AIRWAY-BASED MARKERS

Bronchoscopically Procured Airway Specimens
for DNA and RNA and Protein Analyses It is clear
that if directly procured, histologically dysmorphic airway
specimens yield important clinical information, as discussed
elsewhere in this volume. Newer molecular modalities have
been applied to bronchial biopsy specimens and bronchoalveolar
lavage (BAL). For example, computer-assisted fluorescence
immunophenotyping and interphase cytogenetics as a tool for
the investigation of neoplasms has been applied to BAL in the
region of a nodule with provocative results.
However, it also appears that field carcinogenesis patterns
of transcriptome-wide gene expression can be obtained in the
major airways, which reflect the likelihood that a radiographically
detected peripheral lesion, far remote from the bronchoscope,
is a malignancy. 58,59 Given that cigarette smoke creates
a field of injury throughout the airway, Spira et al. 58,59 sought
to determine if gene expression in histologically normal largeairway
epithelial cells obtained at bronchoscopy from smokers
with suspicion of lung cancer could be used as a lung cancer
biomarker. Using gene-expression profiles from Affymetrix
HG-U133A microarrays, an 80-gene biomarker that distinguished
smokers with and without lung cancer was identified.
When the biomarker was tested on an independent test set, an
accuracy of 83% (80% sensitive, 84% specific) was obtained.
The biomarker profile had 90% sensitivity for stage I cancer
among all subjects. When cytopathology of lower airway cells
obtained at bronchoscopy was combined with the biomarker,
a 95% sensitivity and a 95% negative predictive value (NPV)
were obtained. These findings indicate that gene expression in
cytologically normal large-airway epithelial cells can serve as a
lung cancer biomarker, potentially owing to a cancer-specific
airway-wide response to cigarette smoke. This novel version of
the field cancerization concept is now being tested in less invasively
obtained airway specimens, such as exfoliated buccal and
nasal epithelium. Justification for such an investigation comes
from data that bronchial and nasal epithelium from nonsmokers
were most similar in gene expression when compared with other
epithelial and nonepithelial tissues and with several antioxidant,
detoxification, and structural genes being highly expressed in
both the bronchus and nose. Smoking had a similar effect on
gene expression in nasal epithelium as in the bronchus. 60 The
expression of several detoxification genes was commonly altered
by smoking in all three respiratory epithelial tissues, suggesting a
common airway-wide response to tobacco exposure
Sputum-Based Cytology Although significant skepticism
remains regarding the utility of standard sputum cytology
as a lung cancer detection tool, and after the Mayo lung
project and other NCI-sponsored studies yielded no overall
mortality benefit, it has been long known that proximal and
perhaps slower-growing lesions, such as that of squamous cell
histopathology, could be detected in expectorated sputum with
significant sensitivity (e.g., 30% to 50%). Additionally, automated,
computerized image analysis can augment yield. 61
Sputum for DNA-Based Markers Sputum-based detection
has been studied extensively for somatic genetic and
epigenetic alterations.
For somatic genetic mutations , credible studies have suggested
a plethora of k-ras and p53 mutations found in sputum that correlate,
albeit imperfectly, with lung cancer prevalence, in the casecontrol
context. 62,63 Of the more recent studies, one group has
reported a correlation of FISH-detected HYAL2 and FHIT deletions
in both tumor tissue and sputum in those with a smoking
history in both lung cancer cases and smoking controls. 64 Another
group examined three loci for loss of heterozygosity (LOH) and
microsatellite instability (MSI) within the FHIT locus in sputum,
with sensitivity of 55% and specificity of 82%. 65
DNA Methylation by MSP in Sputa DNA methylation
has been extensively studied in sputum for several presumed
tumor suppressor gene promoters. 66 Among recent studies,
Belinsky et al. 67 describe a case-control study nested within
a high-risk cohort consisting of smokers with 30 pack-year
smoking histories and spirometric airflow obstruction consistent
with COPD. Results suggested qualitative sputum methylation
increased with temporal proximity to the diagnosis, and with the
number of genes methylated progressively conferring increased
risk. Sensitivity and specificity of the 14-gene panel were both
64%. Methylation of three or more genes in sputum was associated
with a 6.5-fold increased risk for lung cancer within 18
months. In stage III lung cancer, a methylation panel in sputum
has been reported to reflect in the tumor itself, with 44% to 72%
PPV and 70% NPV. 32 In most instances, sputum performed
better than serum for the same gene panel (see Chapter 7).
A comprehensive study of quantitative MSP, using
methylation-sensitive probes and primers, was carried out by
Shivapurkar et al. 33 Eleven genes suggested to be silenced by
methylation were examined in lung tumors, adjacent nonmalignant
lung tissues, and sputum. Of all genes, 3-OST-2 ,
DCR-1 , and RASSF1A showed the highest levels in tumors and
the lowest in adjacent nonmalignant tissues. Peripheral blood
mononuclear cells (PBMCs) were uniformly nonmethylated at
the probe sites. For sputum, when 3-OST-2 , RASSF1A, mp16 ,
and APC were combined, significant discriminatory power in
distinguishing cases from controls was apparent (ROC [receiver
operating characteristic] AUC [area under curve] 0.8).
Sputum for RNA-Based Markers There are a few
reports of RNA-based sputum studies in lung cancer detection
that credibly identify the amplicon as RNA. Lacroix et
al. 68 and Jheon et al. 69 reported an ability to amplify several
transcripts (including preproGRP) from sputum, although
in neither study there were no real-time controls employed
to confirm the avoidance of contaminating DNA pseudogene
amplification.
Sputum for Protein-Based Markers Claims of protein
expression in sputum are more easily affirmed as credible
than that for RNA, yet there have been relatively few to
date. By example, the ribonucleoprotein has been suggested
as a sputum-based biomarker by immunocytochemistry, 70
although there has been little follow-up data. A recent study
correlating tumor and sputum telomerase activity found 68%
sensitivity and 90% specificity, for concurrent lung cancer by
sputum analyses, in a case-control context. 71
Exhaled Breath for Volatile Small Compounds
There is a considerable recent literature on the use of the gas
phase of exhaled breath to identify individual volatile components
or complex volatile mixtures that correlate with the
presence of lung cancer. Study design has typically been casecontrol,
such that predictive capacity for incident disease is
largely unknown.
In a very carefully executed study, Wehinger et al. 72 collected
tidal volume breathing mixed expiratory gas samples
into 3-L pounds per cubic foot (pcf ) bags, prior to any subject
diagnostic or therapeutic interventions. They used a proton
transfer reaction mass spectrometry (PTR-MS) approach
to exhaled gas analysis, which avoids preconcentration steps
otherwise required for gas chromatography-based techniques,
but cannot distinguish compounds of the same mass. Among
17 predominantly early-stage lung cancer cases and 170 controls,
mass-to-charge (m/z) 31 or volative organic compound
(VOC)-31 (tentatively protonated formaldehyde) and
VOC-43 (tentatively a protonated fragment of isopropanol)
were the most discriminatory; impressive twofold to threefold
differences of cases versus smoker controls, with some overlap,
were found. Overall test performance, in simulations, showed
sensitivity for detecting lung cancer was 54%, accuracy was
96%, specificity was 99%, PPV (given 5% prevalence) was
90%, and NPV was 96%. ROC curves showed AUCs for
those aged 50 years old of 0.82 and 0.95, respectively.
In some of the initial studies on volatile compounds in the
gas phase that correlate to lung cancer, in a case-control context,
Phillips et al. 73,74 reported significant correlations of gas
chromatography–coupled mass spectrometry (GC-MS) patterns
and the likelihood of having lung cancer. In one study, the
approach yielded a training set sensitivity of 90% and specificity
of 85% for lung cancer. In the test set, 83% sensitivity and
80% specificity was obtained. 74 Distinction from the exhaled
volatile patterns of active or former cigarette smokers without
apparent lung cancer was made. 74 Most recently, Phillips
et al. 75 measured VOCs in 1-L alveolar breath from 193 subjects
with primary lung cancer and 211 controls with a negative
chest CT. Subjects were randomly assigned to a training
set or to a prediction set in a 2:1 split. A fuzzy logic model of
breath biomarkers of lung cancer was constructed in the training
set and then tested in subjects in the prediction set by generating
their typicality scores for lung cancer. Mean typicality
scores employing a 16 VOC models were significantly higher
in lung cancer patients than in the control group ( p 0.0001
in all TNM stages). The model predicted primary lung cancer
with 84.6% sensitivity, 80.0% specificity, and 0.88 AUC of the
ROC curve. Predictive accuracy was similar in TNM stages I
through IV, and was not affected by current or former tobacco
smoking. The predictive model achieved near-maximal performance
with six breath VOCs and was progressively degraded
by random classifiers (Fig. 20.2). Predictions with fuzzy logic
were consistently superior to multilinear analysis. If applied to
a population with 2% prevalence of lung cancer, a screening
breath test would have a NPV of 0.985 and a PPV of 0.163
(true-positive rate 0.277, false-positive rate 0.029).
Another group reported 71% sensitivity and 92% specificity
for lung cancer detection in a case-control setting, using a commercialized
sensor array “electronic nose.” 76 In the sensor array
method, the actual volatile components of the unique signal are
not directly evaluated, although additional workup has revealed
them to be predominantly volatile hydrocarbons. Corroborative
later reports simplifying the sensor array have been reported, 77
although these patterns do not directly identify the responsible
biomarker compounds. A recent uncorroborated study on training
dogs, the ultimate biosensors, in study-blinded fashion, reports
an extraordinary performance in distinguishing exhaled
breath from those with lung cancer versus healthy controls (sensitivity
and specificity both 99%). 78 A recent report that lung
cancer cells in vitro evolve unique volatile metabolites in the gas
phase above the culture dish lends some credence to the cancer
specificity of the evolved exhaled gas detection approach.
For molecules in the condensate (aqueous fraction) of exhaled
breath, small molecules such as IL-2, tumor necrosis factor
(TNF- ), and leptin have been correlated to the presence of
clinically apparent lung cancer cases versus controls. 80 Notable
is that macromolecules directly reflective of carcinogenesis have
also been detected in exhaled breath condensate. It is not intuitively
obvious how large macromolecule markers of lung carcinogenesis,
such as DNA and proteins, can be suspended in the
breath, but there are multiple reports of this phenomenon from
several laboratories, including our own. It is possible that a fraction
of the components of bronchial lining fluid are suspended
merely by entrainment by adjacent high-velocity airflow, or alternatively,
alveolar lining fluid components are suspended by the
agitation inherent in tidal volume inflation and deflation and/or
the occasional opening snap of underinflated or atelectatic airspaces.
Whatever the mechanism, several groups have reported
the presence of DNA in exhaled breath, allowing DNA-based
markers such as specific p53 gene mutations, 81 microsatellite
markers, 82,83 or DNA sequence annotated at high resolution
for methylation. 84,85 In each case, a provocatively higher rate of
carcinogenesis-related DNA aberrancy was detected in the lung
cancer cases versus controls; the sensitivities and specificities are
in the 80% range. Whether these case-control correlations will
translate to meaningful predictive power for risk for lung cancer,
in prospective studies, remains unevaluated.
SURROGATE AIRWAY EPITHELIA
Exfoliated Upper Airway Cells Brush-exfoliated buccal
cells are transcriptionally active and potentially provide an easily
procured window on gene–tobacco interaction that might
be relevant to risk for tobacco-related malignancy. In one casecontrol
study, expression of a candidate gene set of carcinogen
metabolism pathway transcript in buccal epithelium was highly
correlated with those in lung. 86 Genome-wide studies from
other laboratories are underway for both buccal and nasal epithelia.
Additionally, there have been intermittent reports of nasal
epithelial access, in transcriptional and other studies, for gaining
easier access to respiratory epithelium.
There is an impressive array of innovative approaches being
developed for the early detection of lung cancer. Both optical
and molecular-based methods are under study, in case-control
discovery phase endeavors. Validation for a few of the most
promising markers, be they somatic DNA mutation, DNA
methylation, transcript assemblies, proteomic, or other markers,
are underway, although prospective validation is clearly
lacking. Rather than expect a single marker that is definitive
and sensitive for both screening and diagnosis, it might
be more practical to evolve a multistage approach employing
population-wide risk assessment, followed by early disease detection
approaches. The precise formulation of risk assessment
approaches, and even disease detection tools, is in rapid flux. It
is the high potential for public health impact of improvements
in lung cancer diagnosis that has attracted a very considerable
pool of innovations, strategies, and talent to focus on early detection
approaches to this lethal disease.