University of Southern California Dissertations and Theses

Development of DNA methylation biomarkers as an early detection tool for human lung adenocarcinoma

(USC Thesis Other)

Development of DNA methylation biomarkers as an early detection tool for human lung adenocarcinoma

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content DEVELOPMENT OF DNA METHYLATION BIOMARKERS AS AN
EARLY DETECTION TOOL FOR HUMAN LUNG ADENOCARCINOMA

by

Janice Soratorio Galler

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOCHEMISTRY AND MOLECULAR BIOLOGY)

December 2008

Copyright 2008 Janice Soratorio Galler

ii
Dedication
For the past five years, I have been very fortunate to be a part of the Laird
and Laird-Offringa labs. I am so grateful for the opportunity to experience the
rollercoaster ride and adventures of graduate school, for everyone I have met
and for those who have touched my life. Throughout the years, everyone in the
lab has made a lasting impression and taught me some very important lessons.
Thank you for challenging me, helping me realize my potential, cheering me up
when I was down, and supporting me in every way possible. I have made lasting
friendships and many memories will remain for years to come.
To Peter and Ite: Thank you for always being patient and supportive. I am
grateful for being part of your family. I often say I do not know how you do it. To
this day, I am at awe with how you manage your careers and personal lives. I just
hope that in 20 years, Eric and I will have the same relationship you have. You
two are truly inspirational.
To my mom and dad: There are no words to express my gratitude for all
that you’ve done to help my family. I know it hasn’t been easy for the past 4
years, juggling schedules and sacrificing so much so that I could pursue my
dream. There was no one else in the world that I would have trusted with Lauren
and Allison. I am so happy that you play a major part in their lives and I thank you
for the unconditional love and support.

iii
To my brothers, Lew and Emmanuel, and sister, Amabelle (Ryan, Josh,
and Nicole): Thank you for putting up with so much when we were growing up (I
know I wasn’t the easiest person to get along with). And no matter what happens,
you guys have always been there for my family when we needed you. You guys
have made me realize the importance of family and coming home to you made
me forget the bad and focus on the good in my life.
To my husband, Eric: You have always known what to say to me when I
complain about little things. You know how to not just encourage me but also put
me in my place. Thank you for teaching me to be a little tougher. Thank you for
being a wonderful husband and a great father to our children. I am truly blessed
and cannot ask for anything more. I know the past 11 years has not been easy,
but we have made it! I am looking forward to the next chapter in our lives.
To my dear Lauren and Allison: You two are my inspiration. My every
moment revolves around you, even if I was not around at times. Always
remember that mommy loves you so much and I will do anything for you. I wish
for you to stay healthy and happy, enjoy every moment and live life to the fullest!

iv
Acknowledgments
None of this work would have been possible without the help of many
people. First, I’d like to thank Dr. Ite Laird-Offringa for giving me the opportunity
to work on this project and being a great advisor for the past 5 years. Thank you
to the Laird-Offringa lab members, especially Dr. Jeff Tsou, who mentored me
through my rotation and introduced me to the project. When it came to technical
support and discussion, I’d like to thank Dr. Peter Laird and his lab, particularly
Dr. Martin Brena, Dr. Binh Trinh, Dr. Mihaela Campan, Dr. Daniel Weisenberger,
and Tiffany Long. who have always been available to answer questions. This
project involves a great team of people. Thank you to many colleagues and
volunteers who helped collect and process the tissues: Paul Anglim, Courtney
Brennan, Erika Flores, Arzu Oezcelik, and Suhaida Selamat. I greatly appreciate
all the help, support, and guidance.

v
Table of Contents

Dedication ii
Acknowledgments iv
List of Tables vii
List of Figures ix
Abstract x
Chapter 1: Introduction 1
Lung Cancer 1
Early Detection of Lung Cancer 2
Chest radiography, sputum cytology, 3
and low-dose spiral computed tomography
Molecular markers: protein and DNA 6
DNA methylation and non-small cell lung cancer 10
Chapter 2: Identification of a panel of sensitive and specific 13
DNA methylation markers for lung adenocarcinoma
Chapter 2 Abstract 13
Introduction 15
Methods 18
Results/Discussion 22
Conclusion 40
Chapter 3: Development and validation of a panel 43
of 15 DNA methylation-based biomarkers for human
lung adenocarcinoma
Chapter 3 Abstract 43
Introduction 44
Methods 46
Results/Discussion 49
Conclusion 62

vi
Chapter 4: The onset and progression of aberrant 64
DNA methylation of cancer-related genes in lung
adenocarcinoma development
Chapter 4 Abstract 64
Introduction 65
Methods 69
Results/Discussion 70
Conclusion 82
Chapter 5: Discovery of DNA methylation markers using a 84
high-dimensional bead-array
Chapter 5 Abstract 84
Introduction 85
Methods 88
Results/Discussion 91
Conclusion 101
Chapter 6: Discussion 103
Bibliography 111

vii
List of Tables

Table 2.1 Gene name and function for 28 loci studied 24
Table 2.2 Frequency and median PMR values of 27
adenocarcinoma, AdjNTL and NTL tissues for
28 loci
Table 2.3 Performance of top four markers in samples 35
based on gender, race/ethnicity and stage

Table 3.1 Summary of 15 MethyLight primer and probe 50
Sets used in the DNA methylation study

Table 3.2 Summary of the characteristics of the 15 loci 53
Table 3.3 Frequency and median PMR values of 55
adenocarcinoma and AdjNTL tissues for 15 loci

Table 3.4 Summary and sensitivity and specificity of 57
the 15 loci
Table 3.5-1 Validation set 1: summary of frequency, 60
median and p-value of paraffin-embedded samples

Table 3.5-2 Validation set 2: summary of frequency, median 60
and p-value of fresh frozen samples
Table 3.6 Summary of sensitivity and specificity of different 61
DNA methylation marker panel
Table 4.1 MethyLight primer and probe collection for loci 73
analyzed in adjacent non-tumor lung, AAH and
AD tissues
Table 4.2 Summary of frequency and median PMR values 76
of AdjNTL, AAH and AD

viii
Table 5.1 Summary of MethyLight primer and probe used 92
for validation study
Table 5.2 List of top loci identified by Illumina GoldenGate 94
assay
Table 5.3 List of loci specific for lung cancer 97
Table 5.4 Validation study summary of frequency and 100
median PMR values of AD and AdjNTL

ix
List of Figures

Figure 2.1 Graphic representation of PMR values obtained 25
for 28 loci in AD, AdjNTL and NTL
Figure 2.2 Two dimensional hierarchical clustering of 28
samples and loci based on methylation data used
as a continuous variable

Figure 2.3 The distribution of PMR values by group 31
Figure 2.4 Receiver operating characteristic curves for the 32
four top markers
Figure 3.1 Heat map of 200-loci screen of paired AD and 52
AdjNTL samples
Figure 3.2 Panel of 15 DNA methylation biomarkers 54
Figure 3.3 PMR distribution of 15 DNA methylation markers 58
Figure 4.1 Schematic of DNA methylation status of AdjNTL, 74
AAH, and AD
Figure 4.2 DNA methylation pattern of AdjNTL, AAH and AD 77
Figure 4.3 DNA methylation signature of AdjNTL, AAH and 81
AD tissues from one AD patient

Figure 5.1 Methylation status of WBC DNA of 2 “healthy” 99
controls

x
Abstract

Lung cancer is the leading cause of cancer-related death in the United
States. It is estimated that more than 160,000 lung cancer patients will die of the
disease in 2008. Early detection is key to improving patient survival; however,
current screening strategies have low specificity and/or sensitivity in detecting
early lung cancer. Our focus was to study DNA methylation and its role in lung
cancer development and progression with the hope of developing molecular
markers for the early detection of lung adenocarcinoma (AD; the most common
histology of lung cancer). Understanding the timing of DNA methylation changes
during the multi-stage process of AD development could provide a panel of
biomarkers potentially used to detect the disease at earlier stages. We have
developed a panel of 15 loci involved in cell regulatory events that have
significant differential methylation between AD and adjacent non-tumor lung
tissues (AdjNTL) by MethyLight, a quantitative DNA methylation assay. When
used in a panel, all 15 loci showed a sensitivity of 92.85% and specificity of
100%. These tumor-specific DNA methylation markers were used to study
whether the same methylation signatures are present in Atypical Adenomatous
Hyperplasia (AAH, putative precancerous lesion of lung AD). In our analysis of
tissues from AD patients with AAH lesions, we observed variability in the pattern

xi
of methylation between the different loci. Some loci (e.g. CDKN2A) were
aberrantly methylated even in non-cancerous tissues. Other loci started
accumulating methylation marks in AAH (e.g. SFRP1) while other loci do not
become hypermethylated until in its invasive state (RASSF1). The difference in
timing of methylation between the different loci could be a reflection of the roles
of the pathways involved in adeno-carcinogenesis. Additional cancer-specific
DNA methylation markers had recently been identified through high-throughput
Illumina GoldenGate Methylation assay. The onset of DNA methylation of the loci
identified by Illumina GoldenGate assay in AD development remains to be
elucidated. The knowledge of sequential methylation changes in AAH-AD
continuum could yield powerful risk and/or early detection biomarkers for AD. In
addition, our findings could provide insights into the natural history of this lethal
disease.

1
Chapter 1. Introduction

Lung Cancer
Lung cancer is the leading cancer killer for both men and women in the
United States, causing 31% and 26% of cancer fatalities, respectively. With more
than 160,000 deaths estimated in the United States for 2008, lung cancer causes
more deaths than breast, prostate and colon cancer combined (American Cancer
Society 2008). Many epidemiological studies have confirmed the causal
relationship between smoking and the risk of developing lung cancer in both men
and women (Freedman et al 2008, Bae et al 2007, Malila et al 2006, Yun et al
2005, Harris et al 2004). The majority of lung cancers (~85%) arise in current or
previous smokers (Schottenfeld 1996). As tobacco consumption increases in
developing countries such as China, the worldwide incidence of lung cancer is
expected to rise (Zhang and Cai 2003). Interestingly, the American Cancer
Society estimates that in the United States, currently 50% of all new lung cancer
cases occur in former and never smokers. Clearly, other factors such as genetic
predisposition and/or other environmental insults (e.g. asbestos and radon
exposure) also play a role in lung cancer development (reviewed in Alberg et al

2
2007). For the purpose of early detection, it is important to understand the
molecular mechanisms underlying the development of lung cancer to identify at-
risk groups.
Lung cancer is classified into two major groups: small cell lung cancer
(SCLC) and non-small cell lung cancer (NSCLC). Small cell lung cancer, which
accounts for approximately 10-15% of all lung cancer cases, is the most
aggressive form of lung cancer, while the less aggressive NSCLC (~85-90% of
lung cancer cases) (American Cancer Society 2008) is further divided into four
histological subtypes: adenocarcinoma (AD), squamous cell carcinoma (SCC),
large cell carcinoma and a group consisting of neuroendocrine cancers,
carcinoids, etc (Collins et al 2007). Squamous cell carcinoma was the most
frequent subtype of lung cancer due to the widespread use of tobacco smoke in
the early 20
th
century, until relatively recently. Now, the incidence of AD has
surpassed that of SCC (Charloux et al 1997, Pirozynski 2006). Adenocarcinoma
is also the most common subtype in previous and never-smokers (Liu et al 2000)
Thus, the focus of the following chapters will be on AD and the development of
molecular markers for its early detection.
Early Detection of Lung Cancer
The overall five-year survival rate of lung cancer is a dismal 15% (Jemal et
al 2006) because the majority of lung cancer patients are diagnosed at an

3
advanced stage. A recent study by Wisniveskya et al (2005) of patients with
limited disease (stage II) showed a 5-year survival rate of about 50%. In
fact, when the tumor is detected at the earliest stage (IA), there is an improved 5-
year survival of greater than 60% (Mountain et al 1997). Based on these findings,
tumor stage at the time of diagnosis determines overall prognosis of lung cancer
patients. Early detection is the key to improving patient survival. Unfortunately,
most lung cancer cases are diagnosed when curative resection is no longer
possible and are treated with a multimodality approach that may include
radiotherapy and chemotherapy.
Unlike colon, breast and prostate cancer, there are no widely accepted
screening tools for the early detection of lung cancer in high-risk groups. The
importance of screening is evident in prostate cancer. Recent reports suggest
early detection of prostate cancer has led to decreased mortality rate (Bouchardy
et al 2008). Below is a summary of current strategies (e.g. imaging and molecular
marker technologies) that have been tested or are currently being evaluated as
tools for early detection of lung cancer.

Chest radiography, sputum cytology, and low-dose spiral computed
tomography (LDSCT)
The method originally used to examine persons at-risk or suspected of
having lung cancer was chest radiography (x-ray). To complement chest x-ray,

4
sputum cytology has routinely been performed. Cytological evaluation of sputum
of lung cancer patients has been of interest since sputum harbors tumor cells
that can be observed under the microscope. Non-randomized screening studies
have proved chest x-rays with or without sputum cytology to be ineffective in
reducing lung cancer mortality (reviewed in Guessous et al 2007). Although the
performance of the chest x-ray is unacceptable as an early detection tool for lung
cancer, it may have potential in identifying at-risk groups. Abnormalities such as
scarring and fibrosis detected through chest x-rays have been reported to be
associated with increased risk of developing lung cancer (Pinsky et al 2006).
More recently, several studies have been launched to determine the
efficacy of high-resolution imaging modalities such as low dose spiral computed
tomography (LDSCT). LDSCT has been increasing in popularity as the choice for
screening for early cancer. One study that supports the use of LDSCT over chest
x-ray as an early detection tool is the Early Lung Cancer Action Project (ELCAP).
The initial findings of the study showed that 27 lung cancer cases out of 1,000
study subjects were diagnosed by LDSCT compared to only 7 by chest x-ray;
and of the 27 cancers detected, 85% were stage I disease (Henschke et al
1999). Though these preliminary analyses suggest that LDSCT screening
appears superior in detecting early lung cancer, the study also demonstrated
screening by LDSCT yields many false-positive results; 23.3% of participants
exhibited non-calcified pulmonary nodules (Henschke et al 1999). In a 2006

5
update of the ELCAP study, Henschke and colleagues confirmed their initial
findings: after screening more than 1,000 asymptomatic subjects at risk for lung
cancer, LDSCT identified 27 early cancers, and 85% of these were stage I
cancer. Moreover, the 10-year survival rate of these stage I patients was 88%
(Henschke et al 2006). These reports hint at the potential of LDSCT as a
detection method for early lung cancer.
However, since this most recent publication, many have questioned
whether the ELCAP study is flawed. A recent review by Welch et al (2007)
highlighted the weakness of the ELCAP study. In this article, the reasons why the
ELCAP study is flawed were listed: (1) absence of a control group, (2) lack of an
unbiased outcome measure, (3) lack of consideration of previous reports, and
(4) the harms of screening are not addressed (Welch et al 2007). Because
LDSCT gives false positive results, overdiagnosis is a concern because it could
lead to unnecessary and potentially dangerous interventions (e.g. surgery). The
Pittsburgh Lung Screening Study (PLuSS) is another study that appears to
promise early lung cancer detection by LDSCT. Wilson et al (2008) reported that
out of 69 NSCLC cases detected by LDSCT, 40 were stage I cancers at the time
of diagnosis. However, because more stage I cancers are detected than
expected, these could be false positive diagnoses. Also, in the same study, 36
persons with a non-calcified lung nodule had major thoracic surgical procedure
that led to a non-cancer final diagnosis (Wilson et al 2008). Although LDSCT

6
shows potential as a sensitive early detection tool, the lack of specificity of the
technology could result in substantial cost and morbidity. One way in which the
lack of specificity of LDSCT might be addressed is by complementing this
approach with molecular markers.

Molecular markers: protein and DNA
A number of molecular approaches have been devoted to the analysis of
lung cancer, including the profiling of gene expression, protein markers and
genetic and epigenetic changes. Molecular profiling of tumors can not only
provide new insights into the underlying biology of lung cancer, but can also yield
candidate molecular markers for early cancer detection. Such markers might also
be used for classification of tumors as well as assessment of disease
progression, regression and recurrence (Hanash et al 2008).
Progress has been made in the development of protein-based molecular
markers for lung cancer. Studies of gene expression changes in lung cancer
have led to the identification of a number of proteins that are highly expressed in
lung cancer and show promise as early detection tools. Overexpressed proteins
are not only detected in tumor tissue, but are also found in the bodily fluids (e.g.
serum) of lung cancer patients, suggesting they might have value as a non-
invasive detection method. A recent review by Greenberg and Lee (2007)
highlighted some of the serum markers that are currently being evaluated in

7
practice. Carcinoembryonic antigen (CEA), an oncofetal protein, was shown to
be elevated in lung adenocarcinoma tissue (Nonomura et al 1994). The abnormal
expression could be used a marker for detecting adenocarcinoma. CEA has also
been evaluated as prognostic indicator in NSCLC (Fukai et al 2007). In addition
to CEA, other serum protein biomarkers being examined include: Cytokeratin 19
fragment (CYFRA 21-1), Neuron-specific enolase (NSE), Squamous cell
carcinoma antigen, Carbohydrate antigen 125, and Carbohydrate antigen 19-9.
Performance analysis of individual protein markers showed low specificity and/or
sensitivity in detecting lung cancer (Schneider et al 2000). A follow-up study by
Schneider et al (2002) reported a better sensitivity and specificity (92% and 95%,
respectively) when CYFRA 21-1, NSE and C-reactive protein were used in
combination. However, higher levels of protein markers have also been observed
in the serum of cancer-free smokers (Molina et al 2003, Okada et al 2004), thus
somewhat devaluing the potential of these markers as a non-invasive tool for the
early detection of lung cancer. More recently, Han et al (2008) used surface-
enhanced laser desorption/ionization time of flight mass spectrometry (SELDI-
TOF-MS) to profile sera of lung cancer and healthy subjects. The study suggests
that by using pattern recognition, an increased number of lung cancer cases can
be spotted (sensitivity: 91% specificity: 97%). This novel technology has great
potential to perform as an early detection tool for lung cancer, although more

8
comprehensive studies are necessary to evaluate its usefulness in a clinical
setting.
While protein-based assays are valuable, DNA-based markers offer the
distinct advantage that they can be amplified by the polymerase chain reaction
(PCR). In the past decades, there have been extensive efforts to identify genetic
and epigenetic changes that occur during lung carcinogenesis. Such changes not
only provide insight into cancer development and progression, but also provide
potential biomarkers for early detection. DNA methylation, the addition of a
methyl group on a cytosine in a CpG dinucleotide context, has been associated
with cancer development (Vucic et al 2008, Feinberg 2007, Herman and Baylin
2003). Global loss of DNA methylation (hypomethylation) is a hallmark of almost
all human cancers (Herman and Baylin 2003). Simultaneously, hypermethylation
of CpG islands (clusters of CpG dinucleotides near the transcription start sites) is
observed for many genes involved in a diverse range of functions and pathways.
In particular, the promoter regions of tumor-suppressor genes are
hypermethylated leading to their inactivation thus resulting in changes in cell
regulatory events (e.g. cell cycle progression and apoptosis) (Herman and Baylin
2003). The difference in DNA methylation patterns (DNA methylation profiles)
between tumor and non-tumor tissues can be utilized as molecular markers to
distinguish cancer from non-cancer subjects. DNA methylation profiling can also

9
be used in classifying subtypes of cancers and potentially as predictors of
disease outcome and treatment response.
The usefulness of DNA methylation as a non-invasive molecular marker
has been illustrated in studies of biological fluids of cancer patients. Molecular
analysis of nipple aspirate fluid from breast cancer patients showed that the DNA
methylation profile of aspirate fluid DNA generally corresponded with DNA
methylation signatures of the tumors. Furthermore, benign and normal breast
tissue as well as nipple aspirate DNA from healthy women did not bear any DNA
methylation marks (Krassenstein et al 2004). The high specificity suggests that
nipple aspirate fluid could be used as a non-invasive tool to detect breast cancer.
For head and neck cancer, preliminary DNA methylation analyses of salivary
rinses show potential as a non-invasive method for detection as well monitoring
disease recurrence. (Carvalho et al 2008, Franzmann et al 2007, Righini et al
2007, Rosas et al 2001). Perhaps the most ideal source of molecular information
for most cancers is the blood. DNA hypermethylation of several genes had been
observed in circulating free plasma DNA of prostate (Altimari et al 2008, Chuang
et al 2007, Papadopoulou et al 2006, Papadopoulou et al 2004), breast
(Papadopoulou et al 2006, Hoque et al 2006, Skvortsova et al 2006) and
colorectal cancer patients (Lofton-Day et al 2008).

10
DNA methylation and non-small cell lung cancer
Because different types of cancer (even subgroups within a cancer)
exhibit distinct DNA methylation signatures, it is possible to develop cancer-
(sub)type specific methylation profiles. In this section, we will focus DNA
methylation findings on non-small cell lung cancer (mainly adenocarcinoma).
Several research groups using a variety of DNA methylation assays have
identified numerous hypermethylated genes in lung cancer (reviewed in Tsou et
al 2002, Digel and Lubbert 2005, Kerr et al 2007). The list contains genes that
function as tumor suppressors (e.g. CKDN2A, RASSF1, etc), in cell cycle control
(e.g. CDKN2B), DNA repair (e.g. MGMT) and apoptosis (e.g. DAPK). With the
advent of genome-wide DNA methylation analysis of lung cancer, the list of
genes is continuing to grow. Here, we will summarize the data for two
hypermethylated genes that have potential as risk and early detection markers in
human lung adenocarcinoma.
A critical tumor suppressor gene that has been extensively studied in lung
adenocarcinoma is cyclin-dependent kinase inhibitor 2A (melanoma, p16, inhibits
CDK4) or CDKN2A. Hypermethylation of CDKN2A was first described in lung
cancer cell lines more than a decade ago. The distinct methylation of the 5’ CpG
island was associated with gene silencing and was demonstrated to be reversible
upon treatment with 5-aza-deoxycytidine (Merlo et al 1995, Otterson et al 2005).

11
Since then, a multitude of studies have been published on DNA methylation
status of the CDKN2A gene in resected NSCLC tissues (reviewed in Wikman
and Kettunen 2006). In our own analysis, we did not find differential methylation
between tumor and AdjNTL samples at the promoter CpG island of CDKN2A,
probably because the control tissue already had considerable methylation. We
did however find that the exon 2 region of CDKN2A was informative. In the
analysis of 28 tumor-associated loci in 52 AD and 38 AdjNTL samples, we found
significant differential methylation in the exon 2 region of CDKN2A between AD
and non-malignant tissues. Because frequent methylation was also observed in
the non-tumor lung samples, CDKN2A (exon 2) may be suboptimal for early
adenocarcinoma detection from biopsy or fluid samples (Tsou et al 2008, see
Chapter 2). However, several studies from other laboratories demonstrated that
5’ CpG island methylation of the CDKN2A promoter CpG island could be useful
as a marker for early detection, and risk of progression for NSCLC. Although
these reports illustrate the value of CDKN2A hypermethylation as a non-invasive
method of detecting lung cancer, this gene on its own, thus far, has substandard
clinical specificity and sensitivity.
Another well-studied TSG regulated by DNA methylation is Ras
association (RalGDS/AF-6) domain family member (RASSF1). Like CDKN2A,
RASSF1 is frequently methylated in NSCLC (Feng et al 2008, Liu et al 2007,

12
Grote et al 2006, Marsit et al 2006). RASSF1 has been reported to be associated
with AD histology and early stage cancer (stage I) (Liu et al 2007). Lung cancer
patients with tumors methylated in the RASSF1 gene had concordant
methylation in the plasma (Hsu et al 2007), suggesting its potential as a blood-
based biomarker. Patients with positive RASSF1 methylation had significantly
shorter survival than those with negative status (Yanagawa et al 2007).
Identification of DNA methylation markers in lung cancer is leading to a
better understanding of the molecular pathways involved in lung carcinogenesis.
Because an estimated 12,000 genes carry CpG islands, analysis of
hypermethylation in cancer has the potential to generate highly complex and
therefore informative patterns. In order to establish such profiles for human lung
adenocarcinoma, many markers still need to be evaluated. In the next chapters,
we will describe our findings on the usefulness of DNA methylation biomarkers in
distinguishing tumor from non-tumor samples.

13
Chapter 2: Identification of a panel of sensitive and
specific DNA methylation markers for lung
adenocarcinoma

Chapter 2 Abstract
Lung cancer is the number one cancer killer of both men and women in
the United States. Three quarters of lung cancer patients are diagnosed when
their disease has spread regionally or distantly; their 5-year survival is only 15%.
DNA hypermethylation at promoter CpG islands shows great promise as a
cancer-specific marker that might be used to complement visual lung cancer
screening tools such as spiral CT, improving early detection. In lung cancer
patients, such hypermethylation is detectable in a variety of samples ranging
from tumor material to blood and sputum. However, to date the penetrance of
methylation at any single locus has been too low to provide great clinical
sensitivity. Here we used the real-time PCR-based method MethyLight to
examine DNA methylation quantitatively at twenty-eight loci in 51 primary human
lung adenocarcinomas, 38 adjacent non-tumor lung samples, and 11 lung

14
samples from non-lung cancer patients. We identified thirteen loci that show
significant differential methylation levels between tumor and non-tumor lung;
eight of these show highly significant hypermethylation in adenocarcinoma:
CDH13, CDKN2A EX2, CDX2, HOXA1, OPCML, RASSF1, SFPR1, and TWIST1
(p-value of <<0.0001). Methylation of these loci did not differ significantly based
on gender, race, age or tumor stage, indicating their wide applicability as
potential lung adenocarcinoma markers. Using the current tissue collection and
5-fold cross validation, the four most significant loci (CDKN2A EX2, CDX2,
HOXA1 and OPCML) individually distinguish lung adenocarcinoma from non-
cancer lung with a sensitivity of 67-86% and specificity of 74-82%. We applied
random forests to determine a good classifier based on a subset of our loci and
determined that combined use of the same four top markers allows identification
of lung cancer tissue from non-lung cancer tissue with a sensitivity of 94% and a
specificity of 90%. The identification of eight CpG island loci highly specifically
hypermethylated in lung adenocarcinoma provides strong candidates for
evaluation in patient remote media such as plasma and sputum. Of the eight
identified loci, CDKN2A EX2, CDX2, HOXA1 and OPCML merit further
investigation as some of the most promising lung adenocarcinoma markers
identified to date.

15
Introduction
Lung cancer is expected to cause over 160,000 deaths in 2008 –killing
more Americans than cancer of the prostate, breast, colon, rectum and pancreas
combined (American Cancer Society 2008). Lung cancer is clinically classified
into two classes: the aggressive subtype small cell lung cancer (SCLC, ~13% of
cases) and non-small cell lung cancer (NSCLC, the remaining ~87%) (American
Cancer Society 2008). NSCLC is histologically subdivided into four major
subtypes with distinct pathological and molecular characteristics:
adenocarcinoma, squamous cell lung cancer, large cell lung cancer and “other”
(comprising neuroendocrine cancers, carcinoids etc) (Travis et al 1996). Of
these, adenocarcinoma has recently surpassed squamous cell lung cancer as
the most common subtype in the United States, accounting for approximately
40% of NSCLC (Travis et al 1995). The incidence of lung adenocarcinoma is on
the rise in many countries, in particular in women (Chen et al 2005, Yoshimi et al
2003). Adenocarcinoma is also the most common lung cancer subtype in non-
and previous smokers (Liu et al 2000).
The 5-year survival of lung cancer patients is only 15%, largely due to the
fact that three quarters of lung cancer patients are diagnosed when their disease
has spread regionally or distantly (Ries et al 2006). To make an impact on long
term survival better strategies are needed for early detection. Prior experience

16
with chest X-ray, sputum cytology, and fiberoptic examination have failed to
decrease lung cancer patient mortality, although several recent strategies show
promise. Low-dose spiral computed tomography (LDSCT) is one such approach.
It allows detailed imaging of the lung, and can detect very small lesions. Recent
results from the Early Lung Cancer Action Project (ELCAP) indicate that this
approach allows detection of early stage lung cancer (Henschke et al 2006), but
in this and other studies, non-cancerous lesions far outnumber malignancies
(less than 10% of lesions are cancer). In addition, it is unclear whether the early
stage lung cancers identified by LDSCT represent cancers that would ultimately
progress and lead to death. A recent analysis suggests spiral CT screening may
not reduce lung cancer mortality (Bach 2007). Molecular analyses of plasma,
sputum, and bronchioalveolar lavage fluids have also shown promise as
strategies for early detection, but these methods still lack sensitivity (Belinsky
2004). If molecular markers with high sensitivity and specificity for cancers that
will progress can be identified, such markers could be combined with spiral CT to
screen high-risk individuals, allowing molecular detection and visualization of
clinically relevant early lesions. This would greatly increase the chances of
curative resection of lung cancer, while minimizing unnecessary and potentially
life-threatening procedures in patients with benign lesions.
Of the many potential molecular markers, DNA hypermethylation -- an
epigenetic alteration-- shows great promise. DNA hypermethylation occurs in all

17
cancers, frequently leading to gene silencing through methylation of CpG-rich
regions (CpG islands) near the transcriptional start sites of genes (Laird 2003). In
lung cancer patients, such hypermethylation is quantitatively detectable in a
variety of samples ranging from tumor material to blood and sputum (Belinsky
2004). However, to date the penetrance of methylation at any single locus has
not been high enough to provide great clinical sensitivity. Our focus is to increase
the repertoire of sensitive DNA hypermethylation markers for lung cancer, and to
compose a small panel of molecular markers that could be used to detect lung
cancer with high sensitivity and specificity. Given the histopathologic, clinical and
molecular differences between lung cancer subtypes, we believe that markers
should be developed individually for the major histological subtypes. These
markers can later be combined into a lung cancer hypermethylation panel that
can be used for detection of all lung cancers.
Because of its increasing frequency and its preponderance in non- and
previous smokers, we focused first on lung adenocarcinoma (AD). Here we
describe our evaluation of 28 potential DNA methylation markers using primary
human lung adenocarcinoma samples. To ensure that these markers detect
cancer-specific hypermethylation changes, associated with histologically visible
lung cancer (allowing surgical resection), we compared the methylation profiles
of the tumors with histologically normal adjacent lung tissue (AdjNTL) from lung

18
cancer patients. We also examined non-tumor lung from non-cancer patients
(NTL).

Methods
Study subjects
Lung adenocarcinoma and when available adjacent non-tumor lung was
obtained from archival paraffin blocks from 51 subjects who had been treated at
three Los Angeles hospitals: the Los Angeles County Hospital, the USC
University Hospital and the Norris Comprehensive Cancer Center. Clinical
information was missing for 5 patients. Of the rest, 28 were male and 18 were
female, 14 were White of non-Hispanic descent, 14 White of Hispanic descent,
11 were Black, and 7 were of various Asian origins. Ages ranged from 37-82
years old at time of surgery (median: 58 years old). For 32 of these cases, a
separate paraffin block containing histologically verified cancer-free lung was
available. These adjacent non-tumor lung samples were supplemented with 6
additional cancer-free archival samples from lung cancer patients for which the
tumor block was unavailable, and 11 archival non-tumor lung samples from
patients operated for non-cancer reasons, such as pneumothorax or

19
emphysema. All studies were institutionally approved, and the identities of
patients were not made available to laboratory investigators.

Tissue samples and DNA extraction
Hematoxylin and eosin-stained slides were reviewed by an experienced
lung pathologist to support the original classification of the tumor and to select
optimal tumor and non-tumor areas of the specimens. DNA was extracted from
microdissected tumor and non-tumor lung samples via proteinase K digestion
(Laird et al 1999). Briefly, cells were lysed in a solution containing 100 mmol/L
Tris-HCl (pH 8.0), 10 mmol/L EDTA (pH 8.0) 1 mg/mL proteinase K, and 0.05
mg/mL tRNA and incubated at 50
o
C overnight. The DNA was bisulfite converted
as previously described (Eads et al 2000).

Methylation analysis
DNA methylation analysis was done by MethyLight as previously
described (Weisenberger et al 2006). Primer and probe sequences were as
described (Weisenberger et al 2006, Tsou et al 2007, Virmani et al 2002). In
addition to primers and probe sets designed specifically for the gene of interest,
an internal reference primer and probe set designed to analyze Alu repeats (Alu)
was included in the analysis to normalize for input DNA (Weisenberger et al

20
2005). The percentage methylated reference (PMR) was calculated as the
GENE:reference ratio of a sample divided by the GENE:reference ratio of in vitro
methylated (SssI-treated) human white blood cell DNA and multiplying by 100.
Occasionally, PMR values over 100 were observed. This can happen when
genes are very heavily methylated in the cancer sample, while the SssI-treated
sample (in spite of extensive in vitro methylation) is not fully methylated at that
locus. This does not affect the significance of the loci identified in this study, as
the same batch of SssI-treated DNA was used throughout the study.

Statistical analysis
PMR values of AD were compared to AdjNTL and NTL lung as continuous
variables by means of the Wilcoxon rank sum test. For the comparison of paired
AD and AdjNTL samples from the same patients, the Wilcoxon signed rank test
was used. To control the false discovery rate at 5%, a multiple comparisons
threshold was set. It was only applied to those 20 loci for which no previous
information supporting a hypothesis of methylation in lung AD was available
(Benjamini et al 2001). Receiver operating characteristic (ROC) curves were
plotted using the AD vs. all AdjNTL lung PMR values and JMP 6.0 software (SAS
Institute, Cary, NC). The distribution of PMR values by group (AD, AdjNTL and
NTL) were shown using log-transformed data in JMP 6.0. The two-dimensional

21
hierarchical clustering was carried out using JMP 6.0 and log-transformed PMR
values. VHL, which was negative in all specimens, was omitted from the
clustering analysis. Associations between age, gender and race of AD cases
were tested by dichotomizing the subjects either by the presence/absence of
methylation, or, if the samples were frequently methylated, by the median of all
positive PMR values. All statistical tests were two-sided.
To determine which combinations of markers would be most effective to
correctly identify tumor vs. non-tumor samples, we fit a random forest classifier to
the data set, using the R programming language (v 2.5 (Ihaka et al 1996)) and 87
samples and 28 variables (2 AD samples with missing PMR data were omitted,
resulting in 49 AD/38 AdjNTL). Using bootstrap samples of the data, we grew a
forest of 30,000 trees. Splits were determined using a random sample of five
variables and trees were grown until there was only one observation in each leaf.
We determined error rates using the observations that were not used to generate
the trees. For each observation, its outcome was predicted by having the majority
vote from the trees that were generated without the original data point in their
bootstrap sample. These predicted values were compared against the true tissue
type to estimate prediction error.

22
Results/Discussion
Ideal DNA hypermethylation markers for lung adenocarcinoma should
show a high frequency of methylation in tumors as well as methylation levels that
are significantly elevated in tumor compared to non-tumor lung tissue.
Environmental exposures, such as those arising from tobacco smoke, could lead
to higher basal levels of methylation in non-tumor lung (Divine et al 2005), which
might affect the background signal when any resulting markers are applied to
non-invasive molecular analyses of bodily fluids in the future. To ensure the
identification of markers that are more highly methylated in adenocarcinoma
even when compared to heavily exposed but histologically cancer-free lung, we
used adjacent non-tumor lung from lung cancer cases as our cancer-free
comparison. We also examined a number of non-tumor lung (NTL) samples from
non-cancer patients. Quantitative assessment of methylation levels allows a
more detailed evaluation of candidate methylation markers, and of their suitability
for correctly identifying a cancer vs. non-cancer sample. For this reason, we used
the bisulfite conversion based real-time PCR technique, MethyLight, to measure
methylation in tumor and control tissues (Weisenberger et al 2006).
Twenty-eight loci were chosen for evaluation (Table 2.1). The choice of
loci was based on a prescreening of 114 loci carried out on a collection of human
lung cancer cell lines, including 10 adenocarcinoma cell lines (unpublished data,

23
data not shown). We also included many loci that appeared promising based on
previous reports describing their methylation in lung cancer or other cancers, so
that all markers of interest could be compared on one set of tissues using a
single technique and platform. Among others, the 28 loci included CpG islands in
the promoters of TSGs and genes with important roles in cell cycle regulation,
DNA repair, and apoptosis (Table 2.1).
The results of the methylation analyses for the 28 loci in 51 AD, 38
AdjNTL and 11 NTL samples are shown in Figure 2.1. Methylation, expressed as
the percentage methylated reference (PMR (Weisenberger et al 2006)) is
visualized by color coding. Comparison of AD in Figure 2.1 panel A with AdjNTL
in panel B shows that a number of loci are more heavily methylated in AD. The
effect appears to be even more pronounced when AD is compared to NTL from
non-cancer patients. Although the exposure history of these NTL samples is
unknown, their generally lower methylation levels emphasize that these samples
may not be the best controls when searching for loci that show cancer-specific
hypermethylation. Interestingly, for one locus, LZTS1 (leftmost locus in Figure
2.1), the methylation pattern appeared to be reversed; NTL showed the highest
level of methylation, while AD samples were least methylated.
We applied two-dimensional hierarchical clustering to examine the
relationship between the loci and the tumor and non-tumor lung samples (Figure

24

25

Figure 2.1. Graphic representation of PMR values obtained for 28 loci in AD (A),
AdjNTL (B) and NTL (C). Samples are indicated at the left, loci at the top. PMR
values have been categorized as colored boxes denoting no detectable methylation
(blue), methylation below the median of all positive samples of each locus (yellow),
and methylation equal to or above the median (red). The black bar at bottom indicates
loci showing statistically significant differences in methylation levels between tumor
and non-tumor lung.

A
B
C
LZTS1
OPCML
CDX2
HOXA1
CDKN2AEX2
SFRP1
CDH13
TWIST1
RASSF1
SFRP4
SFRP5
ESR1
CDH1
SLC6A20
PGR
MT1A
MT2A
ATM
PTEN
SYK
CDKN2B
CYP1B1
CHFR
APC
SOCS4
MGMT
HMGA1
VHL
ATT32135
ATT05325
ATT47048
ATT90553
ATT57548
ATT28709
ATT83491
ATT73796
ATT45753
ATT64778
ATT35808
ATT34282
ATT20344
ATT88435
ATT98520
ATT14905
ATC58047
ATT22109
ATT40558
ATT60970
ATT93433
ATT50500
ATT27732
ATT50725
ATT68248
ATC65481
ATT29405
ATT67851
ATC50950
ATT83452
ATT99634
ATT06288
ATT98083
ATT13746
ATT70078
ATT18475
ATT40610
ATT68711
ATT77817
ATT88685
ATT40200
ATT86507
ATT58401
ATT36766
ATT67951
ATT90364
ATT29609
ATT11062
ATT99594
ATT67348
ATT87517
XNT32135
XNT05325
XNT47048
XNT57548
XNT83491
XNT73796
XNT45753
XNT64778
XNT35273
XNT14905
XNT68607
XNT22109
XNT40558
XNT93433
XNT50500
XNT73998
XNT05431
XNT27732
XNT24201
XNT67851
XNT83452
XNT99634
XNT06288
XNT98083
XNT13746
XNT70078
XNT18475
XNT40610
XNT77817
XNT88685
XNT40200
XNT86507
XNT67951
XNT90364
XNT76493
XNT29609
XNT11062
XNT87517
XNH21115
XNH73039
XNH57186
XNH40218
XNH16544
XNH80124
XNH35635
XNH17727
XNH08015
XNH45318
XNH22374

26
2.2; VHL was omitted because it showed no methylation in any samples). All but
one of the tumor samples clustered together in a major branch of the dendogram,
while the majority of non-tumor lung samples grouped in a separate cluster. Nine
loci, CDH13, SFRP1, OPCML, TWIST1, SFRP5, CDKN2A EX2, CDX2, HOXA1
and RASSF1, clustered together (bottom right), showing heavier methylation in
the tumor samples.
We next analyzed the statistical significance of the differences in
methylation levels for individual markers and different combinations of tissue
samples (Table 2.2): AD vs. all NTL samples, AD vs. AdjNTL, and AD vs. paired
AdjNTL (32 of the 38 AdjNTL samples were derived from the AD patients in
Figure 2.1A). The paired AdjNTL form an exquisite control for the cancer-specific
nature of the observed methylation changes, as each of these samples conforms
to its tumor sample in patient age, environmental exposure, and genetic
background. To avoid assigning statistical significance to spurious associations,
we incorporated a multiple comparisons threshold for those loci that at time of
analysis lacked any prior data suggesting they might be hypermethylated in lung
adenocarcinoma (Table 2.2, before-last column, (Benjamini et al 1995) see
Materials and Methods for details). Thirteen of the analyzed loci showed
statistically significant differences in methylation when AD samples were
compared to all NTL samples: OPCML, CDX2, HOXA1, CDKN2A EX2, SFRP1,

27

28
Figure 2.2. Two dimensional hierarchical clustering of samples and loci based on
methylation data used as a continuous variable. In the center, methylation levels
are indicated by a color gradient, with the highest methylation levels for each locus
indicated in red and the lowest in deep blue. The Ward hierarchical clustering
method was used to categorize between cancer and non-tumor samples. Sample
IDs are indicated on the left, with AD samples in red, AdjNTL samples in black, and
NTL samples in blue. The relationship of samples is indicated at right in the same
color schematic as the labels. At bottom, the relationship of the loci is indicated.
Note that all eight of the most significant loci cluster at bottom right.

29
CDH13, TWIST1, LZTS1, RASSF1, SFRP4 and 5, ESR1, and CDH1. All of these
except CDH1 remained significant when AD samples were compared to AdjNTL,
while all except LZTS1 remained significant in the comparison of AD to paired
AdjNTL. Because methylation of LZTS1 is reduced in tumors it is not a candidate
for a positive lung adenocarcinoma marker and it was not studied further at this
time. APC methylation was found to be statistically significantly different only
when paired tumor and non-tumor lung samples were compared. This suggests
that basal methylation is high but variable at this locus; elevated methylation in
tumors is likely masked by interpatient variability and only becomes visible when
samples from the same patient are compared. Indeed, Waki and coworkers have
observed frequent methylation of APC in non-cancer lung and other organs
(Waki et al 2003).
Of the thirteen significant loci, OPCML, CDX2, HOXA1, CDKN2A EX2,
SFRP1, CDH13 and TWIST1 show considerable promise as cancer-specific
methylation markers, exhibiting highly significant hypermethylation in tumors
compared to paired non-tumor tissues (p<1x10
-7
, Table 2.2). All eight of these
loci grouped together in the hierarchical clustering (Figure 2.2). The ability of the
top four candidates, CDKN2A EX2, CDX2, HOXA1 and OPCML (all p<1x10
-9
), to
individually identify lung cancer samples was next evaluated. Figure 2.3 shows
the distribution of PMR values in the examined sample collection. Note that for all

30
four markers, the mean value in non-tumor lung from non-cancer patients is
lower than that of adjacent non-tumor lung from lung cancer patients. This
emphasizes the importance of using histologically normal tissue adjacent to lung
cancer for comparison; such tissue may show higher basal methylation levels
while appearing histologically normal, and should be used for comparison with
lung cancer tissue to ensure identification of cancer-specific markers. While all
four markers show increased methylation in adenocarcinoma compared to
adjacent non-tumor tissue, the spread of methylation levels differs, which would
affect their sensitivity and specificity in future detection strategies. The marker
potential of quantitative markers is frequently presented in the form of a receiver
operating characteristic (ROC) curve, in which sensitivity vs. 1-specificity at all
possible cut-off values is plotted. While these DNA methylation markers are
ultimately intended for the non-invasive analysis of patient bodily fluids, a
preliminary indication of their potential to sensitively and specifically detect
cancer could be obtained by plotting ROC curves using the PMR values from the
tumor vs. adjacent non-tumor samples. Figure 2.4 shows that the area under the
curve (AUC, and indicator of marker performance that would be 1 for a marker
showing 100% specificity and sensitivity) is 0.87-0.95 for the four top loci.
Despite the promising AUC values, the sensitivity and specificity of these
top four markers, used individually and determined using the current sample

31
Figure 2.3. The distribution of PMR values by group. Log-transformed PMR values for
AD, AdjNTL and NTL are shown. The mean is shown by the wide horizontal line, and the
top and bottom of the diamond indicate the 95% confidence intervals.

32
Figure 2.4. Receiver operating characteristic curves for the four top markers. All AD and
AdjNTL lung samples for which there was complete methylation data were used for the
analysis.

33
collection in a five-fold cross-validation, was limited: 67-86% and 74-82%
respectively. This supports the notion that DNA hypermethylation markers are
best used in the form of a panel. Because of the costs associated with
quantitative molecular analyses, it would be important to limit the number of
markers included in the panel. To determine which combinations of markers
would be most effective to correctly identify tumor vs. non-tumor samples, we fit
a random forest classifier to the data set, using 87 samples and 28 variables (2
AD samples with missing PMR data were omitted, resulting in 49 AD vs. 38
AdjNTL). Using bootstrap samples of the data, we grew a forest of 30,000 trees.
Splits were determined using a random sample of five variables and trees were
grown until there was only one observation in each leaf. Utilizing all 28 loci, we
estimated a sensitivity of 92% and a specificity of 95%. Using the Gini index from
the random forest classifier (last column, Table 2.2) to measure locus
importance, we restricted our analysis to the most highly ranked variables.
Reducing the locus number to 13 did not affect sensitivity and specificity, and
limiting our markers to the top-ranked four (HOXA1, OPCML, CDKN2AEX2 and
CDX2, which were also the most significant based on our statistical analysis)
resulted in a sensitivity of 94% and a specificity of 90%. Thus, these four markers
appear highly promising DNA hypermethylation markers for development into
non-invasive molecular markers of lung adenocarcinoma, through examination of
DNA shed into bodily fluids such as sputum, bronchioalveolar lavage, or blood.

34
For candidate hypermethylation markers of lung adenocarcinoma, two
important questions arise. First, are these markers hypermethylated in cancer
samples irrespective of the subject’s age, gender and racial/ethnic background?
And secondly, are these markers hypermethylated even in the earliest stages of
lung adenocarcinoma? While the population analyzed in the current study is
small, we reasoned that an indication of the potential of our top four markers to
broadly identify lung adenocarcinoma might be obtained. To address the first
question, we assessed correlations to age and determined whether each of the
four markers showed statistically significant hypermethylation in tumor vs.
adjacent normal tissues in men, women, and all four racial/ethnic groups. We
found no correlation of methylation of CDKN2A EX2, CDX2, HOXA1 and OPCML
with the age. In addition, all four markers remained significantly hypermethylated
in tumor vs. AdjNTL when subjects were stratified by gender or by ethnic group
(p<0.05) (Table 2.3). The only exception was CDKN2A EX2 methylation in Asian
subjects (p=0.11), which may be related to the small sample size but will need to
be further explored.
To address the second question, we determined whether each of the four
markers was significantly hypermethylated in early and later stage tumors, using
paired samples (Table 2.3). We examined stages IA and IB individually, but
grouped stages IIA, IIB, and IIIA (only one paired sample was available for IIA
and IIIA, and none for stage IIIB). Notably, all four markers were significantly

35

race/ethnicity and stage.
p-value*
T ADJ NTL
GENDER
Male (n=28) (n=19)
CDKN2A EX2 199.44 51.88 5.0E-06
CDX2 28.40 4.78 2.0E-05
HOXA1 139.51 5.06 4.6E-05
OPCML 76.62 11.27 1.6E-06
Female (n=14) (n=10)
CDKN2A EX2 189.16 38.64 1.3E-04
CDX2 114.71 3.15 4.0E-05
HOXA1 188.13 1.34 3.1E-06
OPCML 154.32 5.36 3.1E-06
RACE
White Hispanic (n=14) (n=10)
CDKN2A EX2 165.47 37.52 6.0E-04
CDX2 39.49 1.98 2.0E-04
HOXA1 52.98 1.75 7.7E-03
OPCML 142.56 8.17 6.0E-04
White Non-Hispanic (n=14) (n=10)
CDKN2A EX2 314.35 38.00 5.0E-04
CDX2 110.73 4.78 1.1E-03
HOXA1 231.66 4.90 4.6E-05
OPCML 179.79 6.54 4.7E-05
Black (n=11) (n=7)
CDKN2A EX2 194.91 51.88 1.5E-02
CDX2 129.96 9.11 2.4E-02
HOXA1 128.52 6.16 1.1E-02
OPCML 91.15 14.29 5.0E-03
Asian (n=6) (n=4)
CDKN2A EX2 145.56 39.81 1.1E-01
CDX2 24.28 1.78 2.3E-02
HOXA1 117.49 0.99 1.4E-02
OPCML 88.81 4.67 1.4E-02
STAGE
Stage IA (n=12) (n=12)**
CDKN2A EX2 182.57 47.28 4.9E-04
CDX2 99.71 3.15 4.9E-04
HOXA1 143.06 3.25 1.0E-03
OPCML 189.43 7.65 4.9E-04
Stage IB (n=6) (n=6)**
CDKN2A EX2 190.67 61.00 3.1E-02
CDX2 20.62 13.36 3.1E-01
HOXA1 78.98 2.15 6.3E-02
OPCML 76.62 15.16 6.3E-02
Stage IIA/IIB/IIIA*** (n=10) (n=10)**
CDKN2A EX2 208.05 39.81 1.0E-02
CDX2 83.22 1.78 2.0E-03
HOXA1 180.18 6.16 2.0E-03
OPCML 178.54 8.25 2.0E-03
* Mann-Whitney for gender and race; Wilcoxon signed rank test
stage, in which paired adjacent non-tumor samples were used.
** For stage comparisons, paired adjacent samples from same
patients as tumors were used.
*** Only one paired sample was available for stages IIA and IIIA and
none for IIIB, hence IIA/IIB/IIIA were pooled.
Median PMR
Table 3. Performance of top four markers in samples based on gender,
Table 2.3 Performance of top four markers in samples based
on gender , race/ ethnicitiy and stage.

36
hypermethylated in stage IA cancers. CDX2 hypermethylation was significant in
stage IB tumors, but this could be due to the small number of paired samples
(n=6). All markers were also significantly hypermethylated in later stage lung
adenocarcinoma (Stages IIA-IIIA). These analyses indicate that the top four
markers show high potential for identification of lung adenocarcinoma, even in its
earliest stages, an important characteristic if these markers are to be used for
early detection.
If the four top markers were to be used extensively for patient analysis in
the future, it would be useful to determine whether there is any relationship
between their methylation level and survival. We found no significant association
between methylation and survival for the four loci, or any of the other loci studied
(data not shown).
Based on the results of our analyses, four loci that are very strong
candidates for a DNA methylation panel aimed at early lung adenocarcinoma
detection have been identified: CDKN2A EX2, CDX2, HOXA1 and OPCML.
CDNK2A, also referred to as p16
INK4a
, encodes an important cell cycle regulator
that is frequently inactivated in cancer. CDKN2A is one of the first tumor
suppressor genes found to be methylated in a variety of cancers, including lung
cancer (Merlo et al 1995). It is one of the most widely studied hypermethylated
loci, and methylation of its promoter CpG island appears to be a very early event

37
in the development of non-small cell lung cancer (reviewed in Belinsky et al
2004). In fact, methylation of the CDKN2A promoter CpG island has been
observed in the sputum of subjects at risk for lung cancer 3 years prior to
diagnosis (Palmisano et al 2000) and in the sputum of asymptomatic heavy
smokers (Destro et al 2004). A recent analysis of prospectively collected sputum
showed CDKN2A methylation in 39% of cases and 25% of controls; methylation
of this gene was associated with an elevated risk of lung cancer (Belinsky et al
2006). It is thought that methylation observed in the sputum is indicative of field
cancerization of the airways and not necessarily a symptom of a present cancer
(Belinsky et al 2006). Our goal was to identify cancer-specific markers, not risk
markers. We had evaluated methylation of the CDKN2A promoter CpG island as
a cancer indicator, but found substantial methylation in AdjNTL, and no
significant difference between AdjNTL and cancer (data not shown). Based on
cancer-specific hypermethylation of the CDKN2A exon 2 CpG island observed in
colorectal and bladder cancers (Nguyen et al 2001, Salem et al 2000), we tested
this downstream island instead and established that its level of methylation is a
strong indicator of lung adenocarcinoma. While substantial methylation at the
exon 2 CpG island is detected in histologically normal AdjNTL, by comparison,
methylation in adenocarcinoma is highly significantly elevated (p< 1x10E-10).
The detection of CDKN2A methylation in a high fraction of lung cancer
patient plasma samples bodes well for its application to non-invasive detection

38
(Liu et al 2003). Two groups reported an association of methylation of CDKN2A
with poor survival in adenocarcinoma/NSCLC patients (Toyooka et al 2004,
Wang et al 2004), while Divine et al (2005), like us, reported no such association.
The differences between the obtained results might be due to the examination of
a different CpG island or a different population.
Methylation of HOX genes, encoding homeobox transcription factors
involved in embryogenesis and differentiation, had recently been observed in
lung adenocarcinoma and squamous cell lung cancer. In an analysis of eight
adenocarcinomas and matching adjacent lung, substantial methylation of the
HOXA and D clusters was observed (Shiraishi et al 2002). Five cancer samples
showed methylation of HOXA1, while only one AdjNTL sample was methylated at
this locus. In a different study, analysis of a stage I adenocarcinoma and
squamous cell lung carcinoma showed methylation of the HOX clusters, and
examination of the HOXA and D clusters in more detail in squamous cell cancers
and control tissue indicated a methylation frequency of 45-80% for HOXA7-9, but
methylation of HOXA1 was limited (Rauch et al 2007). Neither of these studies
examined a large number of adenocarcinomas, nor were quantitative techniques
used. Here we demonstrate that HOXA1 is a very promising methylation marker
for lung adenocarcinoma. We have also observed methylation of additional HOX
genes (unpublished studies), but HOXA1 appears to be particularly informative.

39
OPCML, encoding an opioid-binding cell adhesion molecule, has been
shown to be frequently methylated in ovarian cancer (Sellar et al 2003) Given
that opioids have demonstrated growth inhibitory and pro-apoptotic effects in
lung cancer cells (Maneckjee et al 1990, Maneckjee et al 1992, Maneckjee et al
1994), it is perhaps not surprising that the OPCML promoter CpG island might be
a target for methylation in lung cancer. Very recently, high throughput
methylation profiling of 11 lung adenocarcinomas and control lung identified a
number of CpG dinucleotides methylated in the cancers (Bibikova et al 2006).
One probe identified methylation in the area covered by the OPCML probe used
here. Although the OPCML locus was not studied in detail in the Bibikova study,
the observed methylation supports the idea that OPCML is a strong candidate
marker in lung adenocarcinoma.
CDX2, another homeobox transcription factor, had been described to be
methylated in squamous esophageal cancer (Guo et al 2007) and colorectal
carcinoma (Kawai et al 2005), but to our knowledge, its methylation in lung
cancer has never been examined. We find it to be methylated in 100% of lung
adenocarcinomas, showing a 10-fold higher median methylation than AdjNTL
tissue (Table 2.2).
Because our primary goal is marker development, here we focused only
on whether loci showed consistent hypermethylation. Whether or not this

40
hypermethylation results in gene inactivation is not relevant for the use of these
loci as DNA methylation markers, and was not determined. However, the
biological consequences of the observed hypermethylation would also be worth
of investigating. While each of the four top-ranked loci is of interest as a DNA
methylation marker, it is as a panel that they promise to be most powerful. To our
knowledge, we are the first to examine CDKN2A EX2, CDX2, HOXA1 and
OPCML in combination. The fact that this marker set allows identification of
cancer specimens in the current tissue collection with a substantially higher
sensitivity and specificity than any previously identified single markers underlines
the importance of developing suitable marker panels.

Conclusions
From a starting panel of 28 methylation loci, we have identified 13 that
show statistically significant methylation differences between lung
adenocarcinoma and non-cancer lung tissue. Of these, 8 showed highly
significant differences. The four most significant markers also ranked as the top
four to be used in a marker panel, as determined by a random forest approach.
Thus, we suggest that CDKN2A EX2, CDX2, HOXA1 and OPCML are the top
candidates from the 28 tested, and should be validated as DNA methylation

41
markers for lung adenocarcinoma. These validations should consist of examining
a sufficiently large number of new subjects representing both genders and all
four major ethnic/racial subgroups in the United States (Whites of non-Hispanic
and Hispanic descent, Blacks, and Asians), as well as early and late stage
cancer. Such studies are currently ongoing. Our analyses of the present sample
collection, which contains modest numbers of representatives from all these
groups, is very encouraging as they suggest that the markers function
independently of subject age, gender or ethnic subgroup, and are positive in
early stage cancer.
The next step would entail the exploration of different methods to measure
these markers non-invasively in early stage lung cancer patients. Potential
“remote” media to be considered are sputum, bronchioalveolar lavage, and blood
plasma, all of which we are in the process of collecting for examination. The
ability of our four-marker panel to clinically detect lung cancer with high sensitivity
and specificity will depend on many factors. A loss of sensitivity might be
foreseen due to the small amounts of DNA shed into the blood of each patient,
but at the same time, an increase in specificity might be expected if tumor DNA is
shed more readily into the bloodstream than DNA from adjacent histologically
normal tissue.
To our knowledge, CDKN2A EX2, CDX2, HOXA1 and OPCML constitute
the strongest lung adenocarcinoma DNA methylation markers identified to date,

42
and we are working on further evaluations of their potential with great
anticipation.

43
Chapter 3: Development and validation of a panel of 15
DNA methylation-based biomarkers for human lung
adenocarcinoma.

Chapter 3 Abstract
Lung cancer is the number one cancer killer in the United States, causing
more deaths than breast, prostate and colon cancer combined. Because there is
no effective screening tool for the early detection of lung cancer, most lung
cancer patients are diagnosed at an advanced stage. Development of novel
strategies for the detection of early cancer is essential in improving patient
survival. One major technological advancement in the scientific field is the
development of molecular markers (e.g. DNA methylation) as non-invasive tools
for cancer detection. Our goal is to develop a panel of DNA methylation markers
that has high sensitivity and specificity for adenocarcinoma (AD). In the previous
chapter, we identified 8 loci showing significantly higher methylation in AD than
adjacent non-tumor lung (AdjNTL) samples. Through primary tissue screening of
200 cancer-related loci through MethyLight, we identified 7 additional loci that
had significant increase in methylation in the AD samples. These markers were

44
combined with the 8 previously reported loci. The panel of 15 DNA methylation
markers showed a sensitivity of 92.85% and specificity of 100%. The validation of
the 15-panel marker by two independent sample collections showed the
sensitivity decreased to 70-78.26%, but specificity remained at 100%. The
superior specificity of the panel suggests its potential as a complement to highly
sensitive imaging technologies such as LDSCT.

Introduction
Lung cancer is the leading cause of cancer-related death in the United
States. The overall 5-year survival of patients with lung cancer is a dismal 15%
despite extensive effort in improvement of diagnosis and treatment (American
Cancer Society 2008). Early detection is the key to increasing lung cancer
patients’ chances for survival; however, effective screening tools for the early
detection of this lethal disease are still lacking. Although more sensitive
diagnostic modalities, such as LDSCT, show potential in detecting much smaller
lung lesions than the conventional chest x-ray, studies have shown that up to half
of the participants in the trials had at least one non-calcified lung nodule, the vast
majority of which were non-cancerous (Jett 2005). Such studies raise concerns
regarding overdiagnosis, which may lead to unnecessary follow-up medical

45
procedures with the associated cost and morbidity as well as the potential for
decreasing the patient’s quality of life. Moreover, the effectiveness of this new
technology in decreasing lung cancer mortality is still in question (Bach 2007,
Welch et al 2007). Therefore, there is an urgent need for the development of a
highly specific test that could potentially be used to complement LDSCT to aid in
the identification of early lung cancer.
In recent years, there has been mounting interest in developing molecular
markers for early lung cancer detection. One very promising marker is DNA
methylation. DNA methylation is the addition of a methyl group in the 5-carbon
position of a cytosine in the context of a CpG dinucleotide. During
carcinogenesis, promoter DNA hypermethylation of CpG islands has been
associated with tumor suppressor gene silencing (Laird 2003). Aberrant DNA
methylation of several cancer-related genes has been detected not only in tumor
material, but also in the blood and sputum of lung cancer patients (Belinsky
2004), showing promise of DNA methylation markers as a minimally-invasive
early detection tool.
Lung cancer is divided into two categories: small cell lung cancer (SCLC)
and non-small cell lung cancer (NSCLC). Small cell lung cancer is the most
aggressive subtype of lung cancer accounting for only 10-15% of all patients
diagnosed with the disease. Non-small cell lung cancer (approximately 85-90%
of lung cancer patients) is further subdivided into different histologies:

46
adenocarcinoma, squamous cell carcinoma, large cell carcinoma, and others
(Travis et al 1996). Our focus is on lung adenocarcinoma, the most frequent
histologic subtype of lung cancer in the United States (Travis et al 1995) as well
as in non- and previous smokers (Liu et al 2000). Our main goal is to identify
DNA hypermethylation markers to compose a small panel of biomarkers that
could be used to detect lung cancer with high sensitivity and specificity. In the
present study, we evaluated a panel of 15 DNA methylation markers using
primary human lung adenocarcinoma (AD) samples and comparing them to
adjacent non-tumor lung (AdjNTL) from lung cancer patients.

Methods
Study subjects
The 43 AD cases used as training set were obtained from LA-USC County
Hospital. The sample collection includes 26 males and 14 females and 22
Caucasians and 18 from other ethnicities (non-Caucasian). The ages of the
patients ranged from 37-82 years old (median 58 years old). Two patients had
missing clinical information.
The first AD validation set was obtained from National Disease Research
Interchange (NDRI) and the Norris Comprehensive Cancer Center Tissue

47
Discard Repository (TDR). Eight paraffin-embedded AD and separate AdjNTL
blocks were used in the study. Study subjects from NDRI were 5 males and 3
females, of which 5 were Caucasian and 2 African American (1 was of unknown
ethnicity). Ages ranged from 54-78 years old at the time of surgery (median age
of 64 years old). Stage information was not available for these subjects. Twelve
additional pairs of archival tumor and AdjNTL tissue blocks of AD cases were
obtained from TDR. All subjects were male and all except 2 were Caucasian (1
Latino and 1 African American). Age and stage information were not provided for
all subjects.
For the second validation study, 23 fresh frozen AD and AdjNTL
specimens from Ontario Tumor Bank (OTB) were used. These lung AD cases
are comprised of 13 females and 10 males with a median age of 68 years old
(range: 51-83 years old; no ethnicity information was given for all subjects)

DNA extraction and bisulfite conversion
To verify the histological classification of the tumor, the paraffin-embedded
sections were stained with hematoxylin and eosin and reviewed by a lung
pathologist. The optimal tumor areas were marked for manual microdissection
and all adjacent non-tumor lung blocks were verified to be tumor-free. DNA was
extracted from the tumor and non-tumor lung specimens as previously described
in Laird et al 1999. Cells were lysed in a solution containing 100 mmol/L Tris-HCl

48
(pH 8.0), 10 mmol/L EDTA (pH 8.0) 1 mg/mL proteinase K, and 0.05 mg/mL
tRNA and incubated at 50
o
C overnight. The DNA was bisulfite converted with the
use of Zymo EZ DNA methylation kit according to manufacturer’s instructions
(Zymo Research Corp, D5001).

Methylation analysis
DNA methylation analysis was done by MethyLight as previously
described in Weisenberger et al 2006. Primer and probe sequences used in this
study are listed in Table 3.1. To normalize for input DNA, an internal reference
primer and probe set designed to analyze Alu repeats (Alu) was included in the
analysis (Weisenberger et al 2005). The percentage methylated reference (PMR)
was calculated as the GENE:reference ratio of a sample divided by the
GENE:reference ratio of in vitro methylated (SssI-treated) human white blood cell
DNA and multiplying by 100.

Statistical analysis
The frequency and median PMR (from samples with positive methylation
values) were calculated to compare the prevalence and level of methylation
between the AD and AdjNTL groups. The distribution of log-transformed PMR
values was plotted using JMP 6.0. The methylation difference between the two
tissue types was calculated by comparing PMR values of AD from AdjNTL

49
tissues by means of the Wilcoxon rank sum test. For the comparison of paired
AD and AdjNTL samples from the same patients, the Wilcoxon signed rank test
was used. All statistical tests were two-sided. Five-fold cross validation was
performed to evaluate the optimal sensitivity and specificity of each locus.
To determine which combinations of markers would be most effective to
correctly identify tumor vs. non-tumor samples, we fit a random forest classifier to
the data set, using the R programming language (v 2.5 (Ihaka et al 1996)) and 73
samples (AD n=42; AdjNTL n=31) and 15 variables. Using bootstrap samples of
the data, we grew a forest of 30,000 trees. Splits were determined using a
random sample of five variables and trees were grown until there was only one
observation in each leaf. We determined error rates using the observations that
were not used to generate the trees. For each observation, its outcome was
predicted by having the majority vote from the trees that were generated without
the original data point in their bootstrap sample. These predicted values were
compared against the true tissue type to estimate prediction error.

Results and Discussion
The goal of this study was to develop a highly specific and sensitive panel
of DNA methylation markers to effectively diagnose human lung AD. The first

50

51
phase of marker development is the discovery of promising candidates by
identifying molecular changes specific for tumor presence. We recently reported
8 loci that are significantly more highly methylated in the AD tissues as compared
to the AdjNTL: OPCML, CDX2, HOXA1, CDKN2A, SFRP1, CDH13, TWIST1,
and RASFF1 (Tsou et al 2007). Limiting the panel to the top four loci (OPCML,
CDX2, HOXA1, CDKN2A) allowed identification of tumor tissue vs. non-tumor
controls with a sensitivity of 94% and specificity of 90%. To improve the
sensitivity and specificity of the assay, we searched for additional DNA
methylation markers.
Using a candidate gene approach, we screened 200 cancer-related loci by
MethyLight using DNA from primary tissues (Figure 3.1). We identified 7
differentially methylated loci that show promise in identifying AD from the control
specimen: 2C35, EYA4, HOXA11, NEUROD1, NEUROD2, PTPRN2, and
TMEFF2. Because the previous analysis used a different batch of SssI-treated
control DNA, inconsistencies in the Percentage Methylated Reference (PMR)
levels between the two studies could generate a problem when putting a panel
together. Therefore, we examined the methylation status of the 15 loci (Table
3.2) in a single experiment. We confirmed frequent methylation of all loci in the
cancer tissues (Frequency > 79%) (Figure 3.2, Table 3.3). For many loci, we also
observed frequent methylation in histologically normal lung tissue of cancer

52
Figure 3.1 Heat map of 200-loci screen of paired AD and AdjNTL samples. Methylation values are represented
as a gradient from no detectable methylation (blue) to high level methylation (dark red). Sample types are
indicated on the left (top panel: AD, bottom panel: AdjNTL). The loci at the top are sorted by difference in median
PMR values between AD and AdjNTL with most informative loci to the left.

53

54

Figure 3.2 Panel of 15 DNA methylation biomarkers. Samples are indicated at the left,
loci at the top. The color gradient represents PMR values from no detectable methyatlion
(blue) to high methylation (dark red). The heatmap demonstrates heavy methylation in
AD samples (top panel) compared with non-tumor controls

55

56
patients, albeit at lower levels. This could indicate that molecular changes occur
very early in lung cancer development. For example, as expected based on
previous reports (Tsou et al 2007, and unpublished data), CDKN2A showed a
high frequency of methylation in the adjacent non-tumor tissue (Table 3.3), but
the tumor samples had much more elevated methylation (PMR of 28.07 vs.
8.11). In fact, for all 15 loci, significant differential methylation between AD and
AdjNTL was found, with p-values ranging from 0.0016 to 1.18E-09 (Table 3.3).
To illustrate the difference between AD and non-tumor lung in more detail, we
plotted the distribution of PMR values in both tissue types. Our results showed
considerably elevated methylation in the diseased tissue, although some overlap
between the two groups is evident (Figure 3.3).
Although the area under the curve (AUC) is utilized as a measure of
marker performance in the clinic, we used it to determine how well our markers
distinguished tumor from non-tumor tissues. The performance of DNA
methylation markers individually was determined by calculating the AUC of each
locus. Homeobox A1 (HOXA1), the highest ranked locus based on Gini index
calculated by random forest analysis (Table 3.3), had an AUC of 0.90. All the
other loci showed an AUC range of 0.72-0.92 (Table 3.4). The results suggest
that each individual locus could contribute to the detection of cancer samples. To
further evaluate the usefulness of each locus as a biomarker, the sensitivity and

57

Table 3.4 Summary and sensitivity and specificity of the 15 loci
HUGO
a
AUC
b
PMR
cutpoint
Sens(%)
c
Spec(%)
d
Sens(%) Spec(%)
HOXA1 0.90 14.53 86 100 79 100
2C35 0.91 18.49 81 100 79 97
CDKN2A 0.85 18.62 71 100 69 97
PTPRN2 0.92 20.70 93 81 88 81
CDX2 0.90 19.53 83 90 81 81
NEUROD1 0.92 23.60 74 97 79 77
HOXA11 0.89 16.70 76 97 71 97
TMEFF2 0.82 16.78 71 100 69 100
SFRP1 0.87 35.59 76 94 71 94
RASSF1 0.72 21.89 57 94 55 94
NEUROD2 0.86 16.95 86 81 83 81
CDH13 0.84 7.60 71 94 69 87
TWIST1 0.79 10.40 64 100 62 97
OPCML 0.84 17.68 76 87 64 87
EYA4 0.78 26.90 62 87 62 84
a
HUGO, Human Genome Organization nomenclature
b
AUC, area under the curve
c
sens, sensitivity
d
spec, specificity
Cross Validation Initial Analysis

58
Figure 3.3 PMR distribution of 15 DNA methylation markers. The log-transformed PMR values and gene
names are indicated on the y-axis. The AD tissues are represented as red open circles (left-hand side) and
AdjNTL from AD patients are represented as blue squares (right-hand side). The mean methylation values
for each group are marked on the plot, showing higher methylation level in AD than histologically normal
tissues.

59
specificity were calculated, using a 5-fold cross validation. After 5-fold cross
validation, HOXA1 had a sensitivity of 79% and specificity of 100% (Table 3.4).
Overall, the specificity of individual markers was relatively high (>77%), however,
sensitivity was lower than desired (55%-88%) (Table 3.4). To increase sensitivity,
loci can be combined into a panel. A random forest analysis indicates that a
panel consisting of all 15 loci had a specificity of 100% and a sensitivity of
92.85%.
Two independent collections of AD and adjacent non-tumor samples were
used to validate the performance of the 15-loci panel. In both test groups, the
markers performed similarly. The results indicate that the tumor tissues are
frequently methylated (test set 1, frequency > 90%; test set 2, frequency > 65%).
The AdjNTL tissues also showed background methylation in all loci. Even if the
control samples had some low level methylation, all loci have much higher
methylation in the AD tissue (p-value > 0.0066; Tables 3.5-1 and 3.5-2). When
the panel of 15 markers was used on the validation samples, the sensitivity
dropped to 70-78.26% while specificity remained at 100% (Table 3.6).
To make a clinical assay more cost effective, a smaller panel would be desirable.
Therefore, we determined sensitivity and specificity of a panel consisting of the
top 10 markers (as ranked by the Gini index from the random forest analysis,
Table 3.3). In the first validation group, the sensitivity decreased to 65% while in
the other sensitivity remained the same at 78.3%. (specificity at 100% for both

60

HUGO p-value
AD AdjNTL AD AdjNTL
HOXA11 100 100 25.60 4.41 3.66E-08
NEUROD1 96 83 14.75 0.86 4.48E-08
2C35 100 91 18.80 1.54 5.97E-08
NEUROD2 100 100 16.17 3.67 2.26E-07
SFRP1 96 91 7.25 0.79 2.26E-07
PTPRN2 96 100 14.86 1.78 3.40E-07
OPCML 96 87 5.93 1.25 5.01E-07
CDH13 91 52 3.88 0.41 7.19E-07
TWIST1 91 57 8.66 0.34 1.12E-06
CDX2 96 100 4.13 1.09 1.47E-06
HOXA1 83 78 25.21 0.67 5.21E-06
EYA4 96 83 3.44 1.82 2.80E-05
CDKN2A EX2 100 100 31.38 11.63 0.0012
RASSF1 65 43 8.83 0.73 0.0017
TMEFF2 100 100 9.46 5.11 0.0066
Frequency (%) Median
Table 3.5-2 Validation set 2: Summary of Frequency, Median and p-value of fresh-
frozen samples
HUGO p-value
AD AdjNTL AD AdjNTL
NEUROD2 100 95 20.06 1.40 1.91E-06
PTPRN2 100 80 31.59 0.34 1.91E-06
2C35 95 65 42.49 0.61 3.81E-06
CDKN2A EX2 100 85 10.33 2.06 3.81E-06
HOXA1 95 70 29.83 0.13 3.81E-06
TWIST1 95 65 42.60 0.08 7.63E-06
HOXA11 95 85 8.18 0.25 9.54E-06
NEUROD1 100 100 28.68 0.83 9.54E-06
OPCML 95 65 11.04 0.20 1.34E-05
CDH13 90 45 3.61 0.00 5.34E-05
SFRP1 100 95 35.65 0.92 1.05E-04
RASSF1 90 70 13.81 0.33 2.67E-04
CDX2 100 90 3.44 0.73 0.002
EYA4 100 100 15.54 0.78 0.003
TMEFF2 100 100 12.22 4.94 0.003
Frequency (%) Median
Table 3.5-1 Validation set 1: Summary of Frequency, Median and p-value of
paraffin- embedded samples

61

62
groups). When the panel was limited to the top 4 ranked loci (Table 3.3), the
sensitivity and specificity of the first test set did not change while in the other a
slight increase in sensitivity (82.6%) was observed (Table 3.6). The instability in
sensitivity could be a result of difference in population between the groups (e.g.
ethnicity). It is important to note that our training set from USC-LA County
Hospital is composed of AD cases of mixed ethnicities: approximately half are
non-Caucasian background. And at least one test set was primarily Caucasians,
with 15 out of 20 cases. Our results suggest the importance of validating our
markers in different ethnicities to determine their value for population-based
screening.

Conclusions
We have developed a panel of highly specific DNA methylation markers
that show potential to serve as a complement to sensitive imaging technologies,
such as LDSCT. Because of lower than expected sensitivity during the validation
analyses, we searched for additional DNA methylation markers through high-
throughput Illumina GoldenGate assay (Chapter 5). In addition, we are in the
process of collecting archival specimens of AD cases from different ethnic
backgrounds (African American, Latino, and Asian). The analysis of these cases

63
will determine whether our panel can be used for screening the population. And
by improving our panel of DNA methylation markers, we are a step closer to
developing an early detection tool for lung cancer.

64
Chapter 4: The onset and progression of aberrant DNA
methylation of cancer-related genes in lung
adenocarcinoma development.

Chapter 4 Abstract
With more than 160,000 lung cancer deaths estimated for 2007 in the
United States, lung cancer will cause more deaths than breast, prostate and
colon cancer combined. The high mortality rate of lung cancer is mainly attributed
to the lack of effective screening tools for early detection of the disease. Studying
the molecular mechanisms underlying lung cancer development and progression
could greatly improve early detection technologies by providing new candidate
markers. One key mechanism associated with tumorigenesis is the inactivation of
tumor suppressor genes by DNA hypermethylation at promoter CpG islands.
Knowledge of the sequential epigenetic changes that occur during the
development of lung cancer could yield powerful DNA methylation markers for
the early detection of lung adenocarcinoma (AD, the most common subtype of
lung cancer) in addition to providing insights into the natural history of the
disease. Using MethyLight, a quantitative DNA methylation assay, we analyzed

65
the DNA methylation of 17 loci in atypical adenomatous hyperplasia (AAH), a
putative precursor lesion of AD, AD and non-tumor adjacent lung tissue from the
same patients. Our results indicate that the onset of methylation varies between
the different loci. Some loci show hypermethylation even in adjacent non-tumor
lung (e.g. CDKN2A), while others show hypermethylation in AAH and AD, but not
in adjacent non-tumor lung (e.g. SFRP1). Yet other loci (e.g. RASSF1) only show
DNA hypermethylation in AD. We are currently examining DNA methylation
status of these loci in a larger sample collection, including many AAH lesions and
bronchioalveolar carcinoma (BAC) samples. BAC is thought to represent
adenocarcinoma in situ, a potential intermediate in the AAH to AD continuum.
The data presented here suggests the occurrence of specific DNA methylation
hits at distinct times in the progression of non-tumor lung to AAH and ultimately
to AD. Understanding the sequential changes in DNA hypermethylation during
the progression of lung cancer could help dissect the natural history of the
disease and could yield more robust biomarkers for the early detection of lung
cancer.

Introduction
Lung cancer, the leading cause of cancer death worldwide, is classified
into two main groups: small cell lung cancer (SCLC) and non-small cell lung

66
cancer (NSCLC). Non–small cell lung cancer, accounting for approximately 85-
90% of all lung cancer cases, is further categorized into the following subtypes:
adenocarcinoma, squamous cell carcinoma, large cell carcinoma, and others
(Travis et al 1996). These classifications are based on histologic as well as
molecular features and are used for evaluating patient prognosis and treatment.
Over the past decade, lung adenocarcinoma prevalence had been steadily
increasing (Kachroo et al 2008) and has surpassed squamous cell carcinoma in
the number of new lung cancer cases diagnoses. Although adenocarcinoma is
considered the most frequent histologic subtype of lung cancer in the United
States (Travis et al 1995), its etiology remains to be elucidated.
There is increasing evidence that the development of invasive lung
adenocarcinoma, like squamous cell carcinoma, may occur in as a multi-stage
pathogenic process. Small non-cancerous lesions have frequently been
observed in the areas surrounding AD tissue. These premalignant foci are
hypothesized to be the initial steps in the development of adenocarcinoma and
are identified as atypical adenomatous hyperplasia (AAH) (reviewed in Kerr et al
2007). The World Health Organization defines AAH as a “localized proliferation of
mild to moderately atypical cells lining involved alveoli and sometimes respiratory
bronchioles, resulting in focal lesions in peripheral alveolated lung, usually less
than 5 mm in diameter” (Dacic et al 2008). Chapman et al (2000) found that AAH

67
is relatively more common in AD patients, which supports it being an AD
precursor.
To date, a number of studies have examined AAH to characterize the
molecular changes that occur in the early stages of adenocarcinoma
progression. Mutational analyses of AAH have shown that EGFR and KRAS
mutations represent early events in AAH-AD pathogenesis (Yoshida et al 2005,
Sakamoto et al 2007, Sartori et al 2008, Soh et al 2008). Atypical adenomatous
hyperplasia lesions that exhibit EGFR mutations lack KRAS mutations (Yoshida
et al 2005). Furthermore, although activating mutations of the KRAS oncogene
are thought to be involved in the risk for the development of AAH to AD (Kohno
et al 2008), Sakamoto et al (2007) suggested that only EGFR-mutant AAHs were
likely to develop into invasive AD. In addition to EGFR and KRAS mutations,
allelic loss (loss of heterozygosity or LOH) occurs at multiple chromosome sites.
Chromosome 3p and 10q, which harbor numerous tumor suppressor genes, are
frequent LOH targets in AAH (Gradowski et al 2007). Although the focus of most
AAH studies has been on genetic changes, more recently there has been
growing interest in the role of epigenetic alterations, particularly aberrant DNA
methylation, in the development of lung adenocarcinoma. Elucidation of
epigenetic alterations is an important aspect of gaining a full understanding the
biology of the disease. In a genome-wide screen for differentially methylated
genes in the earliest stage of adenocarcinoma, ACIN1 was identified to be

68
hypermethylated in an immortalized AAH cell line (Shu et al 2006). Furthermore,
DNA methylation abnormalities in Wnt pathway antagonists and seven other
hallmark cancer genes has been observed in AAH tissues, pointing to a potential
role of DNA methylation in the onset of adeno-carcinogenesis (Licchesi et al
2008, Licchesi et al 2008). In this chapter, we will describe a small pilot study of
four AD cases with associated AAH lesions. Using a sensitive and quantitative
approach (MethyLight), we analyzed the DNA methylation levels of 15 loci that
had been previously shown to be significantly hypermethylated in lung AD
compared to adjacent non-tumor lung from AD patients (Tsou et al 2007 and
unpublished data, see Chapter 3), two additional loci that recently have been
identified to be hypermethylated in AAH (Licchesi et al 2008), and two global
hypomethylation markers (Weisenberger et al 2005). We examined DNA
methylation of these loci in adjacent non-tumor lung, AAH lesions, and AD from
the same patients, with the objective to try to estimate the timing of methylation
of each locus in the progression of non-cancerous to AAH to invasive
adenocarcinoma. A better understanding of the multi-step epigenetic changes in
AD development and progression could lead to improved early detection
strategies and diagnostic procedures and potentially novel targeted therapies.

69
Methods
We studied four AD cases obtained from the University of Aberdeen
Fosterhill Aberdeen, Scotland. All four AD cases had concomitant atypical
adenomatous hyperplasia (AAH) lesions; for one case seven independent AAH
foci were examined.
The paraffin-embedded tumor and non-tumor blocks were sectioned and
hematoxylin and eosin-stained slides evaluated by an expert lung pathologist (Dr.
Keith M. Kerr). Tumor and AAH lesions were marked for manual microdissection.
An independent histologically confirmed cancer-free block of adjacent non-tumor
lung from each patient was used as a control lung tissue.
For each specimen, two 10um sections were manually microdissected.
The cells were lysed in a solution containing 100 mmol/L Tris-HCl (pH 8.0), 10
mmol/L EDTA (pH 8.0) 1 mg/mL proteinase K, and 0.05 mg/mL tRNA and
incubated at 50
o
C overnight. The DNA was bisulfite converted using the Zymo
EZ DNA methylation kit according to manufacturer’s instructions (Zymo Research
Corp, D5001).
The DNA methylation levels of 17 cancer-related genes and 2 global
methylation markers were analyzed using MethyLight (Weisenberger et al 2006).
Primer and probe sequences used are listed in Table 4.1. In addition to primers
and probe sets designed specifically for the gene of interest, an internal

70
reference primer and probe set designed to analyze a CpG-free section of Alu
repeats (Alu) was included in the analysis to normalize for input DNA
(Weisenberger et al 2005). The percentage methylated reference (PMR) was
calculated as the GENE:reference ratio of a sample divided by the
GENE:reference ratio of in vitro methylated (SssI-treated) human white blood cell
DNA and multiplying by 100.

Results and Discussion
The goal of the current study was to further characterize the molecular
mechanism of adenocarcinoma pathogenesis by investigating the DNA
methylation status of 17 cancer-related genes. We examined four AD cases
(each with available cancer, atypical adenomatous hyperplasia (AAH), and
adjacent non-tumor lung tissues) to determine the putative timing at which DNA
methylation marks start accumulating. By using the sensitive and quantitative
assay, MethyLight, we were able to assess the DNA methylation levels of the 17
loci for each lesion. Fifteen of the 17 loci were chosen based on previous studies,
where each locus demonstrated frequent hypermethylation and higher DNA
methylation levels in invasive AD compared with the adjacent non-cancerous
tissues (Tsou et al 2007, and unpublished data, see Chapter 3). The two

71
additional loci, SFRP4 and SFRP5, were chosen based on recently published
DNA methylation data on AAH (Licchesi et al 2008). As expected, our present
study confirms frequent and elevated DNA methylation in the tumor tissues; all
loci exhibited a DNA methylation frequency ranging from 50-100% and showed
median PMR of 2.19-46.07 in the AD samples (Table 4.2). Although the non-
cancerous tissues also appear to have frequent methylation of some loci (0-
100%), the range of the median PMR calculated from all positive samples was
much lower (PMR of 0.16-24.62) compared with tumor tissues (Table 4.2). For
example, we observed that three out of four adjacent non-tumor lung samples
showed cyclin-dependent kinase inhibitor 2A (CDKN2A) gene methylation
(median PMR=24.62; Table 4.2). Though there is substantial CDKN2A
methylation in the adjacent non-tumor tissues, we observed an increase in
methylation in the associated AAH lesions and even more in the AD tissues
(Figure 4.1, Table 4.2). One might argue that the basal methylation level
observed in the adjacent non-tumor lung tissues might be attributable to
infiltration of tumor cells. However, these samples were histologically verified to
be tumor-free. Thus, an alternative explanation is the existence of an epigenetic
field defect in the lungs of AD patients. Indeed, frequent methylation of normal
bronchial margins had been observed in resected NSCLC tumors (Guo et al
2004).

72
In support of our findings, Licchesi et al (2008) recently reported a
progressive increase in DNA methylation frequency of the CDKN2A gene from
non-cancer tissue to AAH and ultimately to invasive AD. CDKN2A plays a key
role in the regulation of the retinoblastoma (RB) cell cycle pathway. Inactivated
CDKN2A causes a disruption of the RB pathway leading to uncontrolled cell
proliferation, a hallmark of cancer development (reviewed by Sherr 1996). For
this reason, CDKN2A methylation might be expected to occur very early on in
lung cancer development. The presence of CDKN2A methylation in AAH
suggests CDKN2A is an early molecular target in the initial stages of
carcinogenesis, and therefore could be used as a biomarker for lung cancer risk.
Indeed, CDKN2A promoter methylation has been observed in the sputum of
high-risk cancer-free individuals (Kersting et al 2000, Destro et al 2004), showing
promise as a non-invasive method of determining lung cancer risk. The data
suggest that CDKN2A is not only marker for risk of lung cancer, but also a
marker for progression.
In addition to CDKN2A, another locus that appears to accumulate DNA
methylation marks in the earliest phase of adenocarcinoma development is the
transmembrane protein with EGF-like and two follistatin-like domains 2
(TMEFF2) gene. Here we show the presence of high-level DNA methylation in
most AAH samples with a median PMR of 22.05 (Figure 4.1 and 4.2, Table 4.2).

73

74

Figure 4.1 Schematic of DNA methylation status of AdjNTL, AAH, and AD. The sample types are
indicated on the left (AdjNTL= top panel, AAH= middle panel, AD=bottom panel). The loci studied are
listed at the top (sorted alphabetically). The gradient represents the level of methylation from no
detectable methylation (blue) to high level methylation (dark red). Heavy methylation is observed in
AD specimens while AAH samples had a variety of pattern.

75
The involvement of TMEFF2 methylation in carcinogenesis was first described in
colon cancer. Interestingly, adenomas, precursors of most colon cancers, show
hypermethylation-based silencing of the TMEFF2 gene (Young et al 2001). In
addition, TMEFF2 methylation has been detected in the stools of patients with
benign colorectal disease (Huang et al 2007). This is the first report of TMEFF2
methylation in precursor lesions of NSCLC. The early methylation of TMEFF2
may represent a target that could be developed as an early detection or risk
marker of lung cancer.
AAH lesions also showed high levels of DNA methylation of two HOX
gene family members. HOX genes, a subgroup of homeobox genes, function in
axis patterning during development. Both homeobox A1 (HOXA1) and homeobox
A11 (HOXA11) genes show an increase in the frequency and level of DNA
methylation during the multi-step transition to invasive AD (Figure 4.2, Table 4.2).
The secreted frizzled-related protein (SFRP) genes have diverse DNA
methylation patterns. SFRP inhibits Wnt signaling by directly binding to Wnt
molecules. Inactivation of Wnt antagonist results in constitutive Wnt signaling
leading to activation of downstream targets, including several oncogenes.
Intriguingly, the SFRP1 gene is found at chromosome 8p21, a site of frequent
LOH in human tumors, including lung cancer (Stoehr et al 2004, Fukui et al 2005,
Shih et al 2006). SFRP1 appears to be methylated in AAH and AD, SFRP4

76

Table 4.2. Summary of frequency and median PMR values of AdjNTL, AAH and AD
Frequency (%)
b
Median
c
AdjNTL
d
AAH
e
AD
f
AdjNTL AAH AD
2C35 75 75 100 11.05 41.89 36.69
CDH13 0 25 50 N/A 3.45 21.18
CDKN2A 75 100 100 24.62 25.60 29.34
CDX2 50 100 100 8.44 5.94 8.28
EYA4 75 75 75 10.20 19.42 5.71
HOXA1 50 100 75 6.84 33.26 22.06
HOXA11 50 100 100 3.44 16.63 29.08
NEUROD1 100 100 100 6.00 9.04 21.95
NEUROD2 75 100 100 2.40 2.19 19.45
OPCML 25 50 100 0.16 12.57 21.57
PTPRN2 100 100 100 4.53 17.43 33.82
RASSF1 50 0 100 0.32 N/A 24.36
SFRP1 50 75 100 8.61 15.42 38.21
SFRP4 75 25 50 0.91 3.10 2.19
SFRP5 75 100 100 5.06 0.23 20.38
TMEFF2 75 100 100 7.54 22.05 25.33
TWIST1 75 75 75 5.55 0.26 46.07
a
HUGO, Human Genome Organization nomenclature sorted alphabetically
b
Percentage of samples with positive methylation value
c
Median percent methylated reference calculated from positive methylation values
d
AdjNTL, Adjacent non-tumor lung from adenocarcinoma patients
e
AAH, Atypical adenomatous hyperplasia from adenocarcinoma patients
f
AD, Adenocarcinoma
HUGO
a

77

Figure 4.2 DNA methylation pattern of AdjNTL, AAH and AD. The PMR values and gene
names are indicated on the y-axis. AdjNTL, AAH, and AD are grouped on the x-axis by
histology. The mean PMR values (green line) showed the methylation difference
between each group.

78
shows very little methylation in any of the samples, while SFRP5 appears to
become methylated in invasive lung AD (Figure 4.2). Licchesi et al (2008) have
also investigated the prevalence of SFRP1, 4, and 5 methylation in adjacent non-
cancer, AAH and AD tissues; unlike our findings, all three loci show similar
patterns of increasing methylation as AD progress. The disparity between the two
findings could be due to the difference in the assays used in the studies
(MethyLight vs. nested MSP). MethyLight is more quantitative than the qualitative
nested MSP. Nested PCR amplifies any detectable signal and could result in a
positive sample even when methylation is very low. The presence of SFRP1
methylation during the earliest phase of lung cancer progression indicates
possible involvement of the Wnt pathway in lung carcinogenesis. Aberrant
activation of the Wnt pathway has been found to occur in many types of cancer.
Wnt ligands are thought to be involved in embryogenesis, regulating cell behavior
and modulating cell-cell interactions (Fuerer et al 2008).
Not all loci previously reported to be hypermethylated in AD exhibited
methylation in the AAH lesions, indicating that the timing of epigenetic hits during
adenocarcinoma development varies from one locus to the next. For example,
while Ras association (RalGDS/AF-6) domain family member 1 (RASSF1) gene
is methylated in all AD cases, none of the AAH lesions show DNA methylation of
RASSF1 (Figure 4.2, Table 4.2). There is a marked increase in the frequency of

79
methylation of RASSF1 in the invasive AD tissues, suggesting RASSF1 may play
a role in the later steps of AAH-AD progression. In contrast to our results,
Licchesi et al (2008) reported that DNA methylation of the RASSF1 gene occurs
early in the development of AD since distinct methylation was already observed
in ~30% of low grade and ~40% of high grade AAH. Again, the two studies
showed inconsistencies, which could be attributed not only the difference in
assay as mentioned earlier, but the smaller sample size of our present analysis
as well. A larger collection of AD cases with concomitant AAH lesions could
reveal the true nature of the onset of DNA methylation in the RASSF1 gene.
Although genes that do not show methylation marks in the earliest phase of AD
development may not be useful as early detection targets, they could still provide
insight into the mechanism by which lung cancer develops.as well. A larger
collection of AD cases with concomitant AAH lesions could reveal the true nature
of the onset of DNA methylation in the RASSF1 gene. Although genes that do
not show methylation marks in the earliest phase of AD development may not be
useful as early detection targets, they could still provide insight into the
mechanism by which lung cancer develops.
DNA hypomethylation is highly associated with cancer development
(Feinberg 2007). Anisowicz et al (2008) showed that non-cancerous tissues of
lung cancer patients have decreased global methylation when compared to non-

80
tumor controls from “healthy” subjects, suggesting global hypomethylation has a
role in lung carcinogenesis. Although it seems that our AdjNTL samples had a
loss of DNA methylation (PMR < 100; Figure 4.2, bottom panel), we had no non-
tumor tissue from non-cancer patients to compare with. Therefore, we cannot
conclude whether there is an early loss in global DNA methylation in our
analysis. In comparison of AdjNTL samples from AAH lesions, it seems that the
same level of methylation is present between the two groups while in invasive
AD, there is a marked decrease in methylation (Figure 4.2, bottom panel). Indeed
there may be two phases of hypomethylation events during the development of
adenocarcinoma. This phenomenon needs to be validated using a larger set of
samples in each phase of lung cancer development. In addition, non-malignant
samples from “healthy” subjects need to be examined.
In addition to studying AAH lesions from 4 individual patients, we studied
seven lesions from one patient. Evidence based on the pattern of X-chromosome
inactivation suggests that individual AAH lesions are monoclonal, but that each
lesion is independent (Niho et al 1999). Our results show a variation in PMR
levels between individual AAH lesions from a single patient. This observation
supports the idea that each lesion is independent (Figure 4.3, open circles). Our
results suggest that DNA methylation of certain key loci occur in early lung
carcinogenesis. In agreement with the Herman laboratory’s studies (Licchesi et al

81

Figure 4.3 DNA methylation signature of AdjNTL, AAH and AD tissues from one AD patient.
The y-axis represents PMR values and the 15 loci studied are found on the x-axis sorted
alphabetically. PMR values are represented as blue box for AdjNTL (n=1), open orange circle
for AAH (n=7), and closed red circle for AD (n=1).

82
2008), we have found that for some loci DNA methylation events occur at the
earliest stages of lung pathogenesis, while other loci appear to not accumulate
any DNA methylation signature until the invasive state.

Conclusions
It is important to note that this is a preliminary study, and our findings need
to be validated in a larger panel of AD cases with AAH lesions. We also
evaluated a limited number of loci. With the advent of new technologies which
allow the analysis of DNA methylation at an almost genome-wide scale (e.g.
Illumina’s Infinium DNA methylation assay), it would be important to do a global
DNA methylation analysis on AAH lesions (if at all possible) to determine when
epigenetic hits occur in different pathways involved in cell cycle regulation,
apoptosis, etc. Although our data provide an indication as to when DNA
methylation occurs in the progression sequence of AdjNTL to AAH to AD, we are
missing information on the putative intermediate state between AAH and AD.
Bronchioalveolar carcinoma (BAC) is defined as a non-invasive lesion of the lung
(Travis et al 1999) and has been suggested to be the intermediary phase
between AAH and AD. If preneoplastic AAHs truly transition to BACs and then to
invasive ADs, then we expect to find methylation in BAC of all loci that exhibit

83
methylation in AAH tissues, and possibly of some loci that appear to become
methylated in AD. With more comprehensive analysis, we will begin to
understand the biology of these lesions and their potential clinical significance.

84
Chapter 5: Discovery of DNA methylation markers using
a high-dimensional bead-array.

Chapter 5 Abstract
DNA hypermethylation at promoter CpG islands is recognized as a key
mechanism for tumor suppressor gene inactivation in cancer. Such abnormal
methylation changes could yield powerful biomarkers for cancer detection. Lung
cancer is the cancer that kills the most Americans every year, causing 30% of all
cancer deaths. Because most lung cancer patients are diagnosed after the
disease has already progressed to an advanced stage (due to lack of effective
screening technology), the overall five-year survival rate is only 15%. DNA
methylation markers could greatly increase specificity of lung cancer screening
techniques. We have developed a panel of 15 DNA methylation biomarkers that
show promise in distinguishing cancer from non-cancer tissues with a sensitivity
of 70% and specificity of 100%. The sensitivity of the assay could be improved
with the addition of more cancer-specific loci. We used the high-throughput
quantitative lllumina GoldenGate assay to identify additional loci that are
frequently and highly methylated in AD compared with AdjNTL. We found that

85
some of the loci we previously identified as hypermethylated in AD by MethyLight
were also identified through this assay (e.g. TWIST1, SFRP1, HOXA11),
validating our previous analysis on a different platform. After applying several
filters and validation through MethyLight, we also identified 6 new markers that
were significantly more methylated in AD than AdjNTL (p< 0.024). We are
currently evaluating whether these markers could improve the sensitivity of our
15-loci DNA methylation marker panel in detecting AD tissues.

Introduction
Lung cancer is the leading cause of cancer-related mortality in the United
States and is expected to cause more than 160,000 deaths in 2008 (American
Cancer Society 2008). Clinically, lung cancer consists of two main categories:
small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC). The
latter accounts for the majority of lung cancer cases (80-90%) and is further
subdivided into different histologies: adenocarcinoma, squamous cell carcinoma,
large cell carcinoma and others (Travis et al 1996). The high mortality rate is
mainly attributed to the inability to detect lung cancer sufficiently early--most lung
cancer cases are detected at advanced stages, which are not suitable for
curative resection. The lack of effective methods for early lung cancer detection

86
has heightened interest in developing new detection strategies. Recently
developed low dose spiral computed tomography (LDSCT) has proven to be
sensitive in detecting small lung lesions, but its effectiveness in reducing lung
cancer mortality is yet to be determined. In addition, there is concern about a
high false positive rate, which could lead to invasive tests and result in increased
cost, morbidity and mortality (Bach 2007).
Molecular markers, particularly DNA methylation, have shown promise for
providing non-invasive detection of lung cancer. Promoter methylation of the
CDKN2A gene has been detected in the serum of lung cancer patients (Belinsky
et al 2007). Furthermore, CDKN2A promoter methylation has been observed in
the sputum of high-risk cancer-free individuals (Kersting et al 2000, Destro et al
2004). However, given the heterogeneity within cancer cases and the incomplete
penetrance of any one marker, a single biomarker would have substandard
clinical specificity and sensitivity. Therefore, we have developed a panel
consisting of 15 tumor-associated genes for human lung adenocarcinoma (the
most common subtype of lung cancer) (unpublished data, see Chapter 3).
Although the 15-locus panel has a specificity of 100%, the sensitivity of the assay
could be improved by the addition of more high penetrance DNA methylation
markers.
The recent development of rapid high-throughput assays has led to a
remarkable increase in knowledge about malignant transformation. In particular,

87
high-density DNA methylation platforms have been extremely valuable in
uncovering cancer-specific DNA methylation signatures. Initially, gel-based
assays such as MSP and COBRA limited DNA methylation analysis to a handful
of loci (Tsou et al 2002). The development of a real-time PCR-based method,
MethyLight, has made analysis of hundreds of loci feasible (Weisenberger et al
2006, and our unpublished results). Recently, more advanced technologies have
been developed for the large-scale evaluation of candidate genes. An example of
such innovation is the Illumina Methylation GoldenGate assay, which allows for
the analysis of 1505 loci, representing more than 800 genes, in a single
experiment.
This high-dimensional assay allows the robust identification of methylated
loci in different cancer cell types. Here, we used the Methylation GoldenGate
assay to identify aberrantly methylated loci that distinguish adenocarcinoma from
non-cancerous lung tissues. We found cancer-specific methylation in loci that
may be involved in lung cancer differentiation and progression. These regions
may play a role in the development of lung cancer and need to be further
investigated to fully understand the biology of this lethal disease.

88
Methods
Study subjects
Five fresh-frozen tumor samples of unidentified adenocarcinoma patients
from the Norris Comprehensive Cancer Center were used for the Illumina
GoldenGate assay. Matched adjacent non-tumor lung from a separate block from
each patient was used as control tissue. No clinical information was given
regarding these patients.
For validation studies, paraffin-embedded tissue blocks from
adenocarcinoma patients (n=8) were acquired from National Disease Research
Interchange (NDRI). All cases had a separate block with adjacent non-tumor lung
control tissue. Study subjects from NDRI were 5 males and 3 females, of which 5
were Caucasian and 2 African American (1 was of unknown ethnicity). Ages
ranged from 54-78 years old at the time of surgery (median age=64 years old).
Stage information was not available for these subjects.
Additional archival adenocarcinoma lung tissue and matching adjacent
non-tumor lung samples (n=12) were obtained from the Norris Comprehensive
Cancer Center Tissue Discard Repository. All subjects were male and all except
2 were Caucasian (1 Latino and 1 African American). Age and stage information
were not provided for all subjects.

89
Tissue samples and DNA isolation
Hematoxylin and eosin-stained slides were reviewed by an experienced
lung pathologist (MNK) to verify the original classification of the tumor and to
select optimal tumor and non-tumor areas of the specimens. DNA was extracted
from microdissected tumor and non-tumor lung samples via proteinase K
digestion (Laird et al 1999). Briefly, cells were lysed in a solution containing 100
mmol/L Tris-HCl (pH 8.0), 10 mmol/L EDTA (pH 8.0) 1 mg/mL proteinase K, and
0.05 mg/mL tRNA and incubated at 50
o
C overnight.

Illumina GoldenGate assay
The detection of DNA methylation is based on the ability to distinguish
cytosine from 5-methylcytosine in the DNA sequence by sodium bisulfite
modification, which converts unmethylated C to U. Bisulfite conversion of 1ug of
genomic DNA was achieved through use of the EZ DNA Methylation kit (Zymo
Research), as described by the manufacturer. A beta-value of 0-1.0 was
reported, signifying 0% to 100% percent methylation for each CpG site. Beta-
values were calculated by subtracting background signal, obtained using
negative controls on the array, and taking the ratio of the methylated signal
intensity against the sum of both methylated and unmethylated signals for each
locus.

90
MethyLight analysis
DNA methylation analysis was done by MethyLight as previously
described (Weisenberger et al 2006). Primer and probe sequences are listed in
Table 5.1. In addition to primers and probe sets designed specifically for the
gene of interest, an internal reference primer and probe set designed to analyze
Alu repeats (Alu) was included in the analysis to normalize for input DNA
(Weisenberger et al 2005). The percentage methylated reference (PMR) was
calculated as the GENE:reference ratio of a sample divided by the
GENE:reference ratio of in vitro methylated (SssI-treated) human white blood cell
DNA and multiplying by 100.

Statistical analysis
To analyze the Illumina GoldenGate data, we first calculated the mean
beta-values of the adenocarcinoma and adjacent non-tumor lung samples for
each locus. The data was filtered using the following criteria: we only considered
loci with (1) a mean beta-value of tumor tissue greater than 0.1, (2) a ratio of the
mean beta-values between the AD and non-tumor tissue greater than or equal to
2, and (3) a mean beta-value difference between the two groups greater than
0.17. The area under the curve (AUC) was calculated using the AD and adjacent
non-tumor lung data to determine marker performance in distinguishing cancer
from non-cancer tissues. The AUC of lung cancer vs. other cancers (combination

91
of breast, ovarian, pancreatic, and colon; data provided by Mihaela Campan,
Sahar Houshdaran, Shirley Oghamian and Toshi Hinoue) was also calculated.
In the validation studies, the frequency and median (of positive
methylation values) were calculated. PMR values of AD were compared to
AdjNTL and as a continuous variable by means of the Wilcoxon signed rank test
using JMP 7.0.1 (SAS Institute, Cary, NC, USA).

Results and Discussion
In an effort to improve our existing panel of molecular markers (see
Chapter 3), we searched for additional loci that are differentially methylated in
adenocarcinoma versus non-tumor samples through the Illumina GoldenGate
assay. To identify the most robust DNA methylation markers, we applied several
filters. The filters used to identify these differentially methylated loci were as
follows: between the two groups, the mean beta-values of the adenocarcinoma
samples should be at least 0.17 higher than the adjacent non-tumor lung and the
ratio between the tumor and non-tumor beta values should be at least 2-fold. In
addition, tumor tissue should show a mean beta-value greater than 0.1 (Table
5.2). In a comparison of 5 tumors with matched adjacent non-tumor tissues from
adenocarcinoma patients, we identified 74 loci that showed elevated methylation

92

93
in the diseased tissues compared to the non-cancer control tissues. A number of
genes previously reported by our group to be hypermethylated in
adenocarcinoma were also identified in this platform such as TWIST1, CDH13,
EYA4, HOXA11, and SFRP1, further validating our previous report (Table 5.2;
marked with asterisk). More importantly, TWIST1 ranked as the top marker in this
assay showing a clear distinction in methylation level between the two groups
(mean difference= 0.53 and fold change= 18.86) (Table 5.2). It is also clear that
many more aberrantly methylated loci were identified through this approach such
as ATP-binding cassette, sub-family B (MDR/TAP), member 1
(ABCB1/MDR1), cyclin D2 (CCND2), Wilms tumor 1 (WT1), and SRY (sex
determining region Y)-box 1 (SOX1) to name a few (Table 5.2). All 74 loci
had an area under the curve (AUC) of >60, indicating these loci could
contribute to distinguishing adenocarcinoma from non-cancerous samples (Table
5.2).
Recently, Bibikova et al (2006) did a similar study in which they analyzed
lung cancer biopsy samples and identified a panel of DNA methylation markers
that distinguished adenocarcinoma from non-cancerous lung tissues using the
same platform. Although a similar approach was taken by our lab, we used a
different chip that contains more than 800 genes compared to 300 from
Bibikova’s study. We found that the top 74 loci we identified in this study had

94

95
some overlap with the top 55 tumor-specific loci from Bibikova et al (2006).
Interestingly, one of the loci that both studies confirmed is CCND2. Cyclin D2, a
member of the D-type cyclins, has been suggested to have a role in cell
cycle regulation and malignant transformation. Indeed, in one study,
CCND2 was aberrantly methylated in NSCLC cell lines and tumor tissues.
The NSCLC cell lines that exhibit methylation showed loss of CCND2
expression, and upon treatment with 5-aza-cytidine, expression was
restored (Virmani et al 2003). Although CCND2 methylation seems to be a
cancer-specific event (0 of 18 nonmalignant lung tissues from cancer
patients were methylation-positive) (Virmani et al 2003), another study
reported that as many as 16% of non-tumor lung specimens were methylated
(Feng et al 2008). The methylation of non-cancer samples could be attributed
changes due to environmental exposure such as smoking. The smoking history
of the NSCLC cases in the other study was not known. Also, two different assays
were utilized between Virmani et al and Feng et al: MSP and MethyLight,
respectively. The difference in the sensitivity of the assays could account for
detecting frequent methylation in the nonmalignant samples in the latter study. In
our current results, CCND2 appears to be a promising DNA methylation marker
(mean difference= 0.19, fold change= 8.77, AUC= 1), however, upon closer
examination, the high mean beta-value of the tumor samples was actually driven

96
by one adenocarcinoma patient who has abnormally high levels of methylation in
most loci (Table 5.2, 4
th
locus on the list). It is important to note that our study
consists of only 5 adenocarcinoma cases, thus, the sensitivity and specificity of
CCND2 could be verified by increasing the number of samples during the
validation studies.
Ideally, a molecular marker that could be developed as a clinical assay for
cancer should not only be tumor-specific (tumor vs. non-tumor) but also cancer-
type (lung cancer vs. other cancers) specific. This would allow such a marker to
be used for non organ-specific (e.g. blood-based) analyses. To address whether
the top tumor-specific loci listed in Table 5.2 are lung cancer-specific, we
compared lung cancer tissues with other cancer tissue types analyzed on the
same platform in the same experiment: breast (n=20), pancreatic (n=9), ovarian
(n=16), colon (n=58) cancers data provided by Mihaela Campan, Shirley
Oghamian, Sahar Houshdaran, and Toshi Hinuoe). To determine how well the
loci can differentiate lung cancer tissues from other malignancies, the AUC was
derived from comparison of lung cancer data to ALL other cancers combined. Of
the 74 loci, 14 loci had an AUC of greater than 0.60, indicating these loci show
promise for detecting lung cancer tissues from other malignancies. Many of the
previously promising of the markers were lost after this filtering, including
TWIST1, WT1, and SOX1. On the other hand, it was exciting to find
ABCB1/MDR1 and HOXA11 remaining on the list (Table 5.3).

97

98
We designed MethyLight reactions for the top lung-cancer specific markers to
validate our findings in a different platform using an independent set of
adenocarcinoma and adjacent non-tumor lung samples. Blood can be tested for
circulating tumor-tissue-derived DNA. However, non-cancer-specific changes in
the DNA found in serum or plasma need to be determined. One factor to
consider is the contamination of serum or plasma of cancer patients by white
blood cell (WBC) DNA. By examining the methylation status of WBC from healthy
subjects, we could discern whether the methylation signature we detect is
cancer-specific. To address this, buffy coat specimens of two “healthy” non-
cancer (age-matched) controls were examined for presence of methylation
(Figure 5.1, unpublished data provided by Mihaela Campan). The five loci
excluded from the validation study exhibited high methylation levels (beta-value
of greater than 0.10) in one or both of the control samples (Figure 5.1, grey
boxes). The following loci were not carried over in the validation studies due to
high level methylation found in the blood of non-cancer patients: HGF_E102_R,
P2RX7_P119_R, HOXA9_E252_R, ASCL2_E76_R, and ASCL2_P360_F.
In the validation analysis of a larger sample panel, we show that most of
the top loci identified by the GoldenGate assay have significantly more
methylation in the tumor than the non-tumor control (p-value of 0.024- 7.63E-06;
Table 5.4). These results confirm that these markers show elevated methylation

99

Figure 5.1 Methylation status of WBC DNA of 2 “healthy” controls. The list of top
markers identified by Illumina GoldenGate assay are on the left sorted by significance.
Shaded boxes represents beta-value > 0.1. Loci with high level of methylation in both
control subjects were excluded from subsequent analyses.

TargetID Control1 Control2
ONECUT2_E96_F
HOXA9_P303_F
HOXA9_P1141_R
HGF_E102_R
HOXA11_P698_F
P2RX7_P119_R
RET_seq_54_S260_F
HOXA9_E252_R
ASCL2_E76_R
IGF2_E134_R
MDR1_seq_42_S300_R
ASCL2_P360_F
ISL1_P379_F
HOXB13_P17_R

100

101
in lung cancer and are promising candidates for lung adenocarcinoma markers.
We anticipate that some of these loci might be included as part of a lung cancer
DNA methylation marker panel, that could potentially be developed into a non-
invasive clinical assay.

Conclusions
With the identification of a number of promising markers, the next step is
to combine these newly identified markers with the panel previously reported
ones to determine if their addition might increase the sensitivity of the assay. In
addition to the novel candidates identified from the GoldenGate analysis, whole-
genome approaches to assess methylation are being developed using
methylation-sensitive restriction enzymes, methyl-binding proteins, and anti-
methylcytosine antibodies (reviewed in Zilberman and Henikoff 2007) in
combination with microarrays or high throughput sequencing. Second generation
sequencing technology enables researchers to study methylation signatures of
different cancer types at single-base resolution. These powerful approaches
could lead to the discovery of new epigenetic mechanisms involved in cancer
development and progression. However, many technical challenges still need to
be resolved before these kinds of techniques generate affordable and reliable

102
data. Given rapid technological advancement, it is anticipated that within a few
years, it will be possible to characterize the entire methylome of lung cancer cells
and controls. Until then, platforms like the GoldenGate (discussed here) and its
successor Illumina Infinium, which can provide information on over 27,000 CpGs
and which will soon be used by the Laird-Offringa lab, will provide the most
robust and affordable information. Progress in the development of new
technology to characterize the methylome could lead to improvements in early
diagnosis, prognosis, and lung cancer therapy.

103
Chapter 6: Discussion

Lung cancer is a major health problem due to its high incidence and
mortality. To decrease lung cancer deaths, a new strategy for early detection is
greatly needed. The previous chapters described our work on the development of
DNA methylation markers for the early detection of lung adenocarcinoma, the
most common histological subtype of lung cancer. We demonstrated that DNA
methylation is a frequent event in lung adenocarcinoma (Chapters 2, 3, 4, and 5).
In our initial analysis of 28 cancer-related loci, we identified 8 loci with
significantly more methylation in the tumor tissue compared with the non-tumor
lung controls from adenocarcinoma and non-cancer patients. Limiting our panel
to the top-ranked 4 loci (HOXA1, OPCML, CDKN2A and CDX2) resulted in a
sensitivity of 94% and a specificity of 90% on the used specimen collection
(Chapter 2). To increase the sensitivity and specificity of our assay, we
developed a panel of 15 DNA methylation biomarkers (8 loci from previous study
and 7 newly identified loci). A panel consisting of all 15 loci showed a sensitivity
of 92.85% and specificity of 100% on the used specimen collection (Chapter 3).
However, sensitivity dropped to below 80% when these markers were applied to
independent sample sets. Additional lung cancer-specific markers identified by
the Illumina GoldgenGate assay are being evaluated to determine whether the
sensitivity of the assay could be improved when additional markers are combined
with the 15-loci panel (Chapter 5, discussed more below).

104
One concern we have is that our sample collections were primarily
comprised of Caucasians. Therefore, it is important to determine whether our
DNA methylation markers function in other ethnic backgrounds (African
American, Latino, and Asian). In Chapter 2, we found statistically significant
differences in the methylation levels between tumor and adjacent tissue for
HOXA1, OPCML, CDKN2A, and CDX2 in Caucasians, Latinos, and African
Americans. The Asian group showed significantly higher levels of methylation in
three out of four loci (CDX2, HOXA1, and OPCML) (Table 2.3). Although our
findings show no statistical difference in methylation between the tumor and non-
tumor tissues for the CDKN2A gene in Asians, our sample size was relatively
small (AD= 6 and AdjNTL=4). Several studies have observed frequent CDKN2A
methylation in Chinese adenocarcinoma patients (Liu et al 2003, Chan et al
2002). In collaboration with visiting researchers (Edward Fabián Carrillo and
Weihong Sun) who provided an extensive collection of Latino and Japanese
samples, we found that most of our biomarkers remained informative in
distinguishing adenocarcinoma from non-tumor tissue (unpublished data).
Because these experiments were run independently, we would like to examine
the methylation status of all ethnic groups in a single experiment. We are
continuing to collect paraffin-embedded samples from adenocarcinoma patients
of different ethnic backgrounds (African American, Latino, and Asian) from the
Norris Comprehensive Cancer Center Tissue Discard Repository. With great
anticipation, we expect that these markers will remain informative in the other
ethnicities.

105
Because most of our analyses were done using invasive cancers, we still
have to determine whether these markers could be detected in early stage
cancer. In Chapter 2, we showed that stage IA tumors already exhibited
significant difference in methylation of CDKN2A, HOXA1, CDX2, and OPCML
between the tumor and non-tumor tissues (Table 2.3). We have not determined
whether the other markers show significantly elevated methylation in the tumor
tissue in early stage of adenocarcinoma. However, other studies have shown
frequent methylation of some of our markers in stage I lung cancer (some
examples are listed below, note that some studies were NSCLC, which include
AD). In two studies of NSCLC patients, frequent aberrant methylation of the
CDH13 was present in 44% and 31.4% of stage I patients (Kim et al 2005, Kim et
al 2007). In another study that focused on adenocarcinoma patients, 32.4% of
stage I cancers showed methylation (Toyooka et al 2004). In the same study,
Toyooka and colleagues also detected CDKN2A and RASSF1 methylation in
17.1% and 31.4% of stage I patients, respectively. One caveat regarding the
comparisons of different studies is the use of different methodologies. For
example, the studies mentioned above used the non-quantitative MSP method,
while all of our analyses were done by MethyLight, which is both quantitative and
more sensitive. Thus, it would be important to validate all 15 loci in a larger
sample collection of early stage tumor tissues (ongoing).
We had the opportunity to collaborate with Dr. Keith Kerr, an expert lung
pathologist from University of Aberdeen to study the role of DNA methylation in
adenocarcinoma development. In a small pilot study of 4 adenocarcinoma cases,

106
we examined the presence of DNA methylation in small lesions thought to be
precursors of lung adenocarcinoma. We compared methylation in these lesions
to methylation in adjacent non-tumor lung and full-blown adenocarcinoma from
the same patients. Although these lesions were not temporally linked, we
hypothesize that they represent the multi-step process of adeno-carcinogenesis.
We found that some DNA methylation loci showed hypermethylation in AdjNTL
samples, others accumulated methylation marks in AAH yet others only showed
methylation in invasive adenocarcinoma (Chapter 4). The early onset of DNA
methylation of some loci suggests that DNA methylation could contribute to the
initiation of adenocarcinoma. The next step of our study is to continue the
analysis of the 15 loci in a larger scale. Preliminary studies by Suhaida Selamat
have confirmed our findings that these loci have different patterns of methylation
in the different steps of lung adenocarcinoma development (unpublished results).
Studies have shown that small nodules less than 1mm (including AAHs)
can be detected by LDSCT screening (Kishi et al 2004). Although LDSCT
screening has increased the number of lung cancer diagnoses, evidence suggest
that many of these early lesions detected may not be precursors of advanced
cancer (Bach et al 2008). A recent review by Peter Bach (2008) raises the
question whether AAHs are the actual premalignant lesions of AD. Bach
proposes that lung carcinogenesis may not occur as a continuum as described
by the current model, but instead he proposes a bipartite model. The challenge is
not just to detect these small nodules, but also to identify the ones that will
progress to full-blown AD. How would one identify lesions that will continue to
become metastatic? DNA methylation studies might help stratify the patients with

107
early stage lung adenocarcinoma or AAH into lower or higher risk groups by
profiling these lesions based on methylation signatures and linking this data to
clinical follow-up information. A recent study by Brock et al. (2008) applied a
classification method to identify patients who have higher risk of recurrence. The
authors conclude that methylation of certain genes could identify cells that could
potentially spread to the lymph nodes, thus could predict recurrence. However,
there are several flaws with this study: (1) markers that are not significant are
combined with p16, which drives the significance, (2) the control population may
not match the cases (the controls have more tumors <3 cm) and in addition, the
range of tumor sizes in controls and cases is not provided, (3) the data of the
validation and training sets were combined. Nevertheless, the data is suggestive
of the fact that methylation of some loci may be a good predictor of progression.
Since we applied a targeted-gene approach to the identification of
hypermethylated loci and had a limited sample size during the screening
process, we may be missing some informative markers. As an example,
transcription factor 21 (TCF21) gene, a candidate tumor suppressor, had been
recently reported to be a good candidate for early detection marker of NSCLC
patients (Shivapurkar et al 2008). A CpG rich-short region within exon 1 that is
completely unmethylated in a normal lung epithelial cell line was identified. This
region is hypermethylated in lung cancer cell lines and primary tumor tissues.
Most interestingly, it was found that the sputa of cancer cases had methylation of
the TCF21 gene, while “healthy” patients (COPD cases) had no to low detectable
methylation, thus illustrating its potential as a biomarker. In our recent screen of

108
200 loci, TCF21 was excluded from our panel. The region that was analyzed by
our group completely overlapped with Shivapurkar et al. In our analysis, although
the adenocarcinoma tissues showed overall higher methylation than the non-
tumor counterpart, there was frequent and high basal methylation level in the
adjacent non-tumor lung (unpublished data). Therefore, we didn’t pursue this
locus and we focused our attention on other loci that performed better. However,
because Shivapurkar illustrated that this locus has potential use in non-invasive
detection using sputum, we need to determine whether this locus could
potentially increase the sensitivity of our assay and whether methylation of
TCF21 gene could also be detected in other biological fluids such as plasma or
serum of lung cancer patients. This is one example of potentially missed
markers. However, since most of our markers have not been tested in remote
media such as sputum or blood, it is possible that the markers we have identified
will outperform many of the markers published to date.
In an effort to expand our marker selection, we turned to new technologies
that allow interrogation of much larger numbers of markers. In Chapter 6, we
describe the use of the Illumina GoldenGate assay, a new high-throughput
technology that enables analysis of 1505 loci (more than 800 genes) in one
experiment. Epigenomic research is now entering an exciting era due to the
development of ultra-high-throughput sequencing technologies. These second
generation sequencing platforms will allow us to study DNA methylation
signatures of different cancers with a single-base resolution. This technology
allowed successful sequencing of bisulfite-treated Arabidopsis DNA (Cokus et al
2008). However, more complex genomes still present challenges such as the

109
difficulty of obtaining sequence information representing the whole genome and
mapping the bisulfite sequence onto a larger genome (Dohm et al 2008).
Because these sequencing platforms generate large data sets, data mining and
bioinformatics analysis have also been challenging. Once optimized, these
platforms could be very powerful tools to achieve a deeper understanding of
epigenetic mechanisms linked to lung cancer.
Now that we have identified a number of very promising DNA methylation
markers, an important question we wish to answer is: are these markers present
in bodily fluids (e.g. bronchioalveolar lavage fluid, sputum, plasma, serum) of
lung cancer patients? To address this question, we examined the plasma of 3 AD
cases and compared it with plasma from 3 non-cancer patients. In this small pilot
study, we were able to detect methylated DNA molecules in the plasma of 1 AD
patient while the non-cancer controls showed very low background methylation
(unpublished data). To confirm our findings, we have obtained a plasma
collection representing stage I and stage II AD patients from Ontario Tumor
Bank. We also obtained tumor and AdjNTL tissues from the same patients, which
were used in our validation studies in Chapter 3. We showed that in this
collection all 15 loci exhibited significantly more methylation in the cancer tissue
than the AdjNTL control (Table 3.3). Thus, we are now poised to determine
whether the methylation signatures match between the tumor and plasma
samples. If so, this will confirm that our 15 DNA methylation biomarkers are good
candidate blood markers. They could then be the basis for a powerful new
approach to non-invasive early detection of lung adenocarcinoma.

110
Another source that might have potential as a non-invasive tool for the
detection of lung cancer is exhaled breath condensate (EBC). There are many
advantages in using EBC: cost, relative ease of collection, a low risk procedure,
and the ability to obtain repeated measurements. The DNA from EBC of lung
cancer patients has been shown to contain p53 mutations (Gessner et al 2004)
and microsatellite alterations in the 3p region (Carpagnano et al 2005,
Carpagnano et al 2008). However, we have not been successful in isolating
enough DNA from EBC of healthy subjects to run PCR reactions. In addition,
though it was reported in the public media some time ago, findings of DNA
methylation in EBC have not yet been reported in the scientific press, raising
concerns about the validity of any obtained data.
Many research laboratories, including our own will continue to interrogate
the epigenome to find additional markers as well as to increase our
understanding of lung cancer etiology. With the identification of our promising 15-
DNA methylation marker panel, and our recent identification of candidate
markers for squamous cell lung cancer (Anglim et al 2008), we have made the
first key steps towards applying these findings into clinical practice. Despite
major challenges, epigenome research has opened the doors to rapid discovery
of novel loci that show promise as risk, early detection and prognostic markers.
These technologies could also identify pathways that may play a role in the
development and progression of lung cancer, which could lead to identification of
novel therapeutic targets. Therefore, the development of molecular markers for
early detection is a crucial investment in the health and longevity of not only lung
cancer patients but also those who are at risk of developing this lethal disease.

111
Bibliography
ACS, Cancer facts & figures-2008. 2008, American Cancer Society: Atlanta.

Alberg A.J., et al., Epidemiology of lung cancer: ACCP evidence-based clinical
practice guidelines (2nd edition).Chest., 2007. 132: p. 29S-55S.

Altimari A., et al., Diagnostic role of circulating free plasma DNA detection in
patients with localized prostate cancer.Am J Clin Pathol. 2008. 129(5): p. 756-62.

Anglim P.P., et al., Identification of a panel of sensitive and specific DNA
methylation markers for squamous cell lung cancer.Mol Cancer. 2008. 7: p. 62.

Anisowicz A., et al., A high-throughput and sensitive method to measure Global
DNA Methylation: Application in Lung Cancer. BMC Cancer. 2008. 8(1): p. 222.

Bach P.B., et al., Computed tomography screening and lung cancer
outcomes.JAMA. 2007. 297(9); p. 953-61.

Bach PB., Is our natural-history model of lung cancer wrong?Lancet Oncol. 2008
9(7): p. 693-7.

Bae J., et al., Cigarette smoking, alcohol consumption, tuberculosis and risk of
lung cancer: the Korean multi-center cancer cohort study.J Prev Med Public
Health., 2007. 40(4): p. 321-8.

Belinsky, S.A., et al., Promoter hypermethylation of multiple genes in sputum
precedes lung cancer incidence in a high-risk cohort. Cancer Res, 2006. 66(6): p.
3338-44.

Belinsky, S.A., Gene-promoter hypermethylation as a biomarker in lung cancer.
Nature Reviews Cancer, 2004. 4: p. 1-11.

Benjamini Y., et al., Controlling the false discovery rate in behavior genetics
research.Behav Brain Res. 2001 125(1-2): p. 279-84.

Bibikova M., et al., High-throughput DNA methylation profiling using universal
bead arrays. Genome Res. 2006. 16(3): p. 383-93.

112
Bouchardy C., et al., Recent trends in prostate cancer mortality show a
continuous decrease in several countries.Int J Cancer., 2008. 123(2): p. 421-9.

Brock M.V., et al., DNA methylation markers and early recurrence in stage I lung
cancer.N Engl J Med., 2008. 358 p:1118-28.

Carpagnano G.E., et al., 3p microsatellite alterations in exhaled breath
condensate from patients with non-small cell lung cancer. Am J Respir Crit Care
Med. 2005. 172(6): p. 738-44.

Carpagnano G.E., et al., 3p microsatellite signature in exhaled breath
condensate and tumor tissue of patients with lung cancer.Am J Respir Crit Care
Med. 2008. 177(3): p. 337-41.

Carvalho A.L, et al., Evaluation of promoter hypermethylation detection in body
fluids as a screening/diagnosis tool for head and neck squamous cell carcinoma.
Clin Cancer Res. 2008. 14(1): p. 97-107.

Chan E.C., et al., Aberrant promoter methylation in Chinese patients with non-
small cell lung cancer: patterns in primary tumors and potential diagnostic
application in bronchoalevolar lavage. Clin Cancer Res. 2002. 8(12): p. 3741-6.

Chapman A.D. and Kerr K.M., The association between atypical adenomatous
hyperplasia and primary lung cancer. Br J Cancer. 2000. 83(5): p. 632-6.

Charloux A., et al., Prognostic value of histology in patients with non-small cell
lung cancer.Lung Cancer., 1997. 17(1): p. 123-34.

Chen, K.-Y., et al., Distribution according to histologic type and outcome by
gender and age group in Taiwanese patients with lung cancinoma. Cancer, 2005.
e-pub ahead of print

Chuang C.K., et al., Hypermethylation of the CpG islands in the promoter region
flanking GSTP1 gene is a potential plasma DNA biomarker for detecting prostate
carcinoma.Cancer Detect Prev. 2007. 31(1): p. 59-63.

Cokus S.J., et al., Shotgun bisulphite sequencing of the Arabidopsis genome
reveals DNA methylation patterning. Nature. 2008. 452(7184): p. 215-9.

Collins L.G., et al., Lung cancer: diagnosis and management.Am Fam Physician.,
2007. 75(1): p. 56-63.

113
Dacic S., Pulmonary preneoplasia. Arch Pathol Lab Med. 2008. 132(7): p. 1073-
8.

Destro, A., et al., K-ras and p16(INK4A)alterations in sputum of NSCLC patients
and in heavy asymptomatic chronic smokers. Lung Cancer, 2004. 44(1): p. 23-
32.

Digel, W. and M. Lubbert, DNA methylation disturbances as novel therapeutic
target in lung cancer: preclinical and clinical results. Crit Rev Oncol Hematol,
2005. 55(1): p. 1-11

Divine, K.K., et al., Multiplicity of abnormal promoter methylation in lung
adenocarcinomas from smokers and never smokers. Int J Cancer, 2005. 114(3):
p. 400-5.

Dohm J.C., et al., Substantial biases in ultra-short read data sets from high-
throughput DNA sequencing.Nucleic Acids Res. 2008 Jul 26. [Epub ahead of
print]

Eads, C.A., et al., MethyLight: a high-throughput assay to measure DNA
methylation. Nucleic Acids Res., 2000. 28: p. E32.

Feinberg A.P., An epigenetic approach to cancer etiology. Cancer J. 2007. 13(1):
p. 70-4.

Feng Q., et al., DNA methylation in tumor and matched normal tissues from non-
small cell lung cancer patients. Cancer Epid. Biom. Prev., 2008. 17: p.645-54.

Franzmann E.J, et al., Soluble CD44 is a potential marker for the early detection
of head and neck cancer. Cancer Epidemiol Biomarkers Prev. 2007. 16(7): p.
1348-55.

Freedman N.D., et al., Cigarette smoking and subsequent risk of lung cancer in
men and women: analysis of a prospective cohort study.Lancet Oncol., 2008.
9(7): p. 649-56.

Fuerer C., et al., Wnt signalling in development and disease. Max Delbrück
Center for Molecular Medicine meeting on Wnt signaling in Development and
Disease. EMBO Rep. 2008. 9(2): p. 134-8.

114
Fukai R., et al., The prognostic value of carcinoembryonic antigen in T1N1M0
and T2N1M0 non-small cell carcinoma of the lung. Eur J Cardiothorac Surg.
2007. 32(3): p. 440-4.

Fukui T., et al., Transcriptional silencing of secreted frizzled related protein 1
(SFRP 1) by promoter hypermethylation in non-small-cell lung cancer.
Oncogene. 2005. 24(41): p. 6323-7.

Gessner C., et al., Detection of p53 gene mutations in exhaled breath
condensate of non-small cell lung cancer patients.Lung Cancer. 2004. 43(2): p.
215-22.

Gradowski J.F., et al., Molecular alterations in atypical adenomatous hyperplasia
occurring in benign and cancer-bearing lungs. Diagn Mol Pathol. 2007. 16(2): p.
87-90.

Greenberg A.K. and Lee M.S., Biomarkers for lung cancer: clinical uses. Curr
Opin Pulm Med., 2007. 13(4): p. 249-5.

Grote H.J., et al., Methylation of RAS association domain family protein 1A as a
biomarker of lung cancer. Cancer. 2006. 108(2): p. 129-34.

Guessous I., et al., Lung cancer screening: current situation and perspective.
Swiss Med Wkly., 2007. 123(21-22): p. 304-11.

Guo M., et al., Epigenetic silencing of CDX2 is a feature of squamous
esophageal cancer. Int J Cancer. 2007. 121(6): p. 1219-26.

Han K.Q., et al., Identification of lung cancer patients by serum protein profiling
using surface-enhanced laser desorption/ionization time-of-flight mass
spectrometry.Am J Clin Oncol. 2008. 31(2): p. 133-9.

Hanash S.M., et al., Mining the plasma proteome for cancer biomarkers. Nature.
2008. 452(7187): p. 571-9.

Harris J.E., et al., Cigarette tar yields in relation to mortality from lung cancer in
the cancer prevention study II prospective cohort, 1982-8.BMJ., 2004. 328(7431):
p. 72.

Henschke C.I., et al., CT screening for lung cancer: prevalence and incidence of
mediastinal masses. Radiology. 2006 239(2): p. 586-90.

115
Henschke C.I., et al., Early lung cancer action project: annual screening using
single-slice helical CT. Ann N Y Acad Sci. 2001. 952: p. 124-34.

Henschke C.I., et al., Early Lung Cancer Action Project: overall design and
findings from baseline screening. Lancet. 1999. 354: p. 99-105.

Herman J.G. and Baylin S.B., Gene silencing in cancer in association with
promoter hypermethylation. N Engl J Med. 2003. 349(21): p. 2042-54.

Hoque M.O., et al., Detection of aberrant methylation of four genes in plasma
DNA for the detection of breast cancer. J Clin Oncol. 2006 24(26): p. 4262-9.

Hsu H.S., et al., Characterization of a multiple epigenetic marker panel for lung
cancer detection and risk assessment in plasma.Cancer., 2007. 110: p.2019-26.

Huang Z.H., et al., Detection of aberrant methylation in fecal DNA as a molecular
screening tool for colorectal cancer and precancerous lesions.World J
Gastroenterol. 2007. 13(6): p. 950-4.

Ihaka R. and Gentleman R., R: a language for data analysis and graphics. J
Comput Graph Statist. 1996. 5: p. 299-314.

Jemal A., et al., Cancer statistics, 2006. CA Cancer J Clin., 2006. 56(2): p. 106-
30.

Jett J.R., Limitations of screening for lung cancer with low-dose spiral computed
tomography. Clin Cancer Res. 2005. 11(13 Pt 2): p. 4988s-4992s.

Kachroo S., et al., Trends in prevalence of prognostic factors and survival in lung
cancer patients from 1985 to 2004 at a tertiary care center. Cancer Detect Prev.
2008. 32(2): p. 101-8.

Kerr K.M., et al., The role of DNA methylation in the development and
progression of lung adenocarcinoma.Dis Markers. 2007. 23(1-2): p. 5-30.

Kersting, M., et al., Differential frequencies of p16INK4a promoter
hypermethylation, p53 mutation, and K-ras mutation in exfoliative material mark
the development of lung cancer in symptomatic chronic smokers. J. Clin. Oncol.,
2000. 18: p. 3221-3229.

116
Kim D.S., et al., Aberrant methylation of E-cadherin and H-cadherin genes in
nonsmall cell lung cancer and its relation to clinicopathologic features.Cancer.
2007. 110(12): p. 2785-92.

Kim J.S., et al., Aberrant methylation of H-cadherin (CDH13) promoter is
associated with tumor progression in primary nonsmall cell lung carcinoma.
Cancer. 2005. 104(9): p. 1825-33.

Kishi K., et al., Small lung tumors with the size of 1cm or less in diameter:
clinical, radiological, and histopathological characteristics.Lung Cancer. 2004.
44(1): p. 43-51.

Kohno T., et al., Association of KRAS polymorphisms with risk for lung
adenocarcinoma accompanied by atypical adenomatous
hyperplasias.Carcinogenesis. 2008. 29(5): p. 957-63.

Krassenstein R., et al., Detection of breast cancer in nipple aspirate fluid by CpG
island hypermethylation Clin Cancer Res. 2004. 10(1 Pt 1): p. 28-32.

Laird, P.W., The Power and the Promise of DNA Methylation Markers. Nature
Reviews Cancer, 2003. 3: p. 253-266.

Licchesi J.D., et al., Epigenetic alteration of Wnt pathway antagonists in
progressive glandular neoplasia of the lung.Carcinogenesis. 2008. 29(5): p. 895-
904.

Licchesi J.D., et al., Promoter hypermethylation of hallmark cancer genes in
atypical adenomatous hyperplasia of the lung. Clin Cancer Res. 2008. 14(9): p.
2570-8.

Liu Y., et al., Promoter methylation of RASSF1A and DAPK and mutations of K-
ras, p53, and EGFR in lung tumors from smokers and never-smokers.BMC
Cancer. 2007. 7: p. 74.

Liu Y., et al., Hypermethylation of p16INK4a in Chinese lung cancer patients:
biological and clinical implications.Carcinogenesis. 2003. 24(12): p. 1897-901.

Liu, N.S., et al., Adenocarcinoma of the lung in young patients; the M.D.
Anderson experience. Cancer, 2000. 88: p. 1837-1841

Lofton-Day C., et al., DNA methylation biomarkers for blood-based colorectal
cancer screening.Clin Chem. 2008. 54(2): p. 414-23.

117
Malila N., et al., Cancer incidence in a cohort of Finnish male smokers.Eur J
Cancer Prev., 2006. 15(2): p. 103-7.

Maneckjee R. and Minna J.D., Nonconventional opioid binding sites mediate
growth inhibitory effects of methadone on human lung cancer cells.Proc Natl
Acad Sci U S A. 1992. 89(4): p. 1169-73.

Maneckjee R. and Minna J.D., Opioid and nicotine receptors affect growth
regulation of human lung cancer cell lines.Proc Natl Acad Sci U S A. 1990. 87(9):
p. 3294-8.

Maneckjee R. and Minna J.D., Opioids induce while nicotine suppresses
apoptosis in human lung cancer cells.Cell Growth Differ. 1994. (10): p. 1033-40.

Marsit C.J., et al., Hypermethylation of RASSF1A and BLU tumor suppressor
genes in non-small cell lung cancer: implications for tobacco smoking during
adolescence. Int J Cancer. 2005. 114(2): p. 219-23.

Merlo A., et al., 5' CpG island methylation is associated with transcriptional
silencing of the tumour suppressor p16/CDKN2/MTS1 in human cancers.Nat
Med. 1995. 1(7): p. 686-92.

Molina R., et al., Tumor markers (CEA, CA 125, CYFRA 21-1, SCC and NSE) in
patients with non-small cell lung cancer as an aid in histological diagnosis and
prognosis. Comparison with the main clinical and pathological prognostic
factors.Tumour Biol. 2003. 24(4): p. 209-18.

Mountain, C.F., Revisions in the international system fro staging lung cancer.
Chest, 1997. 111: p. 1710-1717.

Nguyen C., et al., Susceptibility of nonpromoter CpG islands to de novo
methylation in normal and neoplastic cells. J Natl Cancer Inst. 2001. 93(19): p.
1465-72.

Niho, S., et al., Monoclonality of atypical adenomatous hyperplasia of the lung.
Am. J. Pathol., 1999. 154: p. 249-254.

Nonomura A., et al., Simultaneous detection of intercellular adhesion molecule-1
(CD54) and carcinoembryonic antigen in lung adenocarcinoma.Mod Pathol.
1994. 7(2): p. 155-60.

118
Okada M., et al., Prognostic significance of perioperative serum
carcinoembryonic antigen in non-small cell lung cancer: analysis of 1,000
consecutive resections for clinical stage I disease.Ann Thorac Surg. 2004. 78(1):
p. 216-21.

Otterson G.A., et al., CDKN2 gene silencing in lung cancer by DNA
hypermethylation and kinetics of p16INK4 protein induction by 5-aza
2'deoxycytidine.Oncogene. 1995. 11(6): p. 1211-6.

Palmisano, W.A., et al., Predicting lung cancer by detecting aberrant promoter
methylation in sputum. Cancer Res., 2000. 60: p. 5954-5958.

Papadopoulou E., et al., Cell-free DNA and RNA in plasma as a new molecular
marker for prostate and breast cancer.Ann N Y Acad Sci. 2006. 1075: p. 235-43.

Papadopoulou E., et al., Cell-free DNA and RNA in plasma as a new molecular
marker for prostate cancer.Oncol Res. 2004 14(9): p. 439-45.

Pinsky P.F., et al., Abnormalities on chest radiograph reported in subjects in a
cancer screening trial. Chest. 2006. 130(3): p. 688-93.

Pirozynski M., 100 years of lung cancer. Respir Med., 2006. 100(12): p. 2073-84.

Rauch T., et al., Homeobox gene methylation in lung cancer studied by genome-
wide analysis with a microarray-based methylated CpG island recovery
assay.Proc Natl Acad Sci U S A. 2007. 104(13): p. 5527-32.

Righini C.A., et al., Tumor-specific methylation in saliva: a promising biomarker
for early detection of head and neck cancer recurrence. Clin Cancer Res. 2007.
13(4): p. 1179-85.

Rosas S.L., et al., Promoter hypermethylation patterns of p16, O6-
methylguanine-DNA-methyltransferase, and death-associated protein kinase in
tumors and saliva of head and neck cancer patients. Cancer Res. 2001. 61(3): p.
939-42.

Sakamoto H., et al., Disproportionate representation of KRAS gene mutation in
atypical adenomatous hyperplasia, but even distribution of EGFR gene mutation
from preinvasive to invasive adenocarcinomas. J Pathol. 2007. 212(3): p. 287-94.

Salem C.E., et al., PAX6 methylation and ectopic expression in human tumor
cells. Int J Cancer. 2000. 87(2): p. 179-85.

119
Sartori G., et al., A subset of lung adenocarcinomas and atypical adenomatous
hyperplasia-associated foci are genotypically related: an EGFR, HER2, and K-
ras mutational analysis. Am J Clin Pathol. 2008. 129(2): p. 202-10.

Schneider J., et al., Comparison of the tumor markers tumor M2-PK, CEA,
CYFRA 21-1, NSE and SCC in the diagnosis of lung cancer. Anticancer Res.
2000. 20(6D): p. 5053-8.

Schneider J., et al., Fuzzy logic-based tumor-marker profiles improved sensitivity
in the diagnosis of lung cancer. Int J Clin Oncol. 2002. 7(3): p. 145-51.

Schottenfeld, D., Epidemiology of lung cancer, in Lung cancer: principles and
practice, H.I. Pass, et al., Editors. 1996, Lippincott-Raven Publishers:
Philadelphia. p. 305-321

Sellar G.C., et al., OPCML at 11q25 is epigenetically inactivated and has tumor-
suppressor function in epithelial ovarian cancer. Nat Genet. 2003. 34(3): p. 337-
43.

Sherr C.J., et al., Cancer cell cycles. Science. 1996. 274(5293): p. 1672-7.

Shih Y.L., et al., Promoter methylation of the secreted frizzled-related protein 1
gene SFRP1 is frequent in hepatocellular carcinoma. Cancer. 2006. 107(3): p.
579-90.

Shiraishi, M., et al., HOX gene clusters are hotspots of de novo methylation in
CpG islands of human lung adenocarcinomas. Oncogene, 2002. 21: p. 3659-
3662.

Shivapurkar N., et al., Differential methylation of a short CpG-rich sequence
within exon 1 of TCF21 gene: a promising cancer biomarker assay.Cancer
Epidemiol Biomarkers Prev. 2008. 17(4): p. 995-1000.

Shu, Y., et al., The ACIN1 gene is hypermethylated in early stage lung
adenocarcinoma. J. Thor. Oncol., 2006. 1(2): p. 160-167.

Skvortsova T.E., et al., Cell-free and cell-bound circulating DNA in breast
tumours: DNA quantification and analysis of tumour-related gene methylation. Br
J Cancer. 2006. 94(10): p. 1492-5.

120
Soh J., et al., Sequential molecular changes during multistage pathogenesis of
small peripheral adenocarcinomas of the lung.J Thorac Oncol. 2008. 3(4): p.
340-7.

Stoehr R., et al., Deletions of chromosome 8p and loss of sFRP1 expression are
progression markers of papillary bladder cancer. Lab Invest. 2004. 84(4): p. 465-
78.

Toyooka S., et al., The relationship between aberrant methylation and survival in
non-small-cell lung cancers.Br J Cancer. 2004. 91(4): p. 771-4.

Travis W.D., et al., United States lung carcinoma incidence trends: declining for
most histologic types among males, increasing among females. Cancer. 1996.
77(12): p. 2464-70.

Travis, W.D., L.B. Travis, and S.S. Devesa, Lung Cancer. Cancer, 1995. 75: p.
191-202.

Travis, W.D., T.V. Colby, and B. Corrin, Histological Typing of Lung Cancer and
Pleaural Tumors, in WHO International Histological Classification of Tumors, 3rd
edition. 1999, Springer: Berlin. p. 29.

Tsou J.A., et al., Identification of a panel of sensitive and specific DNA
methylation markers for lung adenocarcinoma.Mol Cancer. 2007. 29; p. 6:70.

Tsou, J.A., et al., DNA methylation analysis: a powerful new tool for lung cancer
diagnosis. Oncogene, 2002. 21: p. 5450-5461.

Virmani A., et al., Aberrant methylation of the cyclin D2 promoter in primary small
cell, nonsmall cell lung and breast cancers. Int J Cancer. 2003. 107(3): p. 341-5.

Virmani, A.K., et al., Hierarchical clustering of lung cancer cell lines using DNA
methylation markers. Cancer Epid. Biom. Prev., 2002. 11: p. 291-297.

Vucic E.A., et al., Epigenetics of cancer progression. Pharmacogenomics. 2008.
9(2): p. 215-34.

Waki, T., et al., Age-related methylation of tumor suppressor and tumor-related
gene: an analysis of autopsy samples. Oncogene, 2003. 22: p. 4128-4133.

121
Wang J., et al., Value of p16INK4a and RASSF1A promoter hypermethylation in
prognosis of patients with resectable non-small cell lung cancer.Clin Cancer Res.
2004. 10(18 Pt 1): p. 6119-25.

Weisenberger D.J., et al., Analysis of repetitive element DNA methylation by
MethyLight.Nucleic Acids Res. 2005. ;33(21): p. 6823-36.

Weisenberger D.J., et al., CpG island methylator phenotype underlies sporadic
microsatellite instability and is tightly associated with BRAF mutation in colorectal
cancer.Nat Genet. 2006. 38(7): p. 787-93.

Welch H.G., et al., Overstating the evidence for lung cancer screening: the
International Early Lung Cancer Action Program (I-ELCAP) study. Arch Intern
Med. 2007 167(21): p. 2289-95.

Wikman H. and Kettunen E., Regulation of the G1/S phase of the cell cycle and
alterations in the RB pathway in human lung cancer. Expert Rev Anticancer Ther.
2006. 6(4): p. 515-30.

Wilson D.O., et al., The Pittsburgh Lung Screening Study (PLuSS): Outcomes
within 3 years of a first CT Scan. Am J Respir Crit Care Med. 2008. [Epub ahead
of print]

Wisnivesky J.P., et al., Radiation therapy for the treatment of unresected stage I-
II non-small cell lung cancer. Chest., 2005. 128(3): p. 1461-7.

Yanagawa N., et al., Promoter hypermethylation of RASSF1A and RUNX3 genes
as an independent prognostic prediction marker in surgically resected non-small
cell lung cancers Lung Cancer., 2007. 58(1): p.131-8.

Yoshida, Y., et al., Mutations of the epidermal growth factor receptor gene in
atypical adenomatous hyperplasia and bronchioloalveolar carcinoma of the lung.
Lung Cancer, 2005. 50(1): p. 1-8.

Yoshimi, I., et al., A comparison of trends in the incidence rate of lung cancer by
histological type in the Osaka Cancer Registry, Japan, and in the Surveillance,
Epidemiology and End Results Program, USA. Jpn. J. Clin. Oncol., 2003. 33: p.
98-104.

Young J., et al., HPP1: a transmembrane protein-encoding gene commonly
methylated in colorectal polyps and cancers.Proc Natl Acad Sci U S A. 2001.
98(1): p. 265-70.

122
Yun Y.H., et al., Cigarette smoking and cancer incidence risk in adult men:
National Health Insurance Corporation Study.Cancer Detect Prev., 2005. 29(1):
p. 15-24.

Zhang H. and Cai B., The impact of tobacco on lung health in
China.Respirology., 2003. 8(1): p. 17-21.

Zilberman D. and Henikoff S., Genome-wide analysis of DNA methylation
patterns. Development. 2007. 134(22): p. 3959-65.

Abstract (if available)

Abstract Lung cancer is the leading cause of cancer-related death in the United States. It is estimated that more than 160,000 lung cancer patients will die of the disease in 2008. Early detection is key to improving patient survival

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Development of DNA methylation based biomarkers for the early detection of squamous cell lung cancer

PDF

DNA methylation changes in the development of lung adenocarcinoma

PDF

The role of DNA methylation in early detection and progression of pancreatic cancer

PDF

DNA methylation as a biomarker in human reproductive health and disease

PDF

CpG methylation profiling in lung cancer cell lines, tumors and non-tumors

PDF

DNA methylation markers for blood-based detection of small cell lung cancer in mouse models

PDF

Identification of DNA methylation markers in diffuse large B-cell lymphoma

PDF

Use of DNA methylation inhibitors for chemotherapy and chemoprevention of cancer

PDF

The expression of human carboxylesterases in normal tissues and cancer cell-lines

PDF

DNA hypermethylation: its role in colorectal tumorigenesis and potential clinical applications

PDF

DNA methylation groups determined by GATA5 gene methylation level are correlated with tumor subtype, sex, smoking status, and body mass index in esophageal and gastric adenocarcinoma

PDF

DNA methylation inhibitors and epigenetic regulation of microRNA expression

PDF

Limit of detection analysis for cell-free DNA methylation using targeted bisulfite sequencing

PDF

LINC00261 induces a G2/M cell cycle arrest and activation of the DNA damage response in lung adenocarcinoma

PDF

Modeling lung adenocarcinoma progression in vitro using immortalized human alveolar epithelial cells

PDF

Development of immunotherapy for small cell lung cancer using iso-aspartylated antigen

PDF

The kinetic study of engineered MBD domain interactions with methylated DNA: insight into binding of methylated DNA by MBD2b

PDF

Early detection of lung cancer by characterizing circulating rare cells using peripheral blood liquid biopsy

PDF

Developing a robust single cell whole genome bisulfite sequencing protocol to analyse circulating tumor cells

PDF

Defining the functional region of LINC00261 in lung adenocarcinoma

Asset Metadata

Creator Galler, Janice Soratorio (author)

Core Title Development of DNA methylation biomarkers as an early detection tool for human lung adenocarcinoma

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Biochemistry and Molecular Biology

Degree Conferral Date 2008-12

Publication Date 10/31/2008

Defense Date 08/29/2008

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

Tag adenocarcinomas,biomarkers,DNA methylation,early detection,lung cancer,OAI-PMH Harvest

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Laird-Offringa, Ite A. (committee chair), Rice, Judd C. (committee member), Siegmund, Kimberly D. (committee member), Yang, Allen S. R. (committee member)

Creator Email galler@usc.edu,janicegaller@hotmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1707

Unique identifier UC1419640

Identifier etd-Galler-2419 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-117489 (legacy record id),usctheses-m1707 (legacy record id)

Legacy Identifier etd-Galler-2419.pdf

Dmrecord 117489

Document Type Dissertation

Rights Galler, Janice Soratorio

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu