Gene expression profiling in head and neck squamous cell carcinoma: Clinical perspectives

Abstract

Background.To date, more than 60 gene expression profiling (GEP) studies have been published in the field of head and neck squamous cell carcinoma (HNSCC) with variable objectives, methods, and results.Methods.The purpose of this study was to present a state-of-the-art review of GEP in HNSCC focusing on the current advances and perspectives for clinical applications.Results.Gene expression signatures have been developed to identify screening and diagnostic molecular markers, to improve tumor staging (cervical lymph node and distant metastasis prediction), to differentiate lung metastasis of HNSCC from primary lung squamous cell carcinoma, to predict tumor response to chemoradiotherapy, and to provide outcome predictors.Conclusion.Some transcriptional signatures that could improve HNSCC management have been identified, but further analyses are required to properly validate and to precisely evaluate their clinical relevance. After an exploratory phase, the completion of large scale projects with stringent methodology is now necessary to transfer GEP from bench to bedside. © 2010 Wiley Periodicals, Inc. Head Neck, 2010

Full text

Page 1

Basic Science Review
Jeffrey N. Myers, MD, PhD, Section Editor
GENE EXPRESSION PROFILING IN HEAD AND NECK
SQUAMOUS CELL CARCINOMA: CLINICAL PERSPECTIVES
Benjamin Lallemant, MD, PhD,1,2,3,4Alexandre Evrard, MD, PhD,2,3,4,5Guillaume Chambon, MD,1
Omar Sabra, MD,1Sophie Kacha, MD,1,2,3,4Jean-Gabriel Lallemant, MD,1
Serge Lumbroso, MD, PhD,2,3,4,5Jean-Paul Brouillet, MD, PhD2,3,4,5
1Service d’ORL et Chirurgie maxillo-faciale, Centre Hospitalier Universitaire de Nı ˆmes,
Place du Pr. Robert DEBRE, 30029 Nı ˆmes CEDEX 9 France. E-mail: benjamin.lallemant@chu-nimes.fr
2Centre National de la Recherche Scientifique, Unite ´ Mixte de Recherche 5203, Institut de Ge ´nomique
Fonctionnelle, Montpellier F-34094, France
3Institut National de la Sante ´ et de la Recherche Me ´dicale, U661 Montpellier F-34094, France
4Universite ´ Montpellier 1, Montpellier F-34094, France
5Laboratoire de Biochimie, Centre Hospitalier Universitaire de Nı ˆmes, Place du Pr. Robert DEBRE,
30029 Nı ˆmes CEDEX 9 France
Accepted 21 April 2010
Published online 14 October 2010 in in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/hed.21491
Abstract:
expression profiling (GEP) studies have been published in the
field of head and neck squamous cell carcinoma (HNSCC)
with variable objectives, methods, and results.
Methods. The purpose of this study was to present a state-
of-the-art review of GEP in HNSCC focusing on the current
advances and perspectives for clinical applications.
Results. Gene expression signatures have been developed
to identify screening and diagnostic molecular markers, to
improve tumor staging (cervical lymph node and distant metasta-
sis prediction), to differentiate lung metastasis of HNSCC from pri-
mary lung squamous cell carcinoma, to predict tumor response
to chemoradiotherapy, and to provide outcome predictors.
Conclusion. Some transcriptional signatures that could
improve HNSCC management have been identified, but further
Background. Todate,morethan 60gene
analyses are required to properly validate and to precisely eval-
uate their clinical relevance. After an exploratory phase, the
completion of large scale projects with stringent methodology
is now necessary to transfer GEP from bench to bedside.
Wiley Periodicals, Inc. Head Neck 32: 1712–1719, 2010
V V
C 2010
Keywords: gene expression profiling; HNSCC; transcriptome;
biomarkers; microarray
Head
(HNSCC) is the fifth most frequent cancer world-
wide.1Its prognosis is globally poor, and curative
treatment implies heavy cosmetic and functional
consequences. The development of new tools for
the management of this disease is, therefore,
urgent. In response to this need, gene expression
profiling (GEP) is a biological approach that has
been extensively used with promising results in
cancer research. With GEP, it is possible to observe
and quantify gene expression deregulation in
andneck squamous cellcarcinoma
Additional Supporting Information may be found in the online version of
this article.
Correspondence to: B. Lallemant
V V
C 2010 Wiley Periodicals, Inc.
1712HNSCC Gene Expression ProfilingHEAD & NECK—DOI 10.1002/hed December2010

Page 2

tumor cells or their surrounding environment. Two
technologies are currently used: DNA microarrays,
which can extensively measure the expression of
thousands of genes, and reverse transcriptase-
quantitative polymerase chain reaction (RT-qPCR),
which provides a more accurate quantification of
the expression of a more limited number of tran-
scripts. Examples of GEP applications in cancer
research include elucidating the changes in the
pattern of gene expression involved in tumorigene-
sis and metastatic progression, identifying biologi-
cal targets for the development of new drugs,
identifying biological pathways involved in tumor
resistance to chemotherapy or radiotherapy, and
improving tumor staging and patient outcome pre-
diction. From a clinical point of view, GEP should
provide more accurate information about the dis-
ease and lead to more personalized and improved
treatment strategies.2
To date, more than 60 GEP studies of human
clinical samples of HNSCC have been published
with variable objectives, methods, and results.
We present a review of this literature focusing
on the current advances and perspectives for
clinical applications of GEP in the field of
HNSCC. We also discuss the appropriate exploi-
tation of this technology to provide efficient bio-
logical tools with the goal to improve the
management of HNSCC.
Gene
Squamous Cell Carcinoma Diagnosis. Due to its
ability to recognize molecular changes directly
related to certain phenotypes, DNA microarray
technology can be thought of as a new type of
increasingly precise microscope. Based on this
rationale, gene expression analysis has been
used to improve cancer diagnosis efficacy. This is
achieved first by identifying transcriptional bio-
markers characteristic of HNSCC cells using a
tumor versus normal tissue gene expression com-
parison; and second by testing the usefulness of
these biomarkers in different clinical situations.
ExpressionProfiles forHead andNeck
Using Gene Expression Profiles to Distinguish Head
and Neck Squamous Cell Carcinoma from Normal
Mucosa. In HNSCC, DNA microarrays were
mainly used to compare gene expression profiles
of HNSCC tissues versus normal matched mu-
cosa. From a fundamental point of view, these
comparisons aimed to unravel the gene regula-
tionand biologicalpathways
HNSCC carcinogenesis. From a clinical point of
view, this approach aimed to identify HNSCC-
implicated in
specific gene expression profiles that could be
used to detect cancerous cells in tissue.
To date, 41 HNSCC versus normal mucosa
GEP studies have been published (for references
see Supplementary file 1). Each study provides
a list of the most frequently and most highly
dysregulated genes in HNSCC tissue as com-
pared with normal mucosa. Yu et al3carried out
a genomic meta-analysis of these studies pub-
lished between 2000 and 2008 and found that
the overlap between the different lists of dysre-
gulated genes was low. The methodological vari-
ability between all these studies certainly plays
an important role in this discrepancy. For
instance, some authors included samples from
differentHNSCClocations,
focused on 1 site specifically; some analyzed
microdissectedtumor
others included the surrounding stroma; the
microarray platforms and the biostatistical anal-
yses that were selected in each study were
highly heterogeneous. Moreover, most studies
relied on a very limited number of samples and,
consequently, none of the HNSCC-specific tran-
scriptional signatures published to date could be
properly validated on large independent cohort
of patients. Despite these drawbacks, some
genes seem to be frequently misregulated and
are candidates as HNSCC-specific transcrip-
tional biomarkers that are worth being vali-
dated on larger independent cohorts of patients
and then being tested for diagnostic clinical
applications as presented below (Table 1).
whereasothers
epithelium,whereas
Clinical Perspectives for Head and Neck Squamous
Cell Carcinoma Transcriptional
Identification of Microscopic Head and Neck Squamous Cell
Biomarkers.
Carcinoma in Lymph Nodes or Surgical Margins.
specific transcriptional
used for the detection of microscopic tumor clus-
ters that are known to be easily overlooked by
conventional histopathological
lymph nodes or in surgical margins. In HNSCC,
micro-metastases in cervical lymph nodes are
responsible for cancer recurrence in the neck of
patients misclassified as N0 and who did not
receive a prophylactic treatment either by neck
dissection or radiotherapy. The detection of
micro-metastases in lymph nodes is, therefore,
no trivial matter, especially in sentinel lymph
nodes whose metastatic status will direct fur-
ther treatment decisions. As suggested by pre-
liminary results, transcriptional biomarkers or
gene expression signatures could improve micro-
HNSCC-
could biomarkersbe
methods in
HNSCC Gene Expression Profiling HEAD & NECK—DOI 10.1002/hed December 20101713

Page 3

metastases diagnostic efficacy in N0 patients
and improve the selection of patients deserving
treatment in the neck.4–6Similarly, local recur-
rence in patients treated by surgery alone may
be due to microscopic cancer cells in the resec-
tion margins overlooked by conventional micros-
copy. Preliminary studies suggest that HNSCC
transcriptional signatures could be used to
detect invaded surgical margins to better select
patients eligible for postoperative treatments.7,8
To date, this potential field of application of
GEP remains unexplored. The main reason is
that the identification and validation of reliable
HNSCC-specific transcriptional biomarkers or
signatures is a prerequisite that is lacking so
far. When such signatures are available, their
clinical value for the identification of micro-
scopic HNSCC tumor clusters in lymph nodes or
surgical margins may promptly be tested.
Identification of Head and Neck Squamous Cell
Carcinoma Cells in Bodily Fluids. During carcino-
genesis, some cancer cells migrate into the
bloodstream or are eliminated into natural cav-
ities such as the bladder, mouth, or intestine.
The genetic changes characteristic of the pres-
ence of degenerated cells (directly derived from
a tumor or in response to a tumor) may be iden-
tifiable in the physiological bodily fluids drain-
ing the affected organ.9
biomarkers representing genetic cancer altera-
tions in the serum, plasma, urine, semen, or sa-
liva of patients has already proven to be feasible
and potentially useful but remains a debated
and challenging approach, as shown by the fol-
lowing results.10–13
RecoveringmRNA
Blood Signatures for the Diagnosis of Head and
Neck Squamous Cell Carcinoma. Li et al12com-
pared the gene expression profiles in the blood
serum of 32 patients with primary T1–T2 N0
M0 oral squamous cell carcinoma (OSCC) and
35 control cases. Among a set of 5 overexpressed
genes, the authors found 2 genes (THSMB and
FTH1) whose combination reached a 91% sensi-
tivity and a 71% specificity for OSCC prediction.
The authors developed a second signature of 6
genes which gave an 84% sensitivity and an
83% specificity. Even if this single study demon-
strated the utility of serum gene GEP for
HNSCC diagnosis, to date, no further study or
validation of these
available.
preliminary dataare
Saliva Signatures for the Diagnosis of Head and
Neck Squamous Cell Carcinoma. HNSCC screen-
ing via diagnostic biomarker detection in saliva
is a promising field of research. To date, most
studies have focused on DNA alterations (muta-
tion, hypermethylation) and only recently on
mRNA detection.13Li et al12used a transcrip-
tome analysis on free-cell saliva. From 10
patients with primary OSCC and 10 control
cases, they identified a list of 1679 genes with
different expressionlevels
with cancer and controls. They validated a set
of 4 HNSCC-deregulated genes in saliva (IL1B,
OAZ1, SAT, and IL8) on a cohort of 32 patients
with OSCC and 32 controls (including those
tested for the transcriptome analysis). The asso-
ciation of these genes in a multivariate model
yielded a 91% sensitivity and 91% specificity for
the prediction of OSCC. A second predictive
model was also generated based on the combina-
tion of 3 genes (IL8, SAT, and H3F3A) that
reached a sensitivity of 90.6% and a specificity
of 90.6%. This study confirms that genetic
betweenpatients
Table 1. List of the most frequently reported dysregulated
genes in microarray studies comparing HNSCC
versus normal mucosa according Yu et al3meta-analysis of
41 transcriptome studies.
Gene
No. of articles
reporting
the gene
Chromosomal
coordinate
Sense of
regulation
in HNSCC
KRT4
KRT5
PLAU
FN1
MAL
MMP1
COL1A2
SPARC
POSTN
IFI6
TGM3
SPP1
ITGA6
KRT13
EMP1
ECM1
TNC
MMP10
MMP3
MMP12
LAMC2
IL8
KRT17
COL5A2
COL4A1
18
16
15
14
14
13
13
13
12
12
12
11
11
11
11
11
10
12q12
12q12
10q24
2q34
2cen-q13
11q22.3
7q22.1
5q31.3
13q13.3
1p35
20q11.2
4q21-q25
2q31.1
17q12-q21
12p12.3
1q21
9q33
11q22.3
11q22.3
11q22.3
1q25-q31
4q13-q21
17q12
2q14-q32
13q34
?
?
þ
þ
?
þ
þ
þ
þ
þ
?
þ
þ
?
?
?
þ
þ
þ
þ
þ
þ
þ
þ
þ
9
9
9
9
9
9
9
9
Abbreviations: HNSCC, head and neck squamous cell carcinoma; þ,
upregulated gene in HNSCC; ?, downregulated gene in HNSCC.
1714 HNSCC Gene Expression ProfilingHEAD & NECK—DOI 10.1002/hedDecember2010

Page 4

transcriptional changes characteristic of pri-
mary HNSCC can be detected in saliva washes
and that despite its supposed fragility in a me-
dium considered as inhospitable, mRNA-based
HNSCC diagnosis or screening biomarkers could
be developed, but confirmatory and validation
studies are still awaited.
Gene Expression Profile and Metastasis Prediction.
In HNSCC, metastasis is the single most impor-
tant predictor of survival.14,15When cervical
lymph node metastases are identified, more
aggressive treatments are proposed either by
surgery or radiotherapy. When distant metasta-
ses are present, the disease is considered incur-
able and palliative treatments are indicated.
Standard metastases screening methods such as
clinical and pathologic examination or morpho-
logic and functional imaging are challenged by
small metastases or micrometastases, which are
frequently overlooked
tools.16Assuming that specific genetic changes
underlie cancer cell capacity to metastasize, dif-
ferences in gene expression between metastatic
and nonmetastatic primary tumors may be
detected. Such molecular profiles could be used
to assess the metastatic potential of tumors on
tissue biopsy to improve TNM staging, resulting
in more appropriate treatments.
by thesediagnostic
Prediction
Fourteen studies have specifically compared
gene expression profiles of HNSCC primary
tumors with or without cervical lymph node me-
tastasis to find signatures that could predict
lymph node status (for references see Supple-
mentary file 2). Roepman et al17provided the
largest studies in which they compared 45 pNþ
versus 37 pN0 HNSCC tumors. They identified
a 102 predictor gene signature, which outper-
formed current diagnostic methods. On an inde-
pendent validation set of 22 samples, this
predictor reached a predictive accuracy of 100%
for N0 status and 77% for Nþ status. On a
cohort of 30 patients, Nguyen et al18identified
85 genes differentially expressed between pNþ
and pN0 tumors. Among these genes, they
selected a subset of 8 transcripts whose predic-
tive accuracy, using an RT-qPCR analysis, was
100% for N0 status and 86% for Nþ status on
an independent validation set of 13 patients.
These studies demonstrate that specific dif-
ferences in gene expression exist between Nþ
and N0 HNSCC and that these differences can
of Cervical Lymph NodeMetastasis.
be used to predict the N status on initial diag-
nostic biopsy. However, these promising results
were obtained on small cohorts with small inde-
pendent validation groups, or even worse, on in-
ternal cross-validation procedures. Hence, the
real clinical value of these predictors remains to
be evaluated. Moreover, the pathologic ‘‘gold
standard’’ used in these studies to define lymph
node status was probably suboptimal. Small me-
tastases and micrometastases could be over-
looked, and patients with pNþ subsequently
were misclassified as pN0. To avoid such mis-
classification, it is necessary to use stringent,
time-consuming, comprehensive pathologic ex-
amination of dissected cervical lymph nodes, at
least in levels I to IV. This highly demanding
methodological procedure is crucial for improv-
ing the reliability of future studies.
Prediction of Distant Metastasis. Four studies
have tried to find gene expression signatures
that could predict the development of distant
metastasis in HNSCC. Braakhuis et al19com-
pared 11 primary HNSCC which gave rise to
distant metastasis during a 3-year follow-up
versus 8 nonmetastatic control cases. They
found that only 150 of the 17,240 tested genes
were differentially expressed between the 2
groups. They found that no gene could be con-
sidered statistically significantly misregulated
and that no metastasis gene signature could be
found in HNSCC. Similarly, Cromer et al20com-
pared 11 nonmetastasizing and 15 metastasizing
hypopharyngeal carcinomas.
report a set of 164 genes potentially involved in
the acquisition of metastatic potential, they also
clearly specify that these genes could not be
used as a predictive metastatic signature. Car-
inci et al21and Giri et al22found comparably
poor results.
As suggested by Braakhuis et al,19HNSCC
differs from other tumor types such as prostate
or breast cancer, in which metastasis signatures
have been identified.23,24Cervical lymph nodes
seem to work as a filter for metastatic cells and
could be a prerequisite step for hematological
spreading. The acquisition by the primary tu-
mor of a metastatic genetic profile is not inevita-
bly associated with the occurrence of distant
metastasis, because this cervical immunologic
fence can prevent or delay their appearance.
This confusion factor likely hampers the identi-
fication of genetic changes
tumors involved in HNSCC metastasis. The
Althoughthey
withinprimary
HNSCC Gene Expression Profiling HEAD & NECK—DOI 10.1002/hedDecember 20101715

Page 5

failure of the described studies may also be
explained by insufficient sample numbers for
appropriate significance testing.
Using Gene Expression Profiles to Distinguish Head
and Neck Squamous Cell Carcinoma Metastasis
from Primary Lung Squamous Cell Carcinoma. In
patients with HNSCC, the presence of solitary
pulmonary nodules is not rare.25Knowledge of
whether this nodule is a second lung primary
tumor or an HNSCC metastasis is crucial for
treatment choice and prognosis. Unfortunately,
in the presence of a solitary pulmonary nodule
of squamous cell origin, the distinction between
primary lung squamous cell carcinoma (lung
SCC) and HNSCC metastasis is impossible. Two
studies have tried to address this issue using
gene expression profiling.
Using a set of 21 primary lung SCC and 31
primary HNSCC, Talbot et al26identified a 500-
gene classifier that correctly distinguished both
tumor groups. Subsequently, when they tested
this signature on a set of 12 squamous cell lung
lesions of unknown etiology, 11 samples were
classified into the lung SCC group and only 1
sample in the HNSCC metastasis group. For the
authors, although no ‘‘gold standard’’ is available,
these results perfectly matched their clinical and
pathologic preliminary opinion about the origin
of these lung lesions. Vachani et al27found a 10-
gene signature derived from 10 primary lung
SCC and 18 primary HNSCC. This signature cor-
rectly predicted the tumor origin with both a sen-
sitivity and specificity of 96%. When it was
applied to the 12 lung lesions previously tested
by Talbot et al,26it also segregated tissue as 11
primary lung SCC and 1 HNSCC metastasis,
with the advantage of relying on a far more lim-
ited number of genes.
These results suggest that it is possible to
identify the origin of histologically similar ma-
lignant lesions based on expression profiling.28
Nonetheless, the analyses of further squamous
cell lung samples paired with in-depth clinical
and/or pathologic characterization concerning
their supposed origin and careful long-term fol-
low-up is crucial for validating and improving
these preliminary signatures.
Gene
Chemotherapy and/or Radiotherapy. HNSCCs are
sensitive to chemotherapy and radiotherapy. For
decades, overwhelming efforts have been made
to determine the most efficient treatment regi-
ExpressionProfiles and Responses to
men for each specific stage of the disease based
on TNM evaluation. This traditional clinical
approach for treatment selection intrinsically
admits that a certain percentage of patients will
not respond to the treatment but will have to
bear its consequences despite its inefficacy.
Hence, the design and implementation of reli-
able treatment response predictors is necessary
to correctly assign appropriate treatments to
specific tumors. To this end, gene expression
profiles have yielded promising results for sev-
eral types of cancer such as breast adenocarci-
noma,bladdercancer,
carcinoma, leukemia, esophageal cancer, or os-
teosarcoma.29–34
The treatment sensitivity of
cancer cells is mediated by various genetic
mechanisms. Thus, the
expression profiles between responders and non-
responders may be identified. Gene expression
signatures can, therefore, be used as a clinical
test on diagnostic biopsies to predict the efficacy
of different treatments relative to tumor biology.
To date, only 3 studies have addressed this spe-
cific issue in relation to HNSCC.
On a cohort of 35 patients with laryngophar-
yngeal carcinoma treated by platinum-based
chemoradiotherapy (26 responders and 9 non-
responders), Ganly et al35identified 17 genes
correlated with locoregional response. Among
these genes, only 4 (MDM2, erb2, H-ras, and
VCAM-1) could be internally validated by RT-
qPCR as potential predictive genes. Pramana
etal36
analyzed 92
patients with advanced HNSCC treated by cis-
platin-basedchemoradiation.
were not able to find a robust predictive classi-
fier. Last, Dumur et al37identified 6 algorithms
based on a 142 probe set that could differentiate
complete responders to chemotherapy or radio-
therapy (8 patients) from nonresponders (6
patients) with an accuracy of prediction ranging
from 93% to 100% on a validation set of 5 addi-
tional patients.
Although these studies provide support for
the use of gene expression profiles in predicting
treatment responses, their results are disap-
pointing. Despite the analysis of a significant
number of samples, no relevant predictive signa-
ture could be found. As suggested by Pramana
et al,36it is possible that biologic differences
between tumors not directly related to treat-
ment response dominate over the fewer and less
obvious differences correlated with chemora-
diation efficacy,which
non-small celllung
differences in gene
diagnosticbiopsies of
These authors
in turnmakes the
1716HNSCC Gene Expression Profiling HEAD & NECK—DOI 10.1002/hedDecember 2010

Page 6

identification of predictive signatures very chal-
lenging. An additional confounding factor could
be the multimodality of the treatments used
(chemotherapyplus
implies 2 distinct biological
action. Ongoing studies that concentrate on the
development of predictive response signatures
for radiotherapy alone or chemotherapy alone
are more likely to be successful. The identifica-
tion of treatment response signatures seems to
be particularly challenging. Nonetheless, this
application of gene expression profiling is of
great clinical interest. Such predictors should be
developed within large-scale therapeutic studies
in which the identification of predictive biologic
factors is performed in parallel with the efficacy
evaluation of the treatment itself.
radiotherapy),which
mechanisms of
Gene
Squamous Cell Carcinoma Outcomes. In HNSCC,
no recurrence and/or survival prediction bio-
markers are available so outcome prediction still
relies onclinical predictors
stage.38Recently, gene expression profiling has
been investigated for outcome prediction to help
clinicians to better stratify patients according to
tumor aggressiveness, to define prognosis, and
to subsequently modulate treatment intensity.
Belbin et al39published the first study that
reported the possible use of gene expression
profiling for outcome prediction in HNSCC.
They identifieda375-gene
divides the 17 patients of their cohort into 2
groups with slightly different (but not statisti-
cally significant) survivals. On a cohort of 60
patients, Chung et al40identified a 582-gene set
that distinguished patients with a high-risk or
low-risk of local recurrence. In another study,
Chung et al41identified a set of 75 genes pre-
dictingsurvival without
et al42developed a 99-gene ‘‘hypoxia signature’’
derived from the profiling of 59 HNSCC tumors
and 11 normal mucosae. When they tested this
signature on the dataset of 60 patients provided
by Chung et al,41they demonstrated that this
‘‘hypoxia signature’’ was an independent prog-
nostic factor of recurrence-free survival outper-
forming the original intrinsic signature initially
proposed by Chung et al.40Ginos et al43identi-
fied a signature associated with recurrent dis-
ease reinforced by genes involved in tumor
invasion and metastasis.
To date, the ability of GEP to provide effec-
tive outcome predictor remains questionable and
Expression Profiles and HeadandNeck
such asTNM
signature that
recurrence.Winter
seems to be very challenging. Many attempts
have been made in different tumor types but
they mostly provided only surrogates of conven-
tional clinical factors with very variable profiles
and accuracies.44,45In HNSCC, global survival
is influenced by several clinical and biological
confounding factors and is unlikely to be pre-
dicted by a mere transcriptional tumor analysis.
In fact, outcome is a general term that encom-
passes many aspects of the evolution and man-
agement of the disease. In the future, it may be
more fruitful to focus GEP studies on more pre-
cise biologically driven tumor behaviors such as
metastases prediction or treatment response
prediction, which in turn may indirectly provide
relevant outcome predictors. Again, such studies
should be carried out on selected cohort of
patients with similar clinical characteristics (eg,
tumor site, TNM stage, tumor differentiation,
and grade), but with clearly different and well-
documented evolutions of their disease.
DISCUSSION
As presented in this review, GEP gave promis-
ing exploratory results in a large spectrum of
clinical applications, but none have been appro-
priately validated. Unfortunately, gene expres-
sionprofileshave still
clinically relevant application for the manage-
ment of HNSCC. Indeed, several methodological
issues remain to be addressed, some specifics for
each clinical issues and some related to the GEP
approach itself. To advance the field, future
gene expression studies should adhere to the fol-
lowing guidelines to overcome some of these
drawbacks: (1) the clinical aim of GEP should
be well-defined and constitute the primary end-
point of the study; (2) why the GEP approach
has been chosen to answer one clinical issue
should be clearly stated; (3) the clinical and path-
ologic characterization and the quality control of
samples must follow a stringent and well stand-
ardized methodology46; (4) GEP studies should
focus on 1 specific HNSCC anatomic subsite and
tobacco and/or alcohol consumers and patients
infected with human papillomavirus should be
studied separately; (5) different statistical meth-
ods can be used to analyze GEP results because
each method has strengths and weaknesses and
may reveal different but still biologically impor-
tant insights. However, the statistical choices for
the analyses should be made a priori and not
adapted to the initial microarray results; and (6)
notprovidedany
HNSCC Gene Expression Profiling HEAD & NECK—DOI 10.1002/hedDecember 2010 1717

Page 7

appropriate validation procedures on large inde-
pendent cohorts are required to confirm gene
expression signature validity.
Inaddition tothese
improvement of microarray technologies (eg,
next-generation high-throughput RNA sequenc-
ing technologies47and new analytical statistical
approaches (eg, integrative knowledge-based net-
work3) will, in the future, accelerate the develop-
ment of clinical applications of GEP. After an
exploratory phase in HNSCC, the completion of
large-scale prospective, multi-institutional clini-
cal trials with stringent methodology is now nec-
essary.This couldbe
designing specific GEP studies or by coupling
GEP to ongoing clinical studies which would gen-
erate large sample sets of well-annotated speci-
mens that are lacking so far. This approach has
been used successfully in breast cancer in which
clinically relevant gene expression signatures
have been developed for different clinical appli-
cations.48Three of these signatures have been
thoroughly validated and are now commercially
available (Oncotype DX, MammaPrint, and Can-
cerTYPE ID). The development of these new bio-
logical tools was a long and tough process that
required the implication of large cooperative
international groups and the support of the
industry. If these examples illustrate the clinical
applicability of the GEP approach in the man-
agement of cancer in general, they also underline
the huge difficulties that accompany the transfer
of this technology from bench to bedside.
recommendations,
achievedeitherby
REFERENCES
1. Sankaranarayanan R, Masuyer E, Swaminathan R, Fer-
lay J, Whelan S. Head and neck cancer: a global perspec-
tive on epidemiology and prognosis. Anticancer Res 1998;
18:4779–4786.
2. van’t Veer LJ, Bernards R. Enabling personalized cancer
medicine through analysis of gene-expression patterns.
Nature 2008;452:564–570.
3. Yu YH, Kuo HK, Chang KW. The evolving transcriptome
of head and neck squamous cell carcinoma: a systematic
review. PLoS One 2008;3:e3215.
4. Garrel R, Dromard M, Costes V, et al. The diagnostic ac-
curacy of reverse transcription-PCR quantification of cy-
tokeratin mRNA in the detection of sentinel lymph node
invasion in oral and oropharyngeal squamous cell carci-
noma: a comparison with immunohistochemistry. Clin
Cancer Res 2006;12:2498–2505.
5. Xu Y, Lefe `vre M, Pe ´rie ´ S, et al. Clinical significance of
micrometastases detection in lymph nodes from head
and neck squamous cell carcinoma. Otolaryngol Head
Neck Surg 2008;139:436–441.
6. Nieuwenhuis EJ, Leemans CR, Kummer JA, Denkers F,
Snow GB, Brakenhoff RH. Assessment and clinical sig-
nificance of micrometastases in lymph nodes of head and
neck cancer patients detected by E48 (Ly-6D) quantita-
tive reverse transcription-polymerase chain reaction.
Lab Invest 2003;83:1233–1240.
7. Eissa S, Kenawy G, Moteleb FA, El-Magkhzangy AN,
Nassar M. Real-time PCR hTERT mRNA pattern in tu-
mor core, edge, resection margin, and lymph nodes in la-
ryngeal tumors: relation to proliferative index and
impact on prognosis. Clin Biochem 2005;38:873–878.
8. Santhi WS, Sebastian P, Varghese BT, Prakash O, Pillai
MR. NF-kappaB and COX-2 during oral tumorigenesis
and in assessment of minimal residual disease in surgi-
cal margins. Exp Mol Pathol 2006;81:123–130.
9. Sidransky D. Nucleic acid-based methods for the detec-
tion of cancer. Science 1997;278:1054–1059.
10. Juusola J, Ballantyne J. Messenger RNA profiling: a pro-
totype method to supplant conventional methods for body
fluid identification. Forensic Sci Int 2003;135:85–96.
11. Johnson PJ, Lo YM. Plasma nucleic acids in the diagno-
sis and management of malignant disease. Clin Chem
2002;48:1186–1193.
12. Li Y, Elashoff D, Oh M, et al. Serum circulating human
mRNA profiling and its utility for oral cancer detection.
J Clin Oncol 2006;24:1754–1760.
13. Li Y, St John MA, Zhou X, et al. Salivary transcriptome
diagnostics for oral cancer detection. Clin Cancer Res
2004;10:8442–8450.
14. de Bree R, Deurloo EE, Snow GB, Leemans CR. Screen-
ing for distant metastases in patients with head and
neck cancer. Laryngoscope 2000;110(3 Pt 1):397–401.
15. Mamelle G, Pampurik J, Luboinski B, Lancar R, Lusin-
chi A, Bosq J. Lymph node prognostic factors in head
and neck squamous cell carcinomas. Am J Surg 1994;
168:494–498.
16. Hermanek P, Sobin LH, Wittekind C. How to improve the
present TNM staging system. Cancer 1999;86:2189–2191.
17. Roepman P, Wessels LF, Kettelarij N, et al. An expres-
sion profile for diagnosis of lymph node metastases from
primary head and neck squamous cell carcinomas. Nat
Genet 2005;37:182–186.
18. Nguyen ST, Hasegawa S, Tsuda H, et al. Identification
of a predictive gene expression signature of cervical
lymph node metastasis in oral squamous cell carcinoma.
Cancer Sci 2007;98:740–746.
19. Braakhuis BJ, Senft A, de Bree R, et al. Expression
profiling and prediction of distant metastases in head
and neck squamous cell carcinoma. J Clin Pathol 2006;
59:1254–1260.
20. Cromer A, Carles A, Millon R, et al. Identification of
genes associated with tumorigenesis and metastatic
potential of hypopharyngeal cancer by microarray analy-
sis. Oncogene 2004;23:2484–2498.
21. Carinci F, Lo Muzio L, Piattelli A, et al. Potential markers
of tongue tumor progression selected by cDNA microarray.
IntJImmunopatholPharmacol2005;18:513–524.
22. Giri U, Ashorn CL, Ramdas L, et al. Molecular signa-
tures associated with clinical outcome in patients with
high-risk head-and-neck
treated by surgery and radiation. Int J Radiat Oncol
Biol Phys 2006;64:670–677.
23. Ramaswamy S, Tamayo P, Rifkin R, et al. Multiclass
cancer diagnosis using tumor gene expression signa-
tures. Proc Natl Acad Sci U S A 2001;98:15149–15154.
24. van’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene
expression profiling predicts clinical outcome of breast
cancer. Nature 2002;415:530–536.
25. Ferlito A, Shaha AR, Silver CE, Rinaldo A, Mondin V. Inci-
dence and sites of distant metastases from head and neck
cancer.ORLJOtorhinolaryngolRelatSpec2001;63:202–207.
26. Talbot SG, Estilo C, Maghami E, et al. Gene expression
profiling allows distinction
squamouscell carcinoma
between primaryand
1718 HNSCC Gene Expression ProfilingHEAD & NECK—DOI 10.1002/hedDecember2010

Page 8

metastatic squamous cell carcinomas in the lung. Cancer
Res 2005;65:3063–3071.
27. Vachani A, Nebozhyn M, Singhal S, et al. A 10-gene clas-
sifier for distinguishing head and neck squamous cell
carcinoma and lung squamous cell carcinoma. Clin Can-
cer Res 2007;13:2905–2915.
28. Bhattacharjee A, Richards WG, Staunton J, et al. Classi-
fication of human lung carcinomas by mRNA expression
profiling reveals distinct adenocarcinoma subclasses.
Proc Natl Acad Sci U S A 2001;98:13790–13795.
29. Zembutsu H, Suzuki Y, Sasaki A, et al. Predicting
response to docetaxel neoadjuvant chemotherapy for
advanced breast cancers through genome-wide gene
expression profiling. Int J Oncol 2009;34:361–370.
30. Ohno R, Nakamura Y. Prediction of response to imatinib
by cDNA microarray analysis. Semin Hematol 2003;40(2
Suppl 2):42–49.
31. Kakiuchi S, Daigo Y, Ishikawa N, et al. Prediction of
sensitivity of advanced non-small cell lung cancers to
gefitinib (Iressa, ZD1839). Hum Mol Genet 2004;13:
3029–3043.
32. Ochi K, Daigo Y, Katagiri T, et al. Prediction of response
to neoadjuvant chemotherapy for osteosarcoma by gene-
expression profiles. Int J Oncol 2004;24:647–655.
33. Takata R, Katagiri T, Kanehira M, et al. Predicting
response to methotrexate, vinblastine, doxorubicin, and
cisplatin neoadjuvant chemotherapy for bladder cancers
through genome-wide gene expression profiling. Clin
Cancer Res 2005;11:2625–2636.
34. Kihara C, Tsunoda T, Tanaka T, et al. Prediction of sen-
sitivity of esophageal tumors to adjuvant chemotherapy
by cDNA microarray analysis of gene-expression profiles.
Cancer Res 2001;61:6474–6479.
35. Ganly I, Talbot S, Carlson D, et al. Identification of
angiogenesis/metastases genes predicting chemoradio-
therapy response in patients with laryngopharyngeal
carcinoma. J Clin Oncol 2007;25:1369–1376.
36. Pramana J, Van den Brekel MW, van Velthuysen ML,
et al. Gene expression profiling to predict outcome after
chemoradiation in head and neck cancer. Int J Radiat
Oncol Biol Phys 2007;69:1544–1552.
37. Dumur CI, Ladd AC, Wright HV, et al. Genes involved
in radiation therapy response in head and neck cancers.
Laryngoscope 2009;119:91–101.
38. Chang SS, Califano J. Current status of biomarkers in
head and neck cancer. J Surg Oncol 2008;97:640–643.
39. Belbin TJ, Singh B, Barber I, et al. Molecular classifica-
tion of head and neck squamous cell carcinoma using
cDNA microarrays. Cancer Res 2002;62:1184–1190.
40. Chung CH, Parker JS, Karaca G, et al. Molecular classi-
fication of head and neck squamous cell carcinomas
using patterns of gene expression. Cancer Cell 2004;5:
489–500.
41. Chung CH, Parker JS, Ely K, et al. Gene expression pro-
files identify epithelial-to-mesenchymal transition and
activation of nuclear factor-kappaB signaling as charac-
teristics of a high-risk head and neck squamous cell car-
cinoma. Cancer Res 2006;66:8210–8218.
42. Winter SC, Buffa FM, Silva P, et al. Relation of a hy-
poxia metagene derived from head and neck cancer to
prognosis of multiple cancers. Cancer Res 2007;67:3441–
3449.
43. Ginos MA, Page GP, Michalowicz BS, et al. Identification
of a gene expression signature associated with recurrent
disease in squamous cell carcinoma of the head and
neck. Cancer Res 2004;64:55–63.
44. De Preter K, Vermeulen J, Brors B, et al. Accurate out-
come prediction in neuroblastoma across independent
data sets using a multigene signature. Clin Cancer Res
2010;16:1532–1541.
45. Sun Z, Yang P. Gene expression profiling on lung cancer
outcome prediction: present clinical value and future
premise. Cancer Epidemiol Biomarkers Prev 2006;15:
2063–2068.
46. Wilkes T, Laux H, Foy CA. Microarray data quality—
review of current developments. Omics 2007;11:1–13.
47. Marguerat S, Wilhelm BT, Ba ¨hler J. Next-generation
sequencing: applications beyond genomes. Biochem Soc
Trans 2008;36(Pt 5):1091–1096.
48. Dunn L, Demichele A. Genomic predictors of outcome
and treatment response in breast cancer. Mol Diagn
Ther 2009;13:73–90.
HNSCC Gene Expression ProfilingHEAD & NECK—DOI 10.1002/hed December 20101719