Transcriptional Regulation of Bim by FoxO3A Mediates
Hepatocyte Lipoapoptosis. J Biol Chem 282, 27141-27154

Li, Z., Srivastava, S., Findlan, R., and Chan, C. (2008) Using Dynamic Gene
Module Map Analysis To Identify Targets That Modulate Free Fatty Acid Induced
Cytotoxicity. Biotechnol Prog 24, 29-37

139

 

CHAPTER 5: PKR Differentially Regulates IRS1
and IRSZ in HepG2 Cells

The work in this chapter has been submitted for publication:
Yang, X., Opperman, M. and Chan, 0., PKR Differentially Regulates IRS1
and lRSZ in HepGZ Cells, submitted to Molecular and Cellular Biology, under

review

Abstract

Irnitially identified to be activated upon virus infection, the double-stranded RNA-
dependent protein kinase (PKR) is best krnown for triggering cell defense responses by
phosphorylating eIF-Za, thus suppressing RNA translation. Here, I uncovered a novel
function of PKR in regulating insulin signaling activity. I found that PKR up-regulates
the inhibitory phosphorylation of IRS1 at Ser312, which is believed to suppress the
activity of IRS1. This effect of PKR on the phosphorylation of IRS1 is mediated by 2
other protein kinases, JNK and IKK. By contrast, PKR regulates IRSZ, another
predominant IRS family protein in the liver, at the transcriptional rather than the post-
translational level, and this effect is mediated by a transcription factor, FoxOl, which
plays a critical role in hyperglycemia during insulin resistance. In summary, I found, for
the first time, that PKR regulates the transmitters of insulin signaling, IRS1 and IRS2,
tlnrough different mechanisms. In addition, I as well as other researchers showed that

PKR activity is down-regulated by insulin through the IRS proteins. Therefore this

140

 

feedback loop between PKR and IRS proteins may have potential implications in the

regulation of insulin signaling and the development of insulin resistance.

Introduction

Insulin signaling, a central signaling pathway that regulates many cellular
activities, such as glucose and lipid metabolism, protein synthesis and degadation, cell
gowth and differentiation (1), has been extensively studied over the past decades. Insulin
signaling is initiated upon binding of insulin to the insulin receptor (IR), a receptor
tyrosine kinase (2), and transmitted intracellularly by the insulin receptor substrates (IRS)
(1, 2). At least 4 of the IR substrates belong to the IRS goup, with IRS1 and IRS2 being
predominant and expressed in most tissues, including the liver (3, 4). Upon
phosphorylation of the tyrosine residues catalyzed by IR, the IRS proteins initiate,
through different binding mechanisms (4), various downstream signal transduction
cascades, including mitogen-activated protein kinase (MAPK) pathways (c-Jun N-
terrninal kinase (JNK), extracellular signal-related kinase (ERK), and p3 8) (5, 6) and
phosphoinositide 3-kinase (PI3 K) (7), which in turn activates Akt/protein kinase B
(Akt/PKB) (8), and atypical protein kinase C (aPKC) (9).

Insulin signaling is sophisticatedly tuned by a large number of regulators.
Pathologically, dysregulation of insulin signaling is closely related to the development of
insulin resistance (10), and contributes to multiple diseases and disorders, such as type 2
diabetes as well as other metabolic, endocrine and cardiovascular disorders (IO-12). At
the molecular level, dysregulation of insulin signaling could occur at several possible

stages, e.g., degadation or mutation of IR (13, 14), inhibitory phosphorylation or

141

 

degadation of IRS (15, 16), or suppression of down-stream signaling molecules, such as
PI-3 kinase or Akt/PKB (reviewed in ref. (17)). However, most of the insulin signaling
disruption and insulin resistance have been attributed to dysregulation of the
phosphorylation of IR and, in particular, IRS (15, 16). [RS proteins mediate intracellular
insulin signaling, through the tyrosine residues, which facilitate recruitment of IRS
substrates and promote insulin signaling (4), and the serine residues, which generally
suppress the activities of IRS by blocking the interaction between IRS and IR (18),
inhibiting the tyrosine phosphorylation of IRS (19), or inducing the degadation of IRS
(20). A number of serine residues have been identified to negatively regulate the activity
of IRS], in particular, Ser307 (equivalent to Ser312 in human IRSl). Ser307 has been
extensively investigated and characterized as a key indicator of inhibitory
phosphorylation of IRS1 and insulin resistance, and confirmed in insulin-resistant rodent
models (21).

Although the mechanisms are not well understood, a number of factors, such as
cerarrnide (22, 23), and Hepatitis C viral (HCV) infection (24, 25), have been associated
with the down-regulation of insulin signaling and development of insulin resistance in
liver cells, potentially through their regulation of IRS1 Ser312 phosphorylation.
Interestingly, these factors share a common link; they both can activate the double-
stranded RNA-dependent protein kinase (PKR). Cerarrnide induces the phosphorylation of
PKR by activating the PKR activator X (RAX) (or its human homolog PACT) (26). HCV
infection induces phosphorylation and activation of PKR, mediated by the HCV core
protein, which directly binds and activates PKR (27). As an anti-viral protein, PKR is

best known for triggering the cell’s defense response and initiating the innate immune

142

 

response by arresting general protein synthesis and inducing apoptosis during virus
infection (28, 29). In addition, as a serine-threonine protein kinase that is ubiquitously
expressed, PKR is actively involved in multiple signaling pathways by regulating the
activities of transcription factors, protein kinases, and phosphatases (30). Therefore, PKR
serves as a signal integator, regulating various cellular activities, such as protein
syntlnesis and degadation, stress and immune responses, and apoptosis (30, 31). Thus far,
it has not been determined whether PKR plays a role in regulating insulin signaling.
Given that ceramide and HCV, botln known as activators of PKR, have been identified as
regulators of insulin signaling and inducers of insulin resistance by promoting inhibitory
phosphorylations of IRS1, I hypothesize that PKR plays a role in the induction of insulin
resistance by phosphorylating IRS1 at Ser312, a key indicator of inhibitory
phosphorylation of IRS1 and insulin resistance.

A number of signaling molecules, such as Ich kinase (IKK) (32), mammalian
target of raparnycin (mTOR) (33), PKCC (34), S6 Kinase 1 (S6Kl) (35) and JNK (36, 37)
have been shown to function as IRS serine kinases that induce insulin resistance by
promoting the inhibitory phosphorylation of IRS (reviewed in (15, 38)). PKR, as a signal
integator of many intracellular signaling events (3 O), has been shown to activate certain
IRS kinases, e.g., IKK (39) and JNK (40, 41). Therefore, I investigated whether PKR is
involved in the induction of inhibitory Ser312 phosphorylation of IRS1, and if so, which
IRS kinase mediates the phosphorylation.

Since IRS1 and IRS2 are both expressed in liver cells, I investigated whether PKR
also affected IRS2. Our results indicated that PKR regulates the gene transcriptiorn, rather

than the post-translational modification, of IRS2. The trarnscription of the IRS2 gene has

143

been shown to be up-regulated by several transcription factors, such as forkhead box 01
(FoxOl) (42, 43) and cAMP response element binding (CREB) (44). I found that PKR
enhanced the expression of IRS2 through FoxOl, which controls the transcription of
IRS2 but not IRS1. PKR did not affect the activity of CREB.

In addition to showing that PKR differentially regulates the two major IRS
proteins, IRS1 and IRS2. I confirmed that insulin suppressed the activity of PKR in liver
cells, mediated by IRS, thereby identifying PKR as a downstream target of insulin
signaling. Indeed, a recent study showed insulin suppressed the activity of PKR in muscle
cells (45). Taken together, our results suggest a novel feedback mechanism between PKR

and the IRS proteins.

Materials and Methods

Cell Culture and Reagents—HepG2 cells were cultured as previously described in
Chapter 2 (46). Human insulin and 2-aminopurine (2-AP) were purchased from Signa-
Aldrich (St. Louis, MO), C2 ceramide, Tautomycetin, SC-514, okadaic acid (0A), PKR
inhibitor, JNK inhibitor (SP600125) and their analogues, used as negative controls, from
EMD Biosciences (San Diego, CA).

Insulin treatment—Human insulin was stocked in HEPES buffer, which was
therefore used in controls for all the experiments with the insulin treatment. I treated the
cells with insulin at the concentrations lower than 1 nM to mimic the physiological
concentrations (47). At 95% confluence, cells were deprived of serum for 16 hours prior
to each experiment and subjected to insulin treatments for the indicated doses and times

at 37 °C in serum-free medium.

144

 

Chemical inhibitors—In the present study, I used the commercially available 2-
AP and PKR inhibitor as one of the tools to elucidate the role of PKR. However, even
though these chemical have been widely used in the studies of PKR for a variety of
systems, the specificity of these inhibitor has not been extensively tested in the literature.
Keeping in mind the potential non-specific targets of these PKR inhibitors, 1 performed
more PKR gene silencing and over-expressing studies to test our hypothesis and draw the
conclusions. The JNK inhibitor, SP600125, IKK inhibitor, SC-514, PPlc inhibitor, TMC,
and PP2A inhibitor, OA are popularly used and proven to be specific to their targets at
the concentrations I used (48-51).

RNA interference for PKR and reverse transfection—Silencer® Validated siRN A
targeting human PKR mRNA and Silencer® Pre-designed siRNAs targeting human
FoxOl, IRS1 and IRS2 mRNAs were purchased fiom Ambion (Austin, TX). The
synthesized oligonucleotides are 5'-GGUGAAGGUAGAUCAAAGATT-3' and 5'-
UCUUUGAUCUACCUUCACCTT-3' for siRNA of PKR, 5'-
CCCAAGAGCAUGCACAAACTT-B' and 5'-GUUUGUGCAUGCUCUUGGGTI‘-3' for
siRNA of IRS1, 5'-GCGAGUACAUCAACAUCGATT-3’ and 5'-
UCGAUGUUGAUGUACUCGCCG-3' for siRNA of IRS2, and 5'-
GCUCAAAUGCUAGUACUAUTT-3' and 5'-AUAGUACUAGCAUUUGAGCTA-3'
for siRNA of FoxOl. reverse transfection of siRNA was performed with the transfection
reagent, Lipofectamine RNAiMAX (Invitrogen) Reverse transfection of siRNA was
performed as described in Chapter 2 (40), with the transfection reagent, Lipofectannine

RNAiMAX (Invitrogen).

145

 

Over-expression of PKR and forward transfection—The PKR plasmid, pCMV6-
XLS-hPKR, and the empty vector, pCMV6-XL5, were purchased from Origene
(Rockville, MD). In general, as described previously in Chapter 2 (40), regular HepG2
cells (Figs. 5.3C, 5.5C) or cells with the FoxOl gene silenced by the siRNA of FoxOI
(Fig. 5.5E) were subjected to forward transfections of the plasmids according to the
Lipofectamine 2000 (Invitrogen) method. After 6 hours of trarnsfection, the cells were
then cultured in regular media for 42 hours, and harvested (Fig. 5.5E) or treated with
other chemicals (Figs. 5.3C, 5.5C) subsequently.

Western blot analysis and immunoprecipitation— HepG2 cell lysate was obtained
and Western Blot analysis was performed as previously described in Chapter 2 (40).
Phospho site-specific anti-IRS1 (Tyr94l), PPPlA (Thr320), and anti-PPPlA antibodies
were purchased from Abcam (Cambridge, MA), phospho site-specific anti-IKKa/B
(Serl 76/ 180), FoxOl (Ser256), anti-biotin, anti-IKKB, and anti-FoxOl antibodies from
Cell Signaling (Danvers, MA), phospho site-specific anti-IRS1 (Ser312), IRS2 (Ser731),
PKR (Thr451), JNK (T183/Y185), anti-IRS], anti-IRS2, anti-PKR, anti-JNK, and anti-
beta actin antibodies from Signa-Aldrich. Secondary anti-rabbit and anti-mouse
antibodies were purchased from Pierce Biotechnology Inc.

Real-time quantitative RT -PCR analysis—Total RNA was extracted from cells
with the RNeasy mini kit (Qiagen, Valencia, CA) and depleted of contaminating DNA
with RNase-free DNase (Qiagen). Equal amounts of total RNA (1 ug) were reverse-
transcribed using an iScript cDNA synthesis kit (Bio-RAD). The first-strand cDNA was
used as a template. The primers used for quantitative RT-PCR analyses of human IRS1

(5'-TCCACCTCGGATTGTCTCTT-3' and 5'-AGGGACTGGAGCCATACTCA-3'),

146

 

hmnan IRS2 (5'-CCACTCGGACAGCTI‘CTTCT-3' and 5'-
AGGATGGTCTCGTGGATGTT-3 '), and human GAPDH (5'-
AACTI'TGGTATCGTGGAAGGA-3' and 5'-CAGTAGAGGCAGGGATGATGT—3')
were synthesized by Operon Biotechnologies, Inc. (Huntsville, AL). RT-PCR was
performed as described previously in Chapter 2 (40), and normalized to GAPDH
expression levels.

PP2A phosphatase activity assay—The PP2A irnmunoprecipitation phosphatase
assay kit purchased from Millipore (Ternecula, CA) was used to measure
dephosphorylation of a phosphopeptide as an index of phosphatase activity. Briefly, the
cells were lysed using tlne phosphatase extraction buffer specified by the assay kit, and
the catalytic subunit of PP2A (PP2A/C) was immunoprecipitated with anti-PP2A-C
supplied in the assay kit. Agarose-bound immune complexes were collected and
resuspended in 80 ul of Ser/Thr buffer with 750 uM of phosphopeptide ((KRpTIRR),
obtained from the kit). The reaction was conducted for 10 rrnin at 30°C in a shaking
incubator. Supematants (25 ul) were transferred in a 96-well plate, and released
phosphate was measured by adding 100 pl of malachite geen phosphate detection
solution. Color was developed for 10 rrnin before reading the plate at 650 nm. The
absorbance of the reactions was corrected by subtracting the absorbance in samples
treated without Ab. Results were expressed as fold change of PP2A activity as compared
with control cells.

Statistical analysis—All experiments were performed at least three times, and

representative results are shown. All data, unless specified, are shown as the mean :t SD.

147

for indicated number of experiments. One-way ANOVA with Student’s t-test were used

to evaluate statistical significances between different treatment goups.

Results

Inhibiting PKR reversed phosphorylation of IRS1 at Ser312 induced by
ceramide—Ceramide has been shown to inhibit insulin signaling and induce insulin
resistance by up-regulating the Ser phosphorylation and blocking Tyr phosphorylation of
IRS (22, 23). I confirmed cerarrnide promotes phosphorylation of IRS1 at Ser312 and
suppresses phosphorylation at Tyn941 in HepG2 cells (Fig. 5.1A). Since cerarrnide also
can activate PKR(26), I hypothesize that PKR mediates the role of ceramide in up-
regulating the serine phosphorylation and down-regulating the tyrosine phosphorylation
of IRS1.

Inhibiting PKR in cerarrnide-treated cells decreased phosphorylation of IRS] at
Ser312 and increased phosphorylation of IRS1 at Tyr94l (Fig. 5.1B). Anotlner inhibitor
of PKR, 2-aminopurine (2-AP), had a similar effect on the phosphorylation of IRS1 (Fig.
5. l C). Taken together, the results of the chemical inhibitors of PKR suggest that the
catalytic function of PKR is involved in inducing the inhibitory serine phosphorylation of
IRS 1 , which suppresses tyrosine phosphorylation. However, considering the potential
non-specific effects of chemical inhibitors, I further performed gene silencing studies
using siRNA of PKR to confirm the role of PKR in regulating the phosphorylation of
IRS1.

Silencing PKR suppressed the serine phosphorylation and enhanced tyrosine

phosphorylation of IRS1—siRNA targeting PKR markedly irnlnibited the gene (40) and

148

 

 

 

protein expressions of PKR (Fig. 5.2) (40). Silencing PKR with this siRNA in HepGZ
cells significantly blocked the Ser312 and amplified the Tyr94l phosphorylation of IRSl

induced by insulin (Fig. 5.2).

A.

p-IRS1 Tyr941 p-IRS1 Tyr941

p-lRS1 Ser312
p-lRS1 Ser312

IRS1
IRS1

p-PKR Thr451
p-PKR Thr451

 

 

 

 

 

PKR PKR
beta Adi" beta Actin
Ceramide (pM) Ceramide
Ctrl GA 2AP
B.
p—lRS1 Tyr941

p-IRS1 Ser312

 

IRS1

p-PKR Thr451

 

PKR

 

 

beta Actin

 

 

 

Ceramide + + + +
PKR Inhibitor (11M) 0 0.5 2 10

Figure 5.1. Effects of ceramide and PKR inhibitors on the phosphorylation of IRSl.
HepGZ cells were exposed to different levels of ceramide for 12 hours (A). Pre-treated
with 10 11M ceramide for 12 hours, HepG2 cells were exposed to different levels of PKR
inhibitor dissolved in DMSO (control) (B) or 10 mM 2—AP dissolved in PBS:glacial
acetic acid (200:1) (GA, control) (C) for another 12 hours. After treatment, the cells were
harvested, and western blot analysis was performed to detect the level of beta Actin and
the total and phosphorylated levels of PKR and IRS1.

149

Figure 5.2. Involvement of PKR in regulating the phosphorylation of [RSL Reverse
transfection of suspended HepG2 cells were performed with scrambled siRNA (negative
control) or siRNA of PKR for 24 hours and the transfected cells were cultured in regular
media for another 24 hours. Cells were then treated with different concentrations of
insulin for 15 minutes and harvested after the treatment. Western blot analysis was
performed to detect the levels of beta actin and PKR, and the total and phosphorylated
levels of IRS1. The phosphorylation levels of IRS1 at Ser312 and Tyr94l were quantified
by normalizing to total IRS1 levels and are expressed as the average of three samples 3:
SD fiom tlnree independent experiments. Student t-test was performed for analyzing the
differences between samples transfected with siPKR and scrambled siRNA (negative
control). *, significantly higher (Tyr94l) or lower (Ser312) than negative control (i.e.,
scrambled siRNA), p<0.01.

150

 

 

WB:

siPKR

 

Neg. Ctrl.

P-IRS1Tyr941—> + -

 

. ‘s 1‘: 3 ..
,-, 1:. :1. .r.. _.-.'. *r;
w- "1'2: - t .
5,. 5:25:53 , .2. .
l - ‘ ,.. -. .: ~ - r .. - . ~ -,:-,-....

p-IRS1 Ser312

 

IRS1

 

 

 

 

beta Actin

 

0.2 0.6 1.0 (15mins)

0.2 0.6 1.0 0

0

Insulin (nM)

     

 

.. %//////x1
. . Meme ._ .

   

l Scrambled

2
0

 

2.5 -

 
 

.1

39.20 22.
5x. .59 a «5.8 $5-..

Insulin (nM)

151

Taken together, both the inhibition and gene silencing studies suggest PKR may
be involved in regulating insulin signaling by inducing phosphorylation of IRS1 at
Ser312 and suppressing phosphorylation at Tyr941. However, it is unclear how PKR
regulates the phosphorylation of IRS1. Co-irnmunoprecipitation showed that PKR did not
directly interact with IRS1 (Fig. 5.3A), suggesting other intermediate signaling molecules
must mediate the effect of PKR on the phosphorylation of IRS1. IRS1 Ser kinases, which
directly phosphorylate the serine residues of IRS], include IKK (32), mTOR (33), PKCl;
(34), S6K1 (35) and JNK (36). PKR has been reported to positively activate the MAPKs,
in particular JNK (40, 41), and IKK (39). Therefore, I hypothesize that PKR induces
phosphorylation of IRS1 at Ser312 through IRS serine kinases, JNK, IKK, or both.

PKR positively regulates JNK and IKK, both of which mediate the effect of PKR
on the Ser phosphorylation of IRS1—I previously showed that PKR co-
immunoprecipitates with JNK and activates JNK in HepG2 cells (40). Here, I show that
IKK also is activated by PKR. Silencing PKR significantly reduced the phosphorylation
of IKKo/B at Ser176/ 180, which indicates H(K activity (Fig. 5.38). To firrtlner confirm
the involvement of IN K and IKK in mediating the effect of PKR on the phosphorylation
of IRS1 at Ser312 and Tyr94l, I over-expressed PKR in HepG2 cells and inhibited the
activity of JNK or IKK. Over-expressing PKR by transfecting the plasmid pCMV6-hPKR
into HepG2 cells enhanced phosphorylation of IRSl at Ser312 and suppressed
phosphorylation at Tyr94l (Fig. 5.3C), supporting our results (Figs. 5.1 and 5.2) that
PKR induces serine phosphorylation of IRS1. Furthermore, in PKR over-expressed cells,
inhibitors of JNK and IKK, SP600125 and $0514, respectively, reduced Ser312

phosphorylation and restored Tyr941 phosphorylation of IRS1 (Fig. 5.3C), suggesting

152

 

 

that both kinases, JNK and IKK, mediate the effect of PKR on the phosphorylation of

IRSl at Ser312, which suppresses tyrosine phosphorylation of IRS1.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

Figure 3
A.
re: IRS1 l8: IRS1 d
lePKR *1 IBzPKR .
IP: NC PKR lP: NC IRS1
B. C.
pCMV pCMV-hPKR
p-lKKor/B Ser176/180: .. _- ‘
~ ~ p-IRS1Tyr941
IKKB .- .1
p-lRS1Ser312
p-JNK T183IY185 ’ RS,

 

 

 

 

JNK p-PKR Thr451

i i i i

 

 

 

NC siPKR

PKR

 

 

 

 

Figure 5.3. Involvement of JNK and IKK in regulating the phosphorylation of IRS].
Confluent HepG2 cells were harvested and immunoprecipitated with anti-PKR, anti-IRS1,
or IgG as negative controls, and western blot analysis was performed to detect the protein
level of IRS1 and PKR (A). Reverse transfection of suspended HepG2 cells was
performed with scramble siRNA (negative control) or siRNA of PKR (siPKR) for 24
hours and the transfected cells were cultured in regular medium for another 24 hours (B).
Cells were harvested, and western blot analysis was performed to detect the total and
phosphorylated levels of JNK and IKK (B). In HepG2 cells, the forward transfection of
empty vector pCMV6-XL5 (pCMV6) or plasmid containing PKR cDNA sequence
(pCMV6-hPKR) was performed and the cells were then treated with the pharmaceutical
inhibitor of JNK, SP600125 (25 uM), inhibitor of IKK, SC-514 (50 uM), or DMSO,
vehicle of these two cherrnicals, for 1 hour (C). After treatments, cells were then harvested
and western blot analysis was performed to detect the total and phosphorylated levels of
IRS1 and PKR(C).

153

Figure 5.4. Involvement of PKR in regulating IRSZ. Confluent HepG2 cells were
treated with 5 uM PKR inhibitor (P1) or its analogue as a negative control (NC) or 10
mM 2-AP dissolved in PBS:glacial acetic acid (200:1) (GA) for 12 hours (A, C). Reverse
transfection of suspended HepG2 cells were performed with scrambled siRNA (negative
control) or siRNA of PKR for 24 hours and the transfected cells were cultured in regular
media for another 24 hours. Cells were then harvested (D) or treated with different
concentrations of insulin for 15 minutes and harvested after the treatment (B). Western
blot analysis was performed to detect the total and phosphorylated levels of IRS2 at
Ser731 (A, B). RT-PCR was performed to detect the gene expression levels of IRSl and
IRS2 in response to the PKR inhibitors (C) or siRNA of PKR (D). Gene expression data
were expressed as the average of nine samples 3: SD from three independent experiments.
The protein levels of IRS2 were quantified by normalizing to beta actin, and the
phosphorylation levels of IRS2 at Ser731 were quantified by normalizing to total IRS2
levels. Botln the protein and phosphorylation levels of IRS2 are expressed as the average
of three samples :t SD fiom three independent experiments. Student t-test was performed
for analyzing the differences between siPKR and scrambled siRNA (negative control) (B,
D) or PKR inhibitors and negative controls. *, significantly lower than negative control
(i.e., scrambled siRNA (B, D) or chemical analogue of the PKR inhibitor (C)), p<0.05. **,
significantly lower than control (GA, solvent of 2-AP), p<0.01.

154

 

 

WB:

p-lRS2 Ser731

IRSZ

beta Actin

NC PI GA 2AP

WB:

siPKR

11131!!!

 

p-IRS2 Ser731

 

 

| R32

 

 

beta Actin

0.2 0.6 1.0 (15mins)

0.2 0.6 1.0 0

0

Insulin (nM)

 

 

ar/

.0

b

b

m

m

C

S

I
its:
1100

39.20 28. «mm.

Insulin (nM)

l Scrambled

   

//

 

2.5 ~

_ m
5 4|
1

$955 2o:
3”: .50: Stew «9:8

2_ a
0.5~

Insulin (nM)

155

Figure 5.4 continued

    

 

 

 

C- r).
1.8 - 1 2 _
I IRS1 ‘ llRS1IlR82
'5 2
5 E 1 ‘ '7
c f; -' A
.2 O) c 0 0.8 -
r: o D
r g -- s
a 0 § 5 0 6 -
x '0 '-
ur a a 2
g 2: ru ,3 0.4 —
E net 0.2 —
0 _
Neg. Ctrt. PKR GA 2-AP NC siPKR
inhibitor

Thus far, I have identified a novel function of PKR in regulating the
phosphorylation of IRS1. In addition to IRS1, I also investigated the potential effect of
PKR on another major IRS family protein, IRSZ, which also mediates insulin signaling in
the liver (3).

PKR up-regulates the protein expression level of IRS2—Both PKR irnhibitors,
PKR inhibitor and 2-AP, down-regulated the protein level of IRS2 (Fig. 5.4A), but not
IRS1 (Fig. 5.1), suggesting that PKR is required for cells to maintain proper protein
expression level of IRSZ. Similarly, silencing the gene expression of PKR also down-
regulated the protein level of IRSZ (Fig. 5.4B), further supporting the effect of PKR on
the protein expression level of IRSZ. Notably, upon inhibiting or silencing PKR, the
phosphorylation of IRSZ at Ser731 varied proportionally to its total protein level (Fig.
5.4A and B), suggesting that PKR is not affecting the phosphorylation of IRS2 at Ser731.
To determine whether PKR transcriptionally regulates the protein level of IRS2, I
measured the mRNA expression levels of IRS1 and IRSZ upon PKR inhibition and gene

silencing. Both the PKR inhibitors and the siRNA of PKR down-regulated the mRNA

156

expression of IRSZ, but not of IRS1 (Fig. 540, suggesting PKR transcriptionally
regulates IRS2.

PKR rap-regulates the protein level of IRSZ through the transcription factor
FoxOI—PKR, a protein kinase, activates several transcription factors, such as IRF-l , p53,
and NF-ch (52, 53), but these have not been shown to regulate the transcription of IRS2.
However, the transcription of IRSZ has been shown to be dependent on the transcription
factor FoxOl in the liver (42, 43) or CREB in pancreatic beta-cells (44). It is not known
whether PKR interacts with either of tlnese two transcription factors. I found that PKR
had no effect on the activity or translocation of CREB in HepG2 cells (not shown).
However, silencing the PKR gene significantly increased the phosphorylation of FoxOl
at Ser256 (Fig. 5.5A). Phosphorylation at Ser256 inhibits the DNA-binding activity of
F 0x01 and its nuclear import by suppressing the nuclear targeting signal on its DNA
binding domain (54, 55). Our results suggest, for the first time, that PKR reduces the
phosphorylation of FoxOl and thereby activates it. As a protein kinase, PKR does not
have the phosphatase activity that is required to dephosphorylate FoxOl . However, PKR
is known to phosphorylate 85611, the regulatory subunit of protein phosphatase 2A
(PP2A), which then activates the catalytic subunit of PP2A (56). PP2A, in turn, is known
to dephosphorylate FoxOl at Ser256 (57). Indeed, silencing PKR significantly
suppressed the activity of PP2A (Fig. 5.5B), thereby confirming that PKR activates PP2A
in HepG2. To further investigate the potential involvement of PP2A in mediating the
dephosphorylation of Ser256 on FoxOl by PKR, I over-expressed PKR in HepG2 cells
and inhibited the activity of PP2A with okadaic acid (OA). Over-expressing PKR by

trarnsfecting the plasmid pCMV6-hPKR into HepG2 cells reduced the phosphorylation of

157

 

 

FoxOl at Ser256 (Fig. 5.5C), supporting our results that PKR induces dephosphorylation
of FoxOl at Ser256 (Fig. 5.5A). More importantly, 0A, a specific PP2A irnhibitor,
restored the serine phosphorylation level of FoxO] in PKR over-expressed cells (Fig.
5.5C). Taken together, PKR dephosphorylates and activates FoxOl , mediated by PP2A.

To confirm the positive effect of FoxOl on the expression of IRS2 in HepG2 cells,
I performed gene silencing study of FoxOl. Silencing the gene expression of FoxO]
significantly reduced the protein level of IRS2, but not IRS1 (Fig. 5.5D), suggesting that
FoxO] controls the expression of IRS2 in HepG2 cells. To firrther confirm that FoxOl
mediates the effect of PKR on IRS2 protein level, I over-expressed PKR in HepG2 cells
as well as silenced FoxOl. Over-expressing PKR in HepGZ cells increased the protein
level of IRS2 (lanes 1 vs. 2 in Fig. 5.5E), whereas silencing FoxOl in control and PKR
over-expressed cells significantly reduced the protein level of IRS2 (Fig. 5.5E),
confirming that PKR up-regulates the protein level of IRS2 through the transcription
factor FoxOl.

In summary, our results suggest two different pathways by which PKR regulates
IRS proteins and thereby insulin signaling (Fig. 5.9). First, PKR induces phosphorylation
of IRS] at Ser312, and suppresses tyrosine phosphorylation of IRS], mediated by IRS
kinases, JNK and IKK. Second, PKR activates a transcription factor, FoxOl, which up-
regulates the gene expression of IRS2. In addition, the activity of PKR was recently
shown to be suppressed by insulin or insulin-like gowth factor 1 (IGFl) in muscle cells
(45). I confirmed a similar effect of insulin on the activity of PKR in HepG2 cells.

Therefore I propose that PKR is a novel downstream target of insulin signaling.

158

 

 

Figure 5.5. Involvement of Fox01 in mediating the effect of PKR on IRS2. Reverse
transfection of suspended HepG2 cells were performed with scrambled siRNA (control)
or siRNA of PKR (A, B) or siRNA of Fox01 (D) for 24 hours and the transfected cells
were cultured in regular media for another 24 hours. The forward transfection of empty
vector pCMV6-XL5 (pCMV6) or plasmid containing PKR cDNA sequence (pCMV6-
hPKR) was performed and the cells were then treated with OA (2 nM) or its vehicle,
ethanol, as a control, for 1 hour (C). Reverse transfection of scramble siRNA (negative
control, lanes 1 and 2) or siRNA of Fox01 (siPKR, lanes 3 and 4) was performed,
followed by the forward transfection of empty vector pCMV6-XL5 (pCMV, lanes 1 and
3) or the plasmid containing PKR cDNA sequence (hPKR, lanes 2 and 4) (B). After
treatment, the cells were harvested, and western blot analysis was performed to detect the
total and phosphorylated levels of Fox01 (A, C) or the total levels of IRS], IRS2, Fox01
and beta actin (D, E). PP2A activity assay was performed to detect the phosphatase
activity of PP2A (B). Student t-test was performed and p values were calculated for
analyzing the differences between the indicated samples.

159

 

 

C.

         

 

1" we- we:
p-Fox01 $256 _’ p-FoxO1 $256 —> was: r"
. > 1 ,
Fox01 { 3 f Fox01
a It — 1.
NC siPKR CMV hPKR hPKR+OA
B.
A 1.2 “ D.
0
O
C
2 IRSZ
0
'U
E IRS1
> .
' RKO1 5 .
O.

 

 

NC siFoxO1

NC siPKR

IRSZ

IRS1

 

PKR an...- , .j. .H' m, d

 

beta Actin

 

 

 

 

NC siFoxO1

Insulin down-regulates the activity of PKR in HepGZ cells, mediated by IRS—
Physiological concentrations of insulin decreased the phosphorylation of PKR at
Thr451in HepG2 cells, in a time and dose dependent manner (Fig. 5.6 A and B),

suggesting that PKR may be a downstream target of the insulin signaling pathway. R85,

160

 

including IRS] and IRS2 in liver cells, have been identified as central transmitters of
insulin signaling (1, 2). To determine whether IRS mediates the suppressive effect of
insulin on PKR phosphorylation, I silenced the expression of IRS] (Fig. 5.7A) or IRSZ
(Fig. 5.7B) in HepG2 cells. The negative effect of insulin on the phosphorylation of PKR,
which was observed in control cells, was abolished in both IRSI- and IRS2-silenced cells
(Fig. 5.7), indicating that IRS] and IRS2 mediate the effect of insulin on the activity of
PKR.

As a tyrosine kinase, IRS does not have Ser/Thr kinase or phosphatase activity.
Therefore an intermediate protein must mediate the dephosphorylation of PKR upon
simulation of IRS by insulin. It is well established that in the liver, insulin up-regulates
the glycogen-targeting subunits of protein phosphatase ] (PP1), i.e., GL and R5 (58, 59),
which activate the catalytic subunit of PP] (PPlc) (60). This process, the up-regulation of
the regulatory subunits and activation of PPlc by insulin, has been suggested to be
mediated by the pathway IRS-PI-3 kinase-Akt (6], 62). As a major eukaryotic protein
serine/threonine phosphatase, PP1 binds to PKR and dephosphorylates PKR at the serine
and threonine residues (63). Therefore, I hypothesize that PP] mediates the
dephosphorylation of PKR induced by insulin. To test this hypothesis, I measured the

phosphorylation of PPlc in response to insulin and studied the inhibition of PPlc.

161

 

A.
WB:

 

p-lRS1Ser312 [ W . p

 

    

 

 

 

 

 

 

beta A011" W

Insulin (nM) 0 0.1 0.2 0.4 1.0 (15 mins)

B.

p-IRS1 Ser312

p-IRS1 Tyr941 —> f

IRS1
p-PKR Thr451 —> .

PKR

 

beta Actin

 

 

 

Insulin (min) 0 5 10 15 30 60 (1.0 nM)

Figure 5.6. Effect of insan on the activity of PKR. HepG2 cells were exposed to
different levels of insulin for 15 rrninutes (A) or lnM of insulin for 5, 10, 15, 30, or 60
rrninutes (B). After treatment, the cells were harvested, and western blot analysis was
performed to detect the total and phosphorylated levels of PKR and IRS].

162

A.
WB:

P-PKRThr451—P -. ,..... Wc eat“ an-

 

PKR "bu—renowned.»

 

IRS2 “wufimum”

 

 

 

IRs1 ‘ u h fl a...

 

 

 

|nsulin(nM) 0 0.2 0.6 1.0 0 0.2 0.6 1.0 (15mins)
Neg. Ctrl. siIRS1

WB:

p-PKRThr451—> "

 

PKR

 

IRSZ

IRS1

beta Actin

|nsulin(nM) 0 0.2 0.6 1.0 0 0.2 0.6 1.0 (15mins)
Neg. Ctrl. silRSZ

 

Figure 5.7. Involvement of IRSl and IRSZ in mediating the effect of insulin on the
phosphorylation of PKR. Reverse transfection of suspended HepG2 cells was
performed with scrambled siRNA (negative control) or siRNA of IRS] (A) or siRNA of
IRS2 (B) for 24 hours and the transfected cells were cultured in regular media for another
24 hours. Cells were then treated with different concentrations of insulin for 15 minutes
and harvested after the treatment. Western blot analysis was performed to detect the total
and phosphorylated level of PKR, and the levels of IRS], IRSZ, and beta Actin.

PP1 mediates the inhibitory effects of insulin on PKR—Insulin suppressed
phosphorylation of PPlc at Thr320, a key residue tlnat upon phosphorylation inhibits the
activity of PPlc (64) (Fig. 5.8). To determine if PP1 mediates the effect of insulin on

PKR, I inhibited PPlc activity with Tautomycetin (TMC), a specific inhibitor of PPlc

163

(48). The phosphorylation of PKR no longer decreases in response to insulin (Fig. 5.8),
suggesting that PP1 mediates the down-regulation of PKR phosphorylation by insulin.
Therefore, I propose that insulin down-regulates the phosphorylation of PKR, mediated

by the pathway from IRS to PP1 (Fig. 5.9).

p-PKR Thr451 —>

 

PKR

 

 

p-PP1c Thr320

 

 

 

PP1c

 

 

beta Actin

 

 

 

 

Insulin (nM) 0 0.6 1.0 0 0.6 1.0 (15 mins)
Neg. Ctrl. TMC

 

 

Figure 5.8. Involvement of PP1 in mediating the effect of insulin on the
phosphorylation of PKR. Confluent HepG2 cells were pre-treated with 5 uM
Tautomycetin (TMC) for 4 hours and exposed to different levels of insulin for 15 nninutes.
After treatment, the cells were harvested, and western blot analysis was performed to

detect the total and the phosphorylated levels of PKR and PPlc, and the level of beta
Actin.

In summary, 1 found that PKR is irnhibited by insulin through the IRS-PP1
pathway, and serves as a downstream substrate of insulin signaling. More importantly,
PKR regulates the central transnnitters of intracellular insulin signaling, the IRS proteins,
through distinct mechanisms. PKR up-regulates phosphorylation of IRS] at Ser312,
through JNK and IKK, thereby suppressing tyrosine phosphorylation of IRS].
Concomitantly, PKR activates the transcription factor, Fox01, up-regulating IRS2
expression. Taken together, these results suggest PKR is involved in insulin signaling

through a feedback mechanism (Fig. 5.9).

164

I

IR

IRS1: p-SerI I lRSZ: protein level
,..—.1 IRS1 IRSZ(

I

Akt

I

IKK JNK PP1
t t I
. + PKR

I IRS 2

PP2A -> Fox01

Figure 5.9. Proposed signaling pathways through which PKR is involved in insulin
signaling network in HepG2 cells. Insulin activates insulin signaling by IR and IRS,
leading to the activation of PP], which in turn dephosphorylates PKR at Thr451. PKR
induces the phosphorylation of IRS] at Ser312 through two other kinases, JNK and IKK.
In addition, by activating PP2A, PKR dephosphorylates a transcription factor, Fox01,
which up-regulates the gene expression of IRS2.

 

 

 

 

 

 

Discussion

In this chapter, I identified novel effects of PKR on 2 major IRS proteins, IRS]
and IRS2, in HepG2 cells. First, PKR up-regulates the phosphorylation of IRS] at Ser312,
which in turn suppresses the tyrosine phosphorylation of IRS]. This effect of PKR is
mediated by JNK and IKK (Fig. 5.3). It is well known that PKR stimulates the
transcription factor NF-ch by activating HéK (39), and this process does not require the
catalytic activity of PKR. Instead, The N-terminus of PKR is responsible for the

activation of IKK (65). As discussed previously (40), PKR has been reported also to play

165

a role in the phosphorylation of the 3 rrnitogen-activated protein kinases (MAPK) (JN K,
ERK, and p38) in the rank order of JNK>p38>ERK (41). Among the three MAPKs, JNK
has been suggested to play a central role in inducing the inhibitory serine phosphorylation
of IRS] (21, 36, 37). I did not test the effects of PKR on the other two less responsive
MAPK proteins, ERK and p3 8, which were also suggested to induce phosphorylation of
IRS] at Ser residues (66, 67). Thus, although JNK may be an important intermediate, it is
not likely the only one involved in mediating the signaling patlnway from PKR to IRS]
phosphorylation. Currently, it remains unclear how PKR interacts with MAPKs, e. g. JNK.
PKR was shown to interact with apoptosis signal-regulating kinase ] (ASKl) (68), a
MAPK kinase kinase (MAPKKK), which phosphorylates SEKl/MKK4 or
MKK3/MKK6 and in turn activates JNK or p38 MAPK, respectively (69). In addition, it
is believed that the activation of the MAPK cascade requires a scaffold protein that
assembles the MAPKKK, MAPKK, and MAPK proteins together into a signaling module
(70). Therefore, the activation of JNK by PKR may require the recruitment of other
proteins, including a scaffold protein that assembles JNK and its upstream activators in
the MAPK signaling pathway. In other words, the PKR—JN K pathway 1 identified does
not preclude other possible intermediates between PKR and JNK.

The phosphorylation of IRS] at Ser312 by PKR provides a potential mechanism
through which ceramide, an activator of PKR and inducer of insulin resistance, promotes
the inhibitory serine phosphorylation of IRS]. Indeed, other activators of PKR have also
been identified to induce insulin resistance and type 2 diabetes. For example, HCV core
protein, which directly binds and activates PKR (27), has been shown to induce insulin

resistance in liver cells by inducing the phosphorylation of IRS] at Ser312 (25). The

166

ability of PKR to promote serine phosphorylation of IRS] provides a possible mechanism
by which PKR mediates HCV infection and the development of insulin resistance and
ultimately type 2 diabetes (24, 25).

Another novel function of PKR is regulating the protein level of IRS2 through the
transcription factor Fox01. Fox01 directs the expression of genes involved in a wide
variety of cellular responses including myogenic gowtln, differentiation, apoptosis, and
stress resistance (71, 72). In liver, FoxOl is regulated by insulin (73) and plays a
significant role in regulating glucose homeostasis and energy metabolism (74). Through
the IRS1-P13K-Akt pathway, insulin induces the phosphorylation of Fox01, thereby
inhibiting its activity (73). In healthy states, the low insulin level during fasting sustains
the activity of Fox01, which facilitates the transcription of key enzymes involved in
gluconeogenesis (75). Fox01 also can up-regulate IRSZ gene expression through a
feedback loop (42, 43), which I also found to occur in HepG2 cells (Fig. 5.5). However in
insulin resistance models, the persistent activation of Fox01, due to disruption of the
IRS-P13 K-Akt pathway, contributes to the development of hyperglycemia and glucose
intolerance (76, 77). Therefore, Fox01, which becomes activated during irnsulin
resistance, is believed to serve as a dominant regulator of hepatic gene expression. The
roles of Fox01 in regulating fasting glucose homeostasis and enhancing hyperglycemia
and glucose intolerance in insulin resistance models begs the question whether PKR
activation is involved in the regulation of glucose metabolism mediated by Fox01.

Here, I identified that PKR regulates IRS] and IRS2 by different mechanisms.
Although the protein structures of the IRS proteins are highly conserved, both animal and

cell studies indicate that IRS] and IRS2 serve complementary, rather than redundant,

167

in: "I"

 

roles in insulin signaling (78-81). It has been suggested that hepatic IRS] and IRS2
control different aspects of hepatic metabolism, with IRS] more closely related to
glucose homeostasis and IRS2 more closely related to lipid metabolism (82). However
more recently, using specific knockouts of liver IRS1 or IRS2, researchers demonstrated
that the two proteins may overlap in their insulin action (7 7 ). Nevertheless, investigators
have claimed that IRS] is stably expressed and frrnctions in the postprandial state, while
IRS2 is highly expressed and activated in the fasted state (77 , 83), which could be due to
elevated activation of Fox01. Taken together, IRS] and IRS2, nevertheless, are believed
to function independently despite these possible overlaps, and the varying effects of PKR
on these two proteins may serve as a potential mecharnism by which the two IRS proteins
are differentially regulated.

I also found that in HepG2 cells the phosphorylation of PKR is down-regulated by
insulin (Fig. 5.6). It is interesting that this down-regulation was abolished upon longer
treatment with insulin (Fig. 5.6B), suggesting, possibly, a negative feedback is regulating
the effect of insulin on PKR. Indeed, insulin signaling is tuned by various negative
feedbacks, e. g., degadation of IR induced by insulin binding (84), negative regulation of
insulin signaling activity by MAPKs (85), mTOR (86), PI-3 kinase (87), etc. These
negative feedbacks modulate the downstream signaling events induced by long-term
treatment of insulin (16).

In summary, I identified that PKR, which is down-regulated by insulin through
the IRS to PP1 pathway, may fimction as a key regulator of insulin signaling (Fig. 5.9).
PKR may inhibit insulin signaling and potentially induce insulin resistance by promoting

the phosphorylation of IRS at Ser312. This firnction of PKR explains perhaps how

168

several PKR activators (e. g., ceramide) induce the inhibitory phosphorylation of IRS],
leading to insulin resistance. On the other hand, PKR activates the transcription factor,
Fox01, which up-regulates the protein expression level of IRS2, with possible
implications on the regulation of glucose homeostasis. Taken together, PKR appears to be
a novel player in the regulation of insulin signaling tlnrough the major transmitters of

insulin signaling, IRS] and IRS2.

169

10.

References

Saltiel, A. R., and Kahn, C. R. (2001) Insulin signalling and the regulation of
glucose and lipid metabolism. Nature 414, 799-806

Patti, M. E., and Kahn, C. R. (1998) The insulin receptor--a critical link in glucose
homeostasis and insulin action. J Basic Clin Physiol Pharmacol 9, 89-109

Thirone, A. C., Huang, C., and Klip, A. (2006) Tissue-specific roles of IRS proteins
in insulin signaling and glucose transport. Trends Endocrinol Metab 17, 72-78

White, M. F. (1998) The IRS-signalling system: a network of docking proteins that
mediate insulin action. Mol Cell Biochem 182, 3-11

Lowenstein, E. J ., Daly, R. J., Batzer, A. G., Li, W., Margolis, B., Lammers, R.,
Ullrich, A., Skolrnik, E. Y., Bar-Sagi, D., and Schlessinger, J. (1992) The SH2 and
SH3 domain-containing protein GRB2 links receptor tyrosine kinases to ras
signaling. Cell 70, 431-442

Skolrnik, E. Y., Lee, C. H., Batzer, A., Vicentini, L. M., Zhou, M., Daly, R., Myers,
M. J., Jr., Backer, J. M., Ullrich, A., White, M. F., and et a1. (1993) The SH2/SH3
domain-containing protein GRB2 interacts with tyrosine-phosphorylated IRS] and
She: implications for insulin control of ras signalling. EMBO J 12, 1929-1936

Backer, J. M., Schroeder, G. G., Kahn, C. R., Myers, M. G., Jr., Wilden, P. A.,
Calnill, D. A., and White, M. F. (1992) Insulin stimulation of phosphatidylinositol
3-kinase activity maps to insulin receptor regions required for endogenous substrate
phosphorylation. J Biol Chem 267 , 1367-1374

Alessi, D. R., James, S. R., Downes, C. P., Holmes, A. B., Gaffrney, P. R., Reese, C.
B., and Cohen, P. (1997) Characterization of a 3-phosphoinositide-dependent

protein kinase which phosphorylates and activates protein kinase Balpha. Curr Biol
7, 261-269

Standaert, M. L., Galloway, L., Kamam, P., Bandyopadhyay, G., Moscat, J., and
F arese, R. V. (1997) Protein kinase C-zeta as a downstream effector of
phosphatidylinositol 3-kinase during insulin stimulation in rat adipocytes. Potential
role in glucose transport. J Biol Chem 272, 30075-30082

Sone, H., Suzuki, H., Takahashi, A., and Yarnada, N. (2001) Disease model:
hyperinsulinemia and insulin resistance. Part A-targeted disruption of insulin
signaling or glucose transport. Trends Mol Med 7, 320-322

170

 

ll.

12.

l3.

14.

15.

16.

17.

l8.

19.

20.

21.

Kahn, B. B. (1998) Type 2 diabetes: when insulin secretion fails to compensate for
insulin resistance. Cell 92, 593-596

Reaven, G. M. (1988) Banting lecture 1988. Role of insulin resistance in human
disease. Diabetes 37, 1595-1607

Irnamura, T., Takata, Y., Sasaoka, T., Takada, Y., Morioka, H., Haruta, T., Sawa,
T., Iwanishi, M., Hu, Y. G., Suzuki, Y., and et al. (1994) Two naturally occurring
mutations in the kinase domain of insulin receptor accelerate degadation of the
insulin receptor and impair the kinase activity. J Biol Chem 269, 31019-31027

McElduff, A., Hedo, J. A., Taylor, S. I., Roth, J ., and Gorden, P. (1984) Insulin
receptor degadation is accelerated in cultured lymphocytes from patients with
genetic syndromes of extreme insulin resistance. J Clin Invest 74, 1366-1374

Zick, Y. (2005) Ser/Thr phosphorylation of IRS proteins: a molecular basis for
irnsulin resistance. Sci SIKE 2005, pe4

Herschkovitz, A., Liu, Y. F., Ilan, E., Ronen, D., Boura-Halfon, S., and Zick, Y.
(2007) Common inhibitory serine sites phosphorylated by IRS-1 kinases, triggered
by insulin and inducers of insulin resistance. J Biol Chem 282, 18018-18027

Taylor, S. I., and Arioglu, E. (1998) Syndromes associated with insulin resistance
and acanthosis nigicans. J Basic Clin Physiol Pharmacol 9, 419-439

Paz, K., Hemi, R., LeRoith, D., Karasik, A., Elhanany, E., Kanety, H., and Zick, Y.
(1997) A molecular basis for insulin resistance. Elevated serine/threonine
phosphorylation of IRS-l and IRS-2 inhibits their binding to the juxtamembrane
region of the insulin receptor and impairs their ability to undergo insulin-induced
tyrosine phosphorylation. J Biol Chem 272, 29911-29918

Hotamisligil, G. S., Peraldi, P., Budavari, A., Ellis, R., White, M. F., and
Spiegelrrnarn, B. M. (1996) IRS-l-mediated inhibition of insulin receptor tyrosine
kinase activity in TNF-alpha- and obesity-induced insulin resistance. Science 271,
665-668

Pederson, T. M., Kramer, D. L., and Rondinone, C. M. (2001) Serine/threonine
phosphorylation of IRS-l triggers its degadation: possible regulation by tyrosine
phosphorylation. Diabetes 50, 24-31

Hirosumi, J., Tuncman, G., Chang, L., Gorgun, C. Z., Uysal, K. T., Maeda, K.,
Karin, M., and Hotamisligil, G. S. (2002) A central role for JNK in obesity and
insulin resistance. Nature 420, 333-336

171

 

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

Miura, A., Kajita, K., Ishizawa, M., Kanoh, Y., Kawai, Y., Natsume, Y., Sakuma,
H ., Yamamoto, Y., Yasuda, K., and Ishizuka, T. (2003) Inhibitory effect of
ceramide on insulin-induced protein kinase Czeta translocation in rat adipocytes.
Metabolism 52, 19-24

Kanety, H., Hemi, R., Papa, M. Z., and Karasik, A. (1996) Sphingomyelinase and
ceramide suppress insulin-induced tyrosine phosphorylation of the insulin receptor
substrate-1. J Biol Chem 271, 9895-9897

Aytug, S., Reich, D., Sapiro, L. E., Berrnsteirr, D., and Begum, N. (2003) Impaired
IRS-1/PI3-kinase signaling in patients with HCV: a mechanism for increased
prevalence of type 2 diabetes. Hepatology 38, 1384-1392

Banerjee, S., Saito, K., Ait-Goughoulte, M., Meyer, K., Ray, R. B., and Ray, R.
(2008) Hepatitis C virus core protein upregulates serine phosphorylation of insulin
receptor substrate-1 and impairs the downstream akt/protein kinase B signaling
pathway for insulin resistance. J Virol 82, 2606-2612

Ruvolo, P. P., Gao, F., Blalock, W. L., Deng, X., and May, W. S. (2001) Ceramide
regulates protein syntlnesis by a novel mechanism involving the cellular PKR
activator RAX. J Biol Chem 276, 11754-11758

Yarn, X. B., Battaglia, S., Boucreux, D., Chen, Z., Brechot, C., and Pavio, N. (2007)
Mapping of the interacting domains of hepatitis C virus core protein and the
double-stranded RNA-activated protein kinase PKR. Virus Res 125, 79-87

Taylor, S. S., Haste, N. M., and Ghosh, G. (2005) PKR and eIF2alpha: integation
of kinase dimerization, activation, and substrate docking. Cell 122, 823-825

Stark, G. R., Kerr, I. M., Williams, B. R., Silverrnan, R. H., and Schreiber, R. D.
(1998) How cells respond to interferons. Annu Rev Biochem 67, 227-264

Williams, B. R. (2001) Signal integation via PKR. Sci ST KE 2001, RE2

Barber, G. N. (2005) The dsRNA-dependent protein kinase, PKR and cell death.
Cell Death Difler 12, 563-570

Gao, Z., Hwang, D., Bataille, F., Lefevre, M., York, D., Quon, M. J., and Ye, J.
(2002) Serine phosphorylation of insulin receptor substrate 1 by inhibitor kappa B
kinase complex. J Biol Chem 277, 48115-48121

Ozes, O. N., Akca, H., Mayo, L. D., Gustin, J. A., Maeharna, T., Dixon, J. E., and
Donner, D. B. (2001) A phosphatidylinositol 3-kinase/Akt/mTOR patlnway

172

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

mediates and PTEN antagonizes tumor necrosis factor inhibition of insulin
signaling through insulin receptor substrate-1. Proc Natl Acad Sci U S A 98, 4640-
4645

Ravichandran, L. V., Esposito, D. L., Chen, J., and Quon, M. J. (2001) Protein
kinase C-zeta phosphorylates irnsulin receptor substrate-1 and impairs its ability to
activate phosphatidylinositol 3-kinase in response to insulin. J Biol Chem 276,
3543-3 549

Um, S. H., Frigerio, F., Watanabe, M., Picard, F., Joaquin, M., Sticker, M.,
Fumagalli, S., Allegini, P. R., Kozrna, S. C., Auwerx, J., and Thomas, G. (2004)
Absence of S6Kl protects against age- and diet-induced obesity while enhancing
insulin sensitivity. Nature 431, 200-205

Aguirre, V., Uchida, T., Yenush, L., Davis, R., and White, M. F. (2000) The c-Jun
NH(2)-terminal kinase promotes insulin resistance druing association with insulin
receptor substrate-l and phosphorylation of Ser(3 07). J Biol Chem 275, 9047-9054

Lee, Y. H., Giraud, J., Davis, R. J., and White, M. F. (2003) c-Jun N-terminal
kinase (JNK) mediates feedback inhibition of the insulin signaling cascade. J Biol
Chem 278, 2896-2902

Zick, Y. (2004) Uncoupling insulin signalling by serine/threonine phosphorylation:
a molecular basis for insulin resistance. Biochem Soc Trans 32, 812-816

Bonnet, M. C., Wei], R., Dam, B., Hovanessian, A. G., and Meurs, E. F. (2000)
PKR stimulates NF -kappaB irrespective of its kinase function by interacting with
the IkappaB kinase complex. Mol Cell Biol 20, 4532-4542

Yang, X., and Chan, C. (2009) Repression of PKR mediates palmitate-induced
apoptosis in HepG2 cells through regulation of Bcl-2. Cell Res

Zhou, H. R., Lau, A. S., and Pestka, J. J. (2003) Role of double-stranded RNA-
activated protein kinase R (PKR) in deoxynivalenol-induced ribotoxic stress
response. T oxicol Sci 74, 335-344

Ide, T., Shirnano, H., Yahagi, N., Matsuzaka, T., Nakakuki, M., Yamamoto, T.,
Nakagawa, Y., Takahashi, A., Suzuki, H., Sone, H., Toyoshima, H., Fukarnizu, A.,
and Yarnada, N. (2004) SREBPs suppress IRS-2-mediated insulin signalling in the
liver. Nat Cell Biol 6, 351-357

Zhang, J ., Ou, J ., Bashmakov, Y., Horton, J. D., Brown, M. S., and Goldstein, J. L.
(2001) Insulin inhibits transcription of IRS-2 gene in rat liver through an insulin

173

45.

46.

47.

48.

49.

50.

51.

52.

response element (IRE) that resembles IRES of other insulin-repressed genes. Proc
Natl Acad Sci U S A 98, 3756-3761

Jhala, U. S., Canettieri, G., Screaton, R. A., Kulkarni, R. N., Krajewski, S., Reed, J .,
Walker, J ., Lin, X., Wlnite, M., and Montrniny, M. (2003) cAMP promotes
pancreatic beta-cell survival via CREB-mediated induction of IRS2. Genes Dev 17,
1575-1580

Eley, H. L., Russell, S. T., and Tisdale, M. J. (2008) Role of the dsRNA-dependent
protein kinase (PKR) in the attenuation of protein loss from muscle by insulin and
insulin-like gowth factor-I (IGF-I). Mol Cell Biochem

Srivastava, S., and Chan, C. (2006) Hydrogen peroxide and hydroxyl radicals
mediate palmitate-induced cytotoxicity to hepatoma cells: Relation to
mitochondrial permeability transition. Free Radic Res 41, 38-49

Gual, P., Gonzalez, T., Gremeaux, T., Barres, R., Le Marchand-Brustel, Y., and
Tanti, J. F. (2003) Hyperosmotic stress inhibits insulin receptor substrate-l function
by distinct mechanisms in 3T3-L1 adipocytes. J Biol Chem 278, 26550-26557

Mitsuhashi, S., Shima, H., Tanuma, N., Matsuura, N., Takekawa, M., Urano, T.,
Kataoka, T., Ubukata, M., and Kikuchi, K. (2003) Usage of tautomycetin, a novel
inhibitor of protein phosphatase 1 (PP1), reveals that PP1 is a positive regulator of
Raf-l in vivo. J Biol Chem 278, 82-88

Shim, J. H., Lee, H. K., Chang, E. J., Chae, W. J., Han, J. H., Han, D. J ., Morio, T.,
Yang, J. J ., Bothwell, A., and Lee, S. K. (2002) Irnmunosuppressive effects of

tautomycetin in vivo and in vitro via T cell-specific apoptosis induction. Proc Natl
Acad Sci USA 99,10617-10622

Kishore, N., Sommers, C., Matlnialagarn, S., Guzova, J., Yao, M., Hauser, S., Huynh,

K., Bonar, S., Mielke, C., Albee, L., Weier, R., Graneto, M., Hanau, C., Perry, T.,
and Tripp, C. S. (2003) A selective H(K-2 inhibitor blocks NF-kappa B-dependent
gene expression in interleukin-1 beta-stimulated synovial fibroblasts. J Biol Chem
278, 32861-32871

Muranyi, A., Gergely, P., Nagy, G. M., and Fekete, M. I. (1997) The possible role
of protein phosphatase 2A in the sodium sensitivity of the receptor binding of
opiate antagonists naloxone and naltrindole. Brain Res Bull 44, 273-279

Kumar, A., Yang, Y. L., Flati, V., Der, S., Kadereit, S., Deb, A., Haque, J ., Reis, L.,
Weissmann, C., and Williams, B. R. (1997) Deficient cytokine signaling in mouse
embryo fibroblasts with a targeted deletion in the PKR gene: role of IRF-1 and NF-
kappaB. EMBO J 16, 406-416

174

 

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

Cuddihy, A. R., Li, S., Tam, N. W., Wong, A. H., Taya, Y., Abraham, N., Bell, J.
C., and Koromilas, A. E. (1999) Double-stranded-RNA-activated protein kinase
PKR enhances transcriptional activation by tumor suppressor p53. Mol Cell Biol 19,
2475-2484

Rena, G., Woods, Y. L., Prescott, A. R., Peggie, M., Unterman, T. G., Williams, M.
R., and Cohen, P. (2002) Two novel phosphorylation sites on FKHR that are
critical for its nuclear exclusion. EMBO J21, 2263-2271

Rena, G., Prescott, A. R., Guo, S., Cohen, P., and Unterman, T. G. (2001) Roles of
the forkhead in rhabdomyosarcoma (F KHR) phosphorylation sites in regulating 14-
3-3 binding, transactivation and nuclear targetting. Biochem J 354, 605-612

Xu, Z., and Williams, B. R. (2000) The B56alpha regulatory subunit of protein
phosphatase 2A is a target for regulation by double-stranded RN A-dependent
protein kinase PKR. Mol Cell Biol 20, 5285-5299

Yan, L., Lavin, V. A., Moser, L. R., Cui, Q., Kanies, C., and Yang, E. (2008) PP2A
regulates the pro-apoptotic activity of FOXO]. J Biol Chem 283, 7411-7420

Browne, G. J ., Delibegovic, M., Keppens, S., Stalmans, W., and Cohen, P. T. (2001)

The level of the glycogen targetting regulatory subunit R5 of protein phosphatase 1
is decreased in the livers of insulin-dependent diabetic rats and starved rats.
Biochem J 360, 449-459

Doherty, M. J., Cadefau, J., Stalmans, W., Bollen, M., and Cohen, P. T. (1998)
Loss of the hepatic glycogen-binding subunit (GL) of protein phosphatase 1
underlies deficient glycogen synthesis in insulin-dependent diabetic rats and in
adrenalectonnized starved rats. Biochem J 333 ( Pt 2), 253-257

Cohen, P. T. (2002) Protein phosphatase 1--targeted in many directions. J Cell Sci
115, 241-256

Bollen, M., Keppens, S., and Stalmans, W. (1998) Specific features of glycogen
metabolism in the liver. Biochem J 336 ( Pt 1), 19-31

Wada, T., Sasaoka, T., Funaki, M., Hori, H., Murakarni, S., Ishiki, M., Haruta, T.,
Asano, T., Ogawa, W., Ishihara, H., and Kobayashi, M. (2001) Overexpression of
SH2-containing inositol phosphatase 2 results in negative regulation of insulin-
induced metabolic actions in 3T3-L1 adipocytes via its 5'-phosphatase catalytic
activity. Mol Cell Biol 21, l 63 3-1 646

175

63.

65.

66.

67.

68.

69.

70.

71.

72.

73.

Tan, S. L., Tareen, S. U., Melville, M. W., Blakely, C. M., and Katze, M. G. (2002)
The direct binding of the catalytic subunit of protein phosphatase 1 to the PKR
protein kinase is necessary but not sufficient for inactivation and disruption of
enzyme dimer formation. J Biol Chem 277, 36109-36117

Dohadwala, M., da Cruz e Silva, E. R, Hall, F. L., Williams, R. T., Carbonaro-Hall,
D. A., Nairrn, A. C., Greengard, P., and Berndt, N. (1994) Phosphorylation and

inactivation of protein phosphatase 1 by cyclin-dependent kinases. Proc Natl Acad
Sci USA 91, 6408-6412

Bonnet, M. C., Daurat, C., Ottone, C., and Meurs, E. F. (2006) The N-terminus of
PKR is responsible for the activation of the NF-kappaB signaling pathway by
interacting with the IKK complex. Cell Signal 18, 1865-1875

Rui, L., Aguirre, V., Kim, J. K., Shulmarn, G. 1., Lee, A., Corbould, A., Dunaif, A.,
and White, M. F. (2001) Insulin/IGF-l and TNF-alpha stimulate phosphorylation of
IRS-l at inhibitory Ser307 via distinct patlnways. J Clin Invest 107, 181-189

Fujishiro, M., Gotoh, Y., Katagiri, H., Sakoda, H., Ogihara, T., Anai, M., Onishi,
Y., Ono, H., Abe, M., Shojima, N., Fukushima, Y., Kikuchi, M., Oka, Y., and
Asano, T. (2003) Three mitogen-activated protein kinases inhibit insulin signaling
by different mechanisms in 3T3-L1 adipocytes. Mol Endocrinol 17, 487-497

Takizawa, T., Tatematsu, C., and Nakanishi, Y. (2002) Double-stranded RNA-
activated protein kinase interacts with apoptosis signal-regulating kinase 1.
Implications for apoptosis signaling pathways. Eur J Biochem 269, 6126-6132

Ichijo, H., Nishida, E., Irie, K., ten Dijke, P., Saitoh, M., Moriguchi, T., Takagi, M.,
Matsumoto, K., Miyazono, K., and Gotoh, Y. (1997) Induction of apoptosis by
ASK], a mammalian MAPKKK that activates SAPK/JNK and p38 signaling
pathways. Science 275, 90-94

Chang, L., and Karin, M. (2001) Mammalian MAP kinase signalling cascades.
Nature 410, 37-40

Buteau, J ., and Accili, D. (2007) Regulation of pancreatic beta-cell function by the
forkhead protein FoxOl. Diabetes Obes Metab 9 Supp12, 140-146

Huang, H., and Tindall, D. J. (2007) Dynamic F oxO transcription factors. J Cell Sci
120, 2479-2487

Barthel, A., Schmoll, D., and Unterman, T. G. (2005) FoxO proteins in insulin
action and metabolism. Trends Endocrinol Metab 16, 183-189

176

 

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

Gross, D. N., van den Heuvel, A. P., and Birnbaum, M. J. (2008) The role of FoxO
in the regulation of metabolism. Oncogene 27, 2320-2336

Puigserver, P., Rhee, J., Donovan, J., Walkey, C. J., Yoon, J. C., Oriente, F.,
Kitarnura, Y., Altomonte, J ., Dong, H., Accili, D., and Spiegehnan, B. M. (2003)
Insulin-regulated hepatic gluconeogenesis through FOXO] -PGC-1 alpha interaction.
Nature 423, 550-555

Samuel, V. T., Choi, C. S., Phillips, T. G., Romanelli, A. J., Geisler, J. G., Bhanot,
S., McKay, R., Monia, B., Shutter, J. R., Lindberg, R. A., Shulrnarn, G. 1., and
Veniant, M. M. (2006) Targeting foxol in mice using antisense oligonucleotide
improves hepatic and peripheral insulin action. Diabetes 55, 2042-2050

Dong, X. C., Copps, K D., Guo, S., Li, Y., Kollipara, R., DePinho, R. A., and
White, M. F. (2008) Inactivation of hepatic Foxol by insulin signaling is required
for adaptive nutrient homeostasis and endocrine gowth regulation. Cell Metab 8,
65-76

Tarnemoto, H., Kadowaki, T., Tobe, K., Yagi, T., Sakura, H., Hayakawa, T.,
Terauchi, Y., Ueki, K., Kaburagi, Y., Satoh, S., and et al. (1994) Insulin resistance
and gowth retardation in rrnice lacking insulin receptor substrate-l. Nature 372,
182-186

Araki, B., Lipes, M. A., Patti, M. B., Bruning, J. C., Haag, B., 3rd, Johnson, R. S.,
and Kahn, C. R. (1994) Alternative pathway of insulin signalling in mice with
targeted disruption of the IRS-1 gene. Nature 372, 186-190

Kido, Y., Burks, D. J ., Withers, D., Bruning, J. C., Kahn, C. R., White, M. F., and
Accili, D. (2000) Tissue-specific insulin resistance in mice with mutations in the
insulin receptor, IRS-1, and IRS-2. J Clin Invest 105, 199-205

Withers, D. J., Gutierrez, J. S., Towery, H., Burks, D. J., Ren, J. M., Previs, S.,
Zhang, Y., Bemal, D., Pons, S., Shulman, G. I., Bonner-Weir, S., and White, M. F.
(1998) Disruption of IRS-2 causes type 2 diabetes in mice. Nature 391, 900-904

Tarniguchi, C. M., Ueki, K., and Kahn, R. (2005) Complementary roles of IRS-1
and IRS-2 in the hepatic regulation of metabolism. J Clin Invest 115, 718-727

Kubota, N., Kubota, T., Itoh, S., Kumagai, H., Kozono, H., Takarnoto, 1.,
Mineyama, T., Ogata, H., Tokuyama, K., Ohsugi, M., Sasako, T., Moroi, M., Sugi,
K., Kakuta, S., Iwakura, Y., Noda, T., Ohnishi, S., Nagai, R., Tobe, K., Terauchi,
Y., Ueki, K., and Kadowaki, T. (2008) Dynamic functional relay between insulin
receptor substrate 1 and 2 in hepatic insulin signaling during fasting and feeding.
Cell Metab 8, 49-64

177

84.

85.

86.

87.

Opara, E., Kutlu, M., van Haeften, T., Kennedy, F. P., and Gerich, J. E. (1988)
Effect of internalization and degadation of insulin on rat adipocyte insulin receptor
binding kinetics. Clin Physiol Biochem 6, 1-11

Kusari, A. B., Byon, J ., Bandyopadhyay, D., Kenner, K. A., and Kusari, J. (1997)
Insulin-induced rrnitogen-activated protein (MAP) kinase phosphatase-1 (MKP-l)
attenuates insulin-stimulated MAP kinase activity: a mechanism for the feedback
inhibition of insulin signaling. Mol Endocrinol 11, 1532-1543

Tremblay, F., and Marette, A. (2001) Arrnino acid and insulin signaling via the
mTOR/p70 S6 kinase pathway. A negative feedback mechanism leading to insulin
resistance in skeletal muscle cells. J Biol Chem 276, 38052-3 8060

Martin, K. A., Merenick, B. L., Ding, M., Fetalvero, K. M., Rzucidlo, E. M., Kozul,
C. D., Brown, D. J., Chiu, H. Y., Shyu, M., Drapeau, B. L., Wagner, R. J., and
Powell, R. J. (2007) Raparnycin promotes vascular smooth muscle cell
differentiation through insulin receptor substrate-l/phosphatidylinositol 3-
kinase/Akt2 feedback signaling. J Biol Chem 282, 36112-36120

178

CHAPTER 6: Conclusion and Discussion

In this thesis, my research focused on the major hepatic cellular activities,
including insulin signaling activity and the saturated FFA-induced cytotoxicity and
apoptosis. Human hepatoblastoma cells (HepG2/C3A) were used for the study. These
human-ori gin cells offer the advantages of ease of culture and experimentation, and have
been shown to retain many hepatospecific functions and tlnerefore suggested as a good
model cell system for hepatocellular functions, such as lipid metabolism (1) and fatty
acid transport (2). Cell lines offer an advantage over primary hepatocytes in that they are
more amenable to genetic manipulation.

In the saturated F FA-induced cytotoxicity and apoptosis studies, two different
types of F FAs were employed, i.e. saturated and monounsaturated fatty acids,
corresponding to the major dietary fractions. Palmitate was chosen as the representative
saturated fatty acid and oleate as the monounsaturated fatty acid. They are the major fatty
acids of their classes found in serum/plasma. While the total concentration of FFAs in
plasma may reach rrnillirnolar range under pathological conditions (3), most studies of
obese/type 2 diabetic patients have reported fatty acid concentration at about 0.7 mM. We
therefore used 0.7 mM or less as the concentrations of palmitate and oleate.

Using the research techniques from molecular and cellular biology, a key protein,
PKR, was identified to be highly involved in both palmitate-induced apoptosis and the
regulation of insulin signaling.

PKR in Palmitate-Induced Apoptosis—Multiple mecharnisms have been

proposed for palmitate-induced cytotoxicity and apoptosis. Chapter 2 of this thesis

179

discusses how PKR, found to be anti-apoptotic in HepG2 cells, plays a central role in
mediating the palrrnitate-initiated pathways that led to hepatic apoptosis. PKR is best
krnown for its pro-apoptotic role by phosphorylating eIF-20. and thereby inhibiting general
protein synthesis (4, 5). In contrast, more recent studies suggest that PKR has an anti-
apoptotic role in certain conditions (6-9). Elevated levels of PKR protein and activity
were observed in human breast cancer cells (6), melanoma cells (7, 8), and hepatitis C
virus (HCV)-related hepatocellular carcinoma (9). However, the mechanism is still
unclear. Research suggests that PKR may suppress apoptosis by activating the NF-ch
pathway before phosphorylating eIF-20t, tlnereby initially inducing cell survival and,
subsequently, cell death in NIH3T3 cells expressing PKR (10). Chapter 2 discusses the
two potential pathways by which PKR mediates the anti-apoptotic signals through Bcl-2.
These results provide a potential mechanism of palrrnitate-induced apoptosis, by
suppressing PKR, in HepG2 cells.

In the first pathway, PKR is involved in controlling the transcription of Bel-2 in
HepG2 cells, mediated by the transcription factor NF-ch. As discussed in Chapter 2, NF-
KB is commonly considered an anti-apoptotic transcription factor (11, 12), which is
consistent with the anti-apoptotic role of PKR in HepG2 cells. However the second anti-
apoptotic patlnway, namely, that PKR up-regulates the phosphorylation of Bcl-2 at Ser70,
is mediated by JNK, a multi-functional kinase considered a “double-edged sword” (13) in
regulating apoptosis. Most of the studies on JNK used a specific chemical inhibitor,
SP600125. As discussed in Chapter 2, the concentrations or treatment lengths of this
inhibitor in palnnitate-related studies play an unclear but important role in determirning the

apoptosis level of the cells, thereby raising concerns of using this inhibitor in the studies

180

 

of the palmitate-induced apoptosis. Furthermore, this particular JNK inhibitor was found
to ineffectively inhibit the activity of JNK in the palnnitate media. Since SP600125 is
hydrophobic, it is possible that the chemical inhibitor may aggegate with palrrnitate,
which is also hydrophobic, or even BSA, the vehicle used to dissolve palmitate, thereby
losing its inhibitory function on JNK. In Liedtke et al. (14), an adenol-viral vector
expressing dominant-negative TAK], which was identified to specifically inhibit JNK
activity, was shown to induce strong apoptosis in Huh7 and HepG2 cells, supporting the
result that JNK is anti-apoptotic in HepG2 cells.

The results in Chapter 2 show for the first time that palmitate treatment decreases
the activity of PKR; however, the mechanism by which palnnitate induces
dephosphorylation of PKR is still unclear. The auto-phosphorylation of PKR requires its
dimerization (15, 16) and binding of ATP to its ATP binding site on the catalytic domain
(17). Another PhD student in our lab is currently investigating the potential interaction
between the free palmitate molecules and the ATP binding site of PKR. Interestingly,
both computational and experimental results indicate that palmitate binds to the ATP
binding site on PKR and thereby disrupts the auto-phosphorylation of PKR.

In summary, an anti-apoptotic role of PKR, mediated by the protein expression
level and phosphorylation status of Bel-2, was identified in HepG2 cells. The
transcription factor NF-ch and the MAP kinase JNK appear to be involved in mediating
the effects of PKR on the protein level and the phosphorylation of Bcl-2, respectively.
Furthermore, palrrnitate was found to suppress these two pathways by inhibiting PKR and

tlnereby attenuate the anti-apoptotic machinery of HepG2 cells.

181

 

 

PKR in Regulating Insulin Signaling—As a serine-threonine protein kinase that
is ubiquitously expressed, PKR is actively involved in multiple signaling pathways by
regulating the activities of transcription factors, protein kinases, and phosphatases (] 8).
Therefore, PKR serves as a signal integator, regulating various cellular activities, such as
protein synthesis and degadation, stress and immune responses, and apoptosis (18, 19).
In Chapter 5 of this thesis, a novel function of PKR in regulating insulin signaling
pathway was identified for HepG2 cells. In detail, PKR differentially regulates the two
major IRS proteins, IRS] and IRSZ, through post-translational modification and
transcriptional regulation, respectively. The involvement of PKR in regulating insulin
signaling may connect the effectors of PKR with insulin signaling. For example,
according to the results in Chapter 5, the inhibitory serine phosphorylation of IRS]
induced by ceramide is attributed to the activation of PKR by cerannide. In addition to
ceramide, PKR is also responsive to many other factors, e.g., HCV infection (20),
endoplasmic reticulum (ER) stress (2] , 22), cytokines such as tumor necrosis factor
(TNF)-01 (23) and interleukin (IL)-1 (24), deoxyrnivalenol (DON, or vorrnitoxin) (25) and
lipopolysaccharide (24). Whether and how PKR is involved in regulating insulin
signaling in response to these factors is worth further investigation.

The function of PKR in regulating the protein level of IRS2 involves the
transcription factor Fox01. Fox01 directs the expression of genes involved in a wide
variety of cellular responses. The gene expression progarn regulated by Fox01 is
considered to be cell-protective and anti-apoptotic (26). For example, Fox01 facilitate
DNA repair by up-regulating GADD45 and DDBl and suppresses cell-cycle by

increasing expression of p27'dp, thereby protecting cells from DNA damage (26). F 0x01

182

also suppresses the rate of aging by up-regulating superoxide dismutase (27), which
repairs oxidative damage (27). The effect of PKR on Fox01 has not been studied
previously. Chapter 5 shows that PKR dephosphorylates and activates FoxOl mediated
by the protein phosphatase PP2A. PKR is best known for its pro-apoptotic role by
phosphorylating eIF-2a and thereby inhibiting general protein synthesis (4, 5). In contrast,
more recent studies including the work in Chapter 2 suggest that PKR has an anti-
ap0ptotic role (8, 28). Due to the cytoprotective and anti-apoptotic roles of Fox01, the
connection identified between PKR and Fox01 also supports the anti-apoptotic function
of PKR, thereby suggesting another interesting research target for PKR-regulated
apoptosis.

In addition to the regulatory function of PKR on insulin signaling, PKR was also
identified as a potential downstream substrate of the insulin signaling pathway. Irnitially
known as a virus responsive protein, PKR has been shown to be involved in regulating a
variety of important cellular activities, such as immune response, general protein
syntlnesis, protein degadation, apoptosis, cell gowtln and proliferation (29), and therefore
associated with diseases including virus infection, e.g., HCV infection (30), cancer (31),
neuron degenerative diseases, e.g., Alzheimer's (32), etc. Therefore, by revealing the
involvement of PKR in insulin signaling pathways, I propose that PKR has a central
position that could correlate insulin (or IGF-l) signaling with other cellular activities,
such as immune response, apoptosis, protein syntlnesis and degadation, some of which
have been already identified to be related to insulin action (33-37). The results in Chapter
5 therefore suggest a promising research direction to identify potential interactions

between insulin signaling and these PKR-related cellular activities.

183

Chapter 5 also showed the abolition of the down-regulation in PKR
phosphorylation by insulin upon longer treatment with insulin (Fig. 5.6B), suggesting a
possible negative feedback in the regulation of PKR by insulin. Indeed, insulin signaling
is tuned by various negative feedbacks, e.g., degadation of IR induced by insulin binding
(38), negative regulation of insulin signaling activity by MAPKs (39), mTOR (40), PI-3
kinase (41), etc. These negative feedbacks modulate the downstream signaling events
induced by long-term treatment of insulin (42). Given that many signaling components of
the feedback pathways in the insulin signaling system have the potential to cross-talk
with PKR, the contribution of each to the dynamic profile of PKR in the insulin network
system is unclear. This complex feedback system represents a good target for
mathematical modeling. Indeed, there have been many mathematical models proposed to
model insulin action, most of which, however, focus on the binding kinetics of insulin
with its receptor, IR (43, 44), and GLUT4 trafficking (45, 46). There are very few models
that describe the kinetics of insulin signaling, incorporating only a limited number of
feedback pathways (47, 48). This is due to the difficulty in quantitatively measuring all
the biochemical and kinetic parameters of the signaling pathways. Thus, an alterrnative
modeling method is needed, which infers the network dynamics in the absence of the
detail kinetic parameters. Currently a Boolean network model is beirng built in our goup
to simulate the insulin signaling network. It is a discrete logic model that solely relies on
the network structure. Preliminary results show that this model, in spite of its simplicity,
is able to reproduce some of the key dynamic features of the insulin signaling system,
thereby suggesting promising applications of this methodology in providing valuable

insights into the execution and regulation of insulin signaling.

184

NAFLD, NASH, and Insulin Resistance—The hepatic cellular activities that are
studied in this thesis, including insulin signaling and saturated FFA-induced cytotoxicity
and apoptosis, are directly related to the initiation and development of the liver disorders
such as NAFLD, NASH, and hepatic insulin resistance. Insulin resistance is very
common in patients who have NAFLD (49), and is believed to play an important role in
the pathogenesis of NAFLD by impairing mitochondrial [ii-oxidation of FFAs and
promoting fat accumulation in liver (50). Indeed, obesity and type 2 diabetes, in which
insulin resistance is usually prevalent, are the two most common risk factors for
developing NAFLD.

Although the development of NASH, the more severe, second stage of NAFLD,
has not been fully understood, several models have been proposed for the pathogenesis of
NASH from NAFLD (51). The prevailing theory is the two-lnit model (52), with insulin
resistance accounting for the first hit, inducing fatty liver, and FFA-induced cytotoxicity
accounting for the second hit, inducing inflammation and cell death (52). Therefore, it is
well-recognized that in the liver, insulin resistance is the cause of fat accumulation,
NAFLD and eventually NASH, rather than the consequence of NAF LD.

In this thesis, a regulatory role of PKR in irnsulin signaling was identified (Chapter
5). Furtlnermore, saturated FFA, i.e., palrrnitate, decreased the phosphorylation of PKR
(Chapter 2). Taken together, these findings raise an interesting possibility for further
study. Namely, can palmitate regulate insulin signaling through PKR? Palmitate has been
identified as an inducer of insulin resistance at the whole-body level (53). In hepatocytes,
palmitate inhibits insulin action by both receptor and post-receptor events (54, 55).

However, it has been recently shown that irnlnibiting insulin signaling by palnritate in

185

hepatoma cells is dependent upon oxidation of fatty acyl-CoA species and requires intact
insulin receptor expression (55). Therefore, the palmitate-induced insulin resistance is
likely due to the combination of different events, in which the down-regulation of IR
expression may play a major role, although the exact mechanism is still not clear. It may
be worthwhile to evaluate the potential involvement of PKR in mediating the induction of
hepatic insulin resistance by palrrnitate.

In summary, a part of the research in this thesis focused on the regulation of
insulin signaling and potentially induction of insulin resistance, which play roles in the
development of type 2 diabetes, as well as fat accrrrnulation in the liver, which is a big
risk factor of NAFLD. The other half of the research was on saturated FFA-induced
cytotoxicity and apoptosis, which are known to contribute to the development of NASH
in NAFLD patients (56, 57). Therefore, the detailed mechanistic information involved in
the regulation of insulin signaling and palmitate-induced cytotoxicity would enhance our
understanding of the pathology of liver-related diseases such as NAFLD, NASH, hepatic
insulin resistance, and type 2 diabetes, and may further identify potential research and
drug targets.

Research Tools in Systems Biology—In addition to the molecular biology studies
on the saturated FFA-induced apoptosis and cytotoxicity, the principles of systems
biology was also applied in this thesis. In collaboration with mathematicians and
computational experts, systems biology methodologies were applied to gain a systems-
level view on lipotoxicity. Phenotype-specific gene networks were reconstructed and
analyzed. The rationale for applying systems biology in studying the cellular activities is

that reconstructing gene networks fiom the gene- gene interactions, which could be

186

extracted fi'om gene expression profiles, could shed light into the possible mechanism
involved in context-specific cellular activities. The assumption behind this is that only a
subset of the genes are responsible for a certain phenotype and the mechanistic
information is enriched in the phenotype-specific gene network. Two strategies (Chapters
3 and 4) were developed for reconstruction of the phenotype-specific gene network. Both
strategies consist of two phases, 1) selection of the phenotype-related genes and 2)
reconstruction of the gene network that is specific to the phenotype. For each phase, we
applied two different approaches to achieve our goal.

For the selection of phenotype-related genes, it is well recognized that regular
statistical feature selection methods are usually computationally expensive and
susceptible to the quality of data (58), thereby requiring integation of additional
information. In the two methods discussed in Chapters 3 and 4, the information
incorporated to facilitate the gene-selection process was obtained from two different
sources, metabolic data (Chapter 3) and prior knnowledge (Chapter 4). Botln approaches
are adaptable to different phenotypes. However, while the first approach integates
metabolic information, which may be more biologically meaningful since it reflects the
real biological process that induces the phenotype, it is more labor-intensive in obtaining
the metabolic information. On the other hand, the second approach, which incorporates
prior knowledge instead of experimental data, provides valuable information in selecting
the genes, but at the same time, the results of this approach largely depend on the source
and quality of the prior knowledge. In addition, some prior knowledge may not be valid

under all biological systems and experimental conditions. Therefore, selecting the source

187

of prior krnowledge and optimizing the weight of the prior knowledge in the gene-
selection process are critical to this method.

Similarly, we viewed the reconstruction of the phenotype-specific gene network
from two different standpoints gounded in different understandings of the biological
system. In Chapter 3, the gene network was composed of gene pairs with high synergy
scores. This network therefore captured genes that interacted synergistically
(cooperatively or antagonistically) on the phenotype. Since the concept of synergy is
intuitively based upon the phenotype, we obtained a single phenotype-specific network
from the gene expression profiles of the different conditions. Therefore, this method does
not require the comparison of different networks across the experimental conditions,
thereby eliminating the noise due to the artificial differences, such as the sample size and
noise level differences, between the experimental conditions. However, this method was
not designed to recover the direct physical interactions between genes and proteins,
thereby limiting the direct experimental validation of synergistic gene pairs in the
synergy network. Nevertheless, this method was able to uncover some novel mecharnistic
information that was not revealed by other methods.

In Chapter 4, gene modules and module interactions, instead of pair-wised gene
interactions, were identified. This method takes advantage of the prior knowledge and
can tune the amount of weight or value to place on the prior knowledge in obtaining
optimal results. Although this module network is not capable of uncovering specific
gene- gene interactions, it successfully recovered functional gene modules, in which genes
involved in certain cellular functions or activities were significantly enriched. Therefore,

this module-interaction network captured the organized modularity in biological network

188

systems. However, this method requires building and comparing the networks under
different conditions, making it susceptible to sampling differences across experimental
conditions. Therefore, a proper normalization process is needed to reduce artifactual
noise.

Thus, both system biology methodologies described in this thesis were able to
capture a systems-level view of the phenotype of interest and provided valuable insiglnts
into potential mechanisms involved in the induction of the different phenotypes. Both
methods suggested some similar findings, which was encouraging. These findings
include the potential involvement in palmitate-induced cytotoxicity of endoplasmic
reticulum (ER) stress, and intracellular metabolism regulation, etc., some of which are
supported by the literature (59-61) and experiments. Interestingly, the gene module
approach also suggested that PKR is involved in regulating apoptosis that contributes to
palmitate-induced cytotoxicity. This finding is consistent with the molecular biology

study of PKR in palnnitate-induced apoptosis (Chapter 2).

Summary

In this thesis, I investigated the hepatocellular activities, such as regulation of
insulin signaling and saturated FFA-induced cytotoxicity, which are krnown to play
central roles in the initiation and development of multiple liver disorders including
NAFLD, NASH, and hepatic insulin resistance. Using molecular and cellular biology
techniques, I identified that 1), PKR mediates the palmitate-induced apoptosis by
regulating Bcl-2, and 2), PKR, affected by insulin, is involved in insulin signaling by

differentially regulating the major IRS proteins. In general, PKR exploits a role as a

189

signaling hub that regulates multiple hepatocellular activities through different signaling
events.

In addition, in collaboration with other researchers, I also applied systems biology
methodologies to obtain furtlner understandings into saturated FFA-induced lipotoxicity.
Both strategies, based on different aspects of the gene network systems, provided
valuable mechanistic information on the phenotype of interest, thereby serving as
powerful research tools that complement the molecular biology study, and suggested

novel research targets and possible mechanisms.

190

10.

References

Cianflone, K. ,Vu, H. ,Zhang, Z, and Sniderrnan, A. D. (1994) Effects of albumin
on lipid synthesis, apo B- 100 secretion, and LDL catabolism in HepG2 cells.
Atherosclerosis 107, 125-135

Guo, W., Huang, N., Cai, J., Xie, W., and Hamilton, J. A. (2006) Fatty acid
transport and metabolism in HepG2 cells. Am J Physiol Gastrointest Liver Physiol
290, G528-534

Artwohl, M., Roden, M., Waldhausl, W., Freudenthaler, A., and Baumgartrner-
Parzer, S. M. (2004) Free fatty acids trigger apoptosis and inhibit cell cycle
progession in human vascular endothelial cells. FASEBJ. 18, 146-148

Taylor, S. S., Haste, N. M., and Ghosh, G. (2005) PKR and eIF2alpha: integation
of kinase dimerization, activation, and substrate docking. Cell 122, 823-825

Stark, G. R., Kerr, 1. M., Williams, B. R., Silvermarn, R. H., and Schreiber, R. D.
(1998) How cells respond to interferons. Annu Rev Biochem 67, 227-264

Kim, S. H., Forman, A. P., Mathews, M. B., and Gunnery, S. (2000) Human breast
cancer cells contain elevated levels and activity of the protein kinase, PKR.
Oncogene 19, 3086-3094

Kim, S. H., Gunnery, S., Choc, J. K., and Mathews, M. B. (2002) Neoplastic
progession in melanoma and colon cancer is associated with increased expression
and activity of the interferon-inducible protein kinase, PKR. Oncogene 21, 8741-
8748

Delgado Andre, N., and De Lucca, F. L. (2007) Knockdown of PKR expression by
RNAi reduces pulmonary metastatic potential of Bl6-F10 melanoma cells in rrnice:
Possible role of NF-kappaB. Cancer Lett 258, 118-125

Hiasa, Y., Karnegaya, Y., Nuriya, H., Onji, M., Kohara, M., Schmidt, E. V., and
Chung, R. T. (2003) Protein kinase R is increased and is firnctional in hepatitis C
virus-related hepatocellular carcinoma. Am J Gastroenterol 98, 2528-2534

Donze, O., Deng, J ., Curran, J ., Sladek, R., Picard, D., and Sonenberg, N. (2004)
The protein kinase PKR: a molecular clock that sequentially activates survival and
death progarns. EMBO J 23, 564-571

191

ll.

12.

l3.

14.

15.

16.

l7.

18.

19.

20.

21.

22.

Kucharczak, J ., Simmons, M. J ., Fan, Y., and Gelinas, C. (2003) To be, or not to be:
NF-kappaB is the answer--role of Rel/NF-kappaB in the regulation of apoptosis.
Oncogene 22, 8961-8982

Li, M., Shillinglaw, W., Henzel, W. J., and Beg, A. A. (200]) The Rela(p65)
subunit of NF-kappaB is essential for irnhibiting double-stranded RNA-induced
cytotoxicity. J Biol Chem 276, 1185-] 194

Liu, J ., and Lin, A. (2005) Role of JNK activation in apoptosis: a double-edged
sword. Cell Res 15, 36-42

Liedtke, C., Plumpe, J ., Kubicka, S., Bradham, C. A., Manns, M. P., Brenner, D. A.,
and Trautwein, C. (2002) Jun kinase modulates tumor necrosis factor-dependent
apoptosis in liver cells. Hepatology 36, 315-325

Dey, M., Cao, C., Dar, A. C., Tamura, T., Ozato, K., Sicheri, F., and Dever, T. E.
(2005) Mechanistic link between PKR dimerization, autophosphorylation, and
eIF2alpha substrate recognition. Cell 122, 901-913

Lemaire, P. A., Lary, J., and Cole, J. L. (2005) Mechanism of PKR activation:
dimerization and kinase activation in the absence of double-stranded RNA. J Mol
Biol 345, 81-90

Dar, A. C., Dever, T. E., and Sicheri, F. (2005) Higher-order substrate recognition
of eIF2alpha by the RNA-dependent protein kinase PKR. Cell 122, 887-900

Williams, B. R. (2001) Signal integation via PKR. Sci STKE 2001, RE2

Barber, G. N. (2005) The dsRNA-dependent protein kinase, PKR and cell death.
Cell Death Difi‘er 12, 563-570

Yan, X. B., Battaglia, S., Boucreux, D., Chen, Z., Brechot, C., and Pavio, N. (2007)
Mapping of the interacting domains of hepatitis C virus core protein and the
double-stranded RNA-activated protein kinase PKR. Virus Res 125, 79-87

Onuki, R., Bando, Y., Suyama, E., Katayarna, T., Kawasaki, H., Baba, T., Tohyama,
M., and Taira, K. (2004) An RNA-dependent protein kinase is involved in
tunicamycin-induced apoptosis and Alzheimer's disease. EMBO J23, 959-968

Langland, J. 0., Cameron, J. M., Heck, M. C., Jancovich, J. K., and Jacobs, B. L.
(2006) Inhibition of PKR by RNA and DNA viruses. Virus Res 119, 100-110

192

rm

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

Meusel, T. R., Kehoe, K. B., and Imani, F. (2002) Protein kinase R regulates
double-stranded RNA induction of TNF-alpha but not IL-1 beta mRN A in human
epithelial cells. J Immunol 168, 6429-6435

Goh, K. C., deVeer, M. J ., and Williams, B. R. (2000) The protein kinase PKR is
required for p38 MAPK activation and the innate immune response to bacterial
endotoxin. EMBO J 19, 4292-4297

Zhou, H. R., Lau, A. S., and Pestka, J. J. (2003) Role of double-stranded RNA-
activated protein kinase R (PKR) in deoxynivalenol-induced ribotoxic stress
response. T oxicol Sci 74, 335-344

van der Horst, A., and Burgering, B. M. (2007) Stressing the role of FoxO proteins
in lifespan and disease. Nat Rev Mol Cell Biol 8, 440-450

Murphy, C. T., McCarroll, S. A., Bargnann, C. 1., Fraser, A., Kamatln, R. S.,
Ahringer, J., Li, H., and Kenyon, C. (2003) Genes that act downstream of DAF-l6
to influence the lifespan of Caenorhabditis elegans. Nature 424, 277-283

Yang, X., and Charn, C. (2009) Repression of PKR mediates palmitate-induced
apoptosis in HepG2 cells through regulation of Bcl-2. Cell Res

Williams, B. R. (1997) Role of the double-stranded RNA-activated protein kinase
(PKR) in cell regulation. Biochem Soc Trans 25, 509-513

Korth, M. J ., and Katze, M. G. (2000) Evading the interferon response: hepatitis C
virus and the interferon-induced protein kinase, PKR. Curr T op Microbiol Immunol
242, 197-224

Shir, A., Friedrich, I., and Levitzki, A. (2003) Tumor specific activation of PKR as
a non-toxic modality of cancer treatment. Semin Cancer Biol 13, 309-314

Peel, A. L. (2004) PKR activation in neurodegenerative disease. J Neuropathol Exp
Neurol 63, 97-105

Grinspoon, S. (2003) Mechanisms and strategies for insulin resistance in acquired
immune deficiency syndrome. Clin Infect Dis 37 Suppl 2, S85-90

Curi, R., Homem-de-Bittencourt, P. 1., Jr., Costa-Rosa, L. F., Femandes, L. C., el-
Razi, S., Seelaender, M. C., and Carpinelli, A. R. (1995) Insulin and prostaglandins
as signals between tumor and the host immune system. Braz J Med Biol Res 28,
773-779

193

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

Keku, T. 0., Lund, P. K., Galanko, J., Simmons, J. G., Woosley, J. T., and Sandler,
R. S. (2005) Insulin resistance, apoptosis, and colorectal adenoma risk. Cancer
Epidemiol Biomarkers Prev 14, 2076-2081

Bolster, D. R., Jefferson, L. S., and Kimball, S. R. (2004) Regulation of protein
synthesis associated with skeletal muscle hypertrophy by insulin-, amino acid- and
exercise-induced signalling. Proc Nutr Soc 63, 351-3 56

Eley, H. L., Russell, S. T., and Tisdale, M. J. (2008) Role of the dsRNA-dependent
protein kinase (PKR) in the attenuation of protein loss from muscle by insulin and
insulin-like gowtln factor-I (IGF-I). Mol Cell Biochem 313, 63-69

Opara, E., Kutlu, M., van Haeften, T., Kennedy, F. P., and Gerich, J. E. (1988)
Effect of internalization and degadation of insulin on rat adipocyte insulin receptor
binding kinetics. Clin Physiol Biochem 6, l-ll

Kusari, A. B., Byon, J., Bandyopadhyay, D., Kenner, K. A., and Kusari, J. (1997)
Insulin-induced nnitogen-activated protein (MAP) kinase phosphatase-1 (MKP-l)
attenuates insulin-stimulated MAP kinase activity: a mechanism for the feedback
inhibition of insulin signaling. Mol Endocrinol 11, 1532-1543

Tremblay, F ., and Marette, A. (2001) Amino acid and insulin signaling via the
mTOR/p70 S6 kinase pathway. A negative feedback mecharnism leading to insulin
resistance in skeletal muscle cells. J Biol Chem 276, 38052-3 8060

Martin, K. A., Merenick, B. L., Dirng, M., Fetalvero, K. M., Rzucidlo, E. M., Kozul,
C. D., Brown, D. J., Chiu, H. Y., Shyu, M., Drapeau, B. L., Wagner, R. J., and
Powell, R. J. (2007) Raparnycin promotes vascular smooth muscle cell
differentiation through insulin receptor substrate-1/phosphatidylinositol 3-
kinase/Akt2 feedback signaling. J Biol Chem 282, 36112-36120

Herschkovitz, A., Liu, Y. F., Ilan, B., Ronen, D., Boura-Halfon, S., and Zick, Y.
(2007) Common inhibitory serine sites phosphorylated by IRS-l kinases, triggered
by insulin and inducers of insulin resistance. J Biol Chem 282, 18018-18027

Wanant, S., and Quon, M. J. (2000) Insulin receptor binding kinetics: modeling and
simulation studies. J Theor Biol 205, 355-364

Quon, M. J ., and Carnpfield, L. A. (1991) A mathematical model and computer
simulation study of insulin receptor regulation. J Theor Biol 150, 59-72

194

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

Holman, G. D., Lo Leggio, L., and Orshman, S. W. (1994) Insulirn-stimulated
GLUT4 glucose transporter recycling. A problem in membrane protein subcellular
trafficking tlnrough multiple pools. J Biol Chem 269, 17516-17524

Quon, M. J. (1994) Advances in kinetic analysis of insulin-stimulated GLUT-4
translocation in adipose cells. Am J Physiol 266, El44-150

Sogard, P., Harlen, M., Svensson, L. T., Zieratln, J. R., and Nilsson, P. (2005)
Integation of mathematical and experimental approaches to resolve insulin
signalling. Acta Physiol Scand 183, 125-126

Sedaghat, A. R., Sherman, A., and Quon, M. J. (2002) A mathematical model of
metabolic insulin signaling pathways. Am J Physiol Endocrinol Metab 283, E1084-
1 101

Hamaguchi, M., Kojima, T., Takeda, N., Nakagawa, T., Tarriguchi, H., Fujii, K.,
Omatsu, T., Nakajima, T., Sarui, H., Shimazaki, M., Kato, T., Okuda, J ., and Ida, K.
(2005) The metabolic syndrome as a predictor of nonalcoholic fatty liver disease.
Ann Intern Med 143, 722-728

Salt, W. B., 2nd (2004) Nonalcoholic fatty liver disease (NAFLD): a
comprehensive review. J Insur Med 36, 27-41

Khashab, M., and Chalasani, N. (2007) Use of insulin sensitizers in NASH.
Endocrinol Metab Clin North Am 36, 1067-1087; xi

Sanyal, A. J., Carnpbell-Sargent, C., Mirshahi, F., Rizzo, W. B., Contos, M. J.,
Sterling, R. K., Luketic, V. A., Shiffrnan, M. L., and Clore, J. N. (2001)
Nonalcoholic steatohepatitis: association of insulin resistance and mitochondrial
abnormalities. Gastroenterology 120, 1183-1192

Thrush, A. B., Heigenhauser, G. J ., Mullen, K. L., Wright, D. C., and Dyck, D. J.
(2008) Palnnitate acutely induces insulin resistance in isolated muscle from obese
but not lean humans. Am J Physiol Regul Integr Comp Physiol 294, R1205-1212

Mordier, S., and Iynedjian, P. B. (2007) Activation of mammalian target of
rapamycin complex 1 and insulin resistance induced by palrrnitate in hepatocytes.
Biochem Biophys Res Commun 362, 206-211

Ruddock, M. W., Stein, A., Landaker, E., Park, J., Cooksey, R. C., McClain, D.,
and Patti, M. E. (2008) Saturated Fatty Acids Inhibit Hepatic Insulin Action by
Modulating Insulin Receptor Expression and Post-Receptor Signalling. J Biochem

195

56.

57.

58.

59.

60.

61.

Wieckowska, A., Papouchado, B. G., Li, Z., Lopez, R., Zein, N. N., and Feldstein,
A. E. (2008) Irncreased hepatic and circulating interleukin-6 levels in human
nonalcoholic steatohepatitis. Am J Gastroenterol 103, 1372-1379

Nehra, V., Angulo, P., Buchmarn, A. L., and Lindor, K D. (2001) Nutritional and
metabolic considerations in the etiology of nonalcoholic steatohepatitis. Dig Dis Sci
46, 2347-23 52

Le Phillip, P., Balnl, A., and Ungar, L. H. (2004) Using prior knowledge to improve
genetic network reconstruction from rrnicroarray data. In Silica Biol 4, 335-353

Guo, W., Wong, S., Xie, W., Lei, T., and Luo, Z. (2007) Palmitate modulates
intracellular signaling, induces endoplasmic reticulum stress, and causes apoptosis

in mouse 3T3-L1 and rat primary preadipocytes. Am J Physiol Endocrinol Metab
293, E576-586

Karaskov, E., Scott, C., Zhang, L., Teodoro, T., Ravazzola, M., and Volchuk, A.
(2006) Chronic palmitate but not oleate exposure induces endoplasmic reticulum

stress, which may contribute to INS-1 pancreatic beta-cell apoptosis.
Endocrinology 147, 3398-3407

Listenberger, L. L., Han, X., Lewis, S. E., Cases, S., Farese, R. V., Jr., Ory, D. S.,
and Schaffer, J. E. (2003) Triglyceride accumulation protects against fatty acid-
induced lipotoxicity. Proc Natl Acad Sci U S A 100, 3077-3082

196

 

 

   

M71111117111111!flilllllllllliilll“