"Molecular and Computational Approaches to Understanding Keloid Scarring"

"Molecular and Computational Approaches to Understanding Keloid Scarring"
MOLECULAR AND COMPUTATIONAL APPROACHES
TO UNDERSTANDING KELOID SCARRING
OOI NICK SERN, BRANDON
(B. Eng. (Hons.), NUS)
A THESIS SUBMITTED
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
GRADUATE PROGRAMME IN BIOENGINEERING
NATIONAL UNIVERSITY OF SINGAPORE
2010
1
ACKNOWLEDGEMENTS
First and foremost, I would like to express my sincerest gratitude to the Graduate
Programme in Bioengineering for giving me the opportunity to pursue my PhD studies.
By having access to the wonderful resources at the National University of Singapore, I
have been greatly facilitated in my quest for knowledge and learning.
I would also like to thank my supervisors, Prof Phan Toan Thang and Prof
Thiagarajan for their invaluable advice and enthusiastic support throughout my
candidature. I have learnt a great deal from my interactions with them and I believe that
this will stand me in good stead for my future endeavors. A big thank you also goes out
to my Thesis Steering Committee comprising of Prof Bay Boon Huat and Dr Martin
Lindsay Buist for their constructive advice and helpful suggestions. Their comments have
helped shape this thesis in more ways than one.
In all areas of work, colleagues play an immense role in the learning and
development of any project. Here I am indebted to Dr Anandaroop Mukhopadhyay, Dr
Masilamani Jeyakumar, Ms Audrey Khoo, Mr Ong Chee Tian, Ms Zhou Yue and Mr Do
Dang Vinh from the Wound Healing and Stem Cell Research Group, and to Dr Geoffrey
Koh and Mr Liu Bing from the Computational Biology Group. It is through them that I
have learnt the in vitro and in silico techniques that were essential to my project. Their
companionship has also been most welcome during the long years of my candidature.
Special thanks also go to Dr Lim Cheh Peng from the Institute of Molecular and Cell
Biology for checking my paper manuscripts, and also for the helpful support in
microarray work and analysis.
2
Last but definitely not least, I would like to thank my family and countless friends
who have supported and encouraged me throughout my PhD years. It is with their
unwavering support and with God‟s grace that I now stand at the brink of completion of
this project.
3
TABLE OF CONTENTS
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xi
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
List of Presentations and Publications . . . . . . . . . . . . . . . . . . . . . . . .
xx
Chapter One: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.1 Backround and motivations for the study . . . . . . . . . . . . . . . .
1
1.2 Approach and methodology . . . . . . . . . . . . . . . . . . . . . . . .
2
1.3 Contributions of the thesis . . . . . . . . . . . . . . . . . . . . . . . . .
4
1.4 Organization of the thesis . . . . . . . . . . . . . . . . . . . . . . . . .
5
Chapter Two: Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 Wound healing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
7
2.1.1 Hemostasis and inflammation . . . . . . . . . . . . . . . . . . 7
2.1.2 Proliferation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8
2.1.3 Remodeling . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8
2.2 Keloid scarring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.2.1 Keloid versus hypertrophic scar . . . . . . . . . . . . . . . .
10
2.2.2 Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.2.3 Clinical presentation . . . . . . . . . . . . . . . . . . . . . .
11
2.2.4 Histopathology . . . . . . . . . . . . . . . . . . . . . . . . . .
12
4
2.2.5 Etiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
2.2.6 Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Chapter Three: Materials and Methods . . . . . . . . . . . . . . . . . . . . .
17
3.1 Media and chemicals . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
3.2 Cell isolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
3.2.1 Keloid keratinocyte and fibroblast database . . . . . . . . . . 18
3.2.2 Keratinocyte culture from keloid scar and normal skin . . . 18
3.2.3 Fibroblast culture from keloid scar and normal skin . . .
19
3.2.4 Cell counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 HDGF experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.1 Immunohistochemistry . . . . . . . . . . . . . . . . . . . . . . 20
3.3.2 Serum stimulation of fibroblasts . . . . . . . . . . . . . . . .
21
3.3.3 Keratinocyte-fibroblast co-culture . . . . . . . . . . . . . . .
21
3.3.4 Treatment of fibroblasts with HDGF . . . . . . . . . . . . .
22
3.3.5 Treatment of keloid co-cultures with inhibitors . . . . . . . 22
3.3.6 Smad-null and Smad-overexpression cell assay . . . . . . . . 23
3.3.7 MTT assay . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
3.3.8 Western blotting . . . . . . . . . . . . . . . . . . . . . . . . .
24
3.3.9 Quantification and statistical analysis . . . . . . . . . . . . .
25
3.4 Microarray experiments . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
3.4.1 Cell culture . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
3.4.2 RNA extraction . . . . . . . . . . . . . . . . . . . . . . . . .
26
3.4.3 cRNA preparation and labeling . . . . . . . . . . . . . . . . .
27
5
3.4.4 Affymetrix chip hybridization and scanning . . . . . . . . . . 27
3.4.5 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.5 Reverse engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5.1 Preparation of additional microarray samples
. . . . . . . . 29
3.5.2 Data preprocessing . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5.3 Application of the fREDUCE algorithm . . . . . . . . . . . . 30
3.5.4 Pathways selected for influence approach . . . . . . . . . .
31
3.5.5 Application of the ARACNE and BANJO algorithms . . .
33
3.5.6 Estimation of the performance of the algorithms . . . . . .
34
Chapter Four: The Role of Hepatoma-derived Growth Factor in Keloid
Pathogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
4.2.1 HDGF expression is increased in keloid scar dermis. . . . .
39
4.2.2 Serum stimulation and epithelial-mesenchymal interactions had no
effect on intracellular HDGF expression . . . . . . . . . . . . 41
4.2.3 Epithelial-mesenchymal interactions in keloid co-culture increased
secretion of HDGF . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.4 Increased keloid fibroblast proliferation upon stimulation with
HDGF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2.5 Treatment of fibroblasts with HDGF activated the ERK pathway,
increased the secretion of VEGF, and decreased the secretion of
collagen I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6
4.2.6 Treatment with mTOR and Sp1 inhibitors did not significantly affect
the production of HDGF . . . . . . . . . . . . . . . . . . . . .
49
4.2.7 Knockout of Smad 2/3 signaling increases intracellular HDGF
expression while knockout of Smad 1 signaling increases
extracellular HDGF expression . . . . . . . . . . . . . . . . . . 50
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
Chapter Five: Genome Wide Transcriptional Profiling of Serum Starved Keloid and
Normal Fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
5.2.1 The time factor did not result in any systematic differences in the
transcriptional profile of the fibroblast cells . . . . . . . . .
66
5.2.2 Genes significantly upregulated in keloid compared to normal
fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
5.2.3 Genes significantly downregulated in keloid compared to normal
fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
5.2.4 Hierarchical clustering and principal components analysis revealed
that genes chosen were capable of distinguishing between keloid and
normal samples . . . . . . . . . . . . . . . . . . . . . . . . . .
72
5.2.5 DAVID analysis suggests a role for immunological factors and
ribosomal proteins in keloid pathogenesis . . . . . . . . . . . 74
5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
7
Chapter Six: Reverse Engineering Gene Networks in Keloid and Normal
Fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
6.2 Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
6.2.1 fREDUCE . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
6.2.2 ARACNE . . . . . . . . . . . . . . . . . . . . . . . . . . . .
96
6.2.3 BANJO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
6.3.1 Binding motifs found from fREDUCE for keloid versus normal
fibroblasts under serum starvation condition . . . . . . . . .
100
6.3.2 Binding motifs found from fREDUCE for keloid versus normal
fibroblasts under serum induced condition . . . . . . . . . .
101
6.3.3 Binding motifs found from fREDUCE for sets C and D suggest
consistent effects from steroid induction for both keloid and normal
fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.3.4 Not many binding motifs found from fREDUCE for sets E and
F. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.3.5 Mean sensitivity performance of BANJO in recovering influence
networks was significantly better than that of ARACNE . . . 106
6.3.6 Transcriptional networks were better suited for reverse engineering
compared to cytokine receptor interactions and intracellular
signaling networks. . . . . . . . . . . . . . . . . . . . . . . . . 107
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
108
8
Chapter Seven: Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
114
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
117
Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
132
A.1 Full list of 181 genes upregulated in keloid compared to normal fibroblasts
using the MAS 5.0 summarization algorithm (P < 0.05) . . . . . . . . 132
A.2 Full list of 290 genes downregulated in keloid compared to normal fibroblasts
using the MAS 5.0 summarization algorithm (P < 0.05) . . . . . . . . 137
A.3 Full list of 86 genes upregulated in keloid compared to normal fibroblasts
using the RMA summarization algorithm (P < 0.05) . . . . . . . . . . 145
A.4 Full list of 258 genes downregulated in keloid compared to normal fibroblasts
using the RMA summarization algorithm (P < 0.05) . . . . . . . . . . 147
A.5 List of genes differentially expressed using both the RMA and MAS 5.0
summarization algorithm (P < 0.05) . . . . . . . . . . . . . . . . . . . . 154
A.6 Cytokine-cytokine receptor interaction from the KEGG database (Benjamini
corrected P-value = 0.094) . . . . . . . . . . . . . . . . . . . . . . . . .
160
A.7 Toll-like receptor signaling pathway from the KEGG database (Benjamini
corrected P-value = 0.246) . . . . . . . . . . . . . . . . . . . . . . . . .
161
9
SUMMARY
Keloid scars are aberrations in the wound healing process, resulting in the
appearance of protrusive crab like extensions growing into normal tissue. They do not
subside with time, and may develop over the most minor of skin wounds, such as insect
bites or acne. Aside from being an aesthetic impediment, keloids are frequently
associated with itchiness, pain and, when involving the skin overlying a joint, restricted
range of motion. To date, none of the known treatment modalities have proven optimal.
In recent years, a systems approach to understanding biology has gained
eminence, in part due to the limitations of a purely reductionist approach in explaining
biological phenomena. However, there are merits to the reductionist approach; much of
what we know of biology today can be attributed to the work of molecular biologists of
the past. In this dissertation, we will adopt both these approaches to tackling the keloid
problem.
In the first part of this thesis, we examined the role played by a novel growth
factor, the hepatoma-derived growth factor (HDGF), in keloid pathogenesis. Using a
combination of immunohistochemical staining and Western blots, we found that secreted
HDGF is increased in the keloid condition and its secretion is modulated by epithelial–
mesenchymal interactions. Furthermore, exogenous HDGF exerts a proliferative effect on
keloid fibroblasts and increases the production of the angiogenic factor VEGF, indicating
that it plays some role in the process of angiogenesis.
With the advent of high throughput technology, researchers are no longer
confined to the study of individual molecules. In the second part of this dissertation, we
utilized the microarray platform to assess the global transcriptional differences between
10
keloid and normal fibroblasts under serum free conditions. Many of the genes that have
been found to be differentially expressed in previous studies were reconfirmed in this
study. In addition, some interesting and novel genes not previously reported were also
discovered. Gene Ontology terms that were found to be significantly enriched include
those relating to immune response, antigen processing and presentation, chemokine and
cytokine activity, extracellular matrix and ribosomal proteins.
In the third part of this thesis, we attempted to reverse engineer gene networks
from microarray expression profiles of keloid and normal fibroblasts. Using a physical
approach to model transcription factor interactions, we discovered some of the binding
motifs that were active in the keloid condition. Furthermore, we used the influence
approach to reverse engineer some of the networks that were found to be significantly
enriched from the second part of this dissertation. Our results indicate that transcriptional
networks were better suited for this process compared to cytokine receptor interactions
and intracellular signaling networks. We also found that the NFKB transcriptional
network that was inferred from normal fibroblast data was more accurate compared to
that inferred from keloid data, suggesting a more robust network in the keloid condition.
The work done in this thesis, utilizing both molecular and computational
approaches, has hopefully advanced our understanding of keloid scarring. In addition, the
results from this study has generated new and promising future areas of research and is a
small step forward to finding a solution to this condition.
11
LIST OF TABLES
5.1
Comparison of different microarray studies . . . . . . . . . . . . . . . . . . 65
5.2
Two-way ANOVA results for determining the contribution of the time and type of
cell on gene expression with probe summarization by MAS 5.0 . . . . . . 66
5.3
Two-way ANOVA results for determining the contribution of the time and type of
cell on gene expression with probe summarization by RMA . . . . . . . . 66
5.4
Top 25 upregulated genes in keloid compared to normal fibroblasts using the
MAS 5.0 summarization algorithm ranked by fold change . . . . . . . . . 68
5.5
Top 25 upregulated genes in keloid compared to normal fibroblasts using the
RMA summarization algorithm ranked by fold change . . . . . . . . . . 68
5.6
Top 25 downregulated genes in keloid compared to normal fibroblasts using the
MAS 5.0 summarization algorithm ranked by fold change . . . . . . . . 70
5.7
Top 25 downregulated genes in keloid compared to normal fibroblasts using the
RMA summarization algorithm ranked by fold change. . . . . . . . . . . 71
5.8
List of Gene Ontology terms that were found to be statistically enriched using the
DAVID Gene Functional Classification Tool with the list of significantly
upregulated genes in keloid as input . . . . . . . . . . . . . . . . . . . . . . 75
5.9
List of Gene Ontology terms that were found to be statistically enriched using the
DAVID Gene Functional Classification Tool with the list of significantly
downregulated genes in keloid as input . . . . . . . . . . . . . . . . . . .
76
5.10
List of downregulated genes in keloid compared to normal fibroblasts involved in
GO term Antigen Processing and Presentation . . . . . . . . . . . . . .
77
5.11
List of upregulated genes in keloid compared to normal fibroblasts involved in
GO term Ribosome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
77
6.1
Binding motifs found from fREDUCE for keloid versus normal fibroblasts under
serum starvation condition . . . . . . . . . . . . . . . . . . . . . . . . . . .
100
6.2
Possible gene targets and TFs found from the TRANSFAC database for top
binding motifs from Table 6.1 . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3
Binding motifs found from fREDUCE for keloid versus normal fibroblasts under
serum induced condition . . . . . . . . . . . . . . . . . . . . . . . . . . . .
102
12
6.4
Possible gene targets and TFs found from the TRANSFAC database for top
binding motifs from Table 6.3 . . . . . . . . . . . . . . . . . . . . . . . .
102
6.5
Binding motifs found from fREDUCE for steroid treated versus control keloid
fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
6.6
Binding motifs found from fREDUCE for steroid treated versus control normal
fibroblasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
6.7
Possible gene targets and TFs found from the TRANSFAC database for top
binding motifs from Table 6.5 and 6.6 . . . . . . . . . . . . . . . . . . . . . 104
6.8
Binding motifs found from fREDUCE for keloid versus normal fibroblasts under
steroid treated condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.9
Possible gene targets and TFs found from the TRANSFAC database for top
binding motifs from Table 6.8 . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.10
PPV and sensitivity results for all data sets run using BANJO and ARACNE
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
13
LIST OF FIGURES
2.1
Schematic representation of different stages of wound repair. . . . . . . . 9
2.2
Keloid formation in different parts of the body and in different patients.
3.1
Co-culture of epidermal keratinocytes and dermal fibroblasts as an in vitro model
to study epithelial-mesenchymal interactions.. . . . . . . . . . . . . . . . . 22
3.2
KEGG pathways used for the influence approach . . . . . . . . . . . . . . 33
4.1
Immunohistochemical staining of keloid and normal tissue for HDGF . . 40
4.2
Western blot of keloid and normal whole tissue extract. . . . . . . . . . . . 41
4.3
Effect of serum and epithelial–mesenchymal interactions on intracellular HDGF
expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4
Expression of HDGF in conditioned media of monocultured and co-cultured cells
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.5
Increased proliferation of keloid fibroblasts treated with recombinant HDGF
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6
Effect of HDGF on the expression of downstream intracellular targets . . 47
4.7
Effect of HDGF on the expression of downstream extracellular targets . . 48
4.8
Effect of mTOR and Sp1 inhibitors on the expression of HDGF . . . . . . 49
4.9
Effect of Smad signaling on intracellular HDGF expression . . . . . . .
51
4.10
Effect of Smad signaling on extracellular HDGF expression . . . . . . .
52
4.11
Schematic representation of the role of HDGF in keloid pathogenesis. . . 59
5.1
Affymetrix GeneChip Expression Array design . . . . . . . . . . . . . . . . 62
5.2
Principal components analysis and hierarchical clustering using the MAS 5.0
algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3
Principal components analysis and hierarchical clustering using the RMA
algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.4
Antigen processing and presentation pathway from the KEGG database
12
78
14
5.5
Ribosome pathway from the KEGG database . . . . . . . . . . . . . . . .
79
6.1
The general strategy for reverse-engineering transcription control systems 94
6.2
Comparison between ARACNE, BANJO, RMA and MAS 5 based on PPV and
sensitivity values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
15
LIST OF SYMBOLS
α-SMA
Alpha-smooth muscle actin
ADAM12
A disintegrin and metalloprotease 12
ANOVA
Analysis of variance
AVEN
Apoptosis caspase activation inhibitor
ATXN1
Ataxin-1
BANJO
Bayesian Network Inference with Java Objects
BDe
Bayesian metric with Dirichlet priors and equivalence
C3
Complement component 3
CADM1
Cell adhesion molecule 1
CDF
Chip definition file
cDNA
Complementary DNA
CFB
Complement factor B
COL1A1
Collagen type I alpha 1
COL5A1
Collagen type V alpha 1
COL5A3
Collagen type V alpha 3
COL15A1
Collagen type XV alpha 1
COL17A1
Collagen type XVII alpha 1
cRNA
Complementary RNA
CTGF
Connective tissue growth factor
CXCL1
Chemokine ligand 1
CXCL2
Chemokine ligand 2
CXCL6
Chemokine ligand 6
16
DAVID
Database for Annotation, Visualization and Integrated
Discovery
DNA
Deoxyribonucleic acid
DPI
Data Processing Inequality
ECM
Extracellular matrix
ECM1
Extracellular matrix protein 1
EPHB4
Ephrin type-B receptor 4
fREDUCE
fast-Regulatory Element Detection Using Correlation with
Expression
GEO
Gene Expression Omnibus
G0S2
G0/G1switch 2
GO
Gene Ontology
GPR137B
G-protein-coupled receptor 137B
GPR153
G-protein-coupled receptor 153
GPSM2
G-protein signaling modulator 2
GRB10
Growth factor receptor-bound protein 10
HATH
Homologous to amino terminus of HDGF
HDGF
Hepatoma-derived growth factor
HIF-1α
Hypoxia induced factor-1α
HLA
Human leukocyte antigen
HOXA11
Homeobox A11
HOXD10
Homeobox D10
HRP
HDGF-related protein
17
HSD11B1
Hydroxysteroid (11-beta) dehydrogenase 1
IFIT1
Interferon-induced protein with tetratricopeptide repeats 1
IFIT3
interferon-induced protein with tetratricopeptide repeats 3
IGF-1
Insulin-like growth factor-1
IGFBP
Insulin-like growth factor binding protein
IGFBP3
Insulin-like growth factor binding protein 3
IHC
Immunohistochemical staining
IL6
Interleukin 6
IL8
Interleukin 8
IL32
Interleukin 32
IVT
In vitro transcription
KEGG
Kyoto Encyclopedia of Genes and Genomes
KF
Keloid fibroblasts
KK
Keloid keratinocytes
KRT19
Keratin 19
LAMA2
Laminin alpha 2
LEDGF
Lens epithelium-derived growth factor
MEMO1
Mediator of cell motility 1
MAPK
Mitogen-activated protein kinase
MAS
Microarray suite
MHC
Major histocompatibility complex
MI
Mutual Information
MMP
Matrix metalloproteinase
18
mRNA
Messenger RNA
mTOR
Mammalian target of rapamycin
MTT
3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium
bromide
MYO1D
Myosin 1D
MYO19
Myosin 19
NF
Normal fibroblasts
NFKB
Nuclear factor kappa-light-chain-enhancer of activated B
cells
NGF
Nerve growth factor
NK
Normal keratinocytes
OAS1
2',5'-oligoadenylate synthetase 1
PAI-1
Plasminogen activator inhibitor-1
PCNA
Proliferating cell nuclear antigen
PCA
Principal components analysis
PCR
Polymerase chain reaction
PDGF
Platelet-derived growth factor
PDGFRB
Platelet-derived growth factor receptor beta
PI3-K
Phosphatidylinositol 3-kinase
POSTN
Periostin
PPV
Positive Predicted Value
PTGES
Prostaglandin E synthase
PTX3
Pentraxin-related gene
19
RAC2
Ras-related C3 botulinum toxin substrate 2
REDUCE
Regulatory Element Detection Using Correlation with
Expression
RMA
Robust multichip analysis
RNA
Ribonucleic acid
RNAP
RNA polymerase
RPS
Ribosomal protein
RSAD2
Radical S-adenosyl methionine domain containing 2
SEM
Standard error of the mean
SEM5A
Semophorin-5A
SFRP1
Secreted frizzled-related protein 1
SLC39A8
Solute carrier family 39 member 8
SOS2
Son of sevenless homolog 2
TAP
Transporter associated with antigen processing
TF
Transcription factor
THBS1
Thrombospondin-1
TNFAIP3
Tumor necrosis factor alpha-induced protein 3
TNFAIP6
Tumor necrosis factor alpha-induced protein 6
TNFSF10
Tumor necrosis factor superfamily member 10
TGF-β
Transforming growth factor beta
VEGF
Vascular endothelial growth factor
WNT5A
Wingless-type MMTV integration site family member 5A
20
LIST OF PRESENTATIONS AND PUBLICATIONS
Biostar 2006: 2nd International Congress on Regenerative Biology, Stuttgart. Hepatomaderived growth factor contributes to keloid pathogenesis via epithelial mesenchymal
interactions and secretion into conditioned media [Poster presentation]
J Cell Mol Med. 2010 Jun; 14(6A):1328-37. Hepatoma-derived growth factor and its role
in keloid pathogenesis. Ooi BNS, Mukhopadhyay A, Masilamani J, Do DV, Lim CP, Cao
XM, Lim IJ, Mao L, Ren HN, Nakamura H, Phan TT. [Published]
Burns. Microarray analysis of serum starved keloid and normal fibroblasts suggest a role
for immunological factors and ribosomal proteins in keloid pathogenesis. BNS Ooi, CP
Lim, XM Cao, TT Phan [Submitted]
Theoretical Biology and Medical Modeling. Insights gained from the reverse engineering
of gene networks in keloid fibroblasts. BNS Ooi, TT Phan, PS Thiagarajan [Submitting]
21
CHAPTER ONE
INTRODUCTION
“The availability of genome sequence is just the beginning. Scientists now want to
understand the genes and the role they play in the prevention, diagnosis and treatment of
disease.”
– Dr Randy Scott, President of Incyte
1.1 Background and motivations for the thesis
Since the discovery of deoxyribonucleic acid (DNA) in the 1950s by Watson and Crick,
biology has moved at a rapid rate. Thanks to the Human Genome Project, we now have in
our possession the complete genome of the human species. Future biomedical research
would involve the application of this knowledge to the understanding of various
biological processes in the hopes of uncovering new methods of treating the numerous
diseases and medical conditions afflicting the human race.
Among the many diseases to beset mankind, keloids do not rank very highly in
the hall of fame. However, the appearance of these large protruding claw-like scars is
bound to elicit shock and distress in most observers due to their unsightly nature.
Furthermore, aside from causing emotional trauma, keloid scars can be painful or itchy,
and may restrict mobility if formed over a joint (Lee et al. 2004). In a study assessing the
quality of life of patients with keloid and hypertrophic scarring, it was demonstrated for
the first time that the quality of life of these patients was reduced due to physical and/or
psychological effect (Bock et al. 2006). The problem is further exacerbated by the fact
22
that there is no particularly effective treatment to date (Tuan & Nichter 1998; Louw
2007). These scars also have a propensity to recur after surgery and have been considered
as benign tumours (English & Shenefelt 1999).
For all the reasons stated above, it would be beneficial if we could discover some
effective method of treating these scars. To this end, an understanding of the molecular
etiology of keloids would be useful. Furthermore, since keloid formation is generally
considered to be a form of abnormal wound healing, any insights gained from this
endeavour would also increase our understanding of the wound healing process.
1.2 Approach and methodology
We have decided to adopt both top down as well as reductionist approaches to
understanding the mechanisms underlying keloid pathology. In the first part of this
dissertation, we investigated the role played by a novel protein in the keloid condition
using molecular biology techniques. While it was found that this molecule, the hepatomaderived growth factor (HDGF) is significantly expressed in keloids, our data also
suggests that it is unlikely that this growth factor is able to induce keloid formation on its
own. Therefore, while a reductionist, in depth study of this molecule would certainly
increase our understanding about keloids, the knowledge gained would only be a small
fraction of the complex mechanisms underlying keloid pathology.
Researchers today are no longer confined to studying one molecule at a time
thanks to the development of various high throughput techniques. These technologies
enable us to have a snapshot of the thousands of molecules present in the cell at any one
23
time. In the second part of this dissertation, we utilized one of these technologies,
specifically the Affymetrix microarray platform, to gain insights into some of the system
level differences between keloid and normal cells. Based on this approach, we would be
able to identify all genes that are significantly different between the two conditions. The
data generated from this study can then be utilized for further research by using a
reductionist approach to study the genes individually, or by extracting biological meaning
through a computational approach.
One way of increasing our biological knowledge is to learn how the different
molecules in the cell are connected. In the third part of the dissertation, we attempt to
reconstruct gene networks using a combination of probabilistic and regression
techniques. There are two general strategies for reverse engineering gene networks – a
physical approach where physical interactions between transcription factors and their
promoters are modeled, and an influence approach where the mechanistic process is
abstracted out as a black box. For the physical approach, we will use the entire
microarray data set for modeling, but for the influence approach, we will focus on small
networks of genes that have been found to be differentially expressed from the second
part of this dissertation. Most attempts at modeling biological networks have been done
using simulated data; our work would highlight some of the issues involved in working
with experimental data. Furthermore, it is hoped that insights gained from this endeavor
would provide some clues about the different transcriptional regulatory mechanisms
present in keloid and normal fibroblasts.
24
1.3 Contributions of the thesis
We first discovered increased expression of HDGF in keloid scars compared to normal
skin. An in vitro study of the role of HDGF using keloid and normal derived cells suggest
that epidermal mesenchymal interactions govern the increased secretion of this growth
factor in the keloid condition. Furthermore, HDGF was found to increase the proliferation
of keloid fibroblasts and was also found to increase the production of the vascular
endothelial growth factor (VEGF). However, one of the hallmarks of keloids is an
increased extracellular matrix production, and HDGF did not seem to contribute to this
aspect of keloid formation.
In the second part of this dissertation, we used the microarray platform in an
attempt to identify groups of genes that can be implicated in the formation of keloids.
While other groups have utilized this technology previously, none had surveryed the
global transcriptional landscape in serum starvation conditions. Furthermore, there was
very little overlap in many of the microarray studies done, and it is hoped that our study
would help identify some of the more consistent differentially expressed genes. Our
results indicate some consistency with previous studies done on keloid fibroblasts. We
also uncovered differentially expressed genes that have not been reported previously, and
enrichment analysis indicate that processes such as immune response, antigen processing
and presentation, chemokine and cytokine activity, extracellular matrix and ribosomal
proteins are among those that are affected in the keloid condition.
In the third part of this dissertation, we attempted to reverse engineer gene
networks using the microarray data that was generated in the second part of the thesis, as
well as any publicly available microarray data on keloid and normal fibroblasts that we
25
could find in the literature. Using the physical approach of correlating expression values
to binding motifs, we found some consensus sequences that were active in the keloid
condition, as well as some sequences that were responsive to steroids, one of the
commonly used treatments for keloids. These consensus sequences are possible
transcription factor binding sites and could be explored for developing future keloid
treatments or to improve the efficacy of current steroid treatments. We also compared
different normalization methods and influence approaches on the reconstruction of
known gene networks taken from the KEGG database that were found to be statistically
enriched in our microarray data. We found that the combination of the Bayesian
algorithm, RMA normalization and transcriptional networks gave the best reconstruction
results and this could serve as a guide for future influence approaches dealing with
experimental data.
1.4 Organization of the thesis
The rest of the thesis is organized as follows. In Chapter Two, background information
on wound healing and keloid scarring is presented. Chapter Three describes the materials
and methods used in both molecular and computational approaches employed in this
study. In Chapter Four, the importance of HDGF in keloid formation is studied using a
combination of cell and molecular techniques. Chapter Five examines the global
transcriptional differences between keloid and normal skin fibroblasts by utilizing the
Affymetrix microarray platform. In Chapter Six, insights obtained from the reverse
26
engineering of keloid and normal fibroblast gene networks are discussed. Conclusions
from the thesis are presented in Chapter Seven.
27
CHAPTER TWO
LITERATURE REVIEW
2.1 Wound healing
To understand the underlying mechanisms involved in pathologic conditions such as
scarring and fibrosis, it is useful to first review what is known about normal tissue
response to injury. Upon wounding, an orderly series of events is triggered, with the final
desired outcome being the restoration of anatomical structure and function. These events
can be grouped into four distinct but overlapping phases, hemostasis, inflammation,
proliferation and remodeling (Mast 1992).
2.1.1 Hemostasis and inflammation
The healing cascade starts with the aggregation of platelets at the wound site and the
release of clotting factors. This results in the formation of a fibrin clot to plug the wound
(Clark 2001). At the same time, a cocktail of growth factors and cytokines are released
from the serum of the disrupted blood vessels and degranulating platelets (Werner &
Grose 2003). Following hemostasis, neutrophils infiltrate into the wound site and
monocytes are activated to become wound macrophages. These inflammatory cells serve
two purposes: firstly as a means of removing foreign material, bacteria and damaged
matrix components by phagocytosis, and secondly as a source of growth factors that are
required to initiate the next phase of the healing process (Sylvia 2003; Diegelmann,
Cohen & Kaplan 1981).
28
2.1.2 Proliferation
In the proliferative phase, the predominant cell in the wound site is the dermal fibroblast
(Stadelmann, Digenis & Tobin 1998). This cell is responsible for producing collagen and
other extracellular matrix components needed to restore structure and function to the
injured tissue. At least 23 different types of collagen have been identified but type I is
predominant in the scar tissue of skin (Prockop & Kivirikko 1995). Also during this
phase, keratinocytes in the epidermis proliferate and migrate from the wound edge
leading to the process of reepithelialization (Santoro & Gaudino 2005). In addition, local
factors in the wound microenvironment such as low pH and reduced oxygen tension
initiate the release of angiogenic factors leading to the migration and proliferation of
endothelial cells (Knighton et al. 1983). Massive angiogenesis leads to the formation of
new blood vessels, and the resulting wound connective tissue is known as granulation
tissue because of the granular appearance of the numerous capillaries (Werner & Grose
2003). Around a week after the wounding has taken place, fibroblasts have differentiated
into myofibroblasts and the wound begins to contract. Myofibroblasts contain the same
kind of actin as found in smooth muscle cells, alpha-smooth muscle actin (α-SMA) to
produce more force during contracture (Hinz 2006).
2.1.3 Remodeling
In the final stage, collagen undergoes cross-linking to improve its strength and stability.
This stage is characterized by continued collagen synthesis and collagen catabolism
finally resulting in a normal scar (Parks 1999). This process requires a balance between
matrix biosynthesis and matrix degradation. A disruption in this balance either due to
29
excessive matrix deposition or decreased matrix degradation leads to keloid and
hypertrophic scars (Raghow 1994).
Figure 2.1: Schematic representation of different stages of wound repair (Werner & Grose
2003). A: 12–24 h after injury the wounded area is filled with a blood clot. Neutrophils invade
into the clot. B: at days 3–7 after injury, macrophages are abundant in the wound tissue.
Endothelial cells migrate into the clot; they proliferate and form new blood vessels. Fibroblasts
migrate into the wound tissue, where they proliferate and deposit extracellular matrix.
Keratinocytes proliferate at the wound edge and migrate above the provisional matrix. C: 1–2 wk
after injury the wound is completely filled with granulation tissue. The wound is completely
covered with a neoepidermis.
30
2.2 Keloid scarring
2.2.1 Keloid versus hypertrophic scars
The term cheloide was coined in 1802 to describe the lateral extensions often observed in
these scars, which resemble the legs of a crab growing into normal tissue (Urioste, Arndt
& Dover 1999). Keloids are commonly compared with hypertrophic scars, and the two
share some similarities such as increased collagen secretion and a similar gross
appearance. However, unlike hypertrophic scars that are confined to the area of injury,
keloids may extend well beyond the confines of the original wound. Furthermore,
hypertrophic scars usually subside with time, whereas keloids continue to evolve over
time, without a quiescent or regressive phase (Nemeth 1993). While hypertrophic scars
usually develop within a few weeks after skin injury, keloids normally show a delayed
onset, normally forming months after skin trauma (Marneros & Krieg 2004).
2.2.2 Epidemiology
It is not well documented how commonly keloids occur in the general population but the
reported incidence range from a high of 16% among adults in Zaire to a low of less than
1% among adults in England (English & Shenefelt 1999). It is widely accepted that darkskinned populations have a higher occurrence of keloids than light-skinned populations,
with the reported incidence ratio between the two groups ranging from 2:1 to 19:1
(Atiyeh, Costagliola & Hayek 2005). Among Asians, keloid incidence appears to be more
common in Chinese (Shaffer, Taylor & Cook-Bolden 2002). Both autosomal dominant
31
and autosomal recessive genetic inheritance have been proposed but not confirmed and
some data suggest familial occurrence (Bloom 1956; Omo-Dare 1975).
A difference in occurrence of keloids based on gender has not been demonstrated
convincingly (Marneros & Krieg 2004). However, most reported cases have occurred in
individuals between 10 and 30 years of age (Rockwell, Cohen & Ehrlich 1989). Hormone
levels are high at this age, indicating that they may have some influence on keloid
formation. This hypothesis is supported by data showing an increased binding of
androgens in keloid tissue (Ford et al. 1983; Schierle, Scholz & Lemperle 1997).
Furthermore, some reports suggest that keloids appear more often in puberty, enlarge
during pregnancy, and decrease in size after menopause (Moustafa, Abdel-Fattah &
Abdel-Fattah 1975). However, other explanations such as increased neo-angiogenesis
during pregnancy are also possible (Seifert & Mrowietz 2009).
2.2.3 Clinical presentation
Keloids are generally considered to be a result of excessive wound healing, although
some also believe these scars to be a type of benign fibrous tumor (Slemp & Kirschner
2006). They are characterized by an overgrowth of dense fibrous tissue coupled with
excessive deposition of extracellular matrix (ECM) components such as collagen and
fibronectin (Rockwell, Cohen & Ehrlich 1989; Babu, Diegelmann & Oliver 1989). They
uniquely affect only humans, and may develop even after the most minor of skin wounds,
such as insect bites or acne (Urioste, Arndt & Dover 1999). Keloids are frequently
associated with itchiness, pain and, when involving the skin overlying a joint, restricted
range of motion (Lee et al. 2004). For unknown reasons, keloids occur more frequently
32
on the chest, shoulders, upper back, back of the neck, and earlobes (Bayat et al. 2004).
Corneal keloidal scarring has also been observed (Shukla, Arora & Arora 1975).
Figure 2.2: Keloid formation in different parts of the body and in different patients (Marneros &
Krieg 2004).
2.2.4 Histopathology
Normal skin contains distinct collagen bundles that run parallel to the epithelial surface.
In hypertrophic scars, collagen bundles are flatter, less demarcated, and are arranged in a
wavy pattern. In keloids, the collagen bundles are thick and are randomly oriented as
swirls and whorls (Rockwell, Cohen & Ehrlich 1989). Keloid formations are
characterized by active angiogenesis and hypoxia (Appleton, Brown & Willoughby
1996). Occlusion of some microvessels by excessive endothelial cells may lead to local
hypoxic conditions and apoptosis (Kischer 1992).
33
2.2.5 Etiology
Several etiological factors for keloids have been proposed, with skin injury being the
most obvious. Spontaneous occurrence of keloids in the absence of trauma is rare
although a few cases have been reported (Shaffer, Taylor & Cook-Bolden 2002).
However, such spontaneous occurrence could be the result of a minor, overlooked trauma
to the skin (Marneros & Krieg 2004). Increased skin tension has also been postulated to
play some role in keloid formation. However, soles and palms which are sites of high
skin tension are rarely sites of keloid formation, and the most affected site reported, the
earlobe, is under minimal tension (Seifert & Mrowietz 2009).
The role of immunologic factors in keloid formation has not been studied in detail
and remains to be elucidated. Immune cell infiltrate in keloids include T lymphocytes and
denditric cells (Santucci et al. 2001) and an increased number of macrophages, epidermal
Langerhans cells and mast cells have been noted as well (Niessen et al. 2004; Smith,
Smith & Finn 1987). Some authors have reported an association with cell membrane
proteins, such as HLA-DRB-16, B-14, and BMW-16 (Datubo-Brown 1990), elevated
tissue levels of IgG, IgA, and IgM (Kischer et al. 1983), and abnormal immune response
to sebum (Yagi, Dafalla & Osman 1979). The sebum hypothesis provides an explanation
for the absence of keloids on anatomical sites lacking sebaceous glands, such as palms
and soles (Seifert & Mrowietz 2009). Dermal injury exposes the pilosebaceous unit to the
systemic circulation, initiating a cell-mediated immune response in persons who retain T
lymphocytes sensitive to sebum. Subsequent release of cytokines, including various
interleukins and TGF-beta, stimulates chemotaxis of mast cells and production of
collagen by fibroblasts. This hypothesis also gives a plausible reason as to why only
34
human beings, the only mammals with true sebaceous glands, are affected by keloid
scarring (Al-Attar et al. 2006).
Several studies have shown that many different cytokines and growth factors are
involved in the formation of keloids. Some of the important molecules that were elevated
in keloids include transforming growth factor beta (TGF-β) (Lee et al. 1999), interleukin6 (IL-6) (Ghazizadeh 2007) and vascular endothelial growth factor (VEGF) (Ong et al.
2007). Keloid fibroblasts were also more responsive in mitogenic assays to plateletderived growth factor (PDGF) (Haisa, Okochi & Grotendorst 1994).
Another possible factor underlying the growth and formation of keloids is their
resistance to apoptosis. Keloid fibroblasts was found to be more resistant to Fas mediated
apoptosis (Chodon et al. 2000) and the overexpression of insulin-like growth factor-1
(IGF-1) receptor inhibited ceramid-induced apoptosis (Ishihara et al. 2000). Furthermore,
decreased expression of proapoptotic genes (Sayah et al. 1999) and increased expression
of apoptotic inhibitors (Messadi et al. 2004) have also been observed in keloid
fibroblasts.
Tissue hypoxia could be another contributory factor to pathogenesis. An increased
level of hypoxia marker, hypoxia induced factor-1α (HIF-1α) was detected in keloid
tissues and hypoxia appears to elevate the expression of plasminogen activator inhibitor-1
(PAI-1) (Zhang et al. 2003). Increased PAI-1 activity correlated with an elevated collagen
expression in fibrin gel cultures of keloid fibroblasts (Tuan et al. 2003). Hypoxia-driven
VEGF is also increased in keloids (Wu et al. 2004).
While most in vitro studies focus on keloid fibroblasts, recent evidence points to
altered interactions between keratinocytes and fibroblasts in keloids. To examine
35
epithelial-mesenchymal cross-talk in skin, experiments using normal or keloid
keratinocytes co-cultured with normal or keloid fibroblasts have been conducted. In such
co-culture systems, keloid keratinocytes promoted the proliferation of keloid fibroblasts
to a greater extent than normal keratinocytes, while the least proliferation was seen in
keloid fibroblasts cultured without any keratinoctyes (Lim et al. 2001; Funayama et al.
2003). Furthermore, co-culturing normal or keloid fibroblasts with keloid keratinocytes
resulted in an increased expression of collagen I and III compared to the non co-cultured
condition (Lim et al. 2002). These data suggest that epithelial-mesenchymal interactions
could contribute to keloid pathogenesis.
2.2.6 Treatment
Like many other diseases, the best treatment for keloids is prevention. Although many
different treatment modalities have been proposed, none have proven to be optimal.
Surgical excision of a keloid is associated with a high recurrence rate and therefore has to
be combined with some other adjunctive therapy. These include compression therapy,
silicone sheeting, cryotherapy, radiation or laser therapy (Slemp & Kirschner 2006; Louw
2007).
Unfortunately, there are drawbacks to many of these methods. Compression
therapy is ultimately limited by the ability to adequately fit the garment to the wounded
area and patient discomfort frequently reduces compliance (Cheng et al. 1984). The
success of silicone sheeting is also limited by patient compliance, and silicone products
may cause adverse effects, including skin maceration and excoriation (Slemp &
Kirschner 2006). Cryotherapy could lead to permanent hypopigmentation resulting from
36
cold sensitivity of melanocytes and is therefore less desirable in patients with darker skin
(Louw 2007). On the other hand, radiation therapy causes hyperpigmentation and carries
the theoretical risk of radiation induced malignancy (Wolfram et al. 2009). The efficacy
of laser treatment has been low with a recurrence rate of 50% (Apfelberg et al. 1989).
Other pharmacologic therapies for reducing the recurrence rate exist, with the
application of corticosteroids being the most well known. Potential side effects of
corticosteroid
injections
include
pain,
skin
atrophy,
telangiectasia
formation,
depigmentation, and infection (Urioste, Arndt & Dover 1999). Treatment with
interferons, which are cytokines secreted by T-helper cells, may help to reduce fibrosis,
but treatment has also been met with some success, but has severe side effects including
fever, chills, night sweats, fatigue, myalgia and headache (Wolfram et al. 2009). 5Fluorouracil is another compound that has been used successfully as an antiproliferative
agent. The injection can be painful however, and purpura and ulcers have been
documented (Wolfram et al. 2009).
The side effects of the above treatments notwithstanding, ultimately, none of the
above methods are completely effective in preventing the recurrence of keloids. Many
attempts have been made to find successful alternatives, with the ultimate direction of
research geared toward understanding scarring at the molecular level in the hope of
obtaining a permanent solution to this problem.
37
CHAPTER THREE
MATERIALS AND METHODS
3.1 Media and chemicals
Dulbecco‟s modified eagle medium (DMEM), Hanks balanced salt solution (HBSS), fetal
calf serum (FCS), streptomycin, penicillin, gentamicin and fungizone were purchased
from Gibco. Keratinocyte growth medium (KGM) was purchased from Clonetics (USA).
Phosphate buffered saline without Ca2+ and Mg2+ (PBS), epidermal growth factor (EGF),
cholera toxin and hydrocortisone were purchased from Sigma Chemical Co (USA).
Dispase II was purchased from Boehringer Mannheim (USA). Rhodamine counter stain
was obtained from Difco (USA). Tris base was purchased from J.T Baker. Triton X-100,
ethylenediaminetetraacetic acid (EDTA), 30% acrylamide/bis solution (37.5:1 2.6%C)
and glycine were purchased from Biorad. Sodium Chloride (NaCl), nonidet P-40 (NP40), sodium dodecyl sulphate, hydrogen peroxide (H2O2), bovine serum albumin (BSA),
tween-20, potassium chloride (KCL), potassium phosphate (K3PO4), magnesium chloride
(MgCl2), MTT [3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide], N,N –
dimethylformamide (DMF) and paraformaldehyde were all purchased from Sigma
Chemical Co (USA). Methanol and acetic acid were purchased from Lab-Scan. RNeasy
kit was bought from Qiagen (Germany) while the GeneChip Eukaryotic Target Labeling
and Control Reagents and arrays were bought from Affymetrix (USA).
38
3.2 Cell isolation
3.2.1 Keloid keratinocyte and fibroblast database
Keratinocytes and fibroblasts were randomly selected from a specimen bank of
keratinocyte/fibroblast strains derived from excised keloid specimens. All patients had
received no previous treatment for the keloids before surgical excision. A full history was
taken and an examination was performed, complete with coloured slide photographic
documentation, before taking informed consent prior to excision. Approval by the
National University of Singapore (NUS) Institutional Review Board (NUS-IRB) was
sought before excision of human tissue and collection of cells.
3.2.2 Keratinocyte culture from keloid scar and normal skin
Excised keloid scar and normal skin specimens were repetitively washed in PBS
containing 150 μg/ml gentamicin and 7.5 μg fungizone, until the washing solution
became clear. The tissue was then divided into pieces of approximately 5mm × 10mm
and the epidermis was scored. Dispase 5mg/ml in HBSS was added and skin was
incubated overnight at 4ºC. The epidermis was carefully scraped off with a scalpel the
next day and placed in trypsin 0.25%/Glucose 0.1%/EDTA 0.02% for 10 min in the
incubator. Trypsin action was quenched by DMEM/10% FCS. The suspended cells were
transferred into tubes and centrifuged at 1000 rpm for 8 min. The cells were seeded in
Keratinocyte Culture Medium (80 ml DMEM supplemented with 20 ml FCS, EGF 10
ng/ml, cholera toxin 1 × 10-9 M and hydrocortisone 0.4 μg/ml) at 1 × 105 cells/cm2 for 24
hrs before changing to Keratinocyte Growth Medium (KGM). The cell strains were
39
maintained and stored at -150ºC. Only cells from second and third passages were used for
the experiments.
3.2.3 Fibroblast culture from keloid scar and normal skin
Remnant dermis from the keloid scar and normal skin were either minced or incubated in
a solution of collagenase type 1 (0.5 mg/ml) and trypsin (0.2 mg/ml) at 37ºC for 6 hrs.
Cells were pelleted and grown in tissue culture flasks. Alternatively the skin tissue
samples were chopped into pieces of 1-2 mm2. The pieces were then transferred to a
100mm tissue culture dish previously coated with a thin layer of DMEM/10%FCS.
Culture medium enough to cover the explants were then added and topped up after 2-3
days. After 4-7 days the fibroblasts outgrew from the tissue. Fibroblast cell strains were
maintained and stored at -150ºC until use. Only cells from the second and third passages
were used for the experiments.
3.2.4 Cell counting
Before cells were seeded into culture flasks for experiments, aliquots of the cell
suspension were mixed with trypan blue in a ratio of 1:4 and counted in a Neubauer‟s
haemocytometer. Non-viable cells will be stained blue while viable cells remain opaque.
Viable cells in the four corner squares of the heamocytometer were counted. Since the
volume of each square is 10-4 cm3 the following formula can be used to calculate the
number of cells in the cell suspension:
Cells per ml
= the average count per square x 4 (dilution factor) x 104
40
Total cell number
= cells per ml x the original volume of fluid from which cell
sample was removed
3.3 HDGF experiments
3.3.1 Immunohistochemistry
Paraffin sections of keloid and normal tissue were dewaxed or deparaffinized in two
changes of xylene followed by re-hyderation in 100%, 95% and 70% ethanol gradient.
Antigens were then retrieved by immersing the slides in 0.01 M citrate buffer, pH 6.0,
heating in a microwave oven (high for 2.5 min., low for 5 min.), cooling at 4°C for 20
min. and washing in water for 5 min. Endogenous peroxidase was blocked in 3% H2O2
and non-specific binding was blocked for 1 hr (CAS block; Zymed Laboratories, South
San Francisco, CA, USA). The sections were incubated with antibodies specific for
HDGF, diluted 1:1000 for 1 hr. After washing, the slides were incubated in anti-mouse
IgG-peroxidase (Zymed) or anti-rabbit IgG-peroxidase (Zymed), diluted 1:500 for 2 hrs,
for HDGF primary antibodies, respectively. The slides were washed in Tris-buffered
NaCl (TBS) or 0.05% Tween-20, pH 7.5, and then with MilliQ H2O (Millipore Corp,
Billerica, MA, USA). The reaction product was developed with 3,3‟-diaminobenzidine
tetrahydrochloride substrate kit (Zymed), and the sections were counterstained with
hematoxylin. All wash steps were carried out in TBS/0.05% Tween-20. The antibodies
were diluted in 1% bovine serum albumin (BSA)/TBS. Non-immune mouse/rabbit
antibody of the appropriate immunoglobulin isotype was used for negative controls.
41
3.3.2 Serum stimulation of fibroblasts
Fibroblasts were seeded in six-well plates at a density of 1 x 104 cells/ml in 10% FCS for
24 hrs and subsequently starved in a serum-free medium for another 48 hrs. After 48 hrs,
the fibroblasts were stimulated by exposure to either 10% FCS or DMEM for 5 days
before being harvested.
3.3.3 Keratinocyte-fibroblast co-culture
Keloid keratinocytes (KK) and normal keratinocytes (NK) obtained from randomly
selected keloid and normal strains were seeded at a density of 1 x 105 cells/cm2 on
Transwell clear polyester membrane inserts with 0.4-µm pore size and 4.5 cm2 (Corning
Incorporated Life Sciences, Acton, MA, USA) area. The cells were maintained for 4 days
in EpiLife medium (Cascade Biologics, OR, USA) until 100% confluent. The medium
was then changed to EpiLife supplemented with increased calcium concentration and the
cells were exposed to the air–liquid interface for another 3 days, allowing the
keratinocytes to stratify and reach terminal differentiation. Keloid fibroblasts (KF) and
normal fibroblasts (NF) were seeded in six well plates at a density of 1 x 105 cells/well in
DMEM/10% FCS for 48 hrs to 80% confluency. Keratinocytes on membrane inserts and
plated fibroblasts were washed twice with PBS before the inserts were placed into the
six-well plates containing fibroblast cultures to initiate KK/KF or NK/NF co-cultures in
fresh serum-free DMEM. Whole-cell extracts and conditioned media were harvested and
analysed separately.
42
Keratinocyte subculture
Fibroblast subculture
Keratinocyte- Fibroblast co-culture
Figure 3.1: Co-culture of epidermal keratinocytes and dermal fibroblasts as an in vitro model to
study epithelial-mesenchymal interactions. Figure courtesy of Dr Anandaroop Mukhopadhyay.
3.3.4 Treatment of fibroblasts with HDGF
KF and NF cells were seeded in 6-well plates at a density of 1 × 104 cells/ml in
DMEM/10% FCS for 24 hrs and then in serum-free medium for another 48 hrs.. The cells
were subsequently treated with 250 ng/ml of recombinant HDGF. Cells without treatment
were used as controls. Whole-cell extracts and conditioned media were harvested at
different time points and subjected to Western blot analysis for different molecular
targets.
3.3.5 Treatment of keloid co-cultures with inhibitors
Keloid co-cultures were established as previously mentioned. The cells were
subsequently treated with varying concentrations of various inhibitors: Rapamycin (0.01
and 2 µM; Calbiochem, CA), WP631 (0.05, 0.1, and 0.2 µM; Calbiochem) and
43
Mitoxantrone (0.05, 0.1 and 0.2 µM; Calbiochem) to investigate their effects on HDGF
expression. Cells without inhibitor treatment were used as controls. After 24 h and 72 h,
conditioned media from the co-cultures was collected and subjected to Western blot
analysis for HDGF expression.
3.3.6 Smad-null and Smad-overexpression cell assay
This assay was used to study the effect of Smad1, 2 and 3 on HDGF expression. Mouse
embryo fibroblasts of wild-type (MEF-wt), Smad 1-/- (S1-/-), Smad 1+/+ (S1+/+), Smad 2−/−
(S2−/−) and Smad 3−/− (S3−/−) were kindly provided by Dr Rik Derynck, UCSF. Cells were
seeded in six-well plates for 24 h, followed by serum-free DMEM starvation for 48 h
before co-culture with keloid keratinocytes. Whole cell extracts and conditioned media
were assayed for HDGF by Western blot.
3.3.7 MTT assay
The MTT assay is a commonly used colorimetric assay to quantify cell numbers. It is
widely used in studies involving cell proliferation or cell toxicity. MTT [3-(4,5Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium] is cleaved by an active succinatetetrazolium reductase system present in the mitochondrial respiratory chain of a living
and metabolizing cell into blue formazon crystals which can be solubilized and their
absorbance measured. The relationship between cell number and absorbance is linear. In
our experimental set up, normal and keloid fibroblasts were seeded in 96 well plates. The
cells were divided into control and treatment groups and were incubated with 10 ul of
MTT (5mg/ml) in 100ul of DMEM to give a final concentration of 0.5mg/ml in each well
44
for about 2hrs. The medium was removed and the blue crystals were solubilized by
Hansen‟s method (Hansen, Nielsen & Berg 1989) using 20% w/v SDS in a solution of
DMF: water (1:1 v/v) and shaking in an orbital shaker. The absorbance of the solution
was then measured directly by using a plate reader at 570nm.
3.3.8 Western blotting
Frozen tissue specimens or cultured fibroblasts under different experimental conditions
were lysed in cell lysis buffer containing 20 mM Tris-HCl (pH 7.5), 1% Triton X-100,
100 mM NaCl, 0.5% Nonidet P-40 and 1 mg/ml protease inhibitor cocktail (Boehringer
Mannheim, Mannheim, Germany). This was followed by centrifugation at 13000 x g for
10 min. Supernatant was collected while the pellet was discarded. Protein concentration
of the tissue extracts were determined by Bradford method. Proteins were then subjected
to Western blot analysis. In total, 50 µg of whole-cell extract was separated by 14% or
8% SDS-PAGE under reducing conditions and electroblotted onto a nitrocellulose
membrane. Blots were incubated with numerous antibodies including mouse and rabbit
anti-HDGF (a gift from Dr Ren Henning, MD Anderson Cancer Centre and Dr H
Nakamura, Hyogo College of Medicine, Japan), mouse anti-PCNA (Santa Cruz, CA,
USA), mouse anti-VEGF, anti-connective tissue growth factor (CTGF), rabbit antip44/p42 mitogen-activated protein kinase (MAPK), mouse anti-phospho p44/p42 MAPK,
rabbit anti-Akt, rabbit anti-phospho Akt (Ser473) (Cell Signaling Technology Inc, USA),
mouse anti-collagen I (Monosan), mouse anti-α-smooth muscle actin (SMA) (Sigma) and
mouse anti-fibronectin (BD Transduction Laboratories). The blots were visualized with a
chemiluminescence-based photoblot system (Amersham Biosciences, Buckinghamshire,
45
UK). For the analysis involving conditioned media, 4 ml of the conditioned media was
concentrated using a Centricon centrifuge (Millipore Corp., MA, USA) and then
subjected to Western blotting.
3.3.9 Quantification of Western blot and statistical analysis
A Bio-Rad gel scanner and densitometer program (Gel-Pro Analyzer ver. 4.5;
MediaCybernetics, Bethesda, MD, USA) was utilized to assess concentrations of the
bands obtained by Western blots. These were measured as total density units. The paired
Student‟s t-test or the Welch‟s t-test was used for all analyses where appropriate. A value
of P < 0.05 was considered to be statistically significant. The error bars denote the
standard error of the mean (SEM). All statistical analyses were done using Microsoft
Excel 2003 (Redmond, WA, USA).
3.4 Microarray experiments
3.4.1 Cell culture
Three different keloid fibroblast samples and three different normal fibroblast samples
that were previously maintained and stored at -150ºC were thawed and used for the
experiments. Fibroblasts were seeded in 15cm dishes at a density of 1 x 104 cells/ml in
10% FCS until confluency and subsequently starved in a serum-free medium for 48 hrs.
After 48 hrs, the serum free medium was replaced and fibroblasts were harvested after
another 24 hrs (day 1), 72 hrs (day 3) and 120 hrs (day 5). Cells were grown and
46
processed in three batches. Each batch consisted of one keloid and one normal sample
harvested at the three different time points. KF1, NF1, KF2 and NF2 were samples from
different patients while KF3 and NF3 were samples from the same patient.
3.4.2 RNA extraction
RNA was extracted using the RNeasy-kit (Qiagen, Hilden, Germany) according to the
manufacturer‟s protocol. Cell culture medium was completely aspirated and 1 ml of
Buffer RLT was added directly to each plate. Plates were scraped and the cell lysate
collected with a cell scraper. Lysate was collected into a microcentrifuge tube and
vortexed for 10 seconds. Lysate was then passed 5 times through a blunt 20-gauge needle
fitted to an RNAse free syringe. 70% ethanol was added to the homogenized lysate in a
1:1 ratio and mixed by pipetting. 700 µl of the sample, including any precipitate that may
have formed, was transferred to an RNeasy spin column placed in a collection tube and
centrifuged for 15 s at ≥ 8000 x g. Successive aliquots of any excess of the sample were
centrifuged in the same RNeasy spin column. Flow-through was discarded after each
centrifugation. After all the sample has been loaded, 700 µl of Buffer RW1 was added to
the RNeasy spin column and centrifuged for 15 s at ≥ 8000 x g to wash the spin column
membrane. Flow-through was discarded and 500 µl of Buffer RPE was added to the
RNeasy spin column and centrifuged for 15 s at ≥ 8000 x g. Flow-through was discarded
and this step was repeated for a longer centrifugation time of 2 min. RNeasy spin column
was then placed in a new 1.5 ml collection tube and 50 µl of RNase free water was added
directly to the membrane. The column was centrifuged for 1 min at ≥ 8000 x g to elute
the RNA. This step was repeated using another 50 µl of RNase free water. All steps were
47
performed at room temperature. Purified RNA was quantified by UV absorbance at 260
and 280 nm on a ND1000 spectrophotometer (NanodropTM, ThermoScientific).
3.4.3 cRNA preparation and labeling
Labeled complementary RNA (cRNA) was produced from 15 µg of total RNA using the
GeneChip One-Cycle Eukaryotic Target Labeling and Control Reagents (Affymetrix,
Santa Clara, USA) according to the manufacturer‟s protocol. Briefly, 15 μg of total RNA
was first reverse transcribed using a T7-Oligo(dT) Promoter Primer in the first-strand
cDNA synthesis reaction. Following RNase H-mediated second-strand cDNA synthesis,
the double-stranded cDNA was purified using the cDNA Cleanup Spin Column and
served as a template in the subsequent in vitro transcription (IVT) reaction. The IVT
reaction was carried out in the presence of T7 RNA Polymerase and a biotinylated
nucleotide analog/ribonucleotide mix for cRNA amplification and biotin labeling. Biotin
labeled cRNA was then purified using the IVT cRNA Cleanup Spin Column, quantified
by UV absorbance at 260 nm on the ND1000 spectrophotometer and fragmented using
the Affymetrix Fragmentation Buffer. All reagents were obtained from Affymetrix.
3.4.4 Affymetrix chip hybridization and scanning
Fragmented cRNA was then hybridized to preequilibrated Affymetrix GeneChip U133A
arrays at 45 °C for 15 hours. The cocktails were removed after hybridization and the
chips were washed and stained using Affymetrix wash buffers and stain cocktails in an
automated fluidic station. The chips were then scanned in a Hewlett-Packard
ChipScanner (Affymetrix, Santa Clara, USA) to detect hybridization signals.
48
3.4.5 Data analysis
Following data collection, preliminary analysis and visualization was done using the
Affymetrix GeneChip Operating Software for an assessment of the quality of the data.
Further statistical analysis was done using Version 10.0.2 of the Genespring GX software
(Agilent, Palo Alto, CA). Normalization and summarization of arrays was done using
both the Microarray Suite (MAS) 5.0 algorithm (default Affymetrix approach;
Affymetrix Users Guide, www.affymetrix.com) and the Robust Multichip Analysis
(RMA) (Bolstad et al. 2003) approach. The MAS 5.0 algorithm is the most widely used
analysis method for GeneChips. The RMA algorithm is an alternative analysis procedure
that is more robust than MAS 5.0 for data with normal errors or long-tailed symmetric
errors. The data was first analyzed by Two-Way Analysis of Variance (ANOVA) to
assess the individual influence of time point (day 1, day 3 or day 5) and cell type (NF or
KF) on gene expression, as well as their net interactive effect. The Welch‟s t-test was
then used to identify genes that were significantly different in keloid compared to normal
fibroblasts. All statistical tests utilized the Benjamini Hochberg method to correct for
multiple testing. Genes that were significantly different (P < 0.05) under both MAS 5.0
summarization and RMA summarization were used for further analysis. Hierarchical
clustering and principal components analysis (PCA) were used to visually verify the
ability of the genes selected to distinguish between keloid and normal cells. Finally, the
list of genes that were significantly different was processed using the Database for
Annotation, Visualization and Integrated Discovery (DAVID) Gene Functional
Classification Tool (Dennis et al. 2003; Huang, Sherman & Lempicki 2009) to identify
biological function associated with these genes.
49
3.5 Reverse engineering
3.5.1 Preparation of additional microarray samples
Two different keloid and two different normal fibroblast samples (all from different
patients) were grown, serum-starved and harvested for RNA after day 1, day 3 and day 5
as in the previous study. In addition, one keloid fibroblast sample was grown, treated
with HDGF and harvested for RNA after 6 hours, day 1 and day 2. All RNA were reverse
transcribed, amplified and labeled using the GeneChip Two-Cycle Eukaryotic Target
Labeling and Control Reagants (Affymetrix, Santa Clara, USA) according to the
manufacturer‟s protocol. Unlike in the previous study where the One-Cycle protocol was
used, the Two-Cycle protocol allowed for smaller amounts of starting RNA at the
expense of longer running time in the form of a second reverse transcription step. As we
had also run out of Affymetrix Genechip U133A arrays, the cheaper and newer
Affymetrix Genechip U133 Plus 2.0 arrays were used to hybridize to the labeled cRNA.
The chips were scanned in a Hewlett-Packard ChipScanner (Affymetrix, Santa Clara,
USA) to detect hybridization signals. Raw microarray data in the form of .CELS files
from Smith et al‟s experiments were also downloaded from the GEO database (Smith et
al. 2008).
3.5.2 Data preprocessing
Following data collection, RMA and MAS 5.0 normalization and summarization were
done using the R Bioconductor package. The four different datasets (original dataset
using U133A arrays, new dataset using U133 Plus 2.0 arrays, HDGF dataset using U133
50
Plus 2.0 arrays and Smith‟s dataset using U133 Plus 2.0 arrays) were normalized and
summarized independently. Two different custom Chip Definition Files (CDF) were used
(Dai et al. 2005). The first CDF was based on the Ensembl Gene database for analysis
with fREDUCE as it is easy to obtain the upstream sequence which is required by
fREDUCE from the Ensembl database. The second was based on the Entrez Gene
database for influence based reverse engineering methods such as BANJO and ARACNE
as these probe mappings allow one to ignore any differential signal due to multiple
probesets and gives a single value for a given gene. In addition, two lists were produced.
In the first list, no filtering was done while in the second list, 25% of the lowly expressed
genes were filtered.
3.5.3 Application of the fREDUCE algorithm
Human genomic sequences 1000 base pairs upstream from the transcriptional start site if
known, or from the initiation codon, were extracted from the Ensembl database (Curwen
et al. 2004). As fREDUCE requires only a single expression dataset and makes use of the
entire genomic dataset (both signal and background), the datasets were compared as
follows: A: Keloid versus normal fibroblasts under serum starvation conditions (only
KF1, KF2, NF1 and NF2 were used to keep the number of samples close to the other
conditions), B: Keloid versus normal fibroblasts under serum conditions, C: Keloid
treated with steroid versus serum induced keloid fibroblasts, D: Normal treated with
steroid versus serum induced normal fibroblasts, E: Keloid versus normal fibroblasts both
treated with steroid and F: Keloid treated with HDGF versus untreated keloid fibroblasts.
The expression value for each gene is represented as the following t-statistic:
51
t 
g
e g  c g
g
Vare
Varc

ne
nc
g
where g is the index over genes, µeg is the mean value of gene g under our condition of
interest, µcg os the mean value of gene g under control conditions, Vareg is the variance of
gene g under our condition of interest, Varcg is the variance of gene g under control
conditions, and ne and nc are the number of samples under our condition of interest and
under control conditions respectively. This statisitic is similar to the z-statistic used by
the fREDUCE creators (Wu et al. 2007). We then ran fREDUCE on the t-statistic for
RMA normalized and MAS 5.0 normalized as well as unfiltered and filtered gene lists on
the basis that a higher t-statistic translates to higher expression. Four different sets of
parameters were run on each replicate: length 6 with 0 IUPAC substitutions, length 6
with 1 IUPAC substitution, length 7 with 0 IUPAC substitutions and length 7 with 1
IUPAC substitution. Top and consistent binding sequences obtained from fREDUCE
above were then searched through the TRANSFAC database (Matys et al. 2003) for
possible gene targets and their corresponding transcription factors. Only gene targets
identified from Homo sapiens were collected, and binding sites for all these targets were
reconfirmed to be located within the 1000 base pair upstream sequences collected from
the Ensemble database previously.
3.5.4 Pathways selected for influence approach
KEGG pathways that were enriched from the previous study were used for the influence
approach. These were the antigen presentation and processing pathway, cytokinecytokine receptor interaction and toll-like receptor signaling pathway. The ribosome
52
pathway was not used as it would be a fully connected graph. Genes that were used as
nodes for modeling were chosen on the basis that there is only one gene representing that
particular node, all other genes will be assumed to be hidden nodes. The following 5
pathways in Figure 3.2 were eventually selected for the influence approach.
A
B
CXCL1
CXCL9
CXCL2
CXCR3
CXCL10
CXCL3
CXCL11
CXCL5
IL8RB
C
CXCL7
CXCL6
CREB
B2M
CIITA
Ii
IL8RA
IL8
E
D
TNFA
IL1B
TLR1
TLR2
RAC1
NFKB
IL6
NFKB
IL8
IKBA
RANTES
F
TLR3
TRIF
TRAF3
TBK1
IRF3
53
G
IP-10
IRF3
MIG
STAT1
I-TAC
Figure 3.2: KEGG pathways used for the influence approach. (A and B) Pathways taken from the
cytokine-cytokine receptor interaction map. (C) Transcriptional pathway taken from antigen
processing and presentation map. (D, E, F and G) Pathways taken from the toll-like receptor
signaling map. Pathways were also chosen such that A and B represent cytokine receptor
interactions, C, E and G represent transcriptional networks and D and F represent intracellular
signaling.
3.5.5 Application of the ARACNE and BANJO algorithms
The full set of data was used for the influence approach. To enable comparison between
the different data sets, gene expression for all the relevant nodes were normalized using
the average of GAPDH and B-actin expression. GAPDH and B-actin were first plotted to
determine their correlation and outliers were removed from the dataset. Three keloid
experiments from the serum starvation U133A dataset did not meet these criteria and was
removed giving a total of 28 keloid experiments and 24 normal experiments. We ran
ARACNE and BANJO on the keloid and normal inputs separately, and also on the MAS
5 and RMA normalized expression values separately. All parameters were left at their
default values. For ARACNE, kernel width and number of bins were automatically
detected by the software while DPI tolerance to remove false positives was set at 0.15.
For BANJO, the Proposer/Searcher strategies were chosen as random local move and
simulated annealing, respectively, and the amount of time BANJO uses to explore the
Bayesian Network space was set to one minute. All the other parameters such as
reannealingTemperature, coolingFactor, and so on, were left with their default values.
54
Parameter values were selected as best values (in terms of network inference accuracy) as
shown by Bansal et al (Bansal et al. 2007). In order to estimate the joint probability
distribution of all variables in the network, BANJO requires discrete data. The data was
therefore discretized into 7 discrete states using the quantile discretization procedure in
the software. Furthermore, as the simulated annealing algorithm in BANJO does not
guarantee a global maximum, the runs were repeated three times and the result with the
highest maximum score was taken.
3.5.6 Estimation of the performance of the algorithms
In order to assess the inference performances we computed the Positive Predicted Value
(PPV) and the Sensitivity scores as described by Bansal et al (Bansal et al. 2007). The
following definitions were used:
TP = Number of True Positives = number of edges in the real network that are
correctly inferred;
FP = Number of False Positives = number of inferred edges that are not in the
real network;
FN = Number of False Negatives = number of edges in the real network that are
not inferred.
The following were then computed:
PPV 
TP
TP  FP
Sensitivit y 
TP
TP  FN
In order to compute the random PPV we considered the expected value of a
55
hypergeometrically distributed random variable whose distribution function and expected
value are, respectively:
Px 
M
CX
E[ x]  M
N M
N
Cn x
Cx
N 1
N
C n 1
n
M
N
Cn
where N = number of possible edges in the network, M = number of true edges and n =
number of predicted edges. Then,
PPV rand 
TPrand
E[ x] M


TP  FP
n
N
All statistical tests are done using the one tailed paired t-test.
56
CHAPTER FOUR
THE ROLE OF HEPATOMA-DERIVED GROWTH FACTOR IN KELOID
PATHOGENESIS
4.1 Introduction
Hepatoma-derived growth factor (HDGF) is a novel heparin-binding protein that was
originally purified from the conditioned media of HuH-7 hepatoma cells (Nakamura et al.
1989). This growth factor was found to have mitogenic activity for a wide variety of
cells, including fibroblasts (Abouzied et al. 2005), endothelial cells (Everett et al. 2004),
renal (Kishima et al. 2002) and lung epithelial cells (Mori et al. 2004), vascular smooth
muscle cells (Everett, Stoops & McNamara 2001) and fetal hepatocytes (Enomoto et al.
2002). A growing number of studies report a possible role of HDGF in the development
of different types of cancers. In particular, it has been implicated in esophageal cancer
(Matsuyama et al. 2001; Yamamoto et al. 2007), pancreatic cancer (Uyama et al. 2006),
hepatocellular carcinoma (Yoshida et al. 2003), melanoma (Bernard et al. 2003), lung
cancer (Ren et al. 2004) and gastric carcinoma (Chang et al. 2007).
HDGF is the first member of the HDGF family of proteins that was discovered to
contain a well conserved N-terminal amino acid sequence, which is called the
homologous to amino terminus of HDGF (HATH) region (Izumoto et al. 1997).
Subsequently, five related proteins have been identified, four of which are named HDGFrelated protein (HRP) -1 to HRP-4 and the fifth of which is named lens epithelium-
57
derived growth factor (LEDGF). Except for their growth factor activity, the functions of
these proteins are largely unknown (Abouzied et al. 2004).
Although the mitogenic effect of HDGF has been proven, the pathway by which it
exerts this proliferative activity is still unclear. Two different pathways have been
proposed. Despite lacking the secretory sequence present in most secretory proteins
(Nakamura et al. 1994), it has been shown that exogenous HDGF could possibly act by
binding to an as yet unknown cell surface receptor, triggering signaling events
downstream that result in increased proliferation (Abouzied et al. 2005). Others have
shown that nuclear localization is required for the mitogenic activity of HDGF (Everett,
Stoops & McNamara 2001; Kishima et al. 2002).
In terms of wound repair, HDGF has been found to be involved in lung
remodeling after injury by promoting the growth of lung epithelial cells (Mori et al.
2004). Among the growth factors responding to vascular wall injury, HDGF is unique, in
that it is not expressed in the vascular wall until injury occurs (Everett et al. 2000).
HDGF gene expression was also increased during retinal pigment epithelial wound repair
(Singh et al. 2001). Furthermore, HDGF has been found to be up-regulated in human
dermal fibroblasts subjected to mechanical stimulation from stressed collagen lattices
(Kessler et al. 2001). Abnormal scarring has been correlated to regions of the body with
higher mechanical force than others (Wang et al. 2006; Aarabi et al. 2007). Therefore,
these results, combined with the fact that HDGF has been implicated in the aberrant
growth of tumours, lead us to speculate that it could also play some role in the formation
of keloids. In unpublished data from microarray experiments done by my supervisor, Prof
58
Phan at Stanford University, HDGF was one of the genes found to be up-regulated in
keloid tissue.
In this study, we investigated the expression and localization of HDGF in vivo by
performing immunohistochemical staining (IHC) and Western blot analysis on keloid and
normal skin tissue. We further studied the expression of HDGF using in vitro models of
normal and keloid fibroblasts subjected to serum stimulation. To examine the effect of
epithelial-mesenchymal interactions on the expression of HDGF, we employed a two
chamber serum free system where keratinocytes on membrane inserts were co-cultured
with the fibroblasts.
In a second set of experiments, we examined the effect of exogenous recombinant
HDGF on the keloid and normal fibroblasts. Cells treated with recombinant HDGF were
assessed for increased proliferation by the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide] assay and by quantifying proliferating cell nuclear antigen
(PCNA) expression. Western blotting was also performed to identify some of the
downstream signaling targets of exogenously applied HDGF.
Finally, to identify some of the upstream signals regulating the expression of
HDGF, we investigated the effect of Sp1 and mammalian target of rapamycin (mTOR)
inhibitors on the secretion of HDGF from fibroblast keratinocyte co-cultures. The effect
of TGF-β signaling on HDGF expression was also examined by assaying Smad-null and
Smad-overexpressing mouse embryo fibroblasts.
59
4.2 Results
4.2.1 HDGF expression is increased in keloid scar dermis
Immunohistochemical labelling showed that HDGF was present in both the epidermis
(Fig. 4.1A) and the dermis (Fig. 4.1B) of normal and keloid tissue. Epidermal staining
intensity was irregular and sample-dependent, with some keloid samples exhibiting
stronger staining, while others exhibiting equal or weaker staining when compared with
their normal counterparts. However, HDGF expression in the dermis was found to be
higher in all keloid samples. This can be more clearly seen at a lower magnification (Fig.
4.1C). In the keloid tissues, almost the whole dermis was stained brown compared with a
significantly smaller area in normal skin. Western blot results from the keloid and normal
whole-tissue extracts reconfirmed these observations. The keloid tissue samples had a
significantly higher expression of HDGF compared with the normal tissue samples (P <
0.05; Fig. 4.2).
60
A
B
Normal skin
Keloid scar
Normal skin
Keloid scar
Normal skin
Keloid scar
C
Figure 4.1: Immunohistochemical staining of keloid and normal tissue for HDGF. Paraffin
sections of normal and keloid tissue were prepared and stained with antibodies against HDGF.
Pictures were taken with magnification at 40X (A, B) and 10X (C). The dermis and the epidermis
are represented by (D) and (E), respectively. In each panel, the inset shows the same tissue
labelled with a non-immune mouse antibody of the appropriate immunoglobulin isotype as a
negative control. HDGF was detected in both the epidermis (A) and the dermis (B) of normal and
keloid tissue. Increased expression was observed in the dermis of keloid tissue compared with the
dermis of normal tissue (B, C).
61
Keloid tissue
1
2
3
4
5
6
Normal tissue
7
8
9
10 11 12 13
HDGF
β-actin
* p = 0.0335 < 0.05
Mean total density (OD)
70000
47910
60000
50000
40000
30000
20000
10000
2112.5
0
Keloid tissue
Normal tissue
Figure 4.2: Western blot of keloid and normal whole tissue extract. In total, 50 µg of tissue
extracts from nine keloid tissue specimens and four normal skin specimens was subjected to
Western blot analysis with antibodies against HDGF. The whole-tissue extracts include both the
epidermis and the dermis. The blots were probed with anti-β-actin antibody to confirm equal
loading. The bar graph represents the mean ± SEM of HDGF levels in the normal and keloid
samples, as quantified by gel densitometry. *indicates statistical significance as assessed by
Welch‟s t-test.
4.2.2 Serum stimulation and epithelial-mesenchymal interactions had no effect on
intracellular HDGF expression
Western blot results indicated that treatment with serum had no significant effect on
intracellular HDGF expression in both normal fibroblasts (NF) and keloid fibroblasts
(KF) (Fig. 4.3). In addition, NFs co-cultured with normal keratinocytes (NK) and KFs cocultured with keloid keratinocytes (KK) did not show any significant difference in HDGF
expression levels when compared with the monocultured controls or when compared with
each other.
62
KF1
NF1
KF2
NF2
10% FCS
-
+
-
-
+
-
-
+
-
-
+
-
Co-culture
-
-
+
-
-
+
-
-
+
-
-
+
67
65
HDGF
β-actin
77
100
80
60
40
20
0
77
75
NF
NK
/
+
FC
S
NF
NF
KF
+
FC
S
KK
/K
F
60
KF
Normalized mean total
density (%)
p > 0.05
Figure 4.3: Effect of serum and epithelial–mesenchymal interactions on intracellular HDGF
expression. Six different strains of keloid/normal fibroblasts were cultured with DMEM, 10%
FCS or co-cultured with keloid/normal keratinocytes for 5 days. In total, 50 µg of total protein
extracts was subjected to Western blot analysis with HDGF antibodies. Two representative strains
are shown. The bar graphs represent the normalized mean ± S.E.M. of HDGF levels in the
different conditions. All blots were probed and normalized with β-actin.
4.2.3 Epithelial-mesenchymal interactions in keloid co-culture increased secretion of
HDGF
Conditioned media obtained when KFs were co-cultured with KKs showed a significant
increase in HDGF compared with conditioned media obtained from monocultured KFs
from day 1 to 5 (P < 0.05 for asterisks; Fig. 4.4A). In contrast, HDGF was undetected
from day 1 to 3, and a weak increase was only detected when NFs were cocultured with
NKs on day 5 (P < 0.05; Fig. 4.4B). Monocultured KFs showed a higher secretion
compared with monocultured NFs and keloid co-cultures showed a higher secretion
63
compared with normal skin cell co-cultures. Monocultured keratinocytes show a
moderately high secretion of HDGF but no significant difference was seen between NKs
and KKs harvested at day 5 (Fig. 4.4C).
*
A
KK/KF
KF
D1
D3 D5 D1 D3 D5
HDGF
Mean total density (OD)
140000
*
115381
120000
*
100000
87247.95
80000
49393.3
60000
30730.4
40000
20000
6696.735
2383.385
0
KF D1
KF D3
KF D5
KK/KF D1 KK/KF D3 KK/KF D5
*
B
14000
11750.2
NF
NK/NF
D1 D3 D5 D1 D3
D5
Mean total density
12000
HDGF
10000
8000
6000
4000
2000
0
0
0
NF D1
NF D3
NF D5
0
0
0
NK/NF D1 NK/NF D3 NK/NF D5
C
KK
HDGF
NK
Mean optical density (OD)
p = 0.57 > 0.05
120000
100000
76737
60040
80000
60000
40000
20000
0
KK
NK
Figure 4.4: Expression of HDGF in conditioned media of monocultured and co-cultured cells.
(A) Conditioned media of keloid fibroblasts monoculture (KF) and keloid fibroblasts cocultured
with keloid keratinocytes (KK/KF) were collected at days 1, 3 and 5. (B) Conditioned media of
normal fibroblast monoculture (NF) and normal fibroblast co-cultured with normal keratinocytes
(NK/NF) were collected at days 1, 3 and 5. Experiments were performed in duplicates. (C)
Conditioned media of seven samples of singly cultured keloid keratinocytes (KK) and normal
keratinocytes (NK) were collected at day 5. Four millilitres of the conditioned media from (A),
(B) and (C) was then concentrated and subjected to Western blot analysis with anti-HDGF
antibody. Representative figures are shown. The bar graphs represent the mean ± S.E.M. of
HDGF levels. * indicates statistical significance as determined by the paired t-test.
64
4.2.4 Increased keloid fibroblast proliferation upon stimulation with HDGF
There was no significant difference in NF proliferation when treated for 72 hrs with
various doses of HDGF (Fig. 4.5A). However, there was a significant dose-dependent
increase of up to ~13% in the proliferation of KFs (P < 0.05 for asterisks; Fig. 4.5B). In
addition, Western blot analysis after 48 hrs showed a significant increase of proliferating
cell nuclear antigen (PCNA) expression in treated KFs compared with untreated keloid
controls, but this effect was not seen in treated NFs compared with untreated normal
controls (P < 0.05; Fig. 4.6A). In total, 250 ng/ml of recombinant HDGF was used for
treatment of both KFs and NFs.
*
B
6
0.3
0
4
0.04
-2.1
2
-2.5
0
-2
-4
-6
-8
dmem
10ng/ml 50ng/ml 100ng/ml 300ng/ml
HDGF concentration
*
% change compared to control
% change compared to control
A
16
14
12
10
8
6
4
2
0
-2
-4
12.8
*
8
4.4
6.5
0
dmem 10ng/ml 50ng/ml 100ng/ml300ng/ml
HDGF concentration
Figure 4.5: Increased proliferation of keloid fibroblasts treated with recombinant HDGF.
Cultures of keloid or normal fibroblasts were grown until 50% confluence and then serum starved
for 48 hrs. The fibroblasts were then treated with HDGF (10, 50, 100 and 300 ng/ml) for 72 hrs
and then subjected to the MTT proliferation assay. Untreated samples were used as control. The
bar graph in (A) represents the mean proliferative response of treated normal fibroblasts as a
percentage of the control. The bar graph in (B) represents the mean proliferative response of
treated keloid fibroblasts as a percentage of the control. * indicates statistical significance
compared with DMEM control as assessed by Student‟s t-test.
65
4.2.5 Treatment of fibroblasts with HDGF activated the ERK pathway, increased
the secretion of VEGF, and decreased the secretion of collagen I
KFs treated with HDGF after 48 hrs showed a significant increase in the expression of
intracellular phospho-extracellular signal regulated kinase (ERK) 1/2 compared with
untreated keloid controls, but this increase was not seen in treated NFs compared with
untreated normal controls (P < 0.05; Fig. 4.6B). Intracellular expression (P < 0.05, Fig.
4.6D) and extracellular secretion (P < 0.05, Fig. 4.7A) of VEGF from both keloid and
normal fibroblasts were significantly increased upon treatment with HDGF. Secretion of
collagen I was downregulated in the conditioned media of treated keloid and normal
fibroblasts compared to untreated controls (p<0.05, Fig. 4.7B). Treatment with HDGF
did not produce any significant difference in the expression of intracellular α-SMA (Fig.
4.6E), extracellular fibronectin (Fig. 4.7C) and extracellular CTGF (Fig. 4.7D).
Furthermore, expression of p-Akt was undetectable in all samples (Fig. 4.6C). At earlier
time points, no significant increase in phospho-ERK was detected (Fig. 4.6F). In total,
β-actin
KF + HDGF
*
5000
4592
4500
Mean total density (OD)
PCNA
KF
NF
A
NF + HDGF
250 ng/ml of recombinant HDGF was used for treatment of both KFs and NFs.
4000
3500
3000
2500
2029
2000
1500
1087
1176
NF
NF + HDGF
1000
500
0
KF
KF + HDGF
66
KF + HDGF
*
Normalized p-ERK against ERK
expression (%)
KF
NF + HDGF
NF
B
p-ERK 1/2
Total
ERK 1/2
β-actin
25
21
20
15
10
3.4
3.15
5
0
0
p-Akt
Total Akt
Normalized p-Akt against Akt expression
(%)
KF + HDGF
KF
NF
C
NF + HDGF
NF
NF +
HDGF
KF
KF +
HDGF
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0
0
NF
NF + HDGF
KF
KF + HDGF
0
KF + HDGF
16046.7
*
18000
16000
14000
11508
9863.47
12000
10000
8000
6000
(OD)
β-actin
*
Mean total density (OD)
VEGF
KF
NF
D
NF + HDGF
β-actin
4004.38
4000
2000
0
NF
NF + HDGF
KF
KF + HDGF
67
KF + HDGF
18000
α-SMA
15159.7
15306
KF
KF + HDGF
16000
14000
12000
9963.38
8755.33
NF
NF + HDGF
10000
8000
6000
(OD)
Mean total density (OD)
KF
NF + HDGF
NF
E
4000
2000
β-actin
0
F
- control
+ HDGF treated
KF
1h
-
NF
6h
+
-
24h
+
-
1h
-
+
6h
-
+
24h
+
-
5.5
4.1
+
p-ERK 1/2
27.6
50
40
30
20
10
0
25.6
25.8
29.1
16.4
KF
KF
1h
h
F1
DG
H
+
6h
13.5
h
h
h
24
24
F6
KF
GF
DG
D
H
+
+H
KF
KF
KF
Normalized p-ERK against
ERK (%)
Normalized p-ERK against
ERK (%)
Total
ERK 1/2
25
20
15
10
5
0
15.6
16.8
h
1h
F1
NF
DG
H
+
NF
3
2.6
h
h
h
6h
24
24
F6
NF
NF
GF
DG
D
H
+
+H
NF
NF
Figure 4.6: Effect of HDGF on the expression of downstream intracellular targets. Normal
fibroblasts and keloid fibroblasts were treated with either DMEM or 250 ng/ml of recombinant
HDGF, harvested after 48 hrs and lysed for Western blot analysis, as described under
experimental procedures. Blots were incubated with anti-PCNA (A), anti-phospho-ERK 1/2 and
total ERK 1/2 (B), anti-phospho-Akt and total Akt (C), anti-VEGF (D) and anti-α-SMA (E)
antibodies The blots were also incubated with anti-β-actin antibody to confirm equal loading. In
another set of experiments, normal fibroblasts and keloid fibroblasts were treated with either
DMEM or HDGF and harvested after 1 hr, 6 hrs and 24 hrs for Western blot analysis. Blots were
incubated with anti-phospho-ERK 1/2 and total ERK 1/2 (F) antibodies. All experiments were
performed in duplicates. Representative figures are shown. The bar graphs represent the mean ±
S.E.M. of protein levels. Phospho-ERK 1/2 was normalized against total ERK 1/2 expression and
phosphor-Akt was normalized against total Akt expression. * indicates statistical significance as
determined by the paired t-test.
68
A
VEGF
10000
Mean total density (OD)
KF + HDGF
KF
NF
NF + HDGF
*
7000
6000
5000
4000
3000
2000
345
271
NF +
HDGF
*
40000
30000
(OD)
16986.3
15761.9
20000
9889.51
10000
KF + HDGF
0
160000
141792
NF + HDGF
KF
KF + HDGF
129044
140000
120000
95389
78985
100000
80000
60000
40000
(OD)
Mean total density (OD)
KF
NF + HDGF
NF
Fibronectin
KF +
HDGF
44471.4
50000
NF
C
KF
*
Mean total density (OD)
KF + HDGF
KF
NF
NF + HDGF
NF
Collagen I
9011
8788
9000
8000
1000
0
B
*
20000
0
NF
NF +
HDGF
6000
D
KF
5386
KF +
HDGF
5618
4000
3021
3000
2580
2000
(OD)
CTGF
Mean total density (OD)
5000
1000
0
NF
NF + HDGF
KF
KF + HDGF
Figure 4.7: Effect of HDGF on the expression of downstream extracellular targets. Normal
fibroblasts and keloid fibroblasts were treated with either DMEM or 250 ng/ml of recombinant
HDGF. After 48 h, four millilitres of conditioned media was concentrated and subjected to
Western blot analysis. Blots were then incubated with anti-VEGF (A), anti-collagen I (B), antifibronectin (C) and anti-CTGF (D) antibodies. All experiments were performed in duplicates.
Representative figures are shown. The bar graphs represent the mean ± S.E.M. of protein levels. *
indicates statistical significance as determined by the paired t-test.
69
4.2.6 Treatment with mTOR and Sp1 inhibitors did not significantly affect the
production of HDGF
Treatment with the mTOR inhibitor, Rapamycin and Sp1 inhibitors, Wp631 and
Mitoxantrone, for 72 hours did significantly affect the production of intracellular (Fig.
4.8A) as well as extracellular (Fig 4.8B) HDGF. Extracellular levels of HDGF appear to
decrease slightly upon treatment with both the mTOR and Sp1 inhibitors but the decrease
was not significant. Intracellular levels of HDGF appear to decrease slightly upon
Mean total density (OD)
+ Wp 0.2
+ Wp 0.1
+ Wp 0.05
+ Mtx 0.2
+ Mtx 0.1
+ Mtx 0.05
70000
+ Rap 2
KF CC
A
+ Rap 0.01
treatment with Sp1 inhibitors, but again this decrease was not significant.
60000
50000
57425
52249
55694.65
58983.85
49698.4
43248.35
53322.2
53349.9
46045.1
40000
30000
20000
10000
0
KF CC + Rap + Rap + Mtx + Mtx + Mtx + Wp + Wp + Wp
0.01
2
0.05
0.1
0.2
0.05
0.1
0.2
HDGF
+ Wp 0.2
+ Wp 0.1
+ Wp 0.05
+ Mtx 0.2
+ Mtx 0.1
50356.6
53083.45
49075.3
43486.05
47667.25
42736.8
44969.6
50000
42216.45
36842.6
40000
60000
Mean total density (OD)
HDGF
+ Rap
2
+ Mtx 0.05
KF CC
B
+ Rap 0.01
β-actin
30000
20000
10000
0
KF CC + Rap + Rap + Mtx + Mtx + Mtx + Wp + Wp + Wp
0.01
2
0.05
0.1
0.2
0.05
0.1
0.2
Figure 4.8: Effect of mTOR and Sp1 inhibitors on the expression of HDGF. mTOR inhibitor
(rapamycin) and Sp1 inhibitors (WP631 and mitoxanthrone) were added to the keloid coculture
for 72 h with indicated concentrations as described in Material and Methods. Untreated cocultures served as controls. (A) For analysis of intracellular HDGF production, KFs were
harvested and lysed for Western blot analysis, as described under experimental procedures. The
blots were also incubated with anti-β-actin antibody to confirm equal loading. (B) For analysis of
secreted HDGF levels, 4 ml of conditioned media was concentrated and subjected to Western blot
analysis with anti-HDGF antibody. All experiments were performed in duplicates. Representative
figures are shown. The bar graphs represent the mean ± S.E.M. of HDGF levels.
70
4.2.7 Knockout of Smad 2/3 signaling increases intracellular HDGF expression
while knockout of Smad 1 signaling increases extracellular HDGF expression
Smad 2-/- (P<0.05, Fig. 4.9C) and Smad 3-/- (P<0.05, Fig. 4.9D) mouse embryo fibroblast
cells had a significantly higher expression of intracellular HDGF compared to their
respective wild type cells. However, intracellular HDGF expression in Smad 1-/- (Fig.
4.9A) and Smad 1+/+ (Fig. 4.9B) cells was not significantly different from the wild type.
Conditioned media from monocultured Smad 1-/- cells had a significantly higher
expression of HDGF compared to the wild type cells and conversely, conditioned media
from monocultured Smad 1+/+ cells had a significantly lower expression of HDGF
compared to the wild type control (P<0.05 for asterisks, Fig. 4.10A). Secretion from
monocultured Smad 2-/- and Smad 3-/- cells does not appear to be significantly different
from wild type controls. Upon co-culturing with keloid keratinocytes, this effect was
abrogated and there was no significant difference in secretion from the Smad co-cultures
compared to the keloid co-culture control (Fig. 4.10B).
71
A
70000
WT (mef)
Mean total density (OD)
S1 -/-
HDGF
β-actin
60000
54944.35
56313.875
S1-/-
WT
50000
40000
30000
20000
10000
0
B
WT (mef)
60000
Mean total density (OD)
S1 +/+
HDGF
β-actin
49831.1
50000
43611.55
40000
30000
20000
10000
0
S1+/+
WT
*
C
70000
WT (mef)
Mean total density (OD)
S2 -/-
HDGF
β-actin
63212.8
60000
53990.26667
50000
40000
30000
20000
10000
0
S2-/-
D
S3 -/-
WT
*
WT (mef)
50000
β-actin
Mean total density (OD)
42812.96
HDGF
40000
31443.0825
30000
20000
10000
0
S3-/-
WT
Figure 4.9: Effect of Smad signaling on intracellular HDGF expression. (A) S1-/-, (B) S1+/+, (C)
S2-/- and (D) S3-/- cells were seeded in six-well plates for 24 h in 10% FCS. Medium was replaced
by serum-free DMEM for another 48 h before co-culturing with keloid keratinocytes. Cells were
harvested for Western blot analysis of HDGF. The bar graphs represent the mean ± S.E.M of
HDGF levels. All blots were probed with β-actin antibody to confirm equal loading. * indicates
statistical significance as assessed by the paired t-test.
72
HDGF
KK
KK
S3 -/- CC KK
KK
S2 -/- CC KK
S1 +/+ CC KK
S1 -/- CC KK
NCC
NCC
S3 -/- NCC
S2 -/- NCC
NCC
S1 +/+ NCC
NCC
S1 -/- NCC
WT NCC
WT CC KK
B
A
HDGF
*
*
100000
82174.5
100000
80000
71901.95
60000
40000
20000
26186.53333
16168.3
10708.77
Mean total density (OD)
Mean total density (OD)
120000
80000
60000
52181.15
57215.05
44370.75
40000
37322.315
20000
285
0
0
WT NCC
S1-/- NCC
S1+/+ NCC S2-/- NCC
S3-/- NCC
WT CC
S1-/- CC
S1+/+ CC
S2-/- CC
S3-/- CC
Figure 4.10: Effect of Smad signaling on extracellular HDGF expression. 4 ml of conditioned
media from S1-/-, S1+/+, S2-/- and S3-/- mouse embryo fibroblasts monoculture (A) and co-cultured
with keloid keratinocytes (B) was collected after 48 h. The samples were then concentrated and
subjected to Western blot analysis with anti-HDGF antibody. Experiments were performed in
duplicates. Representative figures are shown. The bar graphs represent the mean ± S.E.M. of
HDGF levels. * indicates statistical significance as determined by the paired t-test.
4.3 Discussion
HDGF has not been as well studied as other growth factors and cytokines, but there is a
growing wealth of information underlining its importance in cancer. In this paper, we
examined the hypothesis that HDGF could play a role in the dysregulation of wound
73
healing leading to keloid formation. Our data suggest that this growth factor is likely to
be only one of a myriad number of players involved in keloid pathogenesis
Immunohistochemical results from this study indicate that HDGF is quite highly
expressed in the epidermis of both keloid and normal skin tissue. However, in the dermal
layer, there is a higher expression of HDGF in keloid tissue compared with normal skin.
This result led us to focus our efforts on the fibroblasts, which are the major cell type in
the dermis. Preliminary in vitro results demonstrated a significant expression of both
intracellular and extracellular HDGF in monocultured epidermal keratinocytes. We were
unable to detect any significant difference in HDGF expression between KK and NK
samples grown in vitro. However, the very presence of HDGF in these cells suggested
that this growth factor played some as yet unknown role in epidermal biology that is
worth investigating.
The process of cutaneous wound healing can be arbitrarily divided into four
phases - hemostasis, inflammation, proliferation and remodeling. The serum stimulation
model is an in vitro model that can be used to determine the involvement of growth
factors or cytokines in the early stages of the wound healing phase. Fibroblasts interpret
the presence of serum as a physiological wounding signal and would respond to it as if it
had occurred in vivo (Iyer et al. 1999). Our data show no difference in HDGF expression
levels between serum-stimulated and non-serum-stimulated fibroblasts, suggesting the
lack of involvement of HDGF in the early phases of wound healing.
Recent studies from our group and others have shown the importance of paracrine
signalling via epithelial–mesenchymal interactions in keloid pathogenesis (Lim et al.
2001; Funayama et al. 2003). The co-culture experiments were used to assess the effect
74
of this paracrine signaling on HDGF expression. We found no significant difference in
intracellular HDGF expression in whole-cell extracts of co-cultured fibroblasts compared
with the singly cultured fibroblasts. However, conditioned media collected from the
keloid co-culture configuration had significantly higher levels of HDGF compared with
those collected from the singly cultured KFs. Furthermore, secretion of HDGF from the
keloid co-culture configuration was also higher than from normal skin cell co-culture
configuration. When singly cultured, the keratinocytes secreted fairly high amounts of
HDGF but there was no significant difference between KKs and NKs. This suggests that
the secretion of HDGF is somehow modulated by epithelial–mesenchymal interactions.
Taken together, our in vitro results suggest that intracellular HDGF may have
some role to play in the normal process of wound healing, as evident by its high basal
expression in both normal and keloid fibroblasts. However, intracellular HDGF levels
alone cannot account for the higher expression of HDGF that was observed in in vivo
samples of keloid tissue. It appears that the contribution of HDGF to keloid pathogenesis
lies mainly in its secreted form and this secretion is modulated by epithelialmesenchymal interactions. To date, two different pathways have been proposed to
account for the activity of HDGF, either by binding to a cell surface receptor and acting
as a secreted growth factor (Abouzied et al. 2005), or by translocating to the nucleus
where it acts as a nuclear transcription factor (Everett, Stoops & McNamara 2001;
Kishima et al. 2002). Our data suggests that in the formation of keloids, the secretory
pathway of HDGF is of greater importance. However, we are not able to completely rule
out nuclear translocation as it is possible that the increase in intracellular HDGF is too
slight to be detected by Western analysis.
75
HDGF has been reported to act as a potent exogenous mitogen for a wide variety
of cells, including 3T3 fibroblasts (Nakamura et al. 1989; Abouzied et al. 2005) and mice
dermal fibroblasts (Gallitzendoerfer et al. 2008). The susceptibility of KFs to other
mitogenic stimuli has also been previously established (Lim et al. 2001; Phan et al. 2003;
Xia et al. 2004). Therefore, it was not surprising that both our MTT and PCNA results
showed KFs having a better proliferative response to HDGF stimuli compared with their
normal skin counterparts.
Downstream molecular targets of HDGF that were found to be up-regulated after
48 hrs include ERK and VEGF, while those found to be unaffected include alpha-smooth
muscle actin (α-SMA), fibronectin and connective tissue growth factor (CTGF). We were
however unable to detect any increase in ERK phosphorylation at the earlier time points
of 1 hr, 6 hrs and 24 hrs. HDGF has previously been found to induce the phosphorylation
of ERK in human pulmonary endothelial cells (Everett et al. 2004) and gastric epithelial
cells (Mao et al. 2008). However, in 3T3 fibroblasts, it has been reported that
extracellular HDGF does not enter the cell but instead binds to the cell membrane.
Furthermore, it stimulates proliferation but does not activate the ERK signalling pathway
(Abouzied et al. 2005). Our findings that there is no increase in intracellular HDGF and
that extracellular HDGF has mitogenic activity without activating ERK at early time
points suggest that this may also be the case in KFs. However, while the ERK pathway
may not be directly activated by HDGF, it does seem to be somehow involved
downstream of HDGF stimulation. In addition to the ERK pathway, we also tested the
Akt pathway but were unable to detect any phosphorylation of Akt. This is consistent
with the finding in human pulmonary endothelial cells, where no phosphorylation of Akt
76
was detected as well (Everett et al. 2004). Thus, the pathway by which HDGF exerts its
proliferative activity remains as elusive as ever.
The induction of VEGF by HDGF has also been shown previously in the process
of tumourigenesis (Okuda et al. 2003). In the study conducted by Okuda et al., NIH3T3
fibroblasts overexpressing HDGF were found to induce sarcomatous tumours after
injection into nude mice, and the tumour formation was induced mainly by angiogenesis
due to induction of VEGF. We report a similar induction of VEGF by HDGF in both
keloid and normal primary skin fibroblasts, suggesting that HDGF could play some role
in the normal angiogenic process during wound healing. However, in our in vitro coculture models, we observed very little secretion of HDGF in the normal co-culture
experiments. Therefore, while HDGF might induce production of VEGF in both NFs and
KFs, the absence of HDGF in the normal condition limits the production of VEGF
through this mechanism. This result tallies with previous findings from our group. In the
study conducted by Ong et al., keloid tissue was shown to have a higher expression of
VEGF compared with normal tissue, and there was also a significant increase in VEGF
from keloid co-culture compared with normal co-culture conditions (Ong et al. 2007).
These results suggest that the production of VEGF could be tied to the presence or
absence of HDGF.
Interestingly, secretion of collagen I was observed to be downregulated when
HDGF was exogenously applied to the fibroblast cells. Furthermore, there was no
significant difference in α-SMA, fibronectin or CTGF expression between the HDGF
induced cells and the controls. Our results suggest that HDGF induces only a mitogenic
response in the fibroblasts and does not participate in their differentiation into
77
myofibroblasts (of which α-SMA is a marker) or in ECM (collagen I and fibronectin)
production. Inhibition of CTGF has been found to cause a significant reduction in the
number of myofibroblasts in scars and also decreased transcription of types I and III
collagen (Sisco et al. 2008). Therefore, the lack of effect of HDGF on CTGF is consonant
with this hypothesis. Furthermore, Grotendorst et al. have previously reported that
fibroblast cells treated with TGF-β that are proliferating do not express α-SMA or
elevated levels of collagen synthesis (Grotendorst, Rahmanie & Duncan 2004).
Conversely, cells expressing α-SMA do not exhibit DNA synthesis but coexpress higher
levels of types I and III collagen mRNA. The authors concluded that these responses to
TGF-β are mutually exclusive and are controlled by combinatorial signaling pathways
involving not only components of the TGF-β pathway, but also signaling events induced
by other growth factors. HDGF appears to be one of the growth factors involved in
eliciting a mitogenic response from the fibroblasts, but on its own, it may also cause the
downregulation of certain ECM components. It should be noted that these features of
HDGF show some similarity to the insulin-like growth factor I, a potent mitogen that
requires the synergistic effect of TGF-β for ECM protein production (Daian et al. 2003;
Phan et al. 2003).
The Sp1 transcription factor is known to regulate several ECM promoters, and our
group has found that the Sp1 inhibitors Wp631 and mitoxanthrone were able to reduce
the expression of ECM components in KF as well as to inhibit its proliferation
(Mukhopadhyay et al. 2007). mTOR, on the other hand, is a serine/theronine kinase
which has been shown to regulate collagen type I expression via a phosphatidylinositol 3kinase (PI3-K)-independent pathway in human dermal fibroblasts (Shegogue &
78
Trojanowska 2004). The mTOR pathway inhibitor rapamycin is a naturally occurring
antibiotic that has been shown to downregulates the expression of cytoplasmic PCNA,
fibronectin, collagen and a-SMA (Ong et al. 2007). However, we found no significant
effect of both the Sp1 inhibitors as well as the mTOR inhibitor on the production of
HDGF.
Smad proteins are downstream signaling targets of the TGF-β family of growth
factors (Massagué 1998; Derynck & Zhang 2003). Once activated by TGF-β receptors,
Smad 2/3 oligomerizes with Smad 4, and the hetero-oligomeric Smad 2/3-Smad4
complex subsequently translocates from the cytoplasm into the nucleus where it activates
collagen gene transcription (Ghosh et al. 2000). Increased Smad 3 signaling has been
observed in different fibrotic disorders, including keloids (Chin et al. 2001; Phan et al.
2005). Our results show that the suppression of pro-fibrotic TGF-β mediated Smad 2/3
signaling results in an increased basal expression of cytoplasmic HDGF. This is
consistent with our finding that HDGF reduces ECM protein production. We also found a
significant increase in secreted HDGF from monocultured Smad 1-null cells, and a
corresponding decrease in secreted HDGF from Smad 1-overexpressing cells. Very little
is known about the role of Smad 1 in the biology of dermal fibroblasts. However, in lung
fibroblasts, BMP4 mediated Smad 1 signaling has been found to inhibit proliferation and
promote differentiation of the lung fibroblasts (Jeffery et al. 2005). This anti-proliferative
effect of Smad 1 signaling could result in the downregulation of HDGF secretion. Further
investigation into the role of Smad 1 signaling in dermal fibroblasts has to be done to
validate this hypothesis.
79
Figure 4.11 summarizes the main findings of our study. First, we have shown that
in the keloid condition, HDGF acts mainly as a secreted growth factor that is modulated
by epithelial–mesenchymal interactions. Second, exogenous HDGF exerts a proliferative
effect on KFs and is likely to be indirectly involved in ERK signalling. Finally, the
presence of HDGF increases the production of VEGF and indirectly contributes to the
process of angiogenesis. However, it has to be noted that on its own, HDGF appears to
downregulate collagen production and has no effect on fibronectin, α-SMA and CTGF.
Therefore, it is the combination of and interplay between various molecular factors that
ultimately decides the fate of the wound healing process.
Keloid keratinocytes
Epithelial mesenchymal interactions
HDGF
HDGF receptor (?)
p-ERK 1/2
VEGF
Increased proliferation
and vascularization
Keloid fibroblast
Figure 4.11: Schematic representation of the role of HDGF in keloid pathogenesis. Epithelial–
mesenchymal interactions result in an increased secretion of HDGF in keloids. Overproduction of
extracellular HDGF leads to the phosphorylation of ERK 1/2 and increased proliferation of keloid
fibroblasts, most likely through a receptor-mediated pathway. HDGF also stimulates the
fibroblasts to produce VEGF.
80
CHAPTER FIVE
GENOME WIDE TRANSCRIPTIONAL PROFILING OF SERUM
STARVED KELOID AND NORMAL FIBROBLASTS
5.1 Introduction
Since their first reported use in the mid-1990s, microarray technology has been adopted
rapidly within the research community and it is now a standard technique in a molecular
biologist‟s toolbox. The appeal of this technology can be easily understood; with
microarray technology, thousands of genes can be measured simultaneously, giving
researchers a peek into the transcriptional profile of a cell. It has to be noted however,
that unlike in the previous section where protein expression was measured, microarrays
typically measure gene expression levels, specifically messenger RNA (mRNA) levels.
mRNA has to be translated into protein before they become functional, hence requiring a
further processing step that cannot be elucidated using this techonology.
The principle of a microarray experiment is that mRNA from a given cell line or
tissue is used to generate a labelled sample, sometimes termed as the „target‟, which is
hybridized in parallel to a large number of DNA sequences immobilized on a solid
surface in an ordered array (Schena et al. 1995). Although many different microarray
systems have been developed by academic groups and commercial suppliers, the most
commonly used systems today can be divided into two groups, according to the arrayed
material: complementary DNA (cDNA) and oligonucleotide microarrays. The arrayed
material has generally been termed the probe since it is equivalent to the probe used in a
81
northern blot analysis. Probes for cDNA arrays are usually products of the polymerase
chain reaction (PCR) generated from cDNA libraries or clone collections, using either
vector-specific or gene-specific primers, and are printed onto glass slides or nylon
membranes as spots at defined locations. For oligonucleotide arrays, short 20–25mers are
synthesized in situ, either by photolithography onto silicon wafers (high-densityoligonucleotide arrays from Affymetrix) or by ink-jet technology (licensed to Agilent
Technologies) (Schulze & Downward 2001).
Methods based on synthetic oligonucleotides offer the advantage that because
sequence information alone is sufficient to generate the DNA to be arrayed, no timeconsuming handling of cDNA resources is required. Another important difference
between high-density oligonucleotide arrays and spotted arrays lies in the fact that the
high reproducibility of in situ synthesis of oligonucleotide chips allows accurate
comparison of signals generated by samples hybridized to separate arrays. In the case of
spotted arrays, the process of gridding is not accurate enough to allow comparison
between different arrays (Schulze & Downward 2001). However, oligonucleotide chips
are also more expensive compared to their cDNA counterparts.
Comparing between the different oligonucelotide chips, Agilent arrays typically
have a single spot per gene (single probe measurement), whereas Affymetrix arrays
provide multiple measurements: a series of independent or semi-independent
oligonucleotides (the probe set) query each RNA in solution. Affymetrix probe sets are
constructed from a series of perfect-match and paired-mismatch oligonucleotides,
allowing some assessment of non-specific binding and performance of the probes.
Overall, the Affymetrix probe sets provide a variety of measurements that allow robust
82
measures of gene expression. The use of multiple perfect-match and mismatch probes for
each gene enables the development of different methods of interpreting the hybridization
patterns across the probe set and calculating a single „expression level‟ or „signal‟ that
reflect the gene‟s relative expression level (Tumor Analysis Best Practices Working
Group 2004). Due to these reasons, we have decided to use the Affymetrix platform for
our experiments.
Figure 5.1: Affymetrix GeneChip Expression Array design. Each probe pair consists of one
perfect match probe and one mismatch probe.
Other groups have conducted microarray experiments on keloid fibroblasts in an
attempt to identify some of the transcriptional level differences underlying this condition.
The most recent study examined the differential gene expression between keloid and
normal fibroblasts grown in the absence and presence of the steroid hydrocortisone
(Smith et al. 2008). When the fibroblasts were grown only in serum supplemented
medium in the absence of hydrocortisone, 511 genes were found to be expressed at
83
significantly different levels in keloid and normal cells. In the presence of
hydrocortisone, 515 genes were found to be differentially expressed. The study showed
increased expression of several IGF-binding and IGF-binding-related proteins and
decreased expression of a subset of Wnt-pathway inhibitors and multiple IL-1-inducible
genes. Most genes are up- or down-regulated by hydrocortisone to a similar extent in
normal and keloid cells. However, increased expression of CTGF and insulin-like growth
factor binding protein (IGFBP)-3 was observed in keloid fibroblasts only in the presence
of hydrocortisone, suggesting a role for glucocorticoid resistance in the pathogenesis of
keloids. These findings support a role for multiple fibrosis-related pathways in the
pathogenesis of keloids.
Seifert et al. performed a study comparing the gene expression profiles between
different lesional sites of keloids (Seifert et al. 2008). The Affymetrix microarray chip
used in this study covered 38,500 genes. Gene expression patterns in the central part of
keloids involve up-regulation of apoptosis inducing genes such as a disintegrin and
metalloprotease 12 (ADAM12) and ECM degrading genes as matrix metalloproteinase
(MMP) - 19. Overexpression of apoptosis inhibitors such as apoptosis caspase activation
inhibitor (AVEN) and down-regulation of angiogenesis inhibiting genes as pentraxinrelated gene (PTX3) at the active margin of keloids may be responsible for the invasive
character of the keloid margin. The results of the study support the important role of the
biopsy site for research in keloids as these results show that different genes are regulated
in different sites of keloids.
Another study analyzed 22,000 genes in keloid fibroblasts compared with normal
skin fibroblasts using a different Affymetrix chip and revealed 43 up- and 6 down-
84
regulated genes (Satish et al. 2006). The authors described up-regulation of annexin A2,
transgelin, and RPS18 in keloids and they reported for the first time that a few tumorrelated genes were overexpressed in keloid fibroblasts. In this study, the age of the
participating patients were different. Of the three patients, the first was an 8-year-old
male, the second was a 57-year-old and the third was of unknown age. Furthermore, the
site of the biopsy within the keloid was not presented and the race of the individuals and
the reason for keloid development was not recorded.
Chen et al. performed microarray analysis of three keloids after burn injury and
three normal skin samples in Chinese patients using cDNA microarray technology (Chen
et al. 2003). In this study 250 genes were up and 152 genes were down-regulated. The
authors describe differential expression of collagen, fibronectin, proteoglycan, growth
factors, and apoptosis-related genes consistent with the published biochemical and
clinical observations of keloids and found higher expression of TGF-β1 and nerve growth
factor (NGF) in keloids versus normal skin.
A comparison of results from all four of these independent microarray studies was
done by Seifert et al. and interestingly, no overlapping gene expression pattern was found
(Seifert & Mrowietz 2009). Table 5.1 gives an overview of some of the regulated genes.
85
Table 5.1: Comparison of different microarray studies (Seifert & Mrowietz 2009)
For our study, we wanted to examine the transcriptional differences in keloid and
normal fibroblasts in the absence of external signals from serum. It is a well known fact
that keloid fibroblasts have a reduced dependence on serum growth factors as compared
to normal fibroblasts (Russell et al. 1988). One of the aims of this study is to elucidate
some possible reasons for this phenomenon. In addition, we also wanted to examine if
there were any systematic transcriptional level differences between fibroblasts that were
harvested at different time points.
86
5.2 Results
5.2.1 The time factor did not result in any systematic differences in the
transcriptional profile of the fibroblast cells
The Affymetrix GeneChip U133A array is capable of measuring expression levels of
22283 probe sets which represent approximately 18400 gene transcripts and variants. Our
Two-Way ANOVA results indicate that only the type of cell (keloid or normal) resulted
in significant systematic differences in gene expression levels. Both the time factor and
the interactive effect between time and type of cell did not result in any significant
differences in the transcriptional profile of the cells, both when MAS 5.0 was used (Table
5.2) and when RMA was used (Table 5.3)
Table 5.2: Two-way ANOVA results for determining the contribution of the time and type of cell
on gene expression with probe summarization by MAS 5.0
Table 5.3: Two-way ANOVA results for determining the contribution of the time and type of cell
on gene expression with probe summarization by RMA
87
5.2.2 Genes significantly upregulated in keloid compared to normal fibroblasts
ECM and glycoproteins such as collagen type I alpha 1 (COL1A1), collagen type XV
alpha I (COL15A1), extracellular matrix protein 1 (ECM1), thrombospondin-1 (THBS1)
and laminin alpha 2 (LAMA2), signaling molecules such as insulin-like growth factor
binding protein 3 (IGFBP3), platelet-derived growth factor receptor beta (PDGFRB),
wingless-type MMTV integration site family member 5A (WNT5A) and ras-related C3
botulinum toxin substrate 2 (RAC2), as well as transcriptional regulators such as
homeobox D10 (HOXD10) and A11 (HOXA11) were found to be significantly
upregulated in keloid fibroblasts compared to normal fibroblasts in at least one of the top
25 lists (Welch‟s t-test P<0.05, Tables 5.4 and 5.5). Genes found to be upregulated in
both of the top 25 lists include the osteoblast specific factor periostin (POSTN),
cytoskeletal protein keratin 19 (KRT19), cell adhesion molecule 1 (CADM1),
neurodegenerative disorder protein ataxin-1 (ATXN1) and the axonal protein
semophorin-5A (SEM5A) (Welch‟s t-test P<0.05, Tables 5.4 and 5.5). Other notable
genes that were not in the top 25 list include collagen type V alpha 1 (COL5A1), collagen
type V alpha 3 (COL5A3), collagen type XVII alpha 1 (COL17A1), myosins 1D
(MYO1D) and 19 (MYO19), mediator of cell motility 1 (MEMO1), G-protein-coupled
receptors 137B (GPR137B) and 153 (GPR153), G-protein signaling modulator 2
(GPSM2), son of sevenless homolog 2 (SOS2), growth factor receptor-bound protein 10
(GRB10) and Ephrin type-B receptor 4 (EPHB4) (Welch‟s t-test P < 0.05). The full list of
genes that were upregulated with p-value < 0.05 can be found in Appendix A.1 and A.3.
88
Table 5.4: Top 25 upregulated genes in keloid compared to normal fibroblasts using the MAS 5.0
summarization algorithm ranked by fold change
Fold
change
26.2568
19.88896
Gene
Symbol
POSTN
ZIC1
14.57531
10.38634
8.887936
HOXD10
COL15A1
EGR2
7.998279
7.9743
6.770347
HOXA11
CCDC102B
KCNJ6
6.420784
5.80834
5.553577
5.515989
5.489978
JUP ///
KRT19
MAP7
IGFBP3
ADRA2A
RAC2
5.301598
5.168533
4.974008
4.940097
4.889467
4.783432
4.513727
CDYL
NPTX1
PMEPA1
EFNB2
ATXN1
CADM1
SEMA5A
4.476735
WNT5A
4.305683
4.167339
AK5
EVI2A ///
EVI2B
THBS1
HMGCS2
4.105316
4.064728
Gene Title
periostin, osteoblast specific factor
Zic family member 1 (odd-paired homolog,
Drosophila)
homeobox D10
collagen, type XV, alpha 1
early growth response 2 (Krox-20 homolog,
Drosophila)
homeobox A11
coiled-coil domain containing 102B
potassium inwardly-rectifying channel, subfamily J,
member 6
junction plakoglobin /// keratin 19
microtubule-associated protein 7
insulin-like growth factor binding protein 3
adrenergic, alpha-2A-, receptor
ras-related C3 botulinum toxin substrate 2 (rho family,
small GTP binding protein Rac2)
chromodomain protein, Y-like
neuronal pentraxin I
prostate transmembrane protein, androgen induced 1
ephrin-B2
ataxin 1
cell adhesion molecule 1
sema domain, seven thrombospondin repeats (type 1
and type 1-like), transmembrane domain (TM) and
short cytoplasmic domain, (semaphorin) 5A
wingless-type MMTV integration site family, member
5A
adenylate kinase 5
ecotropic viral integration site 2A /// ecotropic viral
integration site 2B
thrombospondin 1
3-hydroxy-3-methylglutaryl-Coenzyme A synthase 2
(mitochondrial)
Corrected
p-value
2.51E-04
0.046789
0.004014
4.25E-04
0.006852
0.033229
0.021884
0.014057
0.014453
0.00785
0.021725
0.011306
0.049297
0.039061
0.015499
0.037054
0.022704
0.003232
0.033839
0.008288
0.0251
0.032527
0.003941
0.003338
0.038054
Table 5.5: Top 25 upregulated genes in keloid compared to normal fibroblasts using the RMA
summarization algorithm ranked by fold change
Fold
change
18.0305
5.40679
3.959037
3.21548
Gene
Symbol
POSTN
IGFBP3
COL15A1
SEMA5A
Gene Title
periostin, osteoblast specific factor
insulin-like growth factor binding protein 3
collagen, type XV, alpha 1
sema domain, seven thrombospondin repeats (type 1
Corrected
p-value
0.006498
0.024325
0.028774
0.013197
89
3.042947
SEMA5A
2.726196
2.631936
2.531559
CADM1
ATXN1
FARP1
2.375326
MICAL2
2.176466
2.129065
ECM1
SLC25A6
2.053115
KCNJ6
1.990826
1.985587
1.959726
1.906954
TBC1D2
CADM1
NXN
MICAL2
1.897066
1.878355
COL1A1
PDGFRB
1.869649
SLC25A6
1.865537
GPSM2
1.852414
1.807992
1.796024
1.764263
LOC644191
///
LOC728937
/// RPS26
CTSB
ODZ3
JUP ///
KRT19
LAMA2
FHOD1
CTDSPL
1.749725
1.747925
SHMT2
HDLBP
1.84265
1.836179
1.835598
and type 1-like), transmembrane domain (TM) and
short cytoplasmic domain, (semaphorin) 5A
sema domain, seven thrombospondin repeats (type 1
and type 1-like), transmembrane domain (TM) and
short cytoplasmic domain, (semaphorin) 5A
cell adhesion molecule 1
ataxin 1
FERM, RhoGEF (ARHGEF) and pleckstrin domain
protein 1 (chondrocyte-derived)
microtubule associated monoxygenase, calponin and
LIM domain containing 2
extracellular matrix protein 1
solute carrier family 25 (mitochondrial carrier; adenine
nucleotide translocator), member 6
potassium inwardly-rectifying channel, subfamily J,
member 6
TBC1 domain family, member 2
cell adhesion molecule 1
nucleoredoxin
microtubule associated monoxygenase, calponin and
LIM domain containing 2
collagen, type I, alpha 1
platelet-derived growth factor receptor, beta
polypeptide
solute carrier family 25 (mitochondrial carrier; adenine
nucleotide translocator), member 6
G-protein signaling modulator 2 (AGS3-like, C.
elegans)
similar to hCG15685 /// similar to 40S ribosomal
protein S26 /// ribosomal protein S26
0.035137
0.001646
0.020822
0.045664
0.037055
0.047073
0.019936
0.021022
0.048158
0.046985
0.002448
0.017212
0.008827
0.002791
0.046273
0.04761
0.038141
cathepsin B
odz, odd Oz/ten-m homolog 3 (Drosophila)
junction plakoglobin /// keratin 19
0.018933
6.59E-04
0.02721
laminin, alpha 2
formin homology 2 domain containing 1
CTD (carboxy-terminal domain, RNA polymerase II,
polypeptide A) small phosphatase-like
serine hydroxymethyltransferase 2 (mitochondrial)
high density lipoprotein binding protein
0.006566
0.014803
0.018992
0.038598
0.037741
5.2.3 Genes significantly downregulated in keloid compared to normal fibroblasts
A host of chemokine factors including chemokine ligands 6 (CXCL6), 1 (CXCL1), and 2
(CXCL2) as well as interleukin 8 (IL8) were among the genes found to be significantly
90
downregulated in keloid compared to normal fibroblasts in both the top 25 lists (Welch‟s
t-test P<0.05, Tables 5.6 and 5.7). In addition, cytokines such as interleukins 6 (IL6) and
32 (IL32) as well as tumor necrosis factor alpha-induced protein 6 (TNFAIP6), 3
(TNFAIP3) and tumor necrosis factor superfamily member 10 (TNFSF10) were also
found to be downregulated in at least one of the top 25 lists (Welch‟s t-test P<0.05,
Tables 5.6 and 5.7). Other interesting genes that were downregulated include matrix
metalloproteinase 2 (MMP2), hydroxysteroid (11-beta) dehydrogenase 1 (HSD11B1),
complement factor B (CFB), complement component 3 (C3), radical S-adenosyl
methionine domain containing 2 (RSAD2), 2',5'-oligoadenylate synthetase 1 (OAS1),
solute carrier family 39 member 8 (SLC39A8), G0/G1switch 2 (G0S2), interferoninduced protein with tetratricopeptide repeats 1 (IFIT1) and 3 (IFIT3), prostaglandin E
synthase (PTGES) as well as secreted frizzled-related protein 1 (SFRP1) (Welch‟s t-test P
< 0.05) . The full list of genes that were downregulated with p-value < 0.05 can be found
in Appendix A.2 and A.4.
Table 5.6: Top 25 downregulated genes in keloid compared to normal fibroblasts using the MAS
5.0 summarization algorithm ranked by fold change
Fold
change
77.61498
Gene
Symbol
CXCL6
73.48209
CXCL1
67.89986
64.39615
49.49322
41.41703
39.93947
32.48277
IL8
CXCL11
HSD11B1
CCL5
CXCL2
RARRES1
29.77878
27.68656
27.15919
RSAD2
PLA2G2A
C2 /// CFB
Gene Title
chemokine (C-X-C motif) ligand 6 (granulocyte
chemotactic protein 2)
chemokine (C-X-C motif) ligand 1 (melanoma growth
stimulating activity, alpha)
interleukin 8
chemokine (C-X-C motif) ligand 11
hydroxysteroid (11-beta) dehydrogenase 1
chemokine (C-C motif) ligand 5
chemokine (C-X-C motif) ligand 2
retinoic acid receptor responder (tazarotene induced)
1
radical S-adenosyl methionine domain containing 2
phospholipase A2, group IIA (platelets, synovial fluid)
complement component 2 /// complement factor B
Corrected
p-value
1.48E-05
7.93E-05
0.006303
0.006303
1.98E-06
0.005142
0.002401
0.004841
0.019981
0.001149
2.15E-05
91
27.04305
26.4216
26.13762
23.80846
23.43326
23.40176
21.83553
21.53614
19.30517
15.52373
15.49821
14.28212
14.04543
13.35598
CXCL5
TNFAIP6
CXCL3
IL32
CP
CXCL10
CHI3L2
IDO1
NTRK2
C3
SLC39A8
G0S2
OAS1
TNFSF10
chemokine (C-X-C motif) ligand 5
tumor necrosis factor, alpha-induced protein 6
chemokine (C-X-C motif) ligand 3
interleukin 32
ceruloplasmin (ferroxidase)
chemokine (C-X-C motif) ligand 10
chitinase 3-like 2
indoleamine 2,3-dioxygenase 1
neurotrophic tyrosine kinase, receptor, type 2
complement component 3
solute carrier family 39 (zinc transporter), member 8
G0/G1switch 2
2',5'-oligoadenylate synthetase 1, 40/46kDa
tumor necrosis factor (ligand) superfamily, member 10
2.44E-04
0.008288
0.005384
0.001636
0.003941
0.032337
5.70E-04
0.004053
0.011306
2.51E-04
3.85E-07
0.003903
0.039061
0.016414
Table 5.7: Top 25 downregulated genes in keloid compared to normal fibroblasts using the RMA
summarization algorithm ranked by fold change
Fold
change
44.34992
Gene
Symbol
CXCL6
41.64695
CXCL1
37.03845
33.61232
29.71772
21.50683
20.34249
19.02187
18.80414
16.76682
14.08095
13.78165
12.01816
11.75815
11.65778
10.88791
9.990352
9.679501
8.604651
8.378408
8.213031
8.204023
8.195308
8.068323
7.890376
C2 /// CFB
HSD11B1
TNFAIP6
CXCL2
TNFAIP6
IL8
SLC39A8
SLC39A8
C3
RSAD2
SOD2
IL8
CCL2
SFRP1
G0S2
IFI44L
CHI3L2
IL6
CA12
OAS1
GCH1
CA12
SOD2
Gene Title
chemokine (C-X-C motif) ligand 6 (granulocyte
chemotactic protein 2)
chemokine (C-X-C motif) ligand 1 (melanoma growth
stimulating activity, alpha)
complement component 2 /// complement factor B
hydroxysteroid (11-beta) dehydrogenase 1
tumor necrosis factor, alpha-induced protein 6
chemokine (C-X-C motif) ligand 2
tumor necrosis factor, alpha-induced protein 6
interleukin 8
solute carrier family 39 (zinc transporter), member 8
solute carrier family 39 (zinc transporter), member 8
complement component 3
radical S-adenosyl methionine domain containing 2
superoxide dismutase 2, mitochondrial
interleukin 8
chemokine (C-C motif) ligand 2
secreted frizzled-related protein 1
G0/G1switch 2
interferon-induced protein 44-like
chitinase 3-like 2
interleukin 6 (interferon, beta 2)
carbonic anhydrase XII
2',5'-oligoadenylate synthetase 1, 40/46kDa
GTP cyclohydrolase 1
carbonic anhydrase XII
superoxide dismutase 2, mitochondrial
Corrected
p-value
9.29E-10
1.89E-05
1.75E-08
1.50E-06
0.008827
5.89E-05
0.011388
0.003057
1.22E-05
7.69E-04
2.94E-04
0.013835
0.002706
0.001761
9.46E-07
0.026217
0.001019
0.04903
0.019936
4.59E-04
2.40E-04
0.015085
2.29E-04
0.00631
0.001861
92
7.244113
IFIT1
7.188362
6.98734
TNFAIP3
IFIT3
6.850272
TNFSF10
interferon-induced protein with tetratricopeptide
repeats 1
tumor necrosis factor, alpha-induced protein 3
interferon-induced protein with tetratricopeptide
repeats 3
tumor necrosis factor (ligand) superfamily, member 10
0.035976
8.07E-05
0.014427
0.03935
5.2.4 Hierarchical clustering and principal components analysis revealed that genes
chosen were capable of distinguishing between keloid and normal samples
When samples were grouped by principal components analysis using all ~23 000 probe
sets in the GeneChip U133A arrays, keloid samples were fairly well separated from
normal samples (Fig. 5.2A), although there was a slight overlap when genes were
summarized using the RMA algorithm (Fig. 5.3A). When samples were grouped using
only probe sets that were found to be significantly different, there was a clear separation
between keloid and normal samples (Figs. 5.2B and 5.3B). The same outcome was also
observed when hierarchical clustering was used. When using the full list of genes, keloid
samples were generally clustered together and normal samples were also generally
clustered together, but again there were some samples that were not clustered accordingly
(Figs. 5.2C and 5.3C). When samples were clustered using only probe sets that were
found to be significantly different, two major clusters were formed with keloid samples in
one cluster and normal samples in the other (Figs. 5.2D and 5.3D).
93
A
C
KF1 D1
KF1 D3
KF1 D5
KF2 D3
KF2 D5
KF2 D1
KF3 D1
NF2 D1
NF2 D3
NF2 D5
NF1 D1
NF1 D3
NF1 D5
KF3 D3
KF3 D5
NF3 D1
NF3 D3
NF3 D5
KF1 D1
KF1 D3
KF1 D5
KF2 D3
KF2 D5
KF3 D1
KF3 D3
KF3 D5
KF2 D1
NF1 D1
NF1 D3
NF1 D5
NF2 D1
NF2 D3
NF2 D5
NF3 D1
NF3 D3
NF3 D5
B
D
Figure 5.2: Principal components analysis and hierarchical clustering using the MAS 5.0
algorithm. (A) PCA of all samples using the full list of genes (B) PCA of all samples using
differentially expressed genes (P<0.05) (C) Hierarchical clustering of all samples using the full
list of genes (D) Hierarchical clustering of all samples using differentially expressed genes
(P<0.05). Red balls and lines denote keloid samples while blue balls and lines denote normal
samples.
94
A
C
KF1 D1
KF1 D3
KF1 D5
KF2 D3
KF2 D5
KF2 D1
KF3 D1
NF2 D1
NF2 D3
NF2 D5
KF3 D3
KF3 D5
NF3 D1
NF3 D3
NF3 D5
NF1 D1
NF1 D3
NF3 D5
KF1 D1
KF1 D3
KF1 D5
KF2 D3
KF2 D5
KF3 D1
KF3 D3
KF3 D5
KF2 D1
NF1 D1
NF3 D1
NF3 D3
NF2 D1
NF3 D5
NF1 D3
NF1 D5
NF2 D3
NF2 D5
B
D
Figure 5.3: Principal components analysis and hierarchical clustering using the RMA algorithm.
(A) PCA of all samples using the full list of genes (B) PCA of all samples using differentially
expressed genes (P<0.05) (C) Hierarchical clustering of all samples using the full list of genes
(D) Hierarchical clustering of all samples using differentially expressed genes (P<0.05). Red balls
and lines denote keloid samples while blue balls and lines denote normal samples.
5.2.5 DAVID analysis suggests a role for immunological factors and ribosomal
proteins in keloid pathogenesis
A total of 18 Gene Ontology (GO) terms were found to be statistically enriched (P <
0.05) using the DAVID Gene Functional Classification Tool with the list of significantly
95
upregulated genes in keloid as inputs (Table 5.8). When the list of significantly
downregulated genes in keloid was used, a total of 35 GO terms were found to be
statistically enriched (P<0.05) (Table 5.9). These terms include biological processes such
as immune response, response to wounding and locomotory behaviour, molecular
functions such as chemokine and cytokine activity and cellular components such as
extracellular matrix and cytosolic ribosome. Two Kyoto Encyclopedia of Genes and
Genomes (KEGG) pathways were also found to be significantly enriched (P<0.05) when
the full set of differentially expressed genes were used. These were the antigen
processing and presentation pathway (Fig. 5.4, Table 5.10) and the ribosome pathway
(Fig. 5.5, Table 5.11). Most of the significantly different genes that were involved in the
antigen processing and presentation pathway were downregulated in keloid while all the
significantly different genes that were involved in the ribosome pathway were
upregulated in keloids.
Table 5.8: List of Gene Ontology terms that were found to be statistically enriched using the
DAVID Gene Functional Classification Tool with the list of significantly upregulated genes in
keloid as input
Category
Term
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_MF_ALL
GOTERM_MF_ALL
GOTERM_CC_ALL
GOTERM_BP_ALL
GOTERM_CC_ALL
GOTERM_CC_ALL
GO:0005830~cytosolic ribosome (sensu Eukaryota)
GO:0005840~ribosome
GO:0044445~cytosolic part
GO:0033279~ribosomal subunit
GO:0003735~structural constituent of ribosome
GO:0005198~structural molecule activity
GO:0030529~ribonucleoprotein complex
GO:0006412~translation
GO:0015935~small ribosomal subunit
GO:0005843~cytosolic small ribosomal subunit (sensu
Eukaryota)
GO:0043232~intracellular non-membrane-bound
organelle
GO:0043228~non-membrane-bound organelle
GO:0003723~RNA binding
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_MF_ALL
Corrected
p-value
5.10E-07
2.10E-06
7.54E-06
8.89E-06
1.52E-05
3.88E-05
8.31E-05
0.00109
0.001301
0.00132
0.001642
0.001642
0.029371
96
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GO:0031012~extracellular matrix
GO:0005737~cytoplasm
GO:0005578~proteinaceous extracellular matrix
GO:0009059~macromolecule biosynthetic process
GO:0044249~cellular biosynthetic process
0.030531
0.0307
0.031002
0.042621
0.04839
Table 5.9: List of Gene Ontology terms that were found to be statistically enriched using the
DAVID Gene Functional Classification Tool with the list of significantly downregulated genes in
keloid as input
Category
Term
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_MF_ALL
GOTERM_MF_ALL
GOTERM_MF_ALL
GOTERM_BP_ALL
GOTERM_MF_ALL
GOTERM_BP_ALL
GO:0006955~immune response
GO:0002376~immune system process
GO:0050896~response to stimulus
GO:0019882~antigen processing and presentation
GO:0006952~defense response
GO:0005125~cytokine activity
GO:0042379~chemokine receptor binding
GO:0008009~chemokine activity
GO:0006954~inflammatory response
GO:0001664~G-protein-coupled receptor binding
GO:0048002~antigen processing and presentation of
peptide antigen
GO:0009611~response to wounding
GO:0005615~extracellular space
GO:0002474~antigen processing and presentation of
peptide antigen via MHC class I
GO:0009607~response to biotic stimulus
GO:0042221~response to chemical stimulus
GO:0042611~MHC protein complex
GO:0007626~locomotory behavior
GO:0005576~extracellular region
GO:0044421~extracellular region part
GO:0006950~response to stress
GO:0009605~response to external stimulus
GO:0042330~taxis
GO:0006935~chemotaxis
GO:0051707~response to other organism
GO:0051704~multi-organism process
GO:0042612~MHC class I protein complex
GO:0009615~response to virus
GO:0000041~transition metal ion transport
GO:0005507~copper ion binding
GO:0005102~receptor binding
GO:0007610~behavior
GO:0046870~cadmium ion binding
GOTERM_BP_ALL
GOTERM_CC_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_CC_ALL
GOTERM_BP_ALL
GOTERM_CC_ALL
GOTERM_CC_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_CC_ALL
GOTERM_BP_ALL
GOTERM_BP_ALL
GOTERM_MF_ALL
GOTERM_MF_ALL
GOTERM_BP_ALL
GOTERM_MF_ALL
Corrected
p-value
2.22E-21
3.44E-19
9.19E-14
2.09E-09
2.10E-09
2.35E-08
2.42E-08
3.61E-08
5.17E-08
1.23E-06
2.09E-06
2.93E-06
5.60E-06
7.17E-06
2.44E-05
5.25E-05
6.93E-05
7.96E-05
1.16E-04
2.63E-04
3.39E-04
3.81E-04
3.93E-04
3.93E-04
3.97E-04
5.05E-04
6.51E-04
0.001549
0.00606
0.008833
0.010917
0.011518
0.023054
97
GOTERM_BP_ALL
GO:0042127~regulation of cell proliferation
0.037833
Table 5.10: List of downregulated genes in keloid compared to normal fibroblasts involved in
GO term Antigen Processing and Presentation
Fold change
2.737904
2.702095
2.648943
2.638515
2.566287
2.421571
2.231546
2.21116
2.182933
2.131292
2.117058
2.042615
1.798085
1.795796
Gene Symbol
HLA-F
HLA-C
HLA-DMA
TAPBPL
TAPBPL
HLA-F
HLA-G
HLA-G
HLA-B
HLA-G
TAPBP
HLA-B
HLA-C
HLA-A /// HLAA29.1 /// HLA-B
/// HLA-G ///
HLA-H /// HLAJ
1.773213
1.691625
1.678448
HLA-C
HLA-C
HLA-B /// MICA
1.520112
1.229484
HLA-A
B2M
Gene Title
major histocompatibility complex, class I, F
major histocompatibility complex, class I, C
major histocompatibility complex, class II, DM alpha
TAP binding protein-like
TAP binding protein-like
major histocompatibility complex, class I, F
major histocompatibility complex, class I, G
major histocompatibility complex, class I, G
major histocompatibility complex, class I, B
major histocompatibility complex, class I, G
TAP binding protein (tapasin)
major histocompatibility complex, class I, B
major histocompatibility complex, class I, C
major histocompatibility complex, class I, A /// major
histocompatibility complex class I HLA-A29.1 /// major
histocompatibility complex, class I, B /// major
histocompatibility complex, class I, G /// major
histocompatibility complex, class I, H (pseudogene) /// major
histocompatibility complex, class I, J (pseudogene)
major histocompatibility complex, class I, C
major histocompatibility complex, class I, C
major histocompatibility complex, class I, B /// MHC class I
polypeptide-related sequence A
major histocompatibility complex, class I, A
beta-2-microglobulin
Table 5.11: List of upregulated genes in keloid compared to normal fibroblasts involved in GO
term Ribosome
Fold change
2.662594
1.877402
1.865347
1.780408
1.764005
1.696983
1.645023
Gene Symbol
MGC87895 ///
RPS14
LOC285053 ///
LOC390354 ///
RPL18A
LOC644191 ///
LOC728937 ///
RPS26
RPL10
RPS16
EIF3A
RPS9
Gene Title
similar to ribosomal protein S14 /// ribosomal protein S14
similar to ribosomal protein L18a /// ribosomal protein L18a
pseudogene /// ribosomal protein L18a
similar to hCG15685 /// similar to 40S ribosomal protein S26
/// ribosomal protein S26
ribosomal protein L10
ribosomal protein S16
eukaryotic translation initiation factor 3, subunit A
ribosomal protein S9
98
1.636263
1.602853
1.597647
1.59623
1.531203
1.523152
1.474366
1.415472
1.371258
1.360637
RPL13
RPS2
RPL4
RPS8
SERP1
RPL13
RPS6
RPL8
SERP1
RPL13
ribosomal protein L13
ribosomal protein S2
ribosomal protein L4
ribosomal protein S8
stress-associated endoplasmic reticulum protein 1
ribosomal protein L13
ribosomal protein S6
ribosomal protein L8
stress-associated endoplasmic reticulum protein 1
ribosomal protein L13
Figure 5.4: Antigen processing and presentation pathway from the KEGG database. List of
differentially expressed genes were submitted to the DAVID Gene Functional Classification Tool
using the full list of Affymetrix U133A genes as background for statistical analysis. The antigen
processing and presentation pathway had a Benjamini corrected P-value of 0.0029. Genes that
were significantly upregulated in the pathway are denoted with a red cross while genes that were
significantly downregulated in the pathway are denoted with a blue cross.
99
Figure 5.5: Ribosome pathway from the KEGG database. List of differentially expressed genes
were submitted to the DAVID Gene Functional Classification Tool using the full list of
Affymetrix U133A genes as background for statistical analysis. The ribosome pathway had a
Benjamini corrected P-value of 0.029. Genes that were significantly upregulated in the pathway
are denoted with a red cross while genes that were significantly downregulated in the pathway are
denoted with a blue cross.
5.3 Discussion
A number of different groups have performed microarray studies on keloid fibroblasts
previously (Chen et al. 2003; Satish et al. 2006; Seifert et al. 2008; Smith et al. 2008) but
none have examined the transcriptional level and time dependent effects of serum
starvation on these cells. In this study, we report that the time factor did not have any
100
significant and systematic effect on gene expression levels of keloid and normal
fibroblasts. This result is consistent with the common assumption that cells left in
minimal media show no major transcriptional level differences. However, it is also
possible that our time points and number of samples are too small to detect any
systematic differences due to the time factor. Under serum starvation conditions, it was
found that there was a greater number of genes downregulated and fewer genes
upregulated in the keloid compared to the normal fibroblasts. Interestingly, this same
observation has also been made when the fibroblasts were left in serum supplemented
media (175 genes upregulated, 559 genes downregulated) or when hydrocortisone
supplemented media was used (221 genes upregulated, 547 genes downregulated) (Smith
et al. 2008). Research in keloid has focused mainly on the upregulation of candidate
genes such as TGF-β and PDGF, but these results suggest that gene downregulation
could also be a very important aspect of keloid formation.
The microarray platform allows for the measurement of thousands of gene
transcripts simultaneously, but its strength could also be its weakness. It is not feasible,
for example, to independently validate all the hundreds of abnormally expressed genes
using other molecular techniques. Furthermore, the vast amounts of data combined with
the limited number of samples make for a profoundly under-determined problem; there is
not enough data to distinguish between many of the different hypothesis that could be
consistent with the data set. Finally, the different methods that can be used for
background correction, normalization and summarization of the arrays can lead to
different results. Some of the more widely used algorithms include the default Affymetrix
MAS 5.0 algorithm (default Affymetrix approach; Affymetrix Users Guide,
101
www.affymetrix.com), the RMA algorithm (Bolstad et al. 2003) and the dChip algorithm
(Li & Wong 2001). At present, there is no consensus as to which method is best, with
different studies giving conflicting results (Bolstad et al. 2003; Shedden et al. 2005; Harr
& Schlötterer 2006; Lim et al. 2007).
In our experiments, we have used the MAS 5.0 algorithm and the RMA algorithm
for normalization and summarization. We found that using the MAS 5.0 algorithm gives
a larger number of differentially expressed genes (471 genes, P < 0.05) compared to the
RMA algorithm (344 genes, P < 0.05). Furthermore, the list of genes that were most
highly upregulated or downregulated were also different when a different algorithm was
used. In total, the intersection between the RMA summarized data and the MAS 5.0
summarized data showed 217 genes to be differentially expressed (Appendix A.5). It
remains unclear which method produces the more accurate and reliable results, but we
have chosen to use the MAS 5.0 summarization algorithm for DAVID analysis by virtue
it producing a larger number of differentially expressed genes. For enrichment analysis, a
larger gene list has higher statistical power resulting in a higher sensitivity to slightly
enriched terms, as well as to more specific terms (Huang, Sherman & Lempicki 2009).
Many of the genes that were found to be significantly different in this study were
similar to those that have been found in other studies previously. ECM proteins such as
COL1A1 and COL5A1 were found to be significantly upregulated in keloids.
Upregulation of collagen is a well known characteristic of keloid fibroblasts and
upregulation of these two collagen types in particular has also been reported in previous
microarray studies (Table 5.1). PDGFRB (Messadi et al. 1998), THBS1 (Chipev et al.
2000) and RAC2 (Witt et al. 2008) were also found to be significantly upregulated in
102
keloids consistent with previous findings. In a microarray array study done by Seifert et
al., it was reported that keratin 18 (KRT18) was upregulated in keloid compared to
normal samples, and that there was upregulation of POSTN and ECM1 in fibroblasts
derived from the deeper part of the keloid compared to the superficial part of the keloid
(Seifert et al. 2008). While we did not see any difference in KRT18 expression levels, we
found a higher expression of KRT19, POSTN and ECM1 in keloid compared to normal
fibroblasts.
IGFBP3 is another important upregulated factor that was found in our study.
Higher expression of IGFBP3 in dermal fibroblasts from patients with systemic sclerosis
(Feghali & Wright 1999), idiopathic pulmonary fibrosis (Yasuoka et al. 2006), and
leiomyomas (Tsibris et al. 2002) indicate that this protein has some role to play in
fibrosis. Furthermore, IGFBP3 is capable of inducing production of ECM components
such as collagen type I and fibronectin in normal primary adult lung fibroblasts (Pilewski
et al. 2005). A known function of IGFBP3 is to bind to IGF-1 (Collett-Solberg & Cohen
1996) but it has also been found to bind to extracellular matrix components, have nuclear
localization signals, and bind to putative receptors on the cell surface (Mohan & Baylink
2002). It is possible that these other functions of IGFBP3 contribute to its role in fibrosis
and could be an interesting future area of study. Increased expression of IGFBP3 has also
been found in other studies of keloids (Satish et al. 2006; Smith et al. 2008).
Our results also show a reduction in the Wnt signaling antagonist SFRP1 as well
as an increase in the WNT5A gene expression levels in keloid compared to normal
fibroblasts. Low levels of SFRP1 in keloid fibroblasts have been reported previously
(Smith et al. 2008) indicating a role for the Wnt pathway in keloid pathogenesis. SFRP1
103
suppresion has also been found in numerous different carcinomas including breast
(Shulewitz et al. 2006), bladder (Marsit et al. 2005), ovarian (Takada et al. 2004), colon
(Caldwell et al. 2004) and prostrate (Lodygin et al. 2005) cancers. Furthermore, it has
also been shown to have an antiproliferative effect on vascular cells (Ezan et al. 2004),
making it a very attractive candidate for further study. SFRP1 suppresses both the
canonical Wnt/β-catenin pathway, as well as the non canonical β-catenin independent
pathway (Yang et al. 2009). WNT5A, on the other hand, is only associated with the noncanonical Wnt signaling pathway (Slusarski, Corces & Moon 1997; Kilian et al. 2003).
The role of WNT5A in cancer remains unclear with some cancers showing an
upregulation of this factor while others show a decrease in expression (McDonald &
Silver 2009). However, it has been found to promote angiogenesis by inducing
endothelial cell proliferation and enhanced cell survival under serum-deprived conditions
(Masckauchán et al. 2006). While the canonical Wnt signaling has been shown to be
upregulated in keloids (Sato 2006), not much is known of the non-canonical pathway.
Expression of some IL-1 responsive genes such as CXCL-1, CXCL-6, MMP-2
and TNFAIP6 were found to be downregulated in keloids and again, this result is
consonant with that obtained in previous studies (Smith et al. 2008; Yeh, Shen & Tai
2009). Lower levels of MMP-2 could be one of the reasons for the accumulation of ECM
components as the MMPs are responsible for ECM degradation. Other inflammatory
cytokines such as IL8 and IL32 were also found to be downregulated, suggesting a role
for inflammation in keloid pathogenesis.
Other interesting genes that were upregulated include those involved in G protein
coupled receptor signaling (GPR137B, GPR153, SOS2), cell motility proteins (MYO1D,
104
MYO19, MEMO1) and tyrosine kinase signaling molecules (EPHB4, SOS2) while those
that were downregulated were involved in apoptosis (PTGES, TNFSF10), production of
cortisol (HSD11B1), activation of the complement pathway of the immune system (CFB,
C3), antiviral defense (RSAD2, OAS1) and interferon induced proteins (IFIT1 and
IFIT3). These genes, as well as a host of other genes in the top 25 list of differentially
expressed genes could be important targets for further study and may be important in
shedding light on the keloid condition. Of particular interest are the proapoptotic genes
TNFSF10 and PTGES which were found to be decreased in keloids. This finding has
never been reported before and could be a result of the serum starvation condition.
Unfortunately, there were some discrepancies in our results compared to previous
studies as well. Smith et al. reported a reduction in multiple homeotic (HOX) genes such
as HOXA11, among others (Smith et al. 2008). However, we found an increased
expression of HOXA11 and HOXD10 in keloid compared to normal fibroblasts. HOX
genes are highly conserved master control genes that play major roles in anteriorposterior development in the embryo but they have also been found to regulate gene
expression in adult differentiated cells, including human dermal fibroblasts (Chang et al.
2002). The differential expression of HOX genes may account for the tumorigenic
phenotype of keloids, but as suggested by Smith et al., it is also possible that these
differences may just be due to the different anatomic sites from which cultures were
isolated (Smith et al. 2008). This could possibly account for the discrepancy in our
results. In addition, these differences could also be due to the different culture conditions
(serum/hydrocortisone instead of serum starvation), different microarray chips (U133A
2.0 instead of U133A), different normalization methods (RMA instead of MAS 5.0),
105
sample to sample variation or false positive results. Our results also showed a
downregulation of IL6, contrary to previous reports where higher expression of IL6 was
found in keloids (Tosa et al. 2005; Ghazizadeh et al. 2007). Overexpression of IL6
related genes in keloid fibroblasts was also not seen in the microarray study done by
Smith et al. and they attributed this to the differences in origin of cultures, culture
conditions, or the microarray platform (Smith et al. 2008). The very same factors could
also be the reasons for our discrepancy.
In general, there are two ways in which microarrays can be used to investigate
problems in cell biology (Schulze & Downward 2001). The first method can be thought
of as a local approach to understanding gene expression changes, where the investigator
is interested only in finding the single change in gene expression that might be the key to
a given alteration in phenotype. This is the method that we have utilized so far, and in
doing so we have come up with a list of potential genes that could be important in
understanding keloid pathophysiology. However, we are never really certain about the
importance of this list; microarray experiments are highly capable of generating long lists
of genes with altered expression, but they provide few clues as to which of these changes
are important in establishing a given phenotype. This deduction is left to the ingenuity of
the experimenter, and the temptation is to stick to familiar genes, or genes that conform
to existing ideas about how the system works, thus resulting in a certain level of
biasedness. In our case, we have tried to eliminate this bias by giving the top 25 lists of
genes that were differentially expressed with p-values less than 0.05, but even so, the list
differs significantly depending on the normalization method used.
106
Another way of working with microarray data is to look at the global picture of
gene expression patterns. Unsupervised methods such as k-means clustering, principalcomponent analysis and self-organizing maps can be used to group closely related genes
or samples together. Using hierarchical clustering and principal-component analysis on
the full set of genes, we found that in general, keloid samples were more similar to each
other compared to normal samples, regardless of the sample type or the time point when
they were harvested. This was despite the fact that the samples were processed in batches,
that is, RNA extraction and hybridization of KF1 and NF1 were done together, KF2 was
paired with NF2 and KF3 was paired with NF3. Batch effects are a very common
problem faced by researchers in the area of microarray studies, particularly when
combining multiple batches of data from different experiments or if an experiment cannot
be conducted all at once (Johnson, Li & Rabinovic 2007). Our results indicate that for our
case, there are minimal systematic batch effects. There was a small degree of overlap
between the keloid and normal samples when the full set of genes was used, but this
could also be due to the fact that KF3 and NF3 were from the same patient. This overlap
was eliminated when the subset containing only genes found to be differentially
expressed was used for the unsupervised methods. This result reconfirms the ability of
the list of genes found to discriminate between keloid and normal fibroblasts.
We were also interested in looking at clusters of related genes that were
differentially expressed to give some biological meaning to the data. Hierarchical
clustering can be used to cluster similarly expressed genes together, but here we have
opted for an enrichment analysis based method instead. The strategy employed here is to
systematically map a large number of interesting genes in a list to the associated
107
biological annotation (e.g., gene ontology terms), and then statistically highlight the most
overrepresented (enriched) biological annotation out of thousands of linked terms and
contents. The DAVID Gene Functional Classification Tool is one of a number of tools
that is capable of performing enrichment analysis, and compared with similar services, it
has some unique features and capabilities, such as an integrated and expanded back-end
annotation database, advanced modular enrichment algorithms and powerful exploratory
ability in an integrated data-mining environment (Huang, Sherman & Lempicki 2009).
For DAVID analysis, we have decided to use the MAS 5.0 algorithm for
summarization and normalization as this produced a larger set of differentially expressed
genes as compared to the RMA algorithm. In general, when it comes to enrichment
analysis, a larger gene list can have higher statistical power resulting in a higher
sensitivity to slightly enriched terms, as well as to more specific terms. Otherwise, the
sensitivity is decreased toward largely enriched terms and broader/general terms (Huang,
Sherman & Lempicki 2009). Huang et al. also gives a checklist of characteristics of a
„good‟ gene list for analysis, and this includes the presence of important marker genes, a
reasonable number of genes ranging from hundreds to thousands and the passing of
important statistical thresholds such as t-tests and fold changes. However, he also states
that important statistical thresholds do not have to be sacrificed (e.g., fold changes>1.1
and P-value < 0.2) to reach a comfortable gene size. For our gene list, we have chosen a
stringent P-value of 0.05 but have not placed any restriction on the fold change as we
believe the stringent P-value would be sufficient in identifying the differentially
expressed genes.
108
One deficiency of the DAVID system is that it does not say in what way the
enriched terms differ when comparing between keloid and normal fibroblast cells, that is
to say, whether for example, immune response is heightened or suppressed in keloid
compared to normal cells. All the system does is to take in a list of genes and determine
which terms are enriched from the list that was received. One way of dealing with this
limitation is to input genes that are upregulated and genes that are downregulated
separately. Since this approach reduces the size of the gene list, this may result in
decreased sensitivity and specificity in the enrichment analysis. However, we have found
that in our case, breaking up the list into upregulated and downregulated genes results in
about the same number and type of terms enriched compared to when the full set of
differentially expressed genes was used as input. Furthermore, this approach has the
added advantage of giving us a rough gauge as to how the enriched terms differ when
comparing keloid and normal cells. However, bearing in mind the multiple and complex
roles that a gene could play in a biological system, and also the antagonistic roles of some
genes, this approach would at best give us a very rough idea about the roles that the
enriched terms play in the keloid condition. For a more accurate treatment, detailed
literature review and further experiments have to be done on each of the enriched terms
found.
When comparing keloid to normal fibroblasts, our results indicate an increased
expression of genes involved in ribosomal activity and extracellular matrix and a
decreased expression of genes involved in immune response, locomotory behaviour and
chemokine and cytokine activity, among others. This is an interesting finding that could
be explored in detail in future studies. We also found two KEGG pathways that were
109
significantly enriched when the full list of differentially expressed genes was used. These
pathways are the antigen processing and presentation pathway and the ribosome pathway.
Although fibroblasts are frequently considered just as structural elements of a
tissue, they represent in fact a dynamic cell population that is actively involved not only
in tissue remodeling, but also in autoimmune and inflammatory processes (Smith 2005).
Human leukocyte antigen (HLA) class I molecules are expressed on the surface of all
nucleated human cells including fibroblasts and their role is to present peptides derived
from endogenous proteins to cytotoxic CD8+ T cells(York & Rock 1996; Grommé &
Neefjes 2002; Cresswell et al. 2005). The reduced levels of these class I molecules
implies that keloid fibroblasts are unable to deal with infections, and this result appears to
be consistent with other findings from this study about the lower levels of immunological
factors in keloids. Interferon gamma has been used in the treatment of keloids, and its
efficacy has been attributed to its ability to downregulate collagen synthesis (Larrabee et
al. 1990). Interferon gamma has been known to also increase HLA class I expression in
dermal fibroblasts (Hengel et al. 1995; Zimmer et al. 2006) and this could possibly be
another reason for its success in treating keloids.
It has been previously reported that ribosomal protein (RPS) L23A, RPS10 and
RPS 18 were upregulated in keloids (Satish et al. 2006). Here we find further support for
the upregulation of ribosomal proteins in keloids. Overexpression of several ribosomal
proteins has been reported in carcinomas of the colon (Pogue-Geile et al. 1991) and
breast (Henry, Coggin & King 1993). There is also evidence suggesting that ribosomal
proteins, in addition to participating in protein synthesis, are likely to be involved in other
extraribosomal functions such as DNA replication, transcription and repair, RNA splicing
110
and modification, cell growth and proliferation, regulation of apoptosis and development,
and cellular transformation (Wool 1996; Lai & Xu 2007). All these make the ribosomal
proteins interesting candidates for study as they may shed further light on the keloid
condition.
In conclusion, this study has shown that many of the genes found to be
differentially expressed in keloid fibroblasts when left in serum starvation media are
similar to what has been reported in previous studies, despite the different conditions
used in the other studies. Furthermore, we found no systematic difference when the cells
were harvested at different time points. This study has also revealed a list of novel
differentially expressed genes that could be utilized for further research, out of which the
antigen presentation pathway and ribosomal proteins appear to be interesting candidates.
111
CHAPTER SIX
REVERSE ENGINEERING GENE NETWORKS IN KELOID AND
NORMAL FIBROBLASTS
6.1 Introduction
The genome plays a central role in the control of cellular processes, such as in the
response of cells to environmental signals, the differentiation of cells during
development, and the replication of the genome preceding cell division. A protein
synthesized from the information contained in a coding region of DNA may function as a
transcription factor binding to regulatory sites elsewhere on the DNA, as an enzyme
catalyzing a metabolic reaction, or as a component of a signal transduction pathway.
With few exceptions, all cells in an organism contain the same genetic material. This
implies that in order to understand how genes are implicated in the control of intracellular
and intercellular processes, the scope should be broadened from sequences of nucleotides
encoding proteins to regulatory systems determining which genes are expressed, when
and where in the organism, and to what extent. The goal of reverse engineering methods
is to infer gene networks from observational data, thus providing insight into the inner
workings of a cell (Hartwell et al. 1999; Schadt, Sachs & Friend 2005).
Reverse engineering strategies generally fall within two broad categories “physical” approaches and “influence” approaches (Gardner & Faith 2010). The
modeling of the physical interactions between transcription factors and their promoters is
what is known as the physical approach to reverse engineering. Gene expression is
112
predominantly controlled by auxiliary proteins called transcription factors (TF). A TF
binds directly to a specific upstream region of the target gene known as the promoter
region, which triggers the enzyme, RNA polymerase, to transcribe DNA to RNA. TFs
can therefore be viewed as a class of specialized proteins that govern the on-off switch of
gene expression through either repressing (down-regulation) or inducing (up-regulation)
its output. An advantage of this strategy is that it enables the use of genome sequence
data, in combination with RNA expression data, to enhance the sensitivity and specificity
of predicted interactions but its limitation is that it cannot describe regulatory control by
mechanisms other than transcription factors.
On the other hand, the influence approach abstracts out this mechanistic process
and instead can be viewed just as an input-output device. In other words, it looks for
transcripts that act as “inputs” whose concentration changes can explain the changes in
“output” transcripts. Such a model does not generally describe physical interactions since
transcription is rarely controlled directly by RNA. Nevertheless, in some cases, the input
transcripts may encode the transcription factors that directly regulate transcription. In
such cases, the influence model may accurately reflect a physical interaction. An
advantage of the influence strategy is that the model can implicitly capture regulatory
mechanisms at the protein and metabolite level that are not physically measured but the
limitation of this approach is that the model can be difficult to interpret in terms of the
physical structure of the cell, and therefore difficult to integrate or extend with further
research. Moreover, the implicit description of hidden regulatory factors may lead to
prediction errors.
113
In addition to these two modeling approaches, reverse engineering methods also
differ in terms of the mathematical formalisms used and can be static or dynamic,
continuous or discrete, linear or nonlinear and deterministic or stochastic (Hache,
Lehrach & Herwig 2009). For the purposes of this study, we have chosen to use both the
physical as well as the influence approach for reconstructing the networks. For the
physical approach, we will use the regression method fREDUCE (fast-Regulatory
Element Detection Using Correlation with Expression) (Wu et al. 2007) with the
objective of identifying important cis-binding motifs and their targets in keloid
fibroblasts. For the influence approach, we will compare the performance of the
information theoretic method ARACNE (Algorithm for the Reconstruction of Accurate
Cellular Networks) (Basso et al. 2005) and the Bayesian package BANJO (Bayesian
Network Inference with Java Objects) (Yu et al. 2004) in uncovering regulatory
interactions
in
keloid
and
normal
fibroblasts.
The
effect
of
different
normalization/summarization methods and lowly expressed probes on gene network
inference is also not clear and will be examined in this system.
Generally, the process of modeling gene regulatory networks consists of a few
main steps: designing experiments that produce maximally informative observations,
developing methodologies for choosing a candidate model that „best‟ fits the
observations, analyzing and validating the model, and using the model to formulate and
test new hypotheses (Goutsias & Lee 2007). Microarray data from the previous study will
be used to learn the networks. However, learning the structure of a gene network using
the influence approach is difficult as the number of possibilities scale exponentially with
the number of variables. Therefore, modeling and testing such large structures would
114
require large amounts of data for accuracy. Due to our limited data, we have decided to
focus on small networks of genes that have been found to be differentially expressed
from the second part of this dissertation. Furthermore, to increase the number of samples,
we will use data from other independent microarray experiments performed in our lab
and also data from Smith et al (2008), which is the only keloid fibroblast data publicly
available at the Gene Expression Omnibus (GEO) database. For the physical approach,
since the binding motif repeats are regressed against the expression levels of each gene, it
is the number of genes that constitute the sample size. Therefore, the full range of genes
is used for this approach instead of the smaller transcriptional networks that have found
to be differentially expressed.
Figure 6.1: The general strategy for reverse-engineering transcription control systems (Gardner
& Faith 2010). (1) The experimenter perturbs cells with various treatments to elicit distinct
responses. (2) After each perturbation, the experimenter measures the expression (concentration)
of many or all RNA transcripts in the cells. (3) A learning algorithm calculates the parameters of
a model that describes the transcription control system underlying the observed responses. The
resulting model may then be used in the analysis and prediction of the control system function.
In total, we have four different treatment conditions (serum-treated, serum-free,
hydrocortisone-treated and HDGF-treated) and two different cell derivations (keloid and
normal) from multiple patients. Although our datasets consist of some time-series, the
gap between each time point is very large (in the order of days) and may lead to
115
inaccurate results if used to infer time-series regulatory networks. Therefore, we have
limited our study to steady state conditions with the assumption that each time point is
statistically independent from others. This is a possibly valid assumption as the sampling
time is very long. Furthermore, the genes were not directly perturbed by knockdown or
overexpression in our experiments and it is very likely that the different conditions used
will result in multiple unknown perturbations. As such, inference algorithms such as
dynamic Bayesian networks (which require numerous closely spaced time points) and
differential equation approaches (which require either time series data or knowledge of
perturbations) cannot be applied in our case.
6.2 Algorithms
6.2.1 fREDUCE
This method is an extension to the REDUCE (Regulatory Element Detection Using
Correlation with Expression) algorithm (Bussemaker, Li & Siggia 2001). REDUCE is a
deterministic method that first enumerates oligonucleotides and then identifies words
whose occurrence in promoter sequences correlate most strongly with expression data.
This procedure is applied iteratively to produce a set of oligonucleotides that produce the
best simultaneous fit to the data.
One weakness of REDUCE is that it can miss weak but biologically significant
variants of the regulator site. Highly degenerate motifs whose individual variants fall
below the detection threshold will be missed altogether. This is particularly the case for
116
regulators in higher mammalian genomes, which can exhibit strong site to site variation
in specificity. However, a straightforward extension of REDUCE using exhaustive
enumeration of degenerate motifs becomes impractical when the motif length or number
of degenerate positions increases as the computation of the Pearson correlation
coefficient which is required for identifying the motifs is computationally laborious.
fREDUCE uses the following strategy to efficiently compute the Pearson
coefficients of the most significant degenerate motifs: 1) A list of degenerate motifs that
can be derived from the sequence data is generated. 2) For each degenerate motif, a
“pseudo-Pearson" coefficient, an estimate of the actual Pearson coefficient can be
calculated. The pseudo-Pearson coefficient is guaranteed to be an upper-bound on the
actual Pearson coefficient and is used as a filter to eliminate most (typically >99.9%) of
the motif list. 3) Actual Pearson coefficients are computed and the top motif is found and
4) The contribution from the top motif is subtracted from the expression data to form a
residual, which is used for subsequent rounds of motif searching. This algorithm has been
shown to outperform many of the other motif finding algorithms, including its
predecessor REDUCE (Wu et al. 2007).
6.2.2 ARACNE
Information-theoretic approaches use a pseudo-distance between probability distributions
called Mutual Information (MI), to compare expression profiles from a set of
microarrays. For each pair of genes (i, j), their MIij is computed and the edge aij = aji is set
to 0 or 1 depending on a significance threshold to which MIij is compared. MI can be
117
used to measure the degree of independence between two genes. Mutual information MIij
between gene i and gene j is computed as:
M ij  H i  H j  H ij
where H, the entropy, is defined as:
n
H k   p( xk ) log( p( x k ))
k 1
The entropy Hk has many interesting properties, specifically it reaches a maximum for
uniformly distributed variables, i.e. the higher the entropy, the more randomly distributed
are gene expression levels across the experiments. From the definition, it follows that MI
becomes zero if the two variable xi and xj are statistically independent (p(xixj) = p(xi)p(xj),
since their joint entropy Hij = Hi + Hj. A higher MI indicates that the two genes are nonrandomly associated to each other. It can be easily shown that MI is symmetric, Mij = Mji,
therefore the network is described by an undirected graph G. The definition of MI
requires each data point, i.e. each experiment, to be statistically independent from the
others, thus information-theoretic approaches, as described here, can deal with steadystate gene expression data set, or with time-series data as long as the sampling time is
long enough to assume that each point is independent from the previous ones. Edges in
networks derived by information-theoretic approaches represent statistical dependences
among gene expression profiles.
ARACNE (Algorithm for the Reconstruction of Accurate Cellular Networks)
belongs to the family of information-theoretic approaches to gene network inference by
implementing the relevance network algorithm (Basso et al. 2005). ARACNE computes
Mij for all pairs of genes i and j in the data set. Mij is estimated using the method of
Gaussian kernel density (Steuer et al. 2002). Once Mij for all gene pairs has been
118
computed, ARACNE excludes all the pairs for which the null hypothesis of mutually
independent genes cannot be ruled out (H0 : MIij = 0). A p-value for the null hypothesis,
computed using Monte Carlo simulations, is associated to each value of the mutual
information. The final step of this algorithm is a pruning step based on the Data
Processing Inequality (DPI) principle that tries to reduce the number of false positives.
This principle asserts that if both (i,j) and (j,k) are directly interacting, and (i,k) is
indirectly interacting through j, then M i ,k  min( M ij , M jk ) . This condition is necessary
but not sufficient, i.e. the inequality can be satisfied even if (i,k) are directly interacting,
therefore the authors acknowledge that by applying this pruning step using DPI they may
be discarding some direct interactions as well.
6.2.3 BANJO
A Bayesian network is a graphical model for probabilistic relationships among a set of
random variables Xi, with i = 1…n. These relationships are encoded in the structure of a
directed acyclic graph G whose vertices (or nodes) are the random variables Xi. The
relationships between variables are described by a joint probability distribution
P(X1,…,Xn) that is consistent with the independence assertions embedded in the graph G
and has the form:
n
P( X 1 ,..., X n )   P( X i  xi || X j  x j ,..., X j  p  x j  p )
i 1
where the p+1 genes on which the probability is conditioned are called the parents of
gene i and represent its regulators, and the joint probability density is expressed as a
product of conditional probabilities by applying the chain rule of probabilities and
independence. This rule is based on Bayes theorem: P(A,B) = P(B||A) * P(A) = P(A||B) *
119
P(B). We observe that the joint probabilty distribution can be decomposed as the product
of conditional probabilities only if the Markov assumption holds, i.e. each variable Xi is
independent of its non-descendants, given its parent in the directed acyclic graph G. In
order to reverse-engineer a Bayesian network model of a gene network we must find the
directed acyclic graph G (i.e. the regulators of each transcript) that best describes the
gene expression data D. This is done by choosing a scoring function that evaluates each
graph G (i.e. a possible network topology) with respect to the gene expression data D,
and then searching for the graph G that maximizes the score.
BANJO is a gene network inference software that has been developed by the
group of Hartemink (Yu et al. 2004). In BANJO, heuristic approaches are used to search
the network space to find the network graph G (Proposer/Searcher module in BANJO).
For each network structure explored, the parameters of the conditional probability density
distribution are inferred and an overall network‟s score is computed using the Bayesian
metric with Dirichlet priors and equivalence (BDe) metric in BANJO‟s Evaluator
module. The output network will be the one with the best score (BANJO‟s Decider
module). BANJO outputs a signed directed graph indicating regulation among genes.
BANJO can analyse both steady-state and time-series data. In the case of steady-state
data, BANJO, as well as the other Bayesian networks algorithms, is not able to infer
networks involving cycles (e.g. feedback or feed-forward loops).
120
6.3 Results
6.3.1 Binding motifs found from fREDUCE for keloid versus normal fibroblasts
under serum starvation condition
Binding motifs found using the gene expression values from set A (keloid versus normal
fibroblasts under serum starvation conditions) are shown in Table 6.1. Highlighted motifs
indicate motifs found in at least two variations of the conditions/parameters. Both MAS5
and RMA normalization as well as filtered and unfiltered gene lists provided hits for the
binding motifs. Of particular note are the binding motifs CGCCGA (found in 5 of the
conditions), GCCGAC (found in 3 of the conditions), CTTCTT (found in 3 of the
conditions) and CACATAT (found in 3 of the conditions). A search through the
TRANSFAC database did not produce any results for the binding motif CACATAT, but
found possible gene targets for CGCCGA (MYB), GCCGAC (ATF2) and CTTCTT
(ADH1) (Table 6.2).
Table 6.1: Binding motifs found from fREDUCE for keloid versus normal fibroblasts under
serum starvation condition
Normalization
Parameters
Binding Motif
p-value
MAS 5
(unfiltered)
Length 7
(0 IUPAC)
CCGGCC
GCCGAC
CGTAGC
CGCBGA
MCGGAA
GCCGAC
CACATAT
CCGGCC
GBCGAC
CACATAT
CGCCGA
CTTCTT
CGTAGC
5.31
1.99
1.22
5.30
1.42
3.35
2.56
1.12
3.56
2.02
2.86
1.25
1.11
Length 7
(1 IUPAC)
RMA
(unfiltered)
Length 7
(0 IUPAC)
Length 7
(1 IUPAC)
MAS 5
(filtered)
Length 7
(0 IUPAC)
121
Length 7
(1 IUPAC)
RMA
(filtered)
Length 7
(0 IUPAC)
Length 7
(1 IUPAC)
CGCCBA
(3.65)
CCTTCYT
(0.27)
CGCCGA
TATACAC
CACAKAT
CGCCGA
CTTCTTV
3.65
0.27
2.58
1.95
2.33
2.03
0.18
Note: P-values are shown as -log10 values.
IUPAC characters M = C/A; Y = C/T; K = T/G; B = C/T/G, V = A/C/G
Table 6.2: Possible gene targets and TFs found from the TRANSFAC database for top binding
motifs from Table 6.1
Binding Motif
CCGGCC
GCCGAC
CGTAGC
CGCCGA
CTTCTT
CACATAT
Possible gene targets
Possible TFs
MC2R (melanocortin 2 receptor)
MT1G (metallothionein 1G)
EPO (erythropoietin)
SURF1 and SURF2 (surfeit 1 and 2)
ATF2 (activating transcription factor 2)
c-myb
ADH1 (alcohol dehydrogenase)
-
SF-1
Tf-LF1 and Tf-LF2
YY1
SP1
MZF-1
-
6.3.2 Binding motifs found from fREDUCE for keloid versus normal fibroblasts
under serum induced condition
No binding motifs were found for unfiltered RMA normalized set B (keloid versus
normal fibroblasts under serum conditions), but binding motifs were found for the other
conditions (Table 6.3). Of particular note is the binding motif GGGGCTC which was
found to be consistent in 4 of the conditions, although all these 4 conditions were using
the MAS 5 normalization. A search through the TRANSFAC database found ADA as a
possible gene with this binding motif (Table 6.4).
122
Table 6.3: Binding motifs found from fREDUCE for keloid versus normal fibroblasts under
serum induced condition
Normalization
Parameters
Binding Motif
p-value
MAS 5
(unfiltered)
Length 7
(0 IUPAC)
CCACACA
GGGGCTC
CCACACA
GGVCTC
AGGCAH
GGGGCTC
2.44
2.19
2.14
1.91
1.30
2.28
GGGGHTC
CGAGRA
GCGCCA
GTCCCG
GTCVCG
CAACGW
2.56
0.11
2.52
1.46
4.29
0.95
Length 7
(1 IUPAC)
MAS 5
(filtered)
RMA
(filtered)
Length 7
(0 IUPAC)
Length 7
(1 IUPAC)
Length 7
(0 IUPAC)
Length 7
(1 IUPAC)
Note: P-values are shown as -log10 values.
IUPAC characters R = A/G; W = T/A; H = A/T/C, V = A/C/G
Table 6.4: Possible gene targets and TFs found from the TRANSFAC database for top binding
motifs from Table 6.3
Binding Motif
CCACACA
GGGGCTC
GTCCCG
Possible gene targets
Possible TFs
ADA (adenosine deaminase)
EGFR (EGF receptor)
ATF2 (activating transcription factor 2)
CCNE1 (cyclin E1)
MET (hepatocyte growth factor receptor)
SP1
SP1
E2F-1
PAX-3
6.3.3 Binding motifs found from fREDUCE for sets C and D suggest consistent
effects from steroid induction for both keloid and normal fibroblasts
Binding motifs were found for set C (keloid treated with steroid versus serum induced
keloid fibroblasts) and D (normal treated with steroid versus serum induced normal
fibroblasts) when fREDUCE was run using parameters length 6 with 0 IUPAC
123
substitutions. Other parameters did not produce any results. Furthermore, results were
only obtained when MAS 5 normalization was used. The effect of hydrocortisone appears
to be realized through the binding motifs GGAGGG and GCCCCC and this was
consistent for both keloid (Table 6.5) and normal (Table 6.6) fibroblasts. A search
through the TRANSFAC database using these binding motifs found a large list of genes
containing these binding motifs, including COL1A2, FN, TGFB1, PDGF1 and IGF2
(Table 6.7). Of particular note is the fact that most of the genes found in this list have SP1
as its transcription factor (Table 6.7).
Table 6.5: Binding motifs found from fREDUCE for steroid treated versus control keloid
fibroblasts
Normalization
Parameters
Binding Motif
p-value
MAS 5
(filtered)
Length 6
(0 IUPAC)
GGAGGG
GCCCCC
CCTGGG
TGTGTG
GGCTGG
CTGTGC
AAACAC
GGWGGG
CCDGGG
CTCCCH
TGTGDG
HACGAA
ACCGCD
CVGTAA
24.62
11
7.33
3.93
3.45
1.73
1.32
30.68
12.92
6.23
4.52
3.63
2.03
0.91
Length 6
(1 IUPAC)
Note: P-values are shown as -log10 values.
IUPAC characters W = T/A; H = A/T/C; V = A/C/G; D = A/T/G
Table 6.6: Binding motifs found from fREDUCE for steroid treated versus control normal
fibroblasts
Normalization
Parameters
Binding Motif
p-value
MAS 5
(filtered)
Length 6
(0 IUPAC)
GCCCCC
GGAGGG
CTGGGG
30.08
19.87
10.31
124
TGGGCC
CCCAGA
AGAACG
TGGGTG
GCGAAA
CCTGAG
5
2.67
2.44
2.25
1.53
1.19
Note: P-values are shown as -log10 values.
Table 6.7: Possible gene targets and TFs found from the TRANSFAC database for top binding
motifs from Tables 6.5 and 6.6
Binding Motif
GGAGGG
GCCCCC
Possible gene targets
Possible TFs
EPO (erythropoietin)
ATF2 (activating transcription factor 2)
RARG (retinoic acid receptor, gamma)
ACTC1 (actin, alpha, cardiac muscle 1)
FN (fibronectin)
c-myc
CEACAM5 (carcinoembryonic antigen-related cell
adhesion molecule 5
CYP17 (cytochrome P450, subfamily XVII)
SFTPB (surfactant protein B)
PDGFA (platelet derived growth factor A chain)
ADA (adenosine deaminase)
SA-ACT (skeletal alpha actin)
MIP (major intrinsic protein of lens fiber)
c-myb
COL1A2 (collagen I alpha 2)
ALDC (aldolase C)
TGFB1 (transforming growth factor beta 1)
apoE (apolipoprotein E)
c-jun
ACTC1 (actin, alpha, cardiac muscle 1)
ATF2 (activating transcription factor 2)
apoB (apolipoprotein B)
GFAP (glial fibrillary acidic protein)
Cyclin D1
Insulin
ALDC (aldolase C)
HRAS (transforming protein p21)
PFKM (muscle phosphofructokinase)
DRD1 (dopamine receptor D1)
IGF2 (insulin-like growth factor 2)
Tf-LF1 and Tf-LF2
SP1
SP1
SP1
Pur factor
SP1
PBX1B
NKX2-1
SP1, WT1
SP1
COUP-TF2
SP1
MZF-1
SP1
SP1 and AP1
SP1
SP1
SP1
NF1, SP1
c-Ets-2
SP1
SP1
WT1
125
6.3.4 Not many binding motifs found from fREDUCE for sets E and F
fREDUCE found few binding motifs for set E (keloid versus normal fibroblasts both
treated with steroid) and no binding motifs for set F (keloid treated with HDGF versus
untreated keloid fibroblasts). Binding motifs for set E were found only when the MAS 5
unfiltered condition and the RMA filtered condition were used (Table 6.8). Furthermore,
binding motifs found in these conditions were not very consistent. A search through the
TRANSFAC database using the top binding motifs from Table 6.8 found EGFR, ADM
and CGA as possible gene targets (Table 6.9).
Table 6.8: Binding motifs found from fREDUCE for keloid versus normal fibroblasts under
steroid treated condition
Normalization
Parameters
Binding Motif
p-value
MAS 5
(unfiltered)
Length 6
(0 IUPAC)
Length 6
(1 IUPAC)
CGCCGC
1.53
CGCCGC
GCGYTT
GGGTTG
CGTTTT
AGCGAC
1.23
1.42
2.75
1.80
1.73
RMA
(filtered)
Length 7
(0 IUPAC)
Note: P-values are shown as -log10 values
IUPAC character Y = C/T
Table 6.9: Possible gene targets and TFs found from the TRANSFAC database for top binding
motifs from Table 6.8
Binding Motif
CGCCGC
GGGTTG
Possible gene targets
Possible TFs
EGFR (EGF receptor)
ADM (adrenomedullin)
CGA (glycoprotein hormone alpha subunit)
SP1
TFAP2A
-
126
6.3.5 Mean sensitivity performance of BANJO in recovering influence networks was
significantly better than that of ARACNE
On average, BANJO was significantly more sensitive compared to ARACNE in
recovering influence networks (Fig. 6.2C). However, there was no significant difference
in average accuracy (PPV) between BANJO and ARACNE (Fig. 6.2A). Furthermore,
there was no significant difference between RMA and MAS 5 normalization both in
terms of mean accuracy (PPV) (Fig. 6.2B) as well as mean sensitivity (Fig. 6.2D)
although p-values were fairly close to 0.05, with RMA being the better choice for both
measures.
A
B
p = 0.422
p = 0.086
0.6
0.6
0.499857143
0.472803571
0.478517857
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0.451464286
0
0
ARACNE (ppv)
RMA (ppv)
BANJO (ppv)
C
D
MAS 5 (ppv)
p = 0.076
* p = 0.002
0.9
0.8
0.7
0.7195
0.748857143
0.8
0.633142857
0.7
0.603785714
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
ARACNE sensitivity
BANJO sensitivity
RMA sensitivity
MAS 5 sensitivity
Figure 6.2: Comparison between ARACNE, BANJO, RMA and MAS 5 based on PPV and
sensitivity values. (A) ARACNE (ppv) compared with BANJO (ppv). (B) RMA (ppv) compared
with MAS 5 (ppv) (C) ARACNE (sensitivity) compared with BANJO (sensitivity) (D) RMA
(sensitivity) compared with MAS 5 (sensitivity). Bar graphs represent mean ± S.E.M values. *
indicates statistical significance as assessed by the paired t-test.
127
6.3.6 Transcriptional networks were better suited for network inference compared
to cytokine receptor interactions and intracellular signaling networks
Transcriptional networks (Fig. 3.2C, E and G) were better suited for network inference
compared to cytokine receptor interactions (Fig. 3.2A and B) and intracellular signaling
networks (Fig. 3.2D and F). The full list of results for all data sets is given in Table 6.10,
with bold typeface indicating performance better than random. In particular, RMA
normalization for transcriptional networks (Networks C, E and G) had consistently better
accuracy (PPV) compared to random, and also had sensitivity values higher than 0.5,
regardless of the algorithm used (Table 6.10). BANJO appears to perform particularly
well for intracellular signaling network F, but did not do so well for network D. For the
NFKB transcriptional network (network E), performance using keloid sets were
consistently lower than performance using normal sets, but there was very little
difference in performance between the keloid and normal datasets for the other networks
(Table 6.10).
Table 6.10: PPV and sensitivity results for all data sets run using BANJO and ARACNE
Data sets
Keloid RMA
Normal RMA
Keloid MAS5
Normal MAS5
Random
A
ARACNE
BANJO
PPV
Se
PPV
Se
0.238 0.556
0.3 0.667
0.1875 0.333 0.3125 0.556
0.214 0.333
0.286 0.667
0.25 0.667
0.304 0.778
0.389
B
ARACNE
BANJO
PPV
Se
PPV
Se
0.6
1
0.4 0.667
0.5 0.667
0.4 0.667
0.75
0.5 0.667
1
0.75
1
0.6
1
0.5
C
Data sets
Keloid RMA
Normal RMA
ARACNE
PPV
Se
0.8
1
1
0.75
D
BANJO
PPV
Se
0.75
0.75
0.8
1
ARACNE
PPV
Se
0.5
0.6
0.5
0.6
BANJO
PPV
Se
0.5
0.8
0.2
0.2
128
Keloid MAS5
Normal MAS5
Random
0.5
0.667
0.25
0.6
0.5
0.5
0.667
0.75
0.5
E
Data sets
Keloid RMA
Normal RMA
Keloid MAS5
Normal MAS5
Random
0.5
0.333
0.625
0.6
0.4
0.5
0.5
1
0.8
F
ARACNE
BANJO
PPV
Se
PPV
Se
0.375
0.364
0.6
0.8
0.444
0.444
0.8
0.8
0.286
0.4
0.273
0.6
0.364
0.333
0.6
0.8
0.333
ARACNE
BANJO
PPV
Se
PPV
Se
0.5
0.375
0.75
1
0.5
0.5
0.75
1
0.429
0.286
0.5
0.75
0.429
0.2
0.25
0.75
0.4
G
Data sets
Keloid RMA
Normal RMA
Keloid MAS5
Normal MAS5
Random
ARACNE
PPV
Se
0.5
0.5
0.667 0.667
0.4 0.333
0.833 0.833
BANJO
PPV
Se
0.714 0.833
0.625 0.833
0.429
0.5
0.5
0.5
0.6
6.4 Discussion
Reverse engineering gene networks from expression data is a considerably difficult
problem, with challenges arising from the nature of the data which is typically noisy, high
dimensional, and significantly undersampled. Most evaluations of reverse engineering
techniques are done on simulated data (Camacho et al. 2007; Hache, Lehrach & Herwig
2009) although some have extended this to small sets of experimental data (Bansal et al.
2007; Cantone et al. 2009). While simulated data can model the high dimensionality as
well as the indeterminacy of the problem accurately, the nature of noise as well as the
underlying function governing the regulatory interactions has to be assumed a priori. A
major problem of working with experimental data, however, is that not enough is known
129
about the real networks and this could lead to difficulties in validating the inferred
networks.
Our results from the physical approach show that MAS 5 normalization was better
suited for the recovery of significant binding motifs as more binding motifs were
obtained when the fREDUCE method was used. However, the results from influence
methods show that RMA is better for the inference of gene networks especially for the
case of transcriptional networks. The performance of different normalization approaches
have been assessed in a previous study by Lim et al. In the study, the Spearman rank
correlation was used to compare between gene expression profile pairs from replicate
samples as well as from samples with randomly permuted probe values (Lim et al. 2007).
The authors found that the GCRMA procedure produced significant correlation artefacts
(false positives), and that the MAS 5 procedure was best suited for the reverse
engineering process. However, for most of their tests, RMA performs similarly, albeit at
a lower level, compared to MAS 5. Here, we report that the performance of RMA or
MAS 5 normalization appears to be dependent on the type of inference done.
The physical approach using the fREDUCE algorithm found binding motifs that
were active in keloid fibroblasts compared to normal fibroblasts under various conditions.
fREDUCE also found binding motifs that were responsive to steroid treatment. One
limitation of the fREDUCE algorithm is that it cannot determine which TFs bind to the
discovered motifs. By manually searching through the TRANSFAC database, we are able
to get some idea about target genes containing these motifs, as well as possible
transcription factors that bind to these motifs. The TRANSFAC database is not complete
130
however, as it is based on published data, therefore undiscovered interactions will not be
reflected in the database.
Our results suggest that steroid treatment affected both keloid and normal
fibroblasts in a similar fashion as the top binding motifs found when these two cell types
were treated with hydrocortisone were the same. Many of the possible gene targets
containing these binding motifs are involved in wound healing, for example fibronectin,
erythropoietin, PDGF, COL1A1 and TGFB. This is consistent with the fact that steroids
are known to have a depressive effect on wound healing (Wicke et al. 2000).
Furthermore, SP1 was the most common TF found for these gene targets. This result
suggests that hydrocortisone exerts its depressive effect on fibroblasts by affecting the
activity of SP1, and could be a future area of research. There were fewer gene targets
found for binding motifs that were active when comparing keloid to normal fibroblasts.
Furthermore, the TFs found for these conditions were also less consistent. This could be
due to the fact that the keloid condition is a result of the effect of multiple TFs, and
unlike the effect of hydrocortisone, no single TF is most responsible.
The success of fREDUCE depends on a number of assumptions regarding the
dynamics of transcription. Most notably, it relates the influence of combinations of TFs
as a log-linear function of RNA levels. Such a highly constrained model may lead to
errors in predictions. Furthermore, it assumes that the 1000 base pairs upstream of the
transcription start site play some role in the regulation of the gene. Despite these
limitations, fREDUCE has been used successfully to discover binding motifs in human
liver tissue (Wu et al. 2007).
131
We used fairly naïve methods in the preprocessing of data for our influence based
inference methods. To enable comparison between multiple datasets, we normalized the
expression values with the average of GAPDH and B-actin expression values for each
individual chip whereas for discretization, we used the 7 bin quantile discretization that is
available in BANJO. More sophisticated discretization techniques (Friedman et al. 2000;
Pe'er et al. 2001; Becquet et al. 2002) might potentially produce better results.
Due to our limited data, it would be unwise to run the influence algorithms on the
full list of genes. The subsets of genes that we selected were based on KEGG pathways
that were found to be significantly enriched by the differentially expressed genes found in
the previous section. We further subdivided the pathways found into three major groups –
cytokine receptor interactions, transcriptional networks and intracellular signaling. These
lists are by no means complete, and there is bound to be many hidden factors and
feedback interactions that were not explicitly modeled or taken into account. However, in
the absence of further biological knowledge to guide us in our selection, this seemed to
be the most logical step to take. Futhermore, it is hoped that the influence methods, being
of the „black box‟ variety, would be able to cope with these deficiencies.
Our results show that both ARACNE and BANJO seem to perform better for
transcriptional networks compared to cytokine receptor interations or intracellular
signaling networks. This makes some intuitive sense as there is a causal link between
transcription factors and their target genes, whereas in cytokine receptor interactions and
in intracellular signaling, there is at most only a correlation (in most cases, there may not
even be a correlation). However, a cellular signaling network has been successfully
reconstructed previously using a Bayesian approach (Sachs et al. 2005). It should be
132
noted however that Sach‟s study measured phosphorylation levels in addition to
expression levels and used the flow cytometry platform instead of microarray expression
data. On a related note, it is worth pointing out that influence methods using microarray
data do not take the actual binding affinities of transcription factors into consideration as
only expression values are used. This could be a source of inaccuracies in the networks
inferred.
Between the two influence methods tested, BANJO produced significantly better
results compared to ARACNE. The superior performance of BANJO for small data sets
with „global‟ perturbations can also be seen in the results of the in silico study done by
Bansal et al. (Bansal et al. 2007). The lower performance of ARACNE could be due to
the small number of data sets used; ARACNE has been recommended to be used on data
sets containing a minimum of 100 microarray expression profiles as this represents an
empirical lower bound on the amount of data needed to estimate the MI reliably
(Margolin et al. 2006). Having said that, none of the PPV values obtained either through
BANJO or ARACNE was able to beat the random score significantly as assessed by the
chi-squared test, although the absolute PPV values were higher in some of the cases.
Achieving statistical significance with a small number of genes requires the difference in
data distributions to be very large and may be too demanding for our small networks.
The fact that performance for the transcriptional network involving NFKB
(nuclear factor kappa-light-chain-enhancer of activated B cells, Set E) using influence
methods was better for normal fibroblasts compared to keloid fibroblasts suggests that the
influence between NFKB and its targets was weaker in keloid fibroblasts, or that there
were more links in the keloid network that were not captured by our simplified diagram.
133
This would imply that targeting of NFKB alone may not be sufficient in reducing the
expression of its targets in keloids. However, more work needs to be done to verify this
hypothesis.
The ability to infer molecular interactions in cellular systems is one of the most
exciting promises of systems biology. As the most widely available high throughput
technology, gene expression microarrays provide a good test set for the application of
inference algorithms that infer dynamic models from static, genome-scale data. However,
the critical assumption underlying this methodology is that mRNA measurements are
predictive of molecular activity. This assumption has been thrown into question as new
studies reveal the substantial role of alternate regulatory mechanisms, such as translation,
post-translational modifications, genetic and epigenetic factors, as well as the
increasingly appreciated regulatory role of non-coding RNAs. Furthermore, data from the
microarray platform is typically noisy, and is also hidden in multiple probes that can be
combined in multiple ways to produce different expression values. Yet in spite of all
these difficulties, the topic of reverse engineering gene networks is surely worth
pursuing, as it provides us with a means of understanding biology not only in terms of the
genes themselves, but also through their interactions.
134
CHAPTER SEVEN
CONCLUSION
“Now they [genes] are trapped in huge colonies, locked inside highly intelligent beings,
moulded by the outside world, communicating with it by complex processes, through
which, blindly, as if by magic, function emerges."
– Denis Noble, Emeritus Professor, University of Oxford
The reductionist approach to biological problems has had tremendous success in the past,
culminating in the discovery of complete genomes of several organisms, including
humans. Thanks to molecular biology, we now understand how linear arrangement of
nucleotides encodes linear arrangement of amino acids and how proteins interact to form
functional groups such as signal transduction and metabolic pathways. In recent years
however, there has been a greater realization about the limitations of reductionist
approaches. Having reduced the biological universe to a myriad of minute parts, we are
now unable to assemble them back together again in manner that increases our biological
understanding. This has led to the development of the nascent field of systems biology,
where methods adapted from math and engineering disciplines are employed to shed light
on the complex cellular networks present in living organisms.
We have utilized both reductionist as well as top down approaches in this
dissertation in an attempt to better understand keloid scarring. We first investigated the
novel growth factor HDGF and its role in the keloid system using cell and molecular
techniques. Our results suggest that this growth factor is upregulated in the keloid
135
condition and has angiogenic and proliferative potential. However, it does not seem to
increase extracellular matrix production and thus HDGF appears to be just one of a
myriad number of players underlying the keloid condition. Targeting HDGF alone would
probably not be sufficient in ameliorating this condition.
In the second part of this dissertation, we decided to take a top down approach in
an attempt to identify groups of genes that can be implicated in the formation of keloids.
Surveying the global transcriptional landscape is now possible with the advent of
microarray technology, although this platform is still far from perfect as it is inherently
noisy. Furthermore, there is still no single standard for the processing of raw data for the
Affymetrix Genechips, and different algorithms result in different expression values for
the same raw data. Despite all these shortcomings, the microarray platform has been
widely used to profile mRNA expression values, and our results indicate consistency with
previous studies done on keloid fibroblasts. We have also uncovered differentially
expressed genes that have not been reported previously, and enrichment analysis indicate
that processes such as immune response, antigen processing and presentation, chemokine
and cytokine activity, extracellular matrix and ribosomal proteins are among those that
are affected in the keloid condition.
In the third part of this dissertation, we have attempted to reverse engineer gene
networks from the collection of microarray expression data that we have collected. Using
the physical approach of correlating expression values to binding motifs, we found some
consensus sequences that were active in the keloid condition, as well as some sequences
that were responsive to steroid treatment. Using influence approaches on experimental
data, we found that the combination of the Bayesian algorithm, RMA normalization and
136
transcriptional networks gave the best results. Furthermore, our results show that the
NFKB transcriptional network inferred from normal fibroblast data was more accurate
than that inferred from keloid data, suggesting a more robust network in the keloid
condition.
We are still far from a solution to the keloid condition, but it is hoped that the
work done here is a small step towards finding this solution. In this thesis, we have
mainly focused on fibroblasts, which is only one of the possible cell types involved in
wound healing. Future work could involve microarray profiling of keratinocytes, as well
as profiling of cells in the co-culture condition to examine the effect of epithelial
mesenchymal interactions on transcriptional networks. Furthermore, novel genes
obtained from our microarray experiments, or those containing the binding motifs that
were active in the various conditions, could be studied in depth through reductionist
approaches. Pathways such as the antigen processing and presentation pathway and the
toll-like receptor signaling pathway keloid and normal fibroblasts could also be examined
in further detail. Finally, improved methods that model alternate regulatory mechanisms
such as post translational modifications and binding affinities could be developed to
improve the accuracy of the reverse engineering process.
137
BIBLIOGRAPHY
Aarabi, S, Bhatt, KA, Shi, Y, Paterno, J, Chang, EI, Loh, SA, Holmes, JW, Longaker,
MT, Yee, H & Gurtner, GC 2007, 'Mechanical load initiates hypertrophic scar formation
through decreased cellular apoptosis', FASEB J, vol. 21, no. 12, pp. 3250-3261.
Abouzied, MM, Baader, SL, Dietz, F, Kappler, J, Gieselmann, V & Franken, S 2004,
'Expression patterns and different subcellular localization of the growth factors HDGF
(hepatoma-derived growth factor) and HRP-3 (HDGF-related protein-3) suggest
functions in addition to their mitogenic activity', Biochem J, vol. 378, no. Pt 1, pp. 169176.
Abouzied, MM, El-Tahir, HM, Prenner, L, Häberlein, H, Gieselmann, V & Franken, S
2005, 'Hepatoma-derived growth factor. Significance of amino acid residues 81-100 in
cell surface interaction and proliferative activity', J Biol Chem, vol. 280, no. 12, pp.
10945-10954.
Al-Attar, A, Mess, S, Thomassen, JM, Kauffman, CL & Davison, SP 2006, 'Keloid
pathogenesis and treatment', Plast Reconstr Surg, vol. 117, no. 1, pp. 286-300.
Apfelberg, DB, Maser, MR, White, DN & Lash, H 1989, 'Failure of carbon dioxide laser
excision of keloids', Lasers Surg Med, vol. 9, no. 4, pp. 382-388.
Appleton, I, Brown, NJ & Willoughby, DA 1996, 'Apoptosis, necrosis, and proliferation:
possible implications in the etiology of keloids', Am J Pathol, vol. 149, no. 5, pp. 14411447.
Atiyeh, BS, Costagliola, M & Hayek, SN 2005, 'Keloid or hypertrophic scar: the
controversy: review of the literature', Ann Plast Surg, vol. 54, no. 6, pp. 676-680.
Babu, M, Diegelmann, R & Oliver, N 1989, 'Fibronectin is overproduced by keloid
fibroblasts during abnormal wound healing', Mol Cell Biol, vol. 9, no. 4, pp. 1642-1650.
Bansal, M, Belcastro, V, Ambesi-Impiombato, A & di Bernardo, D 2007, 'How to infer
gene networks from expression profiles', Mol Syst Biol, vol. 3, pp. 78.
Basso, K, Margolin, AA, Stolovitzky, G, Klein, U, Dalla-Favera, R & Califano, A 2005,
'Reverse engineering of regulatory networks in human B cells', Nat Genet, vol. 37, no. 4,
pp. 382-390.
Bayat, A, Arscott, G, Ollier, WER, Ferguson, MWJ & Mc Grouther, DA 2004,
'Description of site-specific morphology of keloid phenotypes in an Afrocaribbean
population', Br J Plast Surg, vol. 57, no. 2, pp. 122-133.
138
Becquet, C, Blachon, S, Jeudy, B, Boulicaut, JF & Gandrillon, O 2002, 'Strongassociation-rule mining for large-scale gene-expression data analysis: a case study on
human SAGE data', Genome Biol, vol. 3, no. 12, pp. RESEARCH0067.
Bernard, K, Litman, E, Fitzpatrick, JL, Shellman, YG, Argast, G, Polvinen, K, Everett,
AD, Fukasawa, K, Norris, DA, Ahn, NG & Resing, KA 2003, 'Functional proteomic
analysis of melanoma progression', Cancer Res, vol. 63, no. 20, pp. 6716-6725.
Bloom, D 1956, 'Heredity of keloids; review of the literature and report of a family with
multiple keloids in five generations', N Y State J Med, vol. 56, no. 4, pp. 511-519.
Bock, O, Schmid-Ott, G, Malewski, P & Mrowietz, U 2006, 'Quality of life of patients
with keloid and hypertrophic scarring', Arch Dermatol Res, vol. 297, no. 10, pp. 433-438.
Bolstad, BM, Irizarry, RA, Astrand, M & Speed, TP 2003, 'A comparison of
normalization methods for high density oligonucleotide array data based on variance and
bias', Bioinformatics, vol. 19, no. 2, pp. 185-193.
Bussemaker, HJ, Li, H & Siggia, ED 2001, 'Regulatory element detection using
correlation with expression', Nat Genet, vol. 27, no. 2, pp. 167-171.
Caldwell, GM, Jones, C, Gensberg, K, Jan, S, Hardy, RG, Byrd, P, Chughtai, S, Wallis,
Y, Matthews, GM & Morton, DG 2004, 'The Wnt antagonist sFRP1 in colorectal
tumorigenesis', Cancer Res, vol. 64, no. 3, pp. 883-888.
Camacho, D, Vera Licona, P, Mendes, P & Laubenbacher, R 2007, 'Comparison of
reverse-engineering methods using an in silico network', Ann N Y Acad Sci, vol. 1115, pp.
73-89.
Cantone, I, Marucci, L, Iorio, F, Ricci, MA, Belcastro, V, Bansal, M, Santini, S, di
Bernardo, M, di Bernardo, D & Cosma, MP 2009, 'A yeast synthetic network for in vivo
assessment of reverse-engineering and modeling approaches', Cell, vol. 137, no. 1, pp.
172-181.
Chang, HY, Chi, JT, Dudoit, S, Bondre, C, van de Rijn, M, Botstein, D & Brown, PO
2002, 'Diversity, topographic differentiation, and positional memory in human
fibroblasts', Proc Natl Acad Sci U S A, vol. 99, no. 20, pp. 12877-12882.
Chang, KC, Tai, MH, Lin, JW, Wang, CC, Huang, CC, Hung, CH, Chen, CH, Lu, SN,
Lee, CM, Changchien, CS & Hu, TH 2007, 'Hepatoma-derived growth factor is a novel
prognostic factor for gastrointestinal stromal tumors', Int J Cancer, vol. 121, no. 5, pp.
1059-1065.
Chen, W, Fu, X, Sun, X, Sun, T, Zhao, Z & Sheng, Z 2003, 'Analysis of differentially
expressed genes in keloids and normal skin with cDNA microarray', J Surg Res, vol. 113,
no. 2, pp. 208-216.
139
Cheng, JC, Evans, JH, Leung, KS, Clark, JA, Choy, TT & Leung, PC 1984, 'Pressure
therapy in the treatment of post-burn hypertrophic scar--a critical look into its usefulness
and fallacies by pressure monitoring', Burns Incl Therm Inj, vol. 10, no. 3, pp. 154-163.
Chin, GS, Liu, W, Peled, Z, Lee, TY, Steinbrech, DS, Hsu, M & Longaker, MT 2001,
'Differential expression of transforming growth factor-beta receptors I and II and
activation of Smad 3 in keloid fibroblasts', Plast Reconstr Surg, vol. 108, no. 2, pp. 423429.
Chipev, CC, Simman, R, Hatch, G, Katz, AE, Siegel, DM & Simon, M 2000,
'Myofibroblast phenotype and apoptosis in keloid and palmar fibroblasts in vitro', Cell
Death Differ, vol. 7, no. 2, pp. 166-176.
Chodon, T, Sugihara, T, Igawa, HH, Funayama, E & Furukawa, H 2000, 'Keloid-derived
fibroblasts are refractory to Fas-mediated apoptosis and neutralization of autocrine
transforming growth factor-beta1 can abrogate this resistance', Am J Pathol, vol. 157, no.
5, pp. 1661-1669.
Clark, RA 2001, 'Fibrin and wound healing', Ann N Y Acad Sci, vol. 936, pp. 355-367.
Collett-Solberg, PF & Cohen, P 1996, 'The role of the insulin-like growth factor binding
proteins and the IGFBP proteases in modulating IGF action', Endocrinol Metab Clin
North Am, vol. 25, no. 3, pp. 591-614.
Cresswell, P, Ackerman, AL, Giodini, A, Peaper, DR & Wearsch, PA 2005, 'Mechanisms
of MHC class I-restricted antigen processing and cross-presentation', Immunol Rev, vol.
207, pp. 145-157.
Curwen, V, Eyras, E, Andrews, TD, Clarke, L, Mongin, E, Searle, SMJ & Clamp, M
2004, 'The Ensembl automatic gene annotation system', Genome Res, vol. 14, no. 5, pp.
942-950.
Dai, M, Wang, P, Boyd, AD, Kostov, G, Athey, B, Jones, EG, Bunney, WE, Myers, RM,
Speed, TP, Akil, H, Watson, SJ & Meng, F 2005, 'Evolving gene/transcript definitions
significantly alter the interpretation of GeneChip data', Nucleic Acids Res, vol. 33, no. 20,
pp. e175.
Daian, T, Ohtsuru, A, Rogounovitch, T, Ishihara, H, Hirano, A, Akiyama-Uchida, Y,
Saenko, V, Fujii, T & Yamashita, S 2003, 'Insulin-like growth factor-I enhances
transforming growth factor-beta-induced extracellular matrix protein production through
the P38/activating transcription factor-2 signaling pathway in keloid fibroblasts', J Invest
Dermatol, vol. 120, no. 6, pp. 956-962.
Datubo-Brown, DD 1990, 'Keloids: a review of the literature', Br J Plast Surg, vol. 43,
no. 1, pp. 70-77.
140
Dennis, G, Sherman, BT, Hosack, DA, Yang, J, Gao, W, Lane, HC & Lempicki, RA
2003, 'DAVID: Database for Annotation, Visualization, and Integrated Discovery',
Genome Biol, vol. 4, no. 5, pp. P3.
Derynck, R & Zhang, YE 2003, 'Smad-dependent and Smad-independent pathways in
TGF-beta family signalling', Nature, vol. 425, no. 6958, pp. 577-584.
Diegelmann, RF, Cohen, IK & Kaplan, AM 1981, 'The role of macrophages in wound
repair: a review', Plast Reconstr Surg, vol. 68, no. 1, pp. 107-113.
English, RS & Shenefelt, PD 1999, 'Keloids and hypertrophic scars', Dermatol Surg, vol.
25, no. 8, pp. 631-638.
Enomoto, H, Yoshida, K, Kishima, Y, Kinoshita, T, Yamamoto, M, Everett, AD,
Miyajima, A & Nakamura, H 2002, 'Hepatoma-derived growth factor is highly expressed
in developing liver and promotes fetal hepatocyte proliferation', Hepatology, vol. 36, no.
6, pp. 1519-1527.
Everett, AD, Lobe, DR, Matsumura, ME, Nakamura, H & McNamara, CA 2000,
'Hepatoma-derived growth factor stimulates smooth muscle cell growth and is expressed
in vascular development', J Clin Invest, vol. 105, no. 5, pp. 567-575.
Everett, AD, Stoops, T & McNamara, CA 2001, 'Nuclear targeting is required for
hepatoma-derived growth factor-stimulated mitogenesis in vascular smooth muscle cells',
J Biol Chem, vol. 276, no. 40, pp. 37564-37568.
Everett, AD, Narron, JV, Stoops, T, Nakamura, H & Tucker, A 2004, 'Hepatoma-derived
growth factor is a pulmonary endothelial cell-expressed angiogenic factor', Am J Physiol
Lung Cell Mol Physiol, vol. 286, no. 6, pp. L1194-L1201.
Ezan, J, Leroux, L, Barandon, L, Dufourcq, P, Jaspard, B, Moreau, C, Allières, C, Daret,
D, Couffinhal, T & Duplàa, C 2004, 'FrzA/sFRP-1, a secreted antagonist of the WntFrizzled pathway, controls vascular cell proliferation in vitro and in vivo', Cardiovasc
Res, vol. 63, no. 4, pp. 731-738.
Feghali, CA & Wright, TM 1999, 'Identification of multiple, differentially expressed
messenger RNAs in dermal fibroblasts from patients with systemic sclerosis', Arthritis
Rheum, vol. 42, no. 7, pp. 1451-1457.
Ford, LC, King, DF, Lagasse, LD & Newcomer, V 1983, 'Increased androgen binding in
keloids: a preliminary communication', J Dermatol Surg Oncol, vol. 9, no. 7, pp. 545547.
Friedman, N, Linial, M, Nachman, I & Pe'er, D 2000, 'Using Bayesian networks to
analyze expression data', J Comput Biol, vol. 7, no. 3-4, pp. 601-620.
141
Funayama, E, Chodon, T, Oyama, A & Sugihara, T 2003, 'Keratinocytes promote
proliferation and inhibit apoptosis of the underlying fibroblasts: an important role in the
pathogenesis of keloid', J Invest Dermatol, vol. 121, no. 6, pp. 1326-1331.
Gallitzendoerfer, R, Abouzied, MM, Hartmann, D, Dobrowolski, R, Gieselmann, V &
Franken, S 2008, 'Hepatoma-derived growth factor (HDGF) is dispensable for normal
mouse development', Dev Dyn, vol. 237, no. 7, pp. 1875-1885.
Gardner, TS & Faith, JJ 2010, 'Reverse-engineering transcription control networks',
Physics of life reviews, vol. 2, no. 1, pp. 65-88.
Ghazizadeh, M 2007, 'Essential role of IL-6 signaling pathway in keloid pathogenesis', J
Nippon Med Sch, vol. 74, no. 1, pp. 11-22.
Ghazizadeh, M, Tosa, M, Shimizu, H, Hyakusoku, H & Kawanami, O 2007, 'Functional
implications of the IL-6 signaling pathway in keloid pathogenesis', J Invest Dermatol,
vol. 127, no. 1, pp. 98-105.
Ghosh, AK, Yuan, W, Mori, Y & Varga, J 2000, 'Smad-dependent stimulation of type I
collagen gene expression in human skin fibroblasts by TGF-beta involves functional
cooperation with p300/CBP transcriptional coactivators', Oncogene, vol. 19, no. 31, pp.
3546-3555.
Goutsias, J & Lee, NH 2007, 'Computational and experimental approaches for modeling
gene regulatory networks', Curr Pharm Des, vol. 13, no. 14, pp. 1415-1436.
Grommé, M & Neefjes, J 2002, 'Antigen degradation or presentation by MHC class I
molecules via classical and non-classical pathways', Mol Immunol, vol. 39, no. 3-4, pp.
181-202.
Grotendorst, GR, Rahmanie, H & Duncan, MR 2004, 'Combinatorial signaling pathways
determine fibroblast proliferation and myofibroblast differentiation', FASEB J, vol. 18,
no. 3, pp. 469-479.
Hache, H, Lehrach, H & Herwig, R 2009, 'Reverse Engineering of Gene Regulatory
Networks: A Comparative Study', EURASIP journal on bioinformatics & systems
biology, vol. 2009, pp. 617281.
Hache, H, Lehrach, H & Herwig, R 2009, 'Reverse engineering of gene regulatory
networks: a comparative study', EURASIP journal on bioinformatics & systems biology,
pp. 617281.
Haisa, M, Okochi, H & Grotendorst, GR 1994, 'Elevated levels of PDGF alpha receptors
in keloid fibroblasts contribute to an enhanced response to PDGF', J Invest Dermatol,
vol. 103, no. 4, pp. 560-563.
142
Hansen, MB, Nielsen, SE & Berg, K 1989, 'Re-examination and further development of a
precise and rapid dye method for measuring cell growth/cell kill', J Immunol Methods,
vol. 119, no. 2, pp. 203-210.
Harr, B & Schlötterer, C 2006, 'Comparison of algorithms for the analysis of Affymetrix
microarray data as evaluated by co-expression of genes in known operons', Nucleic Acids
Res, vol. 34, no. 2, pp. e8.
Hartwell, LH, Hopfield, JJ, Leibler, S & Murray, AW 1999, 'From molecular to modular
cell biology', Nature, vol. 402, no. 6761 Suppl, pp. C47-C52.
Hengel, H, Esslinger, C, Pool, J, Goulmy, E & Koszinowski, UH 1995, 'Cytokines
restore MHC class I complex formation and control antigen presentation in human
cytomegalovirus-infected cells', J Gen Virol, vol. 76 ( Pt 12), pp. 2987-2997.
Henry, JL, Coggin, DL & King, CR 1993, 'High-level expression of the ribosomal
protein L19 in human breast tumors that overexpress erbB-2', Cancer Res, vol. 53, no. 6,
pp. 1403-1408.
Hinz, B 2006, 'Masters and servants of the force: the role of matrix adhesions in
myofibroblast force perception and transmission', Eur J Cell Biol, vol. 85, no. 3-4, pp.
175-181.
Huang, DW, Sherman, BT & Lempicki, RA 2009, 'Systematic and integrative analysis of
large gene lists using DAVID bioinformatics resources', Nat Protoc, vol. 4, no. 1, pp. 4457.
Ishihara, H, Yoshimoto, H, Fujioka, M, Murakami, R, Hirano, A, Fujii, T, Ohtsuru, A,
Namba, H & Yamashita, S 2000, 'Keloid fibroblasts resist ceramide-induced apoptosis by
overexpression of insulin-like growth factor I receptor', J Invest Dermatol, vol. 115, no.
6, pp. 1065-1071.
Iyer, VR, Eisen, MB, Ross, DT, Schuler, G, Moore, T, Lee, JC, Trent, JM, Staudt, LM,
Hudson, J, Boguski, MS, Lashkari, D, Shalon, D, Botstein, D & Brown, PO 1999, 'The
transcriptional program in the response of human fibroblasts to serum', Science, vol. 283,
no. 5398, pp. 83-87.
Izumoto, Y, Kuroda, T, Harada, H, Kishimoto, T & Nakamura, H 1997, 'Hepatomaderived growth factor belongs to a gene family in mice showing significant homology in
the amino terminus', Biochem Biophys Res Commun, vol. 238, no. 1, pp. 26-32.
Jeffery, TK, Upton, PD, Trembath, RC & Morrell, NW 2005, 'BMP4 inhibits
proliferation and promotes myocyte differentiation of lung fibroblasts via Smad1 and
JNK pathways', Am J Physiol Lung Cell Mol Physiol, vol. 288, no. 2, pp. L370-L378.
143
Johnson, WE, Li, C & Rabinovic, A 2007, 'Adjusting batch effects in microarray
expression data using empirical Bayes methods', Biostatistics, vol. 8, no. 1, pp. 118-127.
Kessler, D, Dethlefsen, S, Haase, I, Plomann, M, Hirche, F, Krieg, T & Eckes, B 2001,
'Fibroblasts in mechanically stressed collagen lattices assume a "synthetic" phenotype', J
Biol Chem, vol. 276, no. 39, pp. 36575-36585.
Kilian, B, Mansukoski, H, Barbosa, FC, Ulrich, F, Tada, M & Heisenberg, CP 2003, 'The
role of Ppt/Wnt5 in regulating cell shape and movement during zebrafish gastrulation',
Mech Dev, vol. 120, no. 4, pp. 467-476.
Kischer, CW 1992, 'The microvessels in hypertrophic scars, keloids and related lesions: a
review', J Submicrosc Cytol Pathol, vol. 24, no. 2, pp. 281-296.
Kischer, CW, Shetlar, MR, Shetlar, CL & Chvapil, M 1983, 'Immunoglobulins in
hypertrophic scars and keloids', Plast Reconstr Surg, vol. 71, no. 6, pp. 821-825.
Kishima, Y, Yamamoto, H, Izumoto, Y, Yoshida, K, Enomoto, H, Yamamoto, M,
Kuroda, T, Ito, H, Yoshizaki, K & Nakamura, H 2002, 'Hepatoma-derived growth factor
stimulates cell growth after translocation to the nucleus by nuclear localization signals', J
Biol Chem, vol. 277, no. 12, pp. 10315-10322.
Knighton, DR, Hunt, TK, Scheuenstuhl, H, Halliday, BJ, Werb, Z & Banda, MJ 1983,
'Oxygen tension regulates the expression of angiogenesis factor by macrophages',
Science, vol. 221, no. 4617, pp. 1283-1285.
Lai, MD & Xu, J 2007, 'Ribosomal proteins and colorectal cancer', Current genomics,
vol. 8, no. 1, pp. 43-49.
Larrabee, WF, East, CA, Jaffe, HS, Stephenson, C & Peterson, KE 1990, 'Intralesional
interferon gamma treatment for keloids and hypertrophic scars', Arch Otolaryngol Head
Neck Surg, vol. 116, no. 10, pp. 1159-1162.
Lee, SS, Yosipovitch, G, Chan, YH & Goh, CL 2004, 'Pruritus, pain, and small nerve
fiber function in keloids: a controlled study', J Am Acad Dermatol, vol. 51, no. 6, pp.
1002-1006.
Lee, TY, Chin, GS, Kim, WJ, Chau, D, Gittes, GK & Longaker, MT 1999, 'Expression of
transforming growth factor beta 1, 2, and 3 proteins in keloids', Ann Plast Surg, vol. 43,
no. 2, pp. 179-184.
Li, C & Wong, WH 2001, 'Model-based analysis of oligonucleotide arrays: expression
index computation and outlier detection', Proc Natl Acad Sci U S A, vol. 98, no. 1, pp.
31-36.
Lim, IJ, Phan, TT, Song, C, Tan, WT & Longaker, MT 2001, 'Investigation of the
144
influence of keloid-derived keratinocytes on fibroblast growth and proliferation in vitro',
Plast Reconstr Surg, vol. 107, no. 3, pp. 797-808.
Lim, IJ, Phan, TT, Bay, BH, Qi, R, Huynh, H, Tan, WTL, Lee, ST & Longaker, MT
2002, 'Fibroblasts cocultured with keloid keratinocytes: normal fibroblasts secrete
collagen in a keloidlike manner', Am J Physiol Cell Physiol, vol. 283, no. 1, pp. C212C222.
Lim, WK, Wang, K, Lefebvre, C & Califano, A 2007, 'Comparative analysis of
microarray normalization procedures: effects on reverse engineering gene networks',
Bioinformatics, vol. 23, no. 13, pp. i282-i288.
Lodygin, D, Epanchintsev, A, Menssen, A, Diebold, J & Hermeking, H 2005, 'Functional
epigenomics identifies genes frequently silenced in prostate cancer', Cancer Res, vol. 65,
no. 10, pp. 4218-4227.
Louw, L 2007, 'The keloid phenomenon: progress toward a solution', Clin Anat, vol. 20,
no. 1, pp. 3-14.
Mao, J, Xu, Z, Fang, Y, Wang, H, Xu, J, Ye, J, Zheng, S & Zhu, Y 2008, 'Hepatomaderived growth factor involved in the carcinogenesis of gastric epithelial cells through
promotion of cell proliferation by Erk1/2 activation', Cancer Sci, vol. 99, no. 11, pp.
2120-2127.
Margolin, AA, Wang, K, Lim, WK, Kustagi, M, Nemenman, I & Califano, A 2006,
'Reverse engineering cellular networks', Nat Protoc, vol. 1, no. 2, pp. 662-671.
Marneros, AG & Krieg, T 2004, 'Keloids--clinical diagnosis, pathogenesis, and treatment
options', J Dtsch Dermatol Ges, vol. 2, no. 11, pp. 905-913.
Marsit, CJ, Karagas, MR, Andrew, A, Liu, M, Danaee, H, Schned, AR, Nelson, HH &
Kelsey, KT 2005, 'Epigenetic inactivation of SFRP genes and TP53 alteration act jointly
as markers of invasive bladder cancer', Cancer Res, vol. 65, no. 16, pp. 7081-7085.
Masckauchán, TNH, Agalliu, D, Vorontchikhina, M, Ahn, A, Parmalee, NL, Li, CM,
Khoo, A, Tycko, B, Brown, AMC & Kitajewski, J 2006, 'Wnt5a signaling induces
proliferation and survival of endothelial cells in vitro and expression of MMP-1 and Tie2', Mol Biol Cell, vol. 17, no. 12, pp. 5163-5172.
Massagué, J 1998, 'TGF-beta signal transduction', Annu Rev Biochem, vol. 67, pp. 753791.
Mast, BA 1992, 'Wound Healing: Biochemical and Clinical Aspects' in The Skin, ed IK
Cohen, B Diegelmann & WJ and Lindblad, W.B. Saunders Company, Philadelphia, pp.
344.
145
Matsuyama, A, Inoue, H, Shibuta, K, Tanaka, Y, Barnard, GF, Sugimachi, K & Mori, M
2001, 'Hepatoma-derived growth factor is associated with reduced sensitivity to
irradiation in esophageal cancer', Cancer Res, vol. 61, no. 15, pp. 5714-5717.
Matys, V, Fricke, E, Geffers, R, Gössling, E, Haubrock, M, Hehl, R, Hornischer, K,
Karas, D, Kel, AE, Kel-Margoulis, OV, Kloos, DU, Land, S, Lewicki-Potapov, B,
Michael, H, Münch, R, Reuter, I, Rotert, S, Saxel, H, Scheer, M, Thiele, S & Wingender,
E 2003, 'TRANSFAC: transcriptional regulation, from patterns to profiles', Nucleic Acids
Res, vol. 31, no. 1, pp. 374-378.
McDonald, SL & Silver, A 2009, 'The opposing roles of Wnt-5a in cancer', Br J Cancer,
vol. 101, no. 2, pp. 209-214.
Messadi, DV, Le, A, Berg, S, Huang, G, Zhuang, W & Bertolami, CN 1998, 'Effect of
TGF-beta 1 on PDGF receptors expression in human scar fibroblasts', Front Biosci, vol.
3, pp. a16-a22.
Messadi, DV, Doung, HS, Zhang, Q, Kelly, AP, Tuan, TL, Reichenberger, E & Le, AD
2004, 'Activation of NFkappaB signal pathways in keloid fibroblasts', Arch Dermatol
Res, vol. 296, no. 3, pp. 125-133.
Mohan, S & Baylink, DJ 2002, 'IGF-binding proteins are multifunctional and act via
IGF-dependent and -independent mechanisms', J Endocrinol, vol. 175, no. 1, pp. 19-31.
Mori, M, Morishita, H, Nakamura, H, Matsuoka, H, Yoshida, K, Kishima, Y, Zhou, Z,
Kida, H, Funakoshi, T, Goya, S, Yoshida, M, Kumagai, T, Tachibana, I, Yamamoto, Y,
Kawase, I & Hayashi, S 2004, 'Hepatoma-derived growth factor is involved in lung
remodeling by stimulating epithelial growth', Am J Respir Cell Mol Biol, vol. 30, no. 4,
pp. 459-469.
Moustafa, MF, Abdel-Fattah, MA & Abdel-Fattah, DC 1975, 'Presumptive evidence of
the effect of pregnancy estrogens on keloid growth. Case report', Plast Reconstr Surg,
vol. 56, no. 4, pp. 450-453.
Mukhopadhyay, A, Khoo, A, Cheong, HH, Chan, SY, Aalami, O, Lim, IJ & Phan, TT
2007, 'Targeting of Sp1 transcription factor: a novel therapeutic approach for keloids, an
in vitro analysis', Exp Dermatol, vol. 16, no. 12, pp. 1023-1031.
Nakamura, H, Izumoto, Y, Kambe, H, Kuroda, T, Mori, T, Kawamura, K, Yamamoto, H
& Kishimoto, T 1994, 'Molecular cloning of complementary DNA for a novel human
hepatoma-derived growth factor. Its homology with high mobility group-1 protein', J Biol
Chem, vol. 269, no. 40, pp. 25143-25149.
Nakamura, H, Kambe, H, Egawa, T, Kimura, Y, Ito, H, Hayashi, E, Yamamoto, H, Sato,
J & Kishimoto, S 1989, 'Partial purification and characterization of human hepatomaderived growth factor', Clin Chim Acta, vol. 183, no. 3, pp. 273-284.
146
Nemeth, AJ 1993, 'Keloids and hypertrophic scars', J Dermatol Surg Oncol, vol. 19, no.
8, pp. 738-746.
Niessen, FB, Schalkwijk, J, Vos, H & Timens, W 2004, 'Hypertrophic scar formation is
associated with an increased number of epidermal Langerhans cells', J Pathol, vol. 202,
no. 1, pp. 121-129.
Okuda, Y, Nakamura, H, Yoshida, K, Enomoto, H, Uyama, H, Hirotani, T, Funamoto, M,
Ito, H, Everett, AD, Hada, T & Kawase, I 2003, 'Hepatoma-derived growth factor
induces tumorigenesis in vivo through both direct angiogenic activity and induction of
vascular endothelial growth factor', Cancer Sci, vol. 94, no. 12, pp. 1034-1041.
Omo-Dare, P 1975, 'Genetic studies on keloid', J Natl Med Assoc, vol. 67, no. 6, pp. 428432.
Ong, CT, Khoo, YT, Mukhopadhyay, A, Do, DV, Lim, IJ, Aalami, O & Phan, TT 2007,
'mTOR as a potential therapeutic target for treatment of keloids and excessive scars', Exp
Dermatol, vol. 16, no. 5, pp. 394-404.
Ong, CT, Khoo, YT, Tan, EK, Mukhopadhyay, A, Do, DV, Han, HC, Lim, IJ & Phan,
TT 2007, 'Epithelial-mesenchymal interactions in keloid pathogenesis modulate vascular
endothelial growth factor expression and secretion', J Pathol, vol. 211, no. 1, pp. 95-108.
Parks, WC 1999, 'Matrix metalloproteinases in repair', Wound Repair Regen, vol. 7, no.
6, pp. 423-432.
Pe'er, D, Regev, A, Elidan, G & Friedman, N 2001, 'Inferring subnetworks from
perturbed expression profiles', Bioinformatics, vol. 17 Suppl 1, pp. S215-S224.
Phan, TT, Lim, IJ, Aalami, O, Lorget, F, Khoo, A, Tan, EK, Mukhopadhyay, A &
Longaker, MT 2005, 'Smad3 signalling plays an important role in keloid pathogenesis via
epithelial-mesenchymal interactions', J Pathol, vol. 207, no. 2, pp. 232-242.
Phan, TT, Lim, IJ, Bay, BH, Qi, R, Longaker, MT, Lee, ST & Huynh, H 2003, 'Role of
IGF system of mitogens in the induction of fibroblast proliferation by keloid-derived
keratinocytes in vitro', Am J Physiol Cell Physiol, vol. 284, no. 4, pp. C860-C869.
Pilewski, JM, Liu, L, Henry, AC, Knauer, AV & Feghali-Bostwick, CA 2005, 'Insulinlike growth factor binding proteins 3 and 5 are overexpressed in idiopathic pulmonary
fibrosis and contribute to extracellular matrix deposition', Am J Pathol, vol. 166, no. 2,
pp. 399-407.
Pogue-Geile, K, Geiser, JR, Shu, M, Miller, C, Wool, IG, Meisler, AI & Pipas, JM 1991,
'Ribosomal protein genes are overexpressed in colorectal cancer: isolation of a cDNA
clone encoding the human S3 ribosomal protein', Mol Cell Biol, vol. 11, no. 8, pp. 3842-
147
3849.
Prockop, DJ & Kivirikko, KI 1995, 'Collagens: molecular biology, diseases, and
potentials for therapy', Annu Rev Biochem, vol. 64, pp. 403-434.
Raghow, R 1994, 'The role of extracellular matrix in postinflammatory wound healing
and fibrosis', FASEB J, vol. 8, no. 11, pp. 823-831.
Ren, H, Tang, X, Lee, JJ, Feng, L, Everett, AD, Hong, WK, Khuri, FR & Mao, L 2004,
'Expression of hepatoma-derived growth factor is a strong prognostic predictor for
patients with early-stage non-small-cell lung cancer', J Clin Oncol, vol. 22, no. 16, pp.
3230-3237.
Rockwell, WB, Cohen, IK & Ehrlich, HP 1989, 'Keloids and hypertrophic scars: a
comprehensive review', Plast Reconstr Surg, vol. 84, no. 5, pp. 827-837.
Russell, SB, Trupin, KM, Rodríguez-Eaton, S, Russell, JD & Trupin, JS 1988, 'Reduced
growth-factor requirement of keloid-derived fibroblasts may account for tumor growth',
Proc Natl Acad Sci U S A, vol. 85, no. 2, pp. 587-591.
Sachs, K, Perez, O, Pe'er, D, Lauffenburger, DA & Nolan, GP 2005, 'Causal proteinsignaling networks derived from multiparameter single-cell data', Science, vol. 308, no.
5721, pp. 523-529.
Santoro, MM & Gaudino, G 2005, 'Cellular and molecular facets of keratinocyte
reepithelization during wound healing', Exp Cell Res, vol. 304, no. 1, pp. 274-286.
Santucci, M, Borgognoni, L, Reali, UM & Gabbiani, G 2001, 'Keloids and hypertrophic
scars of Caucasians show distinctive morphologic and immunophenotypic profiles',
Virchows Arch, vol. 438, no. 5, pp. 457-463.
Satish, L, Lyons-Weiler, J, Hebda, PA & Wells, A 2006, 'Gene expression patterns in
isolated keloid fibroblasts', Wound Repair Regen, vol. 14, no. 4, pp. 463-470.
Sato, M 2006, 'Upregulation of the Wnt/beta-catenin pathway induced by transforming
growth factor-beta in hypertrophic scars and keloids', Acta Derm Venereol, vol. 86, no. 4,
pp. 300-307.
Sayah, DN, Soo, C, Shaw, WW, Watson, J, Messadi, D, Longaker, MT, Zhang, X &
Ting, K 1999, 'Downregulation of apoptosis-related genes in keloid tissues', J Surg Res,
vol. 87, no. 2, pp. 209-216.
Schadt, EE, Sachs, A & Friend, S 2005, 'Embracing complexity, inching closer to reality',
Sci STKE, vol. 2005, no. 295, pp. pe40.
Schena, M, Shalon, D, Davis, RW & Brown, PO 1995, 'Quantitative monitoring of gene
148
expression patterns with a complementary DNA microarray', Science, vol. 270, no. 5235,
pp. 467-470.
Schierle, HP, Scholz, D & Lemperle, G 1997, 'Elevated levels of testosterone receptors in
keloid tissue: an experimental investigation', Plast Reconstr Surg, vol. 100, no. 2, pp.
390-5; discussion 396.
Schulze, A & Downward, J 2001, 'Navigating gene expression using microarrays--a
technology review', Nat Cell Biol, vol. 3, no. 8, pp. E190-E195.
Seifert, O, Bayat, A, Geffers, R, Dienus, K, Buer, J, Löfgren, S & Matussek, A 2008,
'Identification of unique gene expression patterns within different lesional sites of
keloids', Wound Repair Regen, vol. 16, no. 2, pp. 254-265.
Seifert, O & Mrowietz, U 2009, 'Keloid scarring: bench and bedside', Arch Dermatol Res,
vol. 301, no. 4, pp. 259-272.
Shaffer, JJ, Taylor, SC & Cook-Bolden, F 2002, 'Keloidal scars: a review with a critical
look at therapeutic options', J Am Acad Dermatol, vol. 46, no. 2 Suppl Understanding, pp.
S63-S97.
Shedden, K, Chen, W, Kuick, R, Ghosh, D, Macdonald, J, Cho, KR, Giordano, TJ,
Gruber, SB, Fearon, ER, Taylor, JMG & Hanash, S 2005, 'Comparison of seven methods
for producing Affymetrix expression scores based on False Discovery Rates in disease
profiling data', BMC Bioinformatics, vol. 6, pp. 26.
Shegogue, D & Trojanowska, M 2004, 'Mammalian target of rapamycin positively
regulates collagen type I production via a phosphatidylinositol 3-kinase-independent
pathway', J Biol Chem, vol. 279, no. 22, pp. 23166-23175.
Shukla, IM, Arora, NP & Arora, MM 1975, 'Corneal keloid', Indian J Ophthalmol, vol.
23, no. 2, pp. 18-19.
Shulewitz, M, Soloviev, I, Wu, T, Koeppen, H, Polakis, P & Sakanaka, C 2006,
'Repressor roles for TCF-4 and Sfrp1 in Wnt signaling in breast cancer', Oncogene, vol.
25, no. 31, pp. 4361-4369.
Singh, S, Zheng, JJ, Peiper, SC & Mclaughlin, BJ 2001, 'Gene expression profile of
ARPE-19 during repair of the monolayer', Graefes Arch Clin Exp Ophthalmol, vol. 239,
no. 12, pp. 946-951.
Sisco, M, Kryger, ZB, O'Shaughnessy, KD, Kim, PS, Schultz, GS, Ding, XZ, Roy, NK,
Dean, NM & Mustoe, TA 2008, 'Antisense inhibition of connective tissue growth factor
(CTGF/CCN2) mRNA limits hypertrophic scarring without affecting wound healing in
vivo', Wound Repair Regen, vol. 16, no. 5, pp. 661-673.
149
Slemp, AE & Kirschner, RE 2006, 'Keloids and scars: a review of keloids and scars, their
pathogenesis, risk factors, and management', Curr Opin Pediatr, vol. 18, no. 4, pp. 396402.
Slusarski, DC, Corces, VG & Moon, RT 1997, 'Interaction of Wnt and a Frizzled
homologue triggers G-protein-linked phosphatidylinositol signalling', Nature, vol. 390,
no. 6658, pp. 410-413.
Smith, CJ, Smith, JC & Finn, MC 1987, 'The possible role of mast cells (allergy) in the
production of keloid and hypertrophic scarring', J Burn Care Rehabil, vol. 8, no. 2, pp.
126-131.
Smith, JC, Boone, BE, Opalenik, SR, Williams, SM & Russell, SB 2008, 'Gene profiling
of keloid fibroblasts shows altered expression in multiple fibrosis-associated pathways', J
Invest Dermatol, vol. 128, no. 5, pp. 1298-1310.
Smith, TJ 2005, 'Insights into the role of fibroblasts in human autoimmune diseases', Clin
Exp Immunol, vol. 141, no. 3, pp. 388-397.
Stadelmann, WK, Digenis, AG & Tobin, GR 1998, 'Physiology and healing dynamics of
chronic cutaneous wounds', Am J Surg, vol. 176, no. 2A Suppl, pp. 26S-38S.
Steuer, R, Kurths, J, Daub, CO, Weise, J & Selbig, J 2002, 'The mutual information:
detecting and evaluating dependencies between variables', Bioinformatics, vol. 18 Suppl
2, pp. S231-S240.
Sylvia, CJ 2003, 'The role of neutrophil apoptosis in influencing tissue repair', J Wound
Care, vol. 12, no. 1, pp. 13-16.
Takada, T, Yagi, Y, Maekita, T, Imura, M, Nakagawa, S, Tsao, SW, Miyamoto, K,
Yoshino, O, Yasugi, T, Taketani, Y & Ushijima, T 2004, 'Methylation-associated
silencing of the Wnt antagonist SFRP1 gene in human ovarian cancers', Cancer Sci, vol.
95, no. 9, pp. 741-744.
Tosa, M, Ghazizadeh, M, Shimizu, H, Hirai, T, Hyakusoku, H & Kawanami, O 2005,
'Global gene expression analysis of keloid fibroblasts in response to electron beam
irradiation reveals the involvement of interleukin-6 pathway', J Invest Dermatol, vol. 124,
no. 4, pp. 704-713.
Tsibris, JCM, Segars, J, Coppola, D, Mane, S, Wilbanks, GD, O'Brien, WF & Spellacy,
WN 2002, 'Insights from gene arrays on the development and growth regulation of
uterine leiomyomata', Fertil Steril, vol. 78, no. 1, pp. 114-121.
Tuan, TL & Nichter, LS 1998, 'The molecular basis of keloid and hypertrophic scar
formation', Mol Med Today, vol. 4, no. 1, pp. 19-24.
150
Tuan, TL, Wu, H, Huang, EY, Chong, SSN, Laug, W, Messadi, D, Kelly, P & Le, A
2003, 'Increased plasminogen activator inhibitor-1 in keloid fibroblasts may account for
their elevated collagen accumulation in fibrin gel cultures', Am J Pathol, vol. 162, no. 5,
pp. 1579-1589.
Tumor Analysis Best Practices Working Group 2004, 'Expression profiling--best
practices for data generation and interpretation in clinical trials', Nat Rev Genet, vol. 5,
no. 3, pp. 229-237.
Urioste, SS, Arndt, KA & Dover, JS 1999, 'Keloids and hypertrophic scars: review and
treatment strategies', Semin Cutan Med Surg, vol. 18, no. 2, pp. 159-171.
Uyama, H, Tomita, Y, Nakamura, H, Nakamori, S, Zhang, B, Hoshida, Y, Enomoto, H,
Okuda, Y, Sakon, M, Aozasa, K, Kawase, I, Hayashi, N & Monden, M 2006, 'Hepatomaderived growth factor is a novel prognostic factor for patients with pancreatic cancer',
Clin Cancer Res, vol. 12, no. 20 Pt 1, pp. 6043-6048.
Wang, Z, Fong, KD, Phan, TT, Lim, IJ, Longaker, MT & Yang, GP 2006, 'Increased
transcriptional response to mechanical strain in keloid fibroblasts due to increased focal
adhesion complex formation', J Cell Physiol, vol. 206, no. 2, pp. 510-517.
Werner, S & Grose, R 2003, 'Regulation of wound healing by growth factors and
cytokines', Physiol Rev, vol. 83, no. 3, pp. 835-870.
Wicke, C, Halliday, B, Allen, D, Roche, NS, Scheuenstuhl, H, Spencer, MM, Roberts,
AB & Hunt, TK 2000, 'Effects of steroids and retinoids on wound healing', Arch Surg,
vol. 135, no. 11, pp. 1265-1270.
Witt, E, Maliri, A, McGrouther, DA & Bayat, A 2008, 'RAC Activity in Keloid Disease:
Comparative Analysis of Fibroblasts from Margin of Keloid to its Surrounding Normal
Skin', Eplasty, vol. 8, pp. e19.
Wolfram, D, Tzankov, A, Pülzl, P & Piza-Katzer, H 2009, 'Hypertrophic scars and
keloids--a review of their pathophysiology, risk factors, and therapeutic management',
Dermatol Surg, vol. 35, no. 2, pp. 171-181.
Wool, IG 1996, 'Extraribosomal functions of ribosomal proteins', Trends Biochem Sci,
vol. 21, no. 5, pp. 164-165.
Wu, RZ, Chaivorapol, C, Zheng, J, Li, H & Liang, S 2007, 'fREDUCE: detection of
degenerate regulatory elements using correlation with expression', BMC Bioinformatics,
vol. 8, pp. 399.
Wu, Y, Zhang, Q, Ann, DK, Akhondzadeh, A, Duong, HS, Messadi, DV & Le, AD 2004,
'Increased vascular endothelial growth factor may account for elevated level of
plasminogen activator inhibitor-1 via activating ERK1/2 in keloid fibroblasts', Am J
151
Physiol Cell Physiol, vol. 286, no. 4, pp. C905-C912.
Xia, W, Phan, TT, Lim, IJ, Longaker, MT & Yang, GP 2004, 'Complex epithelialmesenchymal interactions modulate transforming growth factor-beta expression in
keloid-derived cells', Wound Repair Regen, vol. 12, no. 5, pp. 546-556.
Yagi, KI, Dafalla, AA & Osman, AA 1979, 'Does an immune reaction to sebum in
wounds cause keloid scars? Beneficial effect of desensitisation', Br J Plast Surg, vol. 32,
no. 3, pp. 223-225.
Yamamoto, S, Tomita, Y, Hoshida, Y, Morii, E, Yasuda, T, Doki, Y, Aozasa, K, Uyama,
H, Nakamura, H & Monden, M 2007, 'Expression level of hepatoma-derived growth
factor correlates with tumor recurrence of esophageal carcinoma', Ann Surg Oncol, vol.
14, no. 7, pp. 2141-2149.
Yang, ZQ, Liu, G, Bollig-Fischer, A, Haddad, R, Tarca, AL & Ethier, SP 2009,
'Methylation-associated silencing of SFRP1 with an 8p11-12 amplification inhibits
canonical and non-canonical WNT pathways in breast cancers', Int J Cancer, vol. 125,
no. 7, pp. 1613-1621.
Yasuoka, H, Zhou, Z, Pilewski, JM, Oury, TD, Choi, AMK & Feghali-Bostwick, CA
2006, 'Insulin-like growth factor-binding protein-5 induces pulmonary fibrosis and
triggers mononuclear cellular infiltration', Am J Pathol, vol. 169, no. 5, pp. 1633-1642.
Yeh, FL, Shen, HD & Tai, HY 2009, 'Decreased production of MCP-1 and MMP-2 by
keloid-derived fibroblasts', Burns, vol. 35, no. 3, pp. 348-351.
York, IA & Rock, KL 1996, 'Antigen processing and presentation by the class I major
histocompatibility complex', Annu Rev Immunol, vol. 14, pp. 369-396.
Yoshida, K, Nakamura, H, Okuda, Y, Enomoto, H, Kishima, Y, Uyama, H, Ito, H,
Hirasawa, T, Inagaki, S & Kawase, I 2003, 'Expression of hepatoma-derived growth
factor in hepatocarcinogenesis', J Gastroenterol Hepatol, vol. 18, no. 11, pp. 1293-1301.
Yu, J, Smith, VA, Wang, PP, Hartemink, AJ & Jarvis, ED 2004, 'Advances to Bayesian
network inference for generating causal networks from observational biological data',
Bioinformatics, vol. 20, no. 18, pp. 3594-3603.
Zhang, Q, Wu, Y, Ann, DK, Messadi, DV, Tuan, TL, Kelly, AP, Bertolami, CN & Le,
AD 2003, 'Mechanisms of hypoxic regulation of plasminogen activator inhibitor-1 gene
expression in keloid fibroblasts', J Invest Dermatol, vol. 121, no. 5, pp. 1005-1012.
Zimmer, J, Poli, A, Andrès, E, Hanau, D, Brons, NHC & Hentges, F 2006, 'Reduced
cytokine-mediated up-regulation of HLA-DR in TAP-deficient fibroblasts', Immunol Lett,
vol. 107, no. 2, pp. 109-118.
152
APPENDICES
A.1 Full list of 181 genes upregulated in keloid compared to normal
fibroblasts using the MAS 5.0 summarization algorithm (P< 0.05)
Fold
change
26.2568
19.88896
Gene Symbol
Gene Title
POSTN
ZIC1
14.57531
10.38634
8.887936
HOXD10
COL15A1
EGR2
7.998279
7.9743
6.770347
HOXA11
CCDC102B
KCNJ6
6.420784
5.80834
5.553577
5.515989
5.489978
JUP /// KRT19
MAP7
IGFBP3
ADRA2A
RAC2
5.301598
5.168533
4.974008
CDYL
NPTX1
PMEPA1
4.940097
4.889467
4.783432
4.513727
EFNB2
ATXN1
CADM1
SEMA5A
4.476735
WNT5A
4.305683
4.167339
AK5
EVI2A /// EVI2B
4.105316
4.064728
THBS1
HMGCS2
4.031227
4.003236
UNC5B
GPSM2
3.983835
FAM155A
periostin, osteoblast specific factor
Zic family member 1 (odd-paired homolog,
Drosophila)
homeobox D10
collagen, type XV, alpha 1
early growth response 2 (Krox-20 homolog,
Drosophila)
homeobox A11
coiled-coil domain containing 102B
potassium inwardly-rectifying channel, subfamily J,
member 6
junction plakoglobin /// keratin 19
microtubule-associated protein 7
insulin-like growth factor binding protein 3
adrenergic, alpha-2A-, receptor
ras-related C3 botulinum toxin substrate 2 (rho
family, small GTP binding protein Rac2)
chromodomain protein, Y-like
neuronal pentraxin I
prostate transmembrane protein, androgen
induced 1
ephrin-B2
ataxin 1
cell adhesion molecule 1
sema domain, seven thrombospondin repeats
(type 1 and type 1-like), transmembrane domain
(TM) and short cytoplasmic domain, (semaphorin)
5A
wingless-type MMTV integration site family,
member 5A
adenylate kinase 5
ecotropic viral integration site 2A /// ecotropic viral
integration site 2B
thrombospondin 1
3-hydroxy-3-methylglutaryl-Coenzyme A synthase
2 (mitochondrial)
Unc-5 homolog B (C. elegans) (UNC5B), mRNA
G-protein signaling modulator 2 (AGS3-like, C.
elegans)
family with sequence similarity 155, member A
Corrected
p-value
2.51E-04
0.046789
0.004014
4.25E-04
0.006852
0.033229
0.021884
0.014057
0.014453
0.00785
0.021725
0.011306
0.049297
0.039061
0.015499
0.037054
0.022704
0.003232
0.033839
0.008288
0.0251
0.032527
0.003941
0.003338
0.038054
0.015673
0.014816
0.025904
153
3.896551
DYSF
3.846086
3.807187
TRIB2
TOX
3.80115
MICAL2
3.658315
3.650868
3.569428
THBS1
CADM1
WNT5A
3.54556
3.420288
THBS1
SEMA5A
3.40271
3.210168
3.02816
3.028158
3.005537
NR2F2
OXTR
ARMC9
SHMT2
MICAL2
2.982032
FARP1
2.942675
2.895509
HCLS1
GPSM2
2.879644
2.782385
2.772433
FRZB
MYO1D
SEPT11
2.743658
2.694598
2.690918
2.662594
2.658657
2.586957
2.515018
2.506182
2.43455
COL1A1
BGN
SYNGR1
MGC87895 ///
RPS14
SYT1
SEPT11
TBC1D2
UBL3
SLC1A4
2.39163
2.37384
2.348127
2.344866
2.331204
COL5A1
ECM1
MBD3
IPO5
SLC25A6
2.273416
2.271414
2.257667
BMP6
COL5A1
PDGFRB
dysferlin, limb girdle muscular dystrophy 2B
(autosomal recessive)
tribbles homolog 2 (Drosophila)
thymocyte selection-associated high mobility
group box
microtubule associated monoxygenase, calponin
and LIM domain containing 2
thrombospondin 1
cell adhesion molecule 1
wingless-type MMTV integration site family,
member 5A
thrombospondin 1
sema domain, seven thrombospondin repeats
(type 1 and type 1-like), transmembrane domain
(TM) and short cytoplasmic domain, (semaphorin)
5A
nuclear receptor subfamily 2, group F, member 2
oxytocin receptor
armadillo repeat containing 9
serine hydroxymethyltransferase 2 (mitochondrial)
microtubule associated monoxygenase, calponin
and LIM domain containing 2
FERM, RhoGEF (ARHGEF) and pleckstrin domain
protein 1 (chondrocyte-derived)
hematopoietic cell-specific Lyn substrate 1
G-protein signaling modulator 2 (AGS3-like, C.
elegans)
frizzled-related protein
myosin ID
CDNA FLJ37154 fis, clone BRACE2026054,
highly similar to SEPTIN 2
collagen, type I, alpha 1
biglycan
synaptogyrin 1
similar to ribosomal protein S14 /// ribosomal
protein S14
synaptotagmin I
septin 11
TBC1 domain family, member 2
ubiquitin-like 3
solute carrier family 1 (glutamate/neutral amino
acid transporter), member 4
collagen, type V, alpha 1
extracellular matrix protein 1
methyl-CpG binding domain protein 3
importin 5
solute carrier family 25 (mitochondrial carrier;
adenine nucleotide translocator), member 6
bone morphogenetic protein 6
collagen, type V, alpha 1
platelet-derived growth factor receptor, beta
0.006904
0.049297
5.70E-05
0.031156
0.019972
0.004192
0.006868
0.026742
0.012037
0.029412
0.028325
0.033839
0.011685
0.014608
0.004134
0.012601
0.022704
0.015673
0.006852
0.002621
0.035314
0.045835
0.025029
0.024206
0.035314
0.025805
0.006247
0.019401
0.04681
9.92E-04
0.039152
0.022235
0.010591
0.021725
0.03454
0.046681
0.001677
154
2.251143
2.250428
2.231361
2.228339
2.222405
2.21538
2.202431
2.189176
2.17394
2.168009
CTSB
MYO19
LAMA2
IPO5
FASN
NAP1L1
FKBP9
FHOD1
LAMA2
SLC25A6
2.160458
2.159276
2.15634
2.119878
MLPH
GLS
KLF5
PABPC4
2.107291
2.098882
NXN
LOC652607 ///
PABPC1 ///
PABPC3 ///
PABPCP5
2.085628
2.065369
2.056679
2.050869
FBLN2
SACS
PTK7
FARP1
2.02876
SPOCK1
2.009869
1.987354
DUSP7
MARCKS
1.971692
1.95389
1.940716
1.934144
1.918422
1.910812
1.893329
NUAK1
POLR1D
CRTAP
LAMA2
ATXN1
GSTM3
ANP32B
1.885341
EEF1D
1.877402
LOC285053 ///
LOC390354 ///
RPL18A
CCNB1IP1
LOC644191 ///
LOC728937 ///
RPS26
RPL35
1.867147
1.865347
1.862667
polypeptide
cathepsin B
myosin XIX
laminin, alpha 2
importin 5
fatty acid synthase
nucleosome assembly protein 1-like 1
FK506 binding protein 9, 63 kDa
formin homology 2 domain containing 1
laminin, alpha 2
solute carrier family 25 (mitochondrial carrier;
adenine nucleotide translocator), member 6
melanophilin
glutaminase
Kruppel-like factor 5 (intestinal)
poly(A) binding protein, cytoplasmic 4 (inducible
form)
nucleoredoxin
similar to Polyadenylate-binding protein 1
(Poly(A)-binding protein 1) (PABP 1) /// poly(A)
binding protein, cytoplasmic 1 /// poly(A) binding
protein, cytoplasmic 3 /// poly(A) binding protein,
cytoplasmic pseudogene 5
fibulin 2
spastic ataxia of Charlevoix-Saguenay (sacsin)
PTK7 protein tyrosine kinase 7
FERM, RhoGEF (ARHGEF) and pleckstrin domain
protein 1 (chondrocyte-derived)
sparc/osteonectin, cwcv and kazal-like domains
proteoglycan (testican) 1
dual specificity phosphatase 7
myristoylated alanine-rich protein kinase C
substrate
NUAK family, SNF1-like kinase, 1
polymerase (RNA) I polypeptide D, 16kDa
cartilage associated protein
laminin, alpha 2
ataxin 1
glutathione S-transferase mu 3 (brain)
acidic (leucine-rich) nuclear phosphoprotein 32
family, member B
eukaryotic translation elongation factor 1 delta
(guanine nucleotide exchange protein)
similar to ribosomal protein L18a /// ribosomal
protein L18a pseudogene /// ribosomal protein
L18a
cyclin B1 interacting protein 1
similar to hCG15685 /// similar to 40S ribosomal
protein S26 /// ribosomal protein S26
ribosomal protein L35
0.012336
0.039896
0.005205
0.010816
0.031733
0.007082
0.049297
0.021585
0.008295
0.023943
0.032937
0.044103
0.008645
0.025147
0.004134
0.022235
0.008295
0.009496
0.01158
0.012556
0.028002
0.004014
0.025545
0.006441
0.041823
0.043513
0.002955
0.028403
0.006366
0.002621
0.005205
0.021725
0.031156
0.025519
0.025168
155
1.861686
1.853484
1.825099
1.822173
1.812249
1.811397
1.806129
1.798827
1.790564
1.780408
1.765466
1.764005
1.753873
1.74401
1.735983
1.729728
1.717712
IPO5
IPO5
CTSB
PTPRG
PARVB
COL1A1
GARS
SHMT2
GPR137B
RPL10
PTPRK
RPS16
AHCYL1
MBNL1
NONO
CALHM2
PPP3CA
1.717474
1.704002
1.702673
GPR153
TST
DLEU1
1.700546
ATF4
1.698355
1.696983
1.695664
1.690349
1.682196
1.669918
1.667722
1.661295
1.660296
DOCK1
EIF3A
PQBP1
OSBPL3
EIF3D
TMEM2
C17orf81
C14orf139
SPARC
1.660139
1.658402
1.645023
1.643602
1.637327
1.636471
1.636263
1.623621
1.621903
1.61924
1.616134
1.602853
1.602017
1.601743
PNN
MTFR1
RPS9
S100A13
C19orf2
SARS
RPL13
FAT1
COL17A1
SOS2
NQO2
RPS2
AP2A2
TK2
importin 5
importin 5
cathepsin B
protein tyrosine phosphatase, receptor type, G
parvin, beta
collagen, type I, alpha 1
glycyl-tRNA synthetase
serine hydroxymethyltransferase 2 (mitochondrial)
G protein-coupled receptor 137B
ribosomal protein L10
protein tyrosine phosphatase, receptor type, K
ribosomal protein S16
S-adenosylhomocysteine hydrolase-like 1
muscleblind-like (Drosophila)
non-POU domain containing, octamer-binding
calcium homeostasis modulator 2
protein phosphatase 3 (formerly 2B), catalytic
subunit, alpha isoform
G protein-coupled receptor 153
thiosulfate sulfurtransferase (rhodanese)
deleted in lymphocytic leukemia 1 (non-protein
coding)
activating transcription factor 4 (tax-responsive
enhancer element B67)
dedicator of cytokinesis 1
eukaryotic translation initiation factor 3, subunit A
polyglutamine binding protein 1
oxysterol binding protein-like 3
eukaryotic translation initiation factor 3, subunit D
transmembrane protein 2
chromosome 17 open reading frame 81
chromosome 14 open reading frame 139
secreted protein, acidic, cysteine-rich
(osteonectin)
pinin, desmosome associated protein
mitochondrial fission regulator 1
ribosomal protein S9
S100 calcium binding protein A13
chromosome 19 open reading frame 2
seryl-tRNA synthetase
ribosomal protein L13
FAT tumor suppressor homolog 1 (Drosophila)
collagen type XVII alpha I
son of sevenless homolog 2 (Drosophila)
NAD(P)H dehydrogenase, quinone 2
ribosomal protein S2
adaptor-related protein complex 2, alpha 2 subunit
thymidine kinase 2, mitochondrial
0.002103
0.02642
0.018044
0.027471
0.006852
0.020984
0.03012
0.001636
0.034176
0.019277
0.047404
0.048597
0.031368
0.040962
0.028002
0.012441
0.037703
0.036174
0.01541
0.009449
0.017043
0.006852
0.006852
0.047404
0.046651
0.044103
0.011306
0.033839
0.042746
0.030453
0.029702
0.016285
0.025283
0.006868
0.036054
0.018299
0.010804
0.039953
0.006247
0.033713
0.033839
0.040053
0.012601
0.040053
156
1.599825
1.597647
1.59623
1.553895
1.553142
1.549388
1.538817
1.531203
1.529446
1.523152
1.518518
1.509025
1.484414
1.474366
1.458746
1.457511
1.449214
MEMO1
RPL4
RPS8
PARVB
HNRNPA0
EPHB4
POLR1E
SERP1
DNAJB6
RPL13
SEC63
HNRNPA3
BTF3
RPS6
GRB10
FBXL5
ACAD8
1.44043
1.43772
1.434125
DCTD
APRT
PDS5A
1.42483
1.419617
HEBP1
SUV420H1
1.415472
1.407764
1.404108
RPL8
RECQL
KDELR2
1.40236
1.392988
MPRIP
ATP5E
1.375108
1.37272
1.371258
1.365867
LUC7L
ARMCX3
SERP1
PPP2R3A
1.360637
1.35391
RPL13
STK24
1.305478
1.275109
1.264861
1.25246
EIF4H
COX4NB
FBXO34
USP22
mediator of cell motility 1
ribosomal protein L4
ribosomal protein S8
parvin, beta
heterogeneous nuclear ribonucleoprotein A0
EPH receptor B4
polymerase (RNA) I polypeptide E, 53kDa
stress-associated endoplasmic reticulum protein 1
DnaJ (Hsp40) homolog, subfamily B, member 6
ribosomal protein L13
SEC63 homolog (S. cerevisiae)
heterogeneous nuclear ribonucleoprotein A3
basic transcription factor 3
ribosomal protein S6
growth factor receptor-bound protein 10
F-box and leucine-rich repeat protein 5
acyl-Coenzyme A dehydrogenase family, member
8
dCMP deaminase
adenine phosphoribosyltransferase
PDS5, regulator of cohesion maintenance,
homolog A (S. cerevisiae)
heme binding protein 1
suppressor of variegation 4-20 homolog 1
(Drosophila)
ribosomal protein L8
RecQ protein-like (DNA helicase Q1-like)
KDEL (Lys-Asp-Glu-Leu) endoplasmic reticulum
protein retention receptor 2
myosin phosphatase Rho interacting protein
ATP synthase, H+ transporting, mitochondrial F1
complex, epsilon subunit
LUC7-like (S. cerevisiae)
armadillo repeat containing, X-linked 3
stress-associated endoplasmic reticulum protein 1
protein phosphatase 2 (formerly 2A), regulatory
subunit B'', alpha
ribosomal protein L13
serine/threonine kinase 24 (STE20 homolog,
yeast)
eukaryotic translation initiation factor 4H
COX4 neighbor
F-box protein 34
ubiquitin specific peptidase 22
0.010405
0.032078
0.037349
0.032078
0.033839
0.016501
0.021725
0.032078
0.033283
0.021854
0.011005
0.005384
0.048855
0.006198
0.004053
0.046681
0.021854
0.024538
0.048855
0.048855
0.01989
0.031156
0.024538
0.031862
0.047112
0.01541
0.049297
0.036591
0.048597
0.035533
0.023437
0.0227
0.049297
0.033851
0.045835
0.026786
0.038322
157
A.2 Full list of 290 genes downregulated in keloid compared to normal
fibroblasts using the MAS 5.0 summarization algorithm (P < 0.05)
Fold
change
77.61498
Gene Symbol
Gene Title
CXCL6
73.48209
CXCL1
67.89986
64.39615
49.49322
41.41703
39.93947
32.48277
IL8
CXCL11
HSD11B1
CCL5
CXCL2
RARRES1
29.77878
RSAD2
27.68656
PLA2G2A
27.15919
27.04305
26.4216
26.13762
23.80846
23.43326
23.40176
21.83553
21.53614
19.30517
15.52373
15.49821
C2 /// CFB
CXCL5
TNFAIP6
CXCL3
IL32
CP
CXCL10
CHI3L2
IDO1
NTRK2
C3
SLC39A8
14.28212
14.04543
13.35598
G0S2
OAS1
TNFSF10
13.23721
11.90575
CCL8
SLC39A8
11.36108
10.95266
10.86976
10.20459
10.02452
9.336419
9.271726
8.931117
GCH1
CCL5
SLC19A3
HERC5
IL6
NRCAM
CCL2
ABCA8
chemokine (C-X-C motif) ligand 6 (granulocyte
chemotactic protein 2)
chemokine (C-X-C motif) ligand 1 (melanoma
growth stimulating activity, alpha)
interleukin 8
chemokine (C-X-C motif) ligand 11
hydroxysteroid (11-beta) dehydrogenase 1
chemokine (C-C motif) ligand 5
chemokine (C-X-C motif) ligand 2
retinoic acid receptor responder (tazarotene
induced) 1
radical S-adenosyl methionine domain containing
2
phospholipase A2, group IIA (platelets, synovial
fluid)
complement component 2 /// complement factor B
chemokine (C-X-C motif) ligand 5
tumor necrosis factor, alpha-induced protein 6
chemokine (C-X-C motif) ligand 3
interleukin 32
ceruloplasmin (ferroxidase)
chemokine (C-X-C motif) ligand 10
chitinase 3-like 2
indoleamine 2,3-dioxygenase 1
neurotrophic tyrosine kinase, receptor, type 2
complement component 3
solute carrier family 39 (zinc transporter), member
8
G0/G1switch 2
2',5'-oligoadenylate synthetase 1, 40/46kDa
tumor necrosis factor (ligand) superfamily,
member 10
chemokine (C-C motif) ligand 8
solute carrier family 39 (zinc transporter), member
8
GTP cyclohydrolase 1
chemokine (C-C motif) ligand 5
solute carrier family 19, member 3
hect domain and RLD 5
interleukin 6 (interferon, beta 2)
neuronal cell adhesion molecule
chemokine (C-C motif) ligand 2
ATP-binding cassette, sub-family A (ABC1),
Corrected
p-value
1.48E-05
7.93E-05
0.006303
0.006303
1.98E-06
0.005142
0.002401
0.004841
0.019981
0.001149
2.15E-05
2.44E-04
0.008288
0.005384
0.001636
0.003941
0.032337
5.70E-04
0.004053
0.011306
2.51E-04
3.85E-07
0.003903
0.039061
0.016414
0.024538
4.73E-05
2.44E-04
0.010804
0.031156
0.005346
5.25E-04
1.22E-05
2.18E-05
0.006852
158
8.870272
8.768786
8.582171
SFRP1
TMEM100
HLA-DPA1
8.130204
8.070362
7.84843
7.806789
7.653281
7.647369
7.607011
7.487806
7.243977
TLR2
SOD2
CTSS
IFI30
MAOB
DTNA
HERC6
TLR3
RARRES3
7.072357
7.021463
6.598767
6.533486
6.41113
6.335001
6.331704
SOD2
TNFAIP3
CA12
CX3CL1
FGL2
CTSS
IFIT2
6.29937
6.179313
LRRN3
IFIT3
6.154707
6.106007
6.092712
6.0468
BIRC3
CA12
CLGN
TRPA1
5.951259
GBP1
5.838885
5.792193
5.715185
5.679222
5.661747
5.640415
5.429267
TNFAIP2
ISG20
HPSE
ABI3BP
CA12
PTGES
GBP1
5.347407
5.341403
IGF1
HLA-DRB1
5.1838
5.170646
5.018916
4.897188
SNX10
CGREF1
NAMPT
SLC39A8
4.840101
CA12
member 8
secreted frizzled-related protein 1
transmembrane protein 100
major histocompatibility complex, class II, DP
alpha 1
toll-like receptor 2
superoxide dismutase 2, mitochondrial
cathepsin S
interferon, gamma-inducible protein 30
monoamine oxidase B
dystrobrevin, alpha
hect domain and RLD 6
toll-like receptor 3
retinoic acid receptor responder (tazarotene
induced) 3
superoxide dismutase 2, mitochondrial
tumor necrosis factor, alpha-induced protein 3
carbonic anhydrase XII
chemokine (C-X3-C motif) ligand 1
fibrinogen-like 2
cathepsin S
interferon-induced protein with tetratricopeptide
repeats 2
leucine rich repeat neuronal 3
interferon-induced protein with tetratricopeptide
repeats 3
baculoviral IAP repeat-containing 3
carbonic anhydrase XII
calmegin
transient receptor potential cation channel,
subfamily A, member 1
guanylate binding protein 1, interferon-inducible,
67kDa
tumor necrosis factor, alpha-induced protein 2
interferon stimulated exonuclease gene 20kDa
heparanase
ABI family, member 3 (NESH) binding protein
carbonic anhydrase XII
prostaglandin E synthase
guanylate binding protein 1, interferon-inducible,
67kDa
insulin-like growth factor 1 (somatomedin C)
MHC class II HLA-DRB3 mRNA (HLADRB3*01012 allele)
sorting nexin 10
cell growth regulator with EF-hand domain 1
nicotinamide phosphoribosyltransferase
solute carrier family 39 (zinc transporter), member
8
carbonic anhydrase XII
0.013944
0.008604
9.92E-04
0.011223
0.001796
2.95E-04
0.004841
0.040797
0.005869
0.031862
0.031156
0.004841
0.004841
1.48E-05
5.70E-05
0.039558
0.046681
0.023641
0.049297
2.64E-04
0.018974
0.013996
0.004215
0.008363
0.022447
0.013892
1.38E-04
0.048855
0.036366
0.036366
4.73E-05
0.004841
0.021585
0.048855
0.021083
0.041947
0.024396
0.013996
5.70E-05
3.62E-04
159
4.819873
4.81897
4.798669
4.769258
4.76155
4.755867
4.618283
4.604176
4.566699
STAT4
CA12
LRRN3
MT1X
NAMPT
MT1M
WTAP
MT1P2
PRRG4
4.551031
ELF3
4.543112
SLC16A4
4.360824
4.127018
4.100276
4.048781
HLA-DRB1 ///
HLA-DRB2 ///
HLA-DRB3 ///
HLA-DRB4 ///
HLA-DRB5 ///
LOC100133484
///
LOC100133661
///
LOC100133811
/// LOC730415
/// RNASE2 ///
ZNF749
TNFAIP3
RRAD
SOD2
MT1E
ZC3H12A
HLA-DRB1 ///
HLA-DRB2 ///
HLA-DRB3 ///
HLA-DRB4 ///
HLA-DRB5 ///
LOC100133484
///
LOC100133661
///
LOC100133811
/// LOC730415
/// RNASE2 ///
ZNF749
ICAM1
DENND2A
SLC39A14
4.039061
4.038336
4.00928
DENND2A
MT1F
AKR1B1
4.342693
4.308981
4.278844
4.184589
4.177286
4.14845
signal transducer and activator of transcription 4
carbonic anhydrase XII
leucine rich repeat neuronal 3
metallothionein 1X
nicotinamide phosphoribosyltransferase
metallothionein 1M
Wilms tumor 1 associated protein
metallothionein 1 pseudogene 2
proline rich Gla (G-carboxyglutamic acid) 4
(transmembrane)
E74-like factor 3 (ets domain transcription factor,
epithelial-specific )
solute carrier family 16, member 4
(monocarboxylic acid transporter 5)
major histocompatibility complex, class II, DR beta
1 /// major histocompatibility complex, class II, DR
beta 2 (pseudogene) /// major histocompatibility
complex, class II, DR beta 3 /// major
histocompatibility complex, class II, DR beta 4 ///
major histocompatibility complex, class II, DR beta
5 /// similar to Major histocompatibility complex,
class II, DR beta 4 /// similar to HLA class II
histocompatibility antigen, DR-W53 beta chain ///
similar to hCG1992647 /// hypothetical protein
LOC730415 /// ribonuclease, RNase A family, 2
(liver, eosinophil-derived neurotoxin) /// zinc finger
protein 749
tumor necrosis factor, alpha-induced protein 3
Ras-related associated with diabetes
superoxide dismutase 2, mitochondrial
metallothionein 1E
zinc finger CCCH-type containing 12A
major histocompatibility complex, class II, DR beta
1 /// major histocompatibility complex, class II, DR
beta 2 (pseudogene) /// major histocompatibility
complex, class II, DR beta 3 /// major
histocompatibility complex, class II, DR beta 4 ///
major histocompatibility complex, class II, DR beta
5 /// similar to Major histocompatibility complex,
class II, DR beta 4 /// similar to HLA class II
histocompatibility antigen, DR-W53 beta chain ///
similar to hCG1992647 /// hypothetical protein
LOC730415 /// ribonuclease, RNase A family, 2
(liver, eosinophil-derived neurotoxin) /// zinc finger
protein 749
intercellular adhesion molecule 1
DENN/MADD domain containing 2A
solute carrier family 39 (zinc transporter), member
14
DENN/MADD domain containing 2A
metallothionein 1F
aldo-keto reductase family 1, member B1 (aldose
0.005418
2.51E-04
4.46E-04
0.035224
0.002078
0.008363
1.48E-05
0.02436
0.039061
0.004657
0.023221
0.041418
4.97E-04
0.011306
2.51E-04
0.018044
0.02642
0.046363
5.70E-05
0.021083
0.002948
0.014614
0.009128
9.92E-04
160
3.957591
SNCA
3.911683
3.886122
3.853862
3.810552
3.690421
3.648437
WWC1
ICAM1
WTAP
SLC15A3
IFI35
FIGF
3.63322
3.570853
3.568043
3.567944
3.566436
MT1E /// MT1H
/// MT1M ///
MT1P2
AMPD3
MT2A
LAP3
SLC11A2
3.555502
3.515179
3.490871
3.436509
3.41227
MARCH3
PRND
HTR2A
CLU
STEAP1
3.410262
3.398285
3.384179
3.362342
IFITM1
PPARG
GFRA1
NFKBIA
3.348567
MMD
3.33894
3.329882
3.321269
NFE2L3
MT1X
SLC11A2
3.299193
3.268607
3.249937
3.216968
3.160602
3.143586
PTGFR
MT1G
NR4A3
PLSCR1
LGALS9
PPP1R12B
3.135891
UCHL1
3.114003
3.099276
3.092455
3.076524
3.03837
2.998312
CHEK2
MT1F
PTGES
CCR1
EDNRB
IL15RA
reductase)
synuclein, alpha (non A4 component of amyloid
precursor)
WW and C2 domain containing 1
intercellular adhesion molecule 1
Wilms tumor 1 associated protein
solute carrier family 15, member 3
interferon-induced protein 35
c-fos induced growth factor (vascular endothelial
growth factor D)
metallothionein 1E /// metallothionein 1H ///
metallothionein 1M /// metallothionein 1
pseudogene 2
adenosine monophosphate deaminase (isoform E)
metallothionein 2A
leucine aminopeptidase 3
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
membrane-associated ring finger (C3HC4) 3
prion protein 2 (dublet)
5-hydroxytryptamine (serotonin) receptor 2A
clusterin
six transmembrane epithelial antigen of the
prostate 1
interferon induced transmembrane protein 1 (9-27)
peroxisome proliferator-activated receptor gamma
GDNF family receptor alpha 1
nuclear factor of kappa light polypeptide gene
enhancer in B-cells inhibitor, alpha
monocyte to macrophage differentiationassociated
nuclear factor (erythroid-derived 2)-like 3
metallothionein 1X
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
prostaglandin F receptor (FP)
metallothionein 1G
nuclear receptor subfamily 4, group A, member 3
phospholipid scramblase 1
lectin, galactoside-binding, soluble, 9
protein phosphatase 1, regulatory (inhibitor)
subunit 12B
ubiquitin carboxyl-terminal esterase L1 (ubiquitin
thiolesterase)
CHK2 checkpoint homolog (S. pombe)
metallothionein 1F
prostaglandin E synthase
chemokine (C-C motif) receptor 1
endothelin receptor type B
interleukin 15 receptor, alpha
0.019277
0.022704
1.48E-05
2.18E-05
0.006868
0.033983
0.043452
0.002621
0.005508
0.022895
0.017355
0.002401
0.003027
0.023981
0.006852
0.036591
0.002103
0.014614
0.025147
0.021725
0.003195
0.004578
0.047112
0.032078
5.85E-05
0.004578
0.036366
0.024206
0.026199
0.028025
0.045835
2.51E-04
0.012037
0.003262
0.031156
0.022235
3.35E-04
0.011685
161
2.938915
STAT1
2.927115
2.921418
SMC2
NME5
2.915037
2.833072
2.778108
2.770222
2.747634
2.737904
2.734728
2.734047
2.71983
2.702095
2.670183
2.65614
2.648943
SAMHD1
CD82
ABLIM1
LAMB3
MT1F
HLA-F
CFLAR
SIRPA
PDLIM4
HLA-C
SLC25A28
AK3L1 ///
AK3L2
HLA-DMA
2.638515
2.607675
2.566287
2.557721
2.556034
2.552846
2.544963
TAPBPL
NOVA1
TAPBPL
CFLAR
DTNA
PDPN
LGALS3BP
2.528988
2.501654
TDRD7
PSME2
2.50114
2.484067
IRAK3
SLC1A1
2.461875
2.458896
2.452654
2.44606
2.442194
2.431164
2.425255
2.424902
CFLAR
CFLAR
HIST1H2BD
PROCR
CFLAR
CYB5A
FILIP1L
PSTPIP2
2.421571
2.408101
HLA-F
CACNA1A
2.38139
2.357674
2.347081
SP100
CFLAR
PPFIBP2
signal transducer and activator of transcription 1,
91kDa
structural maintenance of chromosomes 2
non-metastatic cells 5, protein expressed in
(nucleoside-diphosphate kinase)
SAM domain and HD domain 1
CD82 molecule
actin binding LIM protein 1
laminin, beta 3
metallothionein 1F
major histocompatibility complex, class I, F
CASP8 and FADD-like apoptosis regulator
signal-regulatory protein alpha
PDZ and LIM domain 4
major histocompatibility complex, class I, C
solute carrier family 25, member 28
adenylate kinase 3-like 1 /// adenylate kinase 3like 2
major histocompatibility complex, class II, DM
alpha
TAP binding protein-like
neuro-oncological ventral antigen 1
TAP binding protein-like
CASP8 and FADD-like apoptosis regulator
dystrobrevin, alpha
podoplanin
lectin, galactoside-binding, soluble, 3 binding
protein
tudor domain containing 7
proteasome (prosome, macropain) activator
subunit 2 (PA28 beta)
interleukin-1 receptor-associated kinase 3
solute carrier family 1 (neuronal/epithelial high
affinity glutamate transporter, system Xag),
member 1
CASP8 and FADD-like apoptosis regulator
CASP8 and FADD-like apoptosis regulator
histone cluster 1, H2bd
protein C receptor, endothelial (EPCR)
CASP8 and FADD-like apoptosis regulator
cytochrome b5 type A (microsomal)
filamin A interacting protein 1-like
proline-serine-threonine phosphatase interacting
protein 2
major histocompatibility complex, class I, F
calcium channel, voltage-dependent, P/Q type,
alpha 1A subunit
SP100 nuclear antigen
CASP8 and FADD-like apoptosis regulator
PTPRF interacting protein, binding protein 2 (liprin
0.022235
0.002621
0.044101
0.021854
0.003571
0.009496
0.020989
0.002955
0.003416
0.006366
5.35E-04
0.026742
0.008699
5.80E-04
0.024767
0.038585
9.67E-04
0.036488
8.05E-04
0.016501
5.70E-05
0.008288
0.009862
0.01541
0.004147
0.01197
0.03058
0.041418
0.003262
0.006247
0.004583
0.005876
0.036591
0.003195
0.040425
0.002401
0.025861
0.046017
0.006124
2.44E-04
162
2.331309
PSMB10
2.316866
2.296515
RNF114
SLC11A2
2.289964
PSMB8
2.28674
2.254133
2.246542
2.238959
2.231546
2.227878
2.214319
2.213672
2.21116
2.193658
2.190023
CFLAR
CLIC2
DRAM
CFLAR
HLA-G
FLJ22662
RNF114
ICAM1
HLA-G
GFPT2
LYN
2.182933
2.166235
2.153668
2.148729
2.143614
2.137008
2.133692
2.131292
2.120025
2.117058
2.116495
2.111383
2.100947
2.076569
2.065661
HLA-B
HTATIP2
MICALL2
ZMYM6
SPARCL1
FAM13A1
SIRPA
HLA-G
CYB5A
TAPBP
AKAP7
PON2
NT5E
MDM2
CNDP2
2.047671
2.042615
2.026807
2.016378
2.016181
1.961992
1.959727
TRIM38
HLA-B
CFLAR
FSTL3
KIAA0391 ///
PSMA6
TRIM38
ATRIP ///
TREX1
GLIPR1
NFKBIE
1.954683
C10orf26
1.989287
1.982656
beta 2)
proteasome (prosome, macropain) subunit, beta
type, 10
ring finger protein 114
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
proteasome (prosome, macropain) subunit, beta
type, 8 (large multifunctional peptidase 7)
CASP8 and FADD-like apoptosis regulator
chloride intracellular channel 2
damage-regulated autophagy modulator
CASP8 and FADD-like apoptosis regulator
major histocompatibility complex, class I, G
hypothetical protein FLJ22662
ring finger protein 114
intercellular adhesion molecule 1
major histocompatibility complex, class I, G
glutamine-fructose-6-phosphate transaminase 2
v-yes-1 Yamaguchi sarcoma viral related
oncogene homolog
major histocompatibility complex, class I, B
HIV-1 Tat interactive protein 2, 30kDa
MICAL-like 2
zinc finger, MYM-type 6
SPARC-like 1 (hevin)
family with sequence similarity 13, member A1
signal-regulatory protein alpha
major histocompatibility complex, class I, G
cytochrome b5 type A (microsomal)
TAP binding protein (tapasin)
A kinase (PRKA) anchor protein 7
paraoxonase 2
5'-nucleotidase, ecto (CD73)
Mdm2 p53 binding protein homolog (mouse)
CNDP dipeptidase 2 (metallopeptidase M20
family)
tripartite motif-containing 38
major histocompatibility complex, class I, B
CASP8 and FADD-like apoptosis regulator
follistatin-like 3 (secreted glycoprotein)
KIAA0391 /// proteasome (prosome, macropain)
subunit, alpha type, 6
tripartite motif-containing 38
ATR interacting protein /// three prime repair
exonuclease 1
GLI pathogenesis-related 1
nuclear factor of kappa light polypeptide gene
enhancer in B-cells inhibitor, epsilon
chromosome 10 open reading frame 26
0.049297
0.006302
0.006778
0.039558
0.011567
0.024261
9.40E-04
0.001636
0.006441
7.60E-04
0.007101
0.00542
0.002621
0.036153
0.004192
5.70E-05
0.015877
0.004841
0.001636
0.038945
0.018784
3.77E-04
1.63E-04
0.010129
0.003416
0.039896
0.004134
2.51E-04
0.043542
0.039558
0.01996
7.91E-04
0.035314
0.032078
0.006852
0.017295
0.047112
0.011306
0.022704
0.003262
163
1.951621
1.934228
1.92549
LY6E
PANX1
TBC1D9
1.916083
1.910365
1.896993
1.879147
ATOX1
CSF1
TNIP1
LYN
1.869223
RELB
1.861563
1.858279
1.848508
1.846391
BNIP3
BEX4
ACP2
VPS13D
1.843842
1.814648
1.813575
1.809137
1.799581
1.798085
1.795796
TNFSF12 ///
TNFSF12TNFSF13 ///
TNFSF13
IRF2
PON2
TRIM38
BTG3
HLA-C
HLA-A /// HLAA29.1 /// HLA-B
/// HLA-G ///
HLA-H /// HLAJ
1.795212
1.795001
1.790361
1.773213
1.76255
1.760506
NNMT
IFNGR1
CDC42EP4
HLA-C
SLC30A1
STAT3
1.752857
BASP1
1.744929
1.743621
1.741601
1.734837
1.726347
1.705262
1.69253
1.691625
1.690892
RFTN1
RAB20
BTG3
IFNGR1
FTH1
DNAJA1
STBD1
HLA-C
PSME1
lymphocyte antigen 6 complex, locus E
pannexin 1
TBC1 domain family, member 9 (with GRAM
domain)
ATX1 antioxidant protein 1 homolog (yeast)
colony stimulating factor 1 (macrophage)
TNFAIP3 interacting protein 1
v-yes-1 Yamaguchi sarcoma viral related
oncogene homolog
v-rel reticuloendotheliosis viral oncogene homolog
B
BCL2/adenovirus E1B 19kDa interacting protein 3
brain expressed, X-linked 4
acid phosphatase 2, lysosomal
vacuolar protein sorting 13 homolog D (S.
cerevisiae)
tumor necrosis factor (ligand) superfamily,
member 12 /// TNFSF12-TNFSF13 readthrough
transcript /// tumor necrosis factor (ligand)
superfamily, member 13
interferon regulatory factor 2
paraoxonase 2
tripartite motif-containing 38
BTG family, member 3
major histocompatibility complex, class I, C
major histocompatibility complex, class I, A ///
major histocompatibility complex class I HLAA29.1 /// major histocompatibility complex, class I,
B /// major histocompatibility complex, class I, G ///
major histocompatibility complex, class I, H
(pseudogene) /// major histocompatibility complex,
class I, J (pseudogene)
nicotinamide N-methyltransferase
interferon gamma receptor 1
CDC42 effector protein (Rho GTPase binding) 4
major histocompatibility complex, class I, C
Hbc647 mRNA sequence
signal transducer and activator of transcription 3
(acute-phase response factor)
brain abundant, membrane attached signal protein
1
raftlin, lipid raft linker 1
RAB20, member RAS oncogene family
BTG family, member 3
interferon gamma receptor 1
ferritin, heavy polypeptide 1
DnaJ (Hsp40) homolog, subfamily A, member 1
starch binding domain 1
major histocompatibility complex, class I, C
proteasome (prosome, macropain) activator
subunit 1 (PA28 alpha)
0.02642
0.022964
8.91E-04
0.020647
0.023957
5.80E-04
0.009496
0.017671
0.021585
0.00786
0.028017
0.006904
0.048855
0.008169
0.008863
0.025296
0.024396
1.18E-04
0.018539
0.025147
0.025147
0.049112
0.003262
0.006213
0.008288
0.014614
0.022704
0.022704
0.007903
0.002401
0.019981
0.020755
0.049297
4.98E-04
0.032811
164
1.678448
HLA-B /// MICA
1.671204
RFX5
1.65795
1.656616
1.648751
1.647957
1.641402
1.606447
HSPA4L
P4HA2
CD59
NNMT
DNAJA1
NOC3L
1.597539
1.584331
1.579374
TBCC
TMEM22
IFITM3
1.554896
1.545585
1.545367
DFNA5
C15orf24
LOC284889 ///
MIF
1.541742
1.540308
1.53979
1.520112
1.50069
ENDOD1
BTN2A2
USP25
HLA-A
SGCB
1.487067
1.482268
1.481941
1.481617
1.470603
DDX18
OPTN
CD59
TRAF3
NDUFA9
1.468841
1.462896
1.455167
RBX1
PEX12
MMP2
1.450561
1.397826
1.365859
1.355799
1.347957
1.338753
1.260726
1.258344
1.244612
1.229484
1.190801
ZCCHC6
MPZL1
LARP7
ALAS1
CSTB
PLS3
CHMP4A
FTH1
BUD31
B2M
TANK
major histocompatibility complex, class I, B ///
MHC class I polypeptide-related sequence A
regulatory factor X, 5 (influences HLA class II
expression)
heat shock 70kDa protein 4-like
prolyl 4-hydroxylase, alpha polypeptide II
CD59 molecule, complement regulatory protein
nicotinamide N-methyltransferase
HDJ2 protein
nucleolar complex associated 3 homolog (S.
cerevisiae)
tubulin folding cofactor C
transmembrane protein 22
interferon induced transmembrane protein 3 (18U)
deafness, autosomal dominant 5
chromosome 15 open reading frame 24
hypothetical protein LOC284889 /// macrophage
migration inhibitory factor (glycosylation-inhibiting
factor)
endonuclease domain containing 1
butyrophilin, subfamily 2, member A2
ubiquitin specific peptidase 25
major histocompatibility complex, class I, A
sarcoglycan, beta (43kDa dystrophin-associated
glycoprotein)
DEAD (Asp-Glu-Ala-Asp) box polypeptide 18
optineurin
CD59 molecule, complement regulatory protein
TNF receptor-associated factor 3
NADH dehydrogenase (ubiquinone) 1 alpha
subcomplex, 9, 39kDa
ring-box 1
peroxisomal biogenesis factor 12
matrix metallopeptidase 2 (gelatinase A, 72kDa
gelatinase, 72kDa type IV collagenase)
zinc finger, CCHC domain containing 6
myelin protein zero-like 1
La ribonucleoprotein domain family, member 7
aminolevulinate, delta-, synthase 1
cystatin B (stefin B)
plastin 3 (T isoform)
chromatin modifying protein 4A
ferritin, heavy polypeptide 1
BUD31 homolog (S. cerevisiae)
beta-2-microglobulin
TRAF family member-associated NFKB activator
0.004053
0.037349
0.024306
0.049297
0.008363
0.019674
0.014305
0.022447
0.008288
0.004657
0.008288
0.047112
0.008288
0.044672
0.040086
0.001322
0.033598
0.01797
0.012635
0.012601
0.043452
0.009636
0.022704
0.048855
0.011306
0.020864
0.01541
0.011728
0.047736
0.049297
0.021863
0.043939
0.049297
0.006124
0.004014
0.006352
0.016094
0.007903
165
A.3 Full list of 86 genes upregulated in keloid compared to normal fibroblasts
using the RMA summarization algorithm (P < 0.05)
Fold
change
18.0305
5.40679
3.959037
3.21548
Gene Symbol
Gene Title
POSTN
IGFBP3
COL15A1
SEMA5A
3.042947
SEMA5A
2.726196
2.631936
2.531559
CADM1
ATXN1
FARP1
2.375326
MICAL2
2.176466
2.129065
ECM1
SLC25A6
2.053115
KCNJ6
1.990826
1.985587
1.959726
1.906954
TBC1D2
CADM1
NXN
MICAL2
1.897066
1.878355
COL1A1
PDGFRB
1.869649
SLC25A6
1.865537
GPSM2
1.852414
1.84265
1.836179
1.835598
1.807992
1.796024
1.764263
LOC644191 ///
LOC728937 ///
RPS26
CTSB
ODZ3
JUP /// KRT19
LAMA2
FHOD1
CTDSPL
periostin, osteoblast specific factor
insulin-like growth factor binding protein 3
collagen, type XV, alpha 1
sema domain, seven thrombospondin repeats
(type 1 and type 1-like), transmembrane domain
(TM) and short cytoplasmic domain, (semaphorin)
5A
sema domain, seven thrombospondin repeats
(type 1 and type 1-like), transmembrane domain
(TM) and short cytoplasmic domain, (semaphorin)
5A
cell adhesion molecule 1
ataxin 1
FERM, RhoGEF (ARHGEF) and pleckstrin
domain protein 1 (chondrocyte-derived)
microtubule associated monoxygenase, calponin
and LIM domain containing 2
extracellular matrix protein 1
solute carrier family 25 (mitochondrial carrier;
adenine nucleotide translocator), member 6
potassium inwardly-rectifying channel, subfamily
J, member 6
TBC1 domain family, member 2
cell adhesion molecule 1
nucleoredoxin
microtubule associated monoxygenase, calponin
and LIM domain containing 2
collagen, type I, alpha 1
platelet-derived growth factor receptor, beta
polypeptide
solute carrier family 25 (mitochondrial carrier;
adenine nucleotide translocator), member 6
G-protein signaling modulator 2 (AGS3-like, C.
elegans)
similar to hCG15685 /// similar to 40S ribosomal
protein S26 /// ribosomal protein S26
1.749725
SHMT2
cathepsin B
odz, odd Oz/ten-m homolog 3 (Drosophila)
junction plakoglobin /// keratin 19
laminin, alpha 2
formin homology 2 domain containing 1
CTD (carboxy-terminal domain, RNA polymerase
II, polypeptide A) small phosphatase-like
serine hydroxymethyltransferase 2 (mitochondrial)
Corrected
p-value
0.006498
0.024325
0.028774
0.013197
0.035137
0.001646
0.020822
0.045664
0.037055
0.047073
0.019936
0.021022
0.048158
0.046985
0.002448
0.017212
0.008827
0.002791
0.046273
0.04761
0.038141
0.018933
6.59E-04
0.02721
0.006566
0.014803
0.018992
0.038598
166
1.747925
1.730545
1.719725
1.686518
1.676297
1.667339
HDLBP
COL5A3
PTK7
NONO
NT5DC2
ATF4
1.664031
1.634808
1.627513
1.615317
1.612627
1.597447
1.58056
1.568977
1.562242
1.544172
1.544152
1.531647
1.521308
1.51899
1.513227
1.499035
1.469517
1.468484
1.459514
1.456908
1.443996
1.441773
1.436187
1.43166
1.425044
1.417303
LOC100130624
FAM155A
PQBP1
RPL13
RPS8
HOXA11
MGC87895 ///
RPS14
RPL3
EIF1
RPL13
RPL8
LOC642741
RPS5
C20orf149
EPHB3
RPL13
RPL10L
RPL3
DBN1
RPL3
RPL10
PARVB
RPS16
SHMT2
RPLP2
SLC1A4
1.415545
1.413611
1.408501
1.405121
1.403469
RPL3
YBX1
RPL13
RPS6
PPP2R3A
1.386019
1.379153
1.360253
1.34878
EIF3B
RPS17L4
RPS2
TREX2 ///
UCHL5IP
RPS13
GRB10
RPL9
1.33581
1.32547
1.303224
high density lipoprotein binding protein
collagen, type V, alpha 3
PTK7 protein tyrosine kinase 7
non-POU domain containing, octamer-binding
5'-nucleotidase domain containing 2
activating transcription factor 4 (tax-responsive
enhancer element B67)
hypothetical LOC100130624
family with sequence similarity 155, member A
polyglutamine binding protein 1
ribosomal protein L13
ribosomal protein S8
homeobox A11
similar to ribosomal protein S14 /// ribosomal
protein S14
ribosomal protein L3
eukaryotic translation initiation factor 1
ribosomal protein L13
ribosomal protein L8
similar to ribosomal protein L3
ribosomal protein S5
chromosome 20 open reading frame 149
EPH receptor B3
ribosomal protein L13
ribosomal protein L10-like
ribosomal protein L3
drebrin 1
ribosomal protein L3
ribosomal protein L10
parvin, beta
ribosomal protein S16
serine hydroxymethyltransferase 2 (mitochondrial)
ribosomal protein, large, P2
solute carrier family 1 (glutamate/neutral amino
acid transporter), member 4
ribosomal protein L3
Y box binding protein 1
ribosomal protein L13
ribosomal protein S6
protein phosphatase 2 (formerly 2A), regulatory
subunit B'', alpha
eukaryotic translation initiation factor 3, subunit B
ribosomal protein S17-like 4
ribosomal protein S2
three prime repair exonuclease 2 /// UCHL5
interacting protein
ribosomal protein S13
growth factor receptor-bound protein 10
ribosomal protein L9
0.037741
0.042118
0.041423
0.017365
0.044871
0.004086
0.039839
0.027042
0.003057
1.97E-04
0.023227
0.037859
0.026445
0.024636
0.038827
7.69E-04
0.001713
0.001728
0.010015
0.047073
0.037648
0.002784
0.030904
0.015644
0.021666
0.025091
0.018933
1.48E-04
0.004886
0.002008
0.008827
0.046909
0.016536
0.040024
7.69E-04
0.043219
0.011736
0.020335
0.04918
0.029039
0.014841
0.00654
0.023425
0.046875
167
1.300171
1.276438
1.274737
RPL38
NUDT3
PARD3
1.27258
1.264076
1.261194
1.255499
1.245831
1.23863
RPL27
RPL17
CDYL
RPS9
RP5-1077B9.4
hCG_21078 ///
RPL27A
EIF1
RPS24
LOC100130553
/// RPS18
INPPL1
RPL28
1.22714
1.216908
1.170586
1.161676
1.146715
ribosomal protein L38
nudix-type motif 3
par-3 partitioning defective 3 homolog (C.
elegans)
ribosomal protein L27
ribosomal protein L17
chromodomain protein, Y-like
ribosomal protein S9
invasion inhibitory protein 45
hCG21078 /// ribosomal protein L27a
0.021022
0.04918
0.047977
eukaryotic translation initiation factor 1
ribosomal protein S24
hypothetical protein LOC100130553 /// ribosomal
protein S18
inositol polyphosphate phosphatase-like 1
ribosomal protein L28
0.033374
0.026217
0.02368
0.040303
0.018933
0.049917
0.007091
0.017708
0.002151
0.022554
0.037601
A.4 Full list of 258 genes downregulated in keloid compared to normal
fibroblasts using the RMA summarization algorithm (P < 0.05)
Fold
change
44.34992
Gene Symbol
Gene Title
CXCL6
41.64695
CXCL1
37.03845
C2 /// CFB
33.61232
29.71772
21.50683
20.34249
19.02187
18.80414
HSD11B1
TNFAIP6
CXCL2
TNFAIP6
IL8
SLC39A8
16.76682
SLC39A8
14.08095
13.78165
C3
RSAD2
12.01816
11.75815
11.65778
10.88791
9.990352
9.679501
SOD2
IL8
CCL2
SFRP1
G0S2
IFI44L
chemokine (C-X-C motif) ligand 6 (granulocyte
chemotactic protein 2)
chemokine (C-X-C motif) ligand 1 (melanoma
growth stimulating activity, alpha)
complement component 2 /// complement factor
B
hydroxysteroid (11-beta) dehydrogenase 1
tumor necrosis factor, alpha-induced protein 6
chemokine (C-X-C motif) ligand 2
tumor necrosis factor, alpha-induced protein 6
interleukin 8
solute carrier family 39 (zinc transporter),
member 8
solute carrier family 39 (zinc transporter),
member 8
complement component 3
radical S-adenosyl methionine domain containing
2
superoxide dismutase 2, mitochondrial
interleukin 8
chemokine (C-C motif) ligand 2
secreted frizzled-related protein 1
G0/G1switch 2
interferon-induced protein 44-like
Corrected
p-value
9.29E-10
1.89E-05
1.75E-08
1.50E-06
0.008827
5.89E-05
0.011388
0.003057
1.22E-05
7.69E-04
2.94E-04
0.013835
0.002706
0.001761
9.46E-07
0.026217
0.001019
0.04903
168
8.604651
8.378408
8.213031
8.204023
8.195308
8.068323
7.890376
7.244113
CHI3L2
IL6
CA12
OAS1
GCH1
CA12
SOD2
IFIT1
7.188362
6.98734
TNFAIP3
IFIT3
6.850272
TNFSF10
6.773449
6.696858
6.448252
6.327625
6.259846
6.225736
6.107744
OAS1
NAMPT
IFIH1
CA12
HERC6
CTSS
GBP1
5.973147
IFIT2
5.971855
5.964412
5.887205
5.848494
5.825303
5.720924
5.663569
5.519223
5.507072
5.327916
5.289413
CA12
CA12
HERC5
CXCL3
SOD2
NAMPT
BTN3A2
WTAP
WTAP
CCL5
ABCA8
5.146521
SLC39A14
4.979725
RARRES3
4.976672
4.80152
IFI44
AKR1B1
4.67249
4.650329
4.624918
4.616492
4.304027
4.30355
LAP3
PTGES
TNFAIP2
CCL5
MT1X
NFKBIA
chitinase 3-like 2
interleukin 6 (interferon, beta 2)
carbonic anhydrase XII
2',5'-oligoadenylate synthetase 1, 40/46kDa
GTP cyclohydrolase 1
carbonic anhydrase XII
superoxide dismutase 2, mitochondrial
interferon-induced protein with tetratricopeptide
repeats 1
tumor necrosis factor, alpha-induced protein 3
interferon-induced protein with tetratricopeptide
repeats 3
tumor necrosis factor (ligand) superfamily,
member 10
2',5'-oligoadenylate synthetase 1, 40/46kDa
nicotinamide phosphoribosyltransferase
interferon induced with helicase C domain 1
carbonic anhydrase XII
hect domain and RLD 6
cathepsin S
guanylate binding protein 1, interferon-inducible,
67kDa
interferon-induced protein with tetratricopeptide
repeats 2
carbonic anhydrase XII
carbonic anhydrase XII
hect domain and RLD 5
chemokine (C-X-C motif) ligand 3
superoxide dismutase 2, mitochondrial
nicotinamide phosphoribosyltransferase
butyrophilin, subfamily 3, member A2
Wilms tumor 1 associated protein
Wilms tumor 1 associated protein
chemokine (C-C motif) ligand 5
ATP-binding cassette, sub-family A (ABC1),
member 8
solute carrier family 39 (zinc transporter),
member 14
retinoic acid receptor responder (tazarotene
induced) 3
interferon-induced protein 44
aldo-keto reductase family 1, member B1 (aldose
reductase)
leucine aminopeptidase 3
prostaglandin E synthase
tumor necrosis factor, alpha-induced protein 2
chemokine (C-C motif) ligand 5
metallothionein 1X
nuclear factor of kappa light polypeptide gene
enhancer in B-cells inhibitor, alpha
0.019936
4.59E-04
2.40E-04
0.015085
2.29E-04
0.00631
0.001861
0.035976
8.07E-05
0.014427
0.03935
0.044028
0.001998
0.016536
0.006223
0.009306
8.67E-05
0.003902
0.049917
0.004125
4.59E-04
0.012534
5.04E-04
7.69E-04
0.023627
0.042157
9.55E-05
0.009169
0.002448
0.006956
0.024324
0.003712
0.037647
3.95E-04
0.024205
0.012578
7.11E-04
0.010015
0.020822
4.59E-04
169
4.096058
4.09106
4.075102
4.064827
4.041319
PLSCR1
CLU
MT1E
TNFAIP3
GBP1
3.909078
3.828279
CLU
STEAP1
3.82739
3.806854
3.680742
3.657572
MT2A
IFI35
XAF1
MT1E /// MT1H ///
MT1M /// MT1P2
3.653427
3.645952
CXCL5
PSMB9
3.586836
3.463399
3.378218
IL15RA
ICAM1
UCHL1
3.338111
3.33326
3.289887
3.285261
3.285236
3.279497
MT1X
NFIB
TMEM100
NRCAM
TLR3
SLC1A1
3.275539
SLC11A2
3.262109
3.245122
3.224767
3.219482
3.194404
IL32
LRRN3
MT1F
PTGFR
PARP12
3.130064
3.075876
3.055434
3.049139
2.993775
2.925094
ICAM1
MID1
MT1P2
VCAM1
CP
AMPD3
2.918264
2.903807
2.896697
IFI30
APOL3
SLC11A2
2.892534
STAT1
phospholipid scramblase 1
clusterin
metallothionein 1E
tumor necrosis factor, alpha-induced protein 3
guanylate binding protein 1, interferon-inducible,
67kDa
clusterin
six transmembrane epithelial antigen of the
prostate 1
metallothionein 2A
interferon-induced protein 35
XIAP associated factor 1
metallothionein 1E /// metallothionein 1H ///
metallothionein 1M /// metallothionein 1
pseudogene 2
chemokine (C-X-C motif) ligand 5
proteasome (prosome, macropain) subunit, beta
type, 9 (large multifunctional peptidase 2)
interleukin 15 receptor, alpha
intercellular adhesion molecule 1
ubiquitin carboxyl-terminal esterase L1 (ubiquitin
thiolesterase)
metallothionein 1X
nuclear factor I/B
transmembrane protein 100
neuronal cell adhesion molecule
toll-like receptor 3
solute carrier family 1 (neuronal/epithelial high
affinity glutamate transporter, system Xag),
member 1
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
interleukin 32
leucine rich repeat neuronal 3
metallothionein 1F
prostaglandin F receptor (FP)
poly (ADP-ribose) polymerase family, member
12
intercellular adhesion molecule 1
midline 1 (Opitz/BBB syndrome)
metallothionein 1 pseudogene 2
vascular cell adhesion molecule 1
ceruloplasmin (ferroxidase)
adenosine monophosphate deaminase (isoform
E)
interferon, gamma-inducible protein 30
apolipoprotein L, 3
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
signal transducer and activator of transcription 1,
0.011388
0.026445
0.047073
0.001023
0.021022
0.033121
0.014024
0.029119
0.025091
0.023346
0.021022
0.004989
0.03417
0.011711
1.10E-04
0.004886
0.026959
0.026219
0.04263
7.94E-04
0.039578
7.69E-04
3.06E-04
4.59E-04
0.00287
0.023227
0.027574
0.048335
3.40E-05
0.033374
0.035137
0.047073
8.67E-05
0.010801
0.023627
0.042118
2.81E-04
0.032961
170
2.862865
MMD
2.810585
2.798357
2.72018
2.711745
ABLIM1
DDX60
FILIP1L
GPRC5B
2.703468
2.693878
HIST1H2BD
PSME2
2.689429
2.688505
2.687159
BTN3A2
CFLAR
LGALS3BP
2.650053
2.647512
2.634461
2.629179
2.62574
2.595849
2.594678
2.593336
2.569516
2.54052
2.528263
2.522403
2.520541
2.514324
CXCL5
DRAM
MT1M
UBE2L6
MT1F
CHEK2
CFLAR
NMI
CFLAR
SIRPA
CEBPD
BTN3A3
CYB5A
SLC11A2
2.498402
2.490228
2.454265
2.411988
2.410709
2.408566
2.390128
PALM
RNF114
TRIM38
SLC25A28
STAT4
CYB5A
KIAA0391 ///
PSMA6
SLC15A3
GHR
C1QTNF1
NFIB
PANX1
LRRN3
RGS3
NT5E
MARCH3
CYB5A
TRIM38
2.366543
2.361601
2.354626
2.351021
2.347035
2.336691
2.333014
2.319342
2.311114
2.27852
2.251881
91kDa
monocyte to macrophage differentiationassociated
actin binding LIM protein 1
DEAD (Asp-Glu-Ala-Asp) box polypeptide 60
filamin A interacting protein 1-like
G protein-coupled receptor, family C, group 5,
member B
histone cluster 1, H2bd
proteasome (prosome, macropain) activator
subunit 2 (PA28 beta)
butyrophilin, subfamily 3, member A2
CASP8 and FADD-like apoptosis regulator
lectin, galactoside-binding, soluble, 3 binding
protein
chemokine (C-X-C motif) ligand 5
damage-regulated autophagy modulator
metallothionein 1M
ubiquitin-conjugating enzyme E2L 6
metallothionein 1F
CHK2 checkpoint homolog (S. pombe)
CASP8 and FADD-like apoptosis regulator
N-myc (and STAT) interactor
CASP8 and FADD-like apoptosis regulator
signal-regulatory protein alpha
CCAAT/enhancer binding protein (C/EBP), delta
butyrophilin, subfamily 3, member A3
cytochrome b5 type A (microsomal)
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
Paralemmin
ring finger protein 114
tripartite motif-containing 38
solute carrier family 25, member 28
signal transducer and activator of transcription 4
cytochrome b5 type A (microsomal)
KIAA0391 /// proteasome (prosome, macropain)
subunit, alpha type, 6
solute carrier family 15, member 3
growth hormone receptor
C1q and tumor necrosis factor related protein 1
nuclear factor I/B
pannexin 1
leucine rich repeat neuronal 3
regulator of G-protein signaling 3
5'-nucleotidase, ecto (CD73)
membrane-associated ring finger (C3HC4) 3
cytochrome b5 type A (microsomal)
tripartite motif-containing 38
0.006815
0.026332
0.043219
0.004067
0.016484
0.015237
0.001077
0.025091
0.015237
0.013645
0.015587
0.003902
0.008827
0.03412
9.22E-04
5.84E-05
0.00493
0.044028
5.48E-04
1.38E-05
0.019171
0.048335
0.011736
0.006815
0.009971
0.047073
0.021022
0.001713
3.39E-04
0.015085
0.027574
0.027138
0.001713
0.048335
0.040444
0.03639
0.016484
0.026787
0.002448
0.001642
0.023425
0.032743
171
2.23989
2.235737
CFLAR
PPFIBP2
2.224926
2.22473
2.215581
2.210316
HLA-F
TAPBP
HLA-C
AK3L1 /// AK3L2
2.20811
2.17376
2.169356
2.167533
2.146688
2.13148
2.124391
2.12098
HLA-F
SMC2
CYLD
NOVA1
C10orf26
HLA-B
HTATIP2
SLC39A8
2.108726
2.097694
2.097592
2.092583
2.089171
2.072831
2.0599
TNIP1
SAMHD1
PON2
CFLAR
HLA-B
CFLAR
DHRS3
2.055893
2.048616
2.023844
DKFZP586H2123
CFLAR
HLA-B /// MICA
2.021447
2.007685
2.001211
1.998355
1.981507
BIRC3
PDPN
LY6E
ZC3H12A
HLA-A /// HLAA29.1 /// HLA-B
/// HLA-G /// HLAH /// HLA-J
1.96592
1.942084
1.935575
1.927046
1.924471
1.911548
1.909791
1.896043
1.877315
1.876992
CSF1
HTATIP2
ACP2
PON2
IFNGR1
FAM117A
C14orf159
HLA-G
MT1F
HLA-C
CASP8 and FADD-like apoptosis regulator
PTPRF interacting protein, binding protein 2
(liprin beta 2)
major histocompatibility complex, class I, F
TAP binding protein (tapasin)
major histocompatibility complex, class I, C
adenylate kinase 3-like 1 /// adenylate kinase 3like 2
major histocompatibility complex, class I, F
structural maintenance of chromosomes 2
cylindromatosis (turban tumor syndrome)
neuro-oncological ventral antigen 1
chromosome 10 open reading frame 26
major histocompatibility complex, class I, B
HIV-1 Tat interactive protein 2, 30kDa
solute carrier family 39 (zinc transporter),
member 8
TNFAIP3 interacting protein 1
SAM domain and HD domain 1
paraoxonase 2
CASP8 and FADD-like apoptosis regulator
major histocompatibility complex, class I, B
CASP8 and FADD-like apoptosis regulator
dehydrogenase/reductase (SDR family) member
3
regeneration associated muscle protease
CASP8 and FADD-like apoptosis regulator
major histocompatibility complex, class I, B ///
MHC class I polypeptide-related sequence A
baculoviral IAP repeat-containing 3
podoplanin
lymphocyte antigen 6 complex, locus E
zinc finger CCCH-type containing 12A
major histocompatibility complex, class I, A ///
major histocompatibility complex class I HLAA29.1 /// major histocompatibility complex, class
I, B /// major histocompatibility complex, class I,
G /// major histocompatibility complex, class I, H
(pseudogene) /// major histocompatibility
complex, class I, J (pseudogene)
colony stimulating factor 1 (macrophage)
HIV-1 Tat interactive protein 2, 30kDa
acid phosphatase 2, lysosomal
paraoxonase 2
interferon gamma receptor 1
family with sequence similarity 117, member A
chromosome 14 open reading frame 159
major histocompatibility complex, class I, G
metallothionein 1F
major histocompatibility complex, class I, C
0.023911
0.003755
0.009477
0.002016
0.006519
0.003202
0.00287
0.002998
0.046985
0.04244
0.025091
0.006956
0.026762
0.004086
0.027814
0.047977
0.025091
0.003645
0.002024
0.00197
0.001019
0.037791
8.55E-04
0.018933
0.001713
0.033129
0.027138
5.99E-05
0.006852
0.015085
0.003902
0.006566
0.014803
0.018356
0.001713
0.010796
0.008626
1.10E-04
7.69E-04
172
1.87264
1.87205
1.85887
1.857911
MEIS3P1
TAPBPL
HLA-G
BASP1
1.847917
1.844846
1.843371
1.837641
1.822478
1.811002
1.805608
1.801476
1.785207
1.763702
1.752336
1.748331
1.718806
1.700887
1.683461
1.668382
1.667567
1.651713
MICALL2
SIRPA
GFRA1
HLA-E
DENND2D
TRIM38
NNMT
FTH1
PION
NNMT
HLA-C
NR4A3
PROCR
PDCD5
HSPB8
HLA-C
PDCD1LG2
SLCO3A1
1.641853
SLC1A1
1.629699
1.626722
1.625006
1.61725
1.617092
HLA-G
BTN3A1
HLA-A
LARGE
AK3L1 /// AK3L2
1.592655
1.589381
PION
NFKBIE
1.584692
1.576378
1.571031
RNF8
CD59
LYN
1.563836
MMP2
1.556934
1.533629
NRP2
P2RX4
1.523406
NFKB1
1.522656
SVEP1
1.518637
WWC1
Meis homeobox 3 pseudogene 1
TAP binding protein-like
major histocompatibility complex, class I, G
brain abundant, membrane attached signal
protein 1
MICAL-like 2
signal-regulatory protein alpha
GDNF family receptor alpha 1
major histocompatibility complex, class I, E
DENN/MADD domain containing 2D
tripartite motif-containing 38
nicotinamide N-methyltransferase
ferritin, heavy polypeptide 1
pigeon homolog (Drosophila)
nicotinamide N-methyltransferase
major histocompatibility complex, class I, C
nuclear receptor subfamily 4, group A, member 3
protein C receptor, endothelial (EPCR)
programmed cell death 5
heat shock 22kDa protein 8
major histocompatibility complex, class I, C
programmed cell death 1 ligand 2
solute carrier organic anion transporter family,
member 3A1
solute carrier family 1 (neuronal/epithelial high
affinity glutamate transporter, system Xag),
member 1
major histocompatibility complex, class I, G
butyrophilin, subfamily 3, member A1
major histocompatibility complex, class I, A
like-glycosyltransferase
adenylate kinase 3-like 1 /// adenylate kinase 3like 2
pigeon homolog (Drosophila)
nuclear factor of kappa light polypeptide gene
enhancer in B-cells inhibitor, epsilon
ring finger protein 8
CD59 molecule, complement regulatory protein
v-yes-1 Yamaguchi sarcoma viral related
oncogene homolog
matrix metallopeptidase 2 (gelatinase A, 72kDa
gelatinase, 72kDa type IV collagenase)
neuropilin 2
purinergic receptor P2X, ligand-gated ion
channel, 4
nuclear factor of kappa light polypeptide gene
enhancer in B-cells 1
sushi, von Willebrand factor type A, EGF and
pentraxin domain containing 1
WW and C2 domain containing 1
0.007091
0.001292
0.003202
0.026445
7.91E-04
0.002908
0.031214
0.042803
0.009234
0.014427
0.033657
0.032417
0.026445
0.015237
0.001075
0.010796
0.027362
0.039578
0.04761
0.016484
0.004125
0.00287
0.014024
0.006852
0.0442
0.002743
0.013456
0.033481
0.003955
0.019276
0.016484
4.56E-04
0.029499
5.99E-05
0.02645
0.021666
0.042118
0.011711
0.045278
173
1.499967
1.497518
1.493991
1.489115
LSAMP
APOL2
PGK1
GBA /// GBAP
1.488427
KHDC1 /// SPA17
1.482473
PSTPIP2
1.466329
POLD3
1.449279
1.446802
DTNA
RELB
1.445625
ELF3
1.440517
CYP27A1
1.439601
1.429019
1.406325
HYPK
ACP6
TFDP2
1.397198
1.388266
1.38517
1.382151
CTSS
B2M
CSTB
SNCA
1.368093
1.367311
1.349224
1.321152
PDLIM4
NFE2L1
FTH1
ARHGEF10L
1.320407
1.313343
1.308686
1.307239
1.272
1.266423
LGALS8
FTHP1
EDNRB
HLA-F
SLC19A3
SLC11A2
1.25115
1.249708
RBKS
CAND2
1.246255
1.238541
1.232883
1.225532
1.196416
1.174331
SNX11
C6orf64
ZMIZ2
CSF1
NFIB
TFDP2
1.155878
WDR48
limbic system-associated membrane protein
apolipoprotein L, 2
phosphoglycerate kinase 1
glucosidase, beta; acid (includes
glucosylceramidase) /// glucosidase, beta; acid,
pseudogene
KH homology domain containing 1 /// sperm
autoantigenic protein 17
proline-serine-threonine phosphatase interacting
protein 2
polymerase (DNA-directed), delta 3, accessory
subunit
dystrobrevin, alpha
v-rel reticuloendotheliosis viral oncogene
homolog B
E74-like factor 3 (ets domain transcription factor,
epithelial-specific )
cytochrome P450, family 27, subfamily A,
polypeptide 1
Huntingtin interacting protein K
acid phosphatase 6, lysophosphatidic
transcription factor Dp-2 (E2F dimerization
partner 2)
cathepsin S
beta-2-microglobulin
cystatin B (stefin B)
synuclein, alpha (non A4 component of amyloid
precursor)
PDZ and LIM domain 4
nuclear factor (erythroid-derived 2)-like 1
ferritin, heavy polypeptide 1
Rho guanine nucleotide exchange factor (GEF)
10-like
lectin, galactoside-binding, soluble, 8
ferritin, heavy polypeptide pseudogene 1
endothelin receptor type B
major histocompatibility complex, class I, F
solute carrier family 19, member 3
solute carrier family 11 (proton-coupled divalent
metal ion transporters), member 2
ribokinase
cullin-associated and neddylation-dissociated 2
(putative)
sorting nexin 11
chromosome 6 open reading frame 64
zinc finger, MIZ-type containing 2
colony stimulating factor 1 (macrophage)
nuclear factor I/B
transcription factor Dp-2 (E2F dimerization
partner 2)
WD repeat domain 48
0.009696
0.047977
0.042095
0.036143
0.004067
0.008121
0.025058
0.037974
0.009818
0.02046
0.04263
0.019813
0.039972
0.006852
0.046985
0.024654
0.017681
0.006815
0.041423
0.032417
0.013128
0.030511
0.044328
0.03731
0.046273
0.039578
0.031469
0.018933
0.018933
0.020736
0.020092
0.044823
0.033374
0.013754
0.047977
0.029499
0.006116
174
1.146091
1.131624
1.106074
1.104509
SP100
TLR1
CXCL5
TAF1B
1.101118
BRCA2
SP100 nuclear antigen
toll-like receptor 1
chemokine (C-X-C motif) ligand 5
TATA box binding protein (TBP)-associated
factor, RNA polymerase I, B, 63kDa
breast cancer 2, early onset
0.015085
0.033657
0.044823
0.037317
0.027138
A.5 List of genes differentially expressed using both the RMA and MAS 5.0
summarization algorithm (P < 0.05)
Regulation
(Keloid vs
Normal)
down
Gene
Symbol
Gene Title
CXCL6
down
CXCL1
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
IL8
HSD11B1
IL8
CCL5
CXCL2
RSAD2
C2 /// CFB
CXCL5
TNFAIP6
CXCL3
IL32
CP
CHI3L2
TNFAIP6
C3
SLC39A8
CXCL5
G0S2
OAS1
TNFSF10
SLC39A8
GCH1
CCL5
SLC19A3
HERC5
IL6
NRCAM
CCL2
ABCA8
chemokine (C-X-C motif) ligand 6 (granulocyte chemotactic
protein 2)
chemokine (C-X-C motif) ligand 1 (melanoma growth stimulating
activity, alpha)
interleukin 8
hydroxysteroid (11-beta) dehydrogenase 1
interleukin 8
chemokine (C-C motif) ligand 5
chemokine (C-X-C motif) ligand 2
radical S-adenosyl methionine domain containing 2
complement component 2 /// complement factor B
chemokine (C-X-C motif) ligand 5
tumor necrosis factor, alpha-induced protein 6
chemokine (C-X-C motif) ligand 3
interleukin 32
ceruloplasmin (ferroxidase)
chitinase 3-like 2
tumor necrosis factor, alpha-induced protein 6
complement component 3
solute carrier family 39 (zinc transporter), member 8
chemokine (C-X-C motif) ligand 5
G0/G1switch 2
2',5'-oligoadenylate synthetase 1, 40/46kDa
tumor necrosis factor (ligand) superfamily, member 10
solute carrier family 39 (zinc transporter), member 8
GTP cyclohydrolase 1
chemokine (C-C motif) ligand 5
solute carrier family 19, member 3
hect domain and RLD 5
interleukin 6 (interferon, beta 2)
neuronal cell adhesion molecule
chemokine (C-C motif) ligand 2
ATP-binding cassette, sub-family A (ABC1), member 8
175
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
SFRP1
TMEM100
SOD2
CTSS
IFI30
HERC6
TLR3
RARRES3
SOD2
TNFAIP3
CA12
CTSS
IFIT2
LRRN3
IFIT3
BIRC3
CA12
GBP1
TNFAIP2
CA12
PTGES
GBP1
NAMPT
SLC39A8
CA12
STAT4
CA12
LRRN3
MT1X
NAMPT
MT1M
WTAP
MT1P2
ELF3
down
down
down
down
down
down
down
down
down
down
down
down
down
TNFAIP3
SOD2
MT1E
ZC3H12A
ICAM1
SLC39A14
MT1F
AKR1B1
SNCA
WWC1
ICAM1
WTAP
SLC15A3
secreted frizzled-related protein 1
transmembrane protein 100
superoxide dismutase 2, mitochondrial
cathepsin S
interferon, gamma-inducible protein 30
hect domain and RLD 6
toll-like receptor 3
retinoic acid receptor responder (tazarotene induced) 3
superoxide dismutase 2, mitochondrial
tumor necrosis factor, alpha-induced protein 3
carbonic anhydrase XII
cathepsin S
interferon-induced protein with tetratricopeptide repeats 2
leucine rich repeat neuronal 3
interferon-induced protein with tetratricopeptide repeats 3
baculoviral IAP repeat-containing 3
carbonic anhydrase XII
guanylate binding protein 1, interferon-inducible, 67kDa
tumor necrosis factor, alpha-induced protein 2
carbonic anhydrase XII
prostaglandin E synthase
guanylate binding protein 1, interferon-inducible, 67kDa
nicotinamide phosphoribosyltransferase
solute carrier family 39 (zinc transporter), member 8
carbonic anhydrase XII
signal transducer and activator of transcription 4
carbonic anhydrase XII
leucine rich repeat neuronal 3
metallothionein 1X
nicotinamide phosphoribosyltransferase
metallothionein 1M
Wilms tumor 1 associated protein
metallothionein 1 pseudogene 2
E74-like factor 3 (ets domain transcription factor, epithelialspecific )
tumor necrosis factor, alpha-induced protein 3
superoxide dismutase 2, mitochondrial
metallothionein 1E
zinc finger CCCH-type containing 12A
intercellular adhesion molecule 1
solute carrier family 39 (zinc transporter), member 14
metallothionein 1F
aldo-keto reductase family 1, member B1 (aldose reductase)
synuclein, alpha (non A4 component of amyloid precursor)
WW and C2 domain containing 1
intercellular adhesion molecule 1
Wilms tumor 1 associated protein
solute carrier family 15, member 3
176
down
down
down
down
down
down
IFI35
MT1E ///
MT1H ///
MT1M ///
MT1P2
AMPD3
MT2A
LAP3
SLC11A2
down
down
down
down
down
MARCH3
CLU
STEAP1
GFRA1
NFKBIA
down
down
down
MMD
MT1X
SLC11A2
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
PTGFR
NR4A3
PLSCR1
UCHL1
CHEK2
MT1F
EDNRB
IL15RA
STAT1
SMC2
SAMHD1
ABLIM1
MT1F
HLA-F
CFLAR
SIRPA
HLA-C
SLC25A28
AK3L1 ///
AK3L2
TAPBPL
NOVA1
DTNA
PDPN
LGALS3BP
PSME2
down
SLC1A1
interferon-induced protein 35
metallothionein 1E /// metallothionein 1H /// metallothionein 1M ///
metallothionein 1 pseudogene 2
adenosine monophosphate deaminase (isoform E)
metallothionein 2A
leucine aminopeptidase 3
solute carrier family 11 (proton-coupled divalent metal ion
transporters), member 2
membrane-associated ring finger (C3HC4) 3
Clusterin
six transmembrane epithelial antigen of the prostate 1
GDNF family receptor alpha 1
nuclear factor of kappa light polypeptide gene enhancer in B-cells
inhibitor, alpha
monocyte to macrophage differentiation-associated
metallothionein 1X
solute carrier family 11 (proton-coupled divalent metal ion
transporters), member 2
prostaglandin F receptor (FP)
nuclear receptor subfamily 4, group A, member 3
phospholipid scramblase 1
ubiquitin carboxyl-terminal esterase L1 (ubiquitin thiolesterase)
CHK2 checkpoint homolog (S. pombe)
metallothionein 1F
endothelin receptor type B
interleukin 15 receptor, alpha
signal transducer and activator of transcription 1, 91kDa
structural maintenance of chromosomes 2
SAM domain and HD domain 1
actin binding LIM protein 1
metallothionein-1F
major histocompatibility complex, class I, F
CASP8 and FADD-like apoptosis regulator
signal-regulatory protein alpha
major histocompatibility complex, class I, C
solute carrier family 25, member 28
adenylate kinase 3-like 1 /// adenylate kinase 3-like 2
TAP binding protein-like
neuro-oncological ventral antigen 1
dystrobrevin, alpha
Podoplanin
lectin, galactoside-binding, soluble, 3 binding protein
proteasome (prosome, macropain) activator subunit 2 (PA28
beta)
solute carrier family 1 (neuronal/epithelial high affinity glutamate
transporter, system Xag), member 1
177
down
down
down
down
down
down
down
down
down
down
down
down
CFLAR
HIST1H2BD
PROCR
CFLAR
CYB5A
FILIP1L
PSTPIP2
HLA-F
CFLAR
PPFIBP2
RNF114
SLC11A2
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
down
CFLAR
DRAM
CFLAR
HLA-G
LYN
HLA-B
HTATIP2
MICALL2
SIRPA
HLA-G
CYB5A
TAPBP
PON2
NT5E
TRIM38
HLA-B
CFLAR
KIAA0391
/// PSMA6
TRIM38
NFKBIE
down
down
down
down
down
down
down
down
down
down
down
down
down
C10orf26
LY6E
PANX1
CSF1
TNIP1
RELB
ACP2
PON2
TRIM38
HLA-C
HLA-A ///
HLA-A29.1
/// HLA-B ///
HLA-G ///
CASP8 and FADD-like apoptosis regulator
histone cluster 1, H2bd
protein C receptor, endothelial (EPCR)
CASP8 and FADD-like apoptosis regulator
cytochrome b5 type A (microsomal)
filamin A interacting protein 1-like
proline-serine-threonine phosphatase interacting protein 2
major histocompatibility complex, class I, F
CASP8 and FADD-like apoptosis regulator
PTPRF interacting protein, binding protein 2 (liprin beta 2)
ring finger protein 114
solute carrier family 11 (proton-coupled divalent metal ion
transporters), member 2
CASP8 and FADD-like apoptosis regulator
damage-regulated autophagy modulator
CASP8 and FADD-like apoptosis regulator
major histocompatibility complex, class I, G
v-yes-1 Yamaguchi sarcoma viral related oncogene homolog
major histocompatibility complex, class I, B
HIV-1 Tat interactive protein 2, 30kDa
MICAL-like 2
signal-regulatory protein alpha
major histocompatibility complex, class I, G
cytochrome b5 type A (microsomal)
TAP binding protein (tapasin)
paraoxonase 2
5'-nucleotidase, ecto (CD73)
tripartite motif-containing 38
major histocompatibility complex, class I, B
CASP8 and FADD-like apoptosis regulator
KIAA0391 /// proteasome (prosome, macropain) subunit, alpha
type, 6
tripartite motif-containing 38
nuclear factor of kappa light polypeptide gene enhancer in B-cells
inhibitor, epsilon
chromosome 10 open reading frame 26
lymphocyte antigen 6 complex, locus E
pannexin 1
colony stimulating factor 1 (macrophage)
TNFAIP3 interacting protein 1
v-rel reticuloendotheliosis viral oncogene homolog B
acid phosphatase 2, lysosomal
paraoxonase 2
tripartite motif-containing 38
major histocompatibility complex, class I, C
major histocompatibility complex, class I, A /// major
histocompatibility complex class I HLA-A29.1 /// major
histocompatibility complex, class I, B /// major histocompatibility
complex, class I, G /// major histocompatibility complex, class I, H
178
down
down
down
down
down
down
down
down
down
down
down
down
down
down
up
up
up
up
up
HLA-H ///
HLA-J
NNMT
HLA-C
BASP1
IFNGR1
FTH1
HLA-C
HLA-B ///
MICA
CD59
NNMT
HLA-A
MMP2
up
up
up
up
CSTB
FTH1
B2M
POSTN
COL15A1
HOXA11
KCNJ6
JUP ///
KRT19
IGFBP3
ATXN1
CADM1
SEMA5A
up
up
up
GPSM2
FAM155A
MICAL2
up
up
CADM1
SEMA5A
up
up
SHMT2
MICAL2
up
FARP1
up
up
up
up
MGC87895
/// RPS14
TBC1D2
ECM1
SLC25A6
up
PDGFRB
(pseudogene) /// major histocompatibility complex, class I, J
(pseudogene)
nicotinamide N-methyltransferase
major histocompatibility complex, class I, C
brain abundant, membrane attached signal protein 1
interferon gamma receptor 1
ferritin, heavy polypeptide 1
major histocompatibility complex, class I, C
major histocompatibility complex, class I, B /// MHC class I
polypeptide-related sequence A
CD59 molecule, complement regulatory protein
nicotinamide N-methyltransferase
major histocompatibility complex, class I, A
matrix metallopeptidase 2 (gelatinase A, 72kDa gelatinase,
72kDa type IV collagenase)
cystatin B (stefin B)
ferritin, heavy polypeptide 1
beta-2-microglobulin
periostin, osteoblast specific factor
collagen, type XV, alpha 1
homeobox A11
potassium inwardly-rectifying channel, subfamily J, member 6
junction plakoglobin /// keratin 19
insulin-like growth factor binding protein 3
ataxin 1
cell adhesion molecule 1
sema domain, seven thrombospondin repeats (type 1 and type 1like), transmembrane domain (TM) and short cytoplasmic domain,
(semaphorin) 5A
G-protein signaling modulator 2 (AGS3-like, C. elegans)
family with sequence similarity 155, member A
microtubule associated monoxygenase, calponin and LIM domain
containing 2
cell adhesion molecule 1
sema domain, seven thrombospondin repeats (type 1 and type 1like), transmembrane domain (TM) and short cytoplasmic domain,
(semaphorin) 5A
serine hydroxymethyltransferase 2 (mitochondrial)
microtubule associated monoxygenase, calponin and LIM domain
containing 2
FERM, RhoGEF (ARHGEF) and pleckstrin domain protein 1
(chondrocyte-derived)
similar to ribosomal protein S14 /// ribosomal protein S14
TBC1 domain family, member 2
extracellular matrix protein 1
solute carrier family 25 (mitochondrial carrier; adenine nucleotide
translocator), member 6
platelet-derived growth factor receptor, beta polypeptide
179
up
up
up
CTSB
FHOD1
SLC25A6
up
up
up
up
up
up
up
up
up
NXN
PTK7
LAMA2
LOC644191
///
LOC728937
/// RPS26
PARVB
COL1A1
SHMT2
NONO
ATF4
up
up
up
up
up
up
up
up
RPS9
RPL13
RPS8
RPL13
RPS6
RPL8
PPP2R3A
RPL13
cathepsin B
formin homology 2 domain containing 1
solute carrier family 25 (mitochondrial carrier; adenine nucleotide
translocator), member 6
Nucleoredoxin
PTK7 protein tyrosine kinase 7
laminin, alpha 2
similar to hCG15685 /// similar to 40S ribosomal protein S26 ///
ribosomal protein S26
parvin, beta
collagen, type I, alpha 1
serine hydroxymethyltransferase 2 (mitochondrial)
non-POU domain containing, octamer-binding
activating transcription factor 4 (tax-responsive enhancer element
B67)
ribosomal protein S9
ribosomal protein L13
ribosomal protein S8
ribosomal protein L13
ribosomal protein S6
ribosomal protein L8
protein phosphatase 2 (formerly 2A), regulatory subunit B'', alpha
ribosomal protein L13
180
A.6 Cytokine-cytokine receptor interaction from the KEGG database
(Benjamini corrected P-value = 0.094)
181
A.7 Toll-like receptor signaling pathway from the KEGG database
(Benjamini corrected P-value = 0.246)
182
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Related manuals

Download PDF

advertisement