Published Ahead of Print on September 8, 2015 as 10.1200/JCO.2015.62.

1086
The latest version is at http://jco.ascopubs.org/cgi/doi/10.1200/JCO.2015.62.1086

JOURNAL OF CLINICAL ONCOLOGY R E V I E W A R T I C L E

Genetic Landscape of Human Papillomavirus–Associated

Head and Neck Cancer and Comparison to
Tobacco-Related Tumors
D. Neil Hayes, Carter Van Waes, and Tanguy Y. Seiwert
D. Neil Hayes, The University of North
Carolina, Chapel Hill, NC; Carter Van A B S T R A C T
Waes, National Institute on Deafness
and Other Communication Disorders, Head and neck cancer is the fifth most common cancer worldwide. It is often amenable to curative
Bethesda, MD; and Tanguy Y. Seiwert, intent therapy when localized to the head and neck region, but it carries a poor prognosis when it
The University of Chicago, Chicago, IL. is recurrent or metastatic. Therefore, initial treatment decisions are critical to improve patient
Published online ahead of print at survival. However, multimodality therapy used with curative intent is toxic. The balance between
www.jco.org on September 8, 2015. offering intensive versus tolerable and function-preserving therapy has been thrown into sharp
Supported by National Cancer Institute relief with the recently described epidemic of human papillomavirus–associated head and neck
Grants No. U24 CA143848, U24 squamous cell carcinomas characterized by improved clinical outcomes compared with smoking-
CA143848-02S1, and U10 CA181009- associated head and neck tumors. Model systems and clinical trials have been slow to address the
01, and by the Bobby F. Garrett Fund
clinical questions that face the field to date. With this as a background, a host of translational
for Head and Neck Cancer Research (to
D.N.H.). C.V.W. is supported by
studies have recently reported the somatic alterations in head and neck cancer and have
National Institute on Deafness and highlighted the distinct genetic and biologic differences between viral and tobacco-associated
Other Communication Disorders Intra- tumors. This review seeks to summarize the main findings of studies, including The Cancer
mural Projects No. ZIA-DC-000016, Genome Atlas, for the clinician scientist, with a goal of leveraging this new knowledge toward the
ZIA-DC-000073, and ZIA-DC-000074. betterment of patients with head and neck cancer.
Authors’ disclosures of potential
conflicts of interest are found in the J Clin Oncol 33. © 2015 by American Society of Clinical Oncology
article online at www.jco.org. Author
contributions are found at the end of
this article. the Surveillance, Epidemiology, and End Results
INTRODUCTION
Corresponding author: D. Neil Hayes,
(SEER) program, do not routinely record it. Adding
MD, The University of North Carolina, Head and neck squamous cell carcinoma (HNSCC) to the knowledge gap is a lack of consensus of diag-
Chapel Hill, Lineberger Comprehensive
is the fifth most common nonskin cancer world- nostic reagents to detect HPV. Variability in the
Cancer Center, CB 7295, Chapel Hill,
NC 27599; e-mail: hayes@med.unc.edu. wide; the annual incidence is 600,000 cases, and assays used across studies limits generalizability and
50,000 cases are diagnosed annually in the United comparability of results.11,12
© 2015 by American Society of Clinical
Oncology States.1,2 Although HNSCC historically was viewed Although morphologic characteristics, such as
as a tobacco- and alcohol-related cancer, infection basal squamous histology, poor differentiation, and
0732-183X/15/3399-1/$20.00
with high-risk human papillomaviruses (HPVs) absent keratin formation, are associated with HPV
DOI: 10.1200/JCO.2015.62.1086
during the past decade has emerged as a risk factor infection, the diagnosis of HPV is made through
for a substantial fraction of HNSCCs.3 The so-called molecular testing. A mystery of the disease is that no
HPV epidemic is most prominent in patients of precursor lesion on which to screen patients during
Northern European and North American origin, the long latency has been reported, despite the as-
whereas other areas of the world still see few HPV- sumption that viral exposure occurs via sexual con-
associated tumors.4-6 In studies that include an as- tact in young adulthood. Screening HPV molecular
sessment of HPV status, a shift in the traditional risk assays, including polymerase chain reaction– based
factors is nearly universally noted: viral-associated techniques, risk false-positive detection of endemic
tumors are seen in younger populations that have viral infections not associated with cancer, whereas
lymph node–positive oropharyngeal carcinomas specific in situ hybridization (ISH) assays often have
and lower rates of smoking.7 Having said this, most lower sensitivity.13 A useful proxy for HPV infection
patients with HPV-associated oropharyngeal carci- in tumors is by immunohistochemistry (IHC) as-
noma are still men in their 50s who have a smoking sessment of p16, the protein product of the gene
history, and true nonsmokers constitute only a CDKN2A.14 As shown in Figure 1, p16 is a repressor
minority.1,3,8-10 Unfortunately, a precise character- of the D cyclins that, in turn, partially phosphorylate
ization of patients with HNSCC by HPV status is the retinoblastoma tumor-suppressor protein (RB1).
challenging. First, because HPV is not yet part of the In the setting of viral infection, expression of the E7
staging system in HNSCC, tumor registries, such as viral oncoprotein mediates degradation of RB1, and

© 2015 by American Society of Clinical Oncology 1

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
Copyright 2015 by American Society of Clinical Oncology
 Hayes, Van Waes, and Seiwert

E7
G0
M

E2F1 p16
RB
E6 Fig 1. Role of E6 and E7 in the cell cycle
G1 CyclinD pathways and gene alterations as a func-
G2
tion of human papillomavirius (HPV) tumor
CDK4/CDK6 status. Genes altered in HPV-positive or
viral oncogenes (red) or in HPV-negative
tumors (black). RB, retinoblastoma tumor-
p53 suppressor protein.
p
p CyclinE
S

RB CDK2
p21

E2F1

pathway feedback results in overexpression of the cell senescence and/or ISH was seen, although analysis was limited because of small
pathways, including p16.15 The detection of p16 is a proxy for RB1 loss numbers and the use of p16 IHC for nonoropharynx tumors.24 The
through any mechanism, including mutations of RB1 that are not phase III trial of cetuximab in combination with radiation favored
mediated by E7 or HPV. Therefore, p16 detection can result in the outcomes in patients with oropharynx tumors in the cetuximab treat-
false assumption that E7 is expressed and, by extension, that the tumor ment arm.18 In a retrospective subset analysis, patients with p16-
is HPV positive.16 For instances of high pretest probability for HPV, positive and p16-negative oropharynx tumors benefited from
such as in squamous cell carcinomas of the oropharynx, the true- cetuximab.25 However, analysis that used HPV nucleic acid testing by
positive rate for p16 as an indication of HPV is high.14 When the ISH did not show a clear benefit, which suggests that p16 expression,
pretest probability is lower, such as in the oral cavity, the true-positive rather than HPV status, may be important. In short, controversy
rate decreases to 10%-41%, which renders p16 IHC an ineffective remains as to whether the combination of cetuximab with radiation
diagnostic tool. If it is assumed that 5% of nonoropharynx tumors seems to benefit patients with HPV-positive oropharyngeal cancer. By
are HPV positive, as many as 1,500 cases per year outside the orophar- contrast, the data are stronger that patients who are p16 positive for
ynx may be difficult to assess with p16. Ultimately, direct tests of E6 HPV may benefit, although such patients include some HPV-
and E7 may prove superior, but these are not yet routine.17 negative, yet p16-positive, tumors. At a molecular level, published
Although the demographic shifts may be modest between pa- studies suggest an anticorrelation between staining for the EGFR re-
tients who are HPV positive and HPV negative, clinical outcomes are ceptor and benefit of cetuximab.20 HPV-positive tumors have gener-
dramatically different. Although patients with HPV commonly pres- ally shown low or absent levels of EGFR protein expression or EGFR
ent with advanced nodal disease and clinical stage, prognosis is favor- gene amplification.1,10 Unfortunately, preclinical models fail to clarify
able compared with patients without HPV.3 Improved outcomes are the picture, because HPV-positive cell lines of HNSCC remain in early
seen across treatment modalities, including chemotherapy, radiation, development. Ongoing trials, such as Radiation Therapy Oncology
chemoradiotherapy, and potentially also surgery, which have led to Group study RTOG 1016 (NCT01302834), may clarify some contro-
proposals to incorporate HPV status in staging.9 Although outcomes versies about cetuximab. Other studies leverage the dramatically
are better, the implications for selection of patient therapy by HPV improved outcomes for patients with HPV-positive tumors by de-
status are not yet clear, because few existing clinical trials assessed HPV escalating therapy and potentially decreasing toxicity.26
status.18-20 Controversies are most pronounced in molecularly targeted In parallel with lessons learned by clinical trials, rapidly emerg-
therapy, for which little conclusive published data exist, although retro- ing advances in tumor characterization, including DNA sequencing,
spective studies have been presented in scientific meetings. Two ran- RNA sequencing, miRNA characterization, epigenetic characteriza-
domized, phase III studies that used anti– epidermal growth factor tion, and proteomics, are in sight and have been widely reviewed
receptor (EGFR) therapy either combined with chemotherapy or as a elsewhere.1,27-34 Data from large cooperative endeavors, such as The
single agent versus methotrexate suggest less activity of EGFR agents in Cancer Genome Atlas (TCGA), and individual efforts, such as the
p16-positive tumors.21,22 An additional phase II study indicated a 0% Chicago HNC Genomics cohorts, have illuminated the genome of
response rate for single-agent anti-EGFR therapy in HPV-associated HNSCC in a manner that reveals the most commonly altered genes
HNSCC according to highly accurate E6/E7 RNA-based HPV test- and coordinated patterns of genome alteration, which could be lever-
ing.23 Conversely, in the EXTREME (Erbitux in First-Line Treatment aged to accelerate our understanding of the disease, improve person-
of Recurrent or Metastatic Head and Neck Cancer) study, a nonsig- alization of existing therapies to increase effectiveness, and propose
nificant trend toward benefit in patients who were positive by p16 accelerated strategies for novel therapeutics.1,10

2 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Genetic Landscape of Head and Neck Cancer

positive more so than HPV-negative tumors display a clear prom-

1.0
HPV+ PIK3CA inence of amplification of 3q at the locus for the squamous lineage
HPV− SOX2
CCND1 transcription factors SOX2, TP63, and the oncogene PIK3CA.
E2F1 PIK3CA activation promotes activation of cell cycle, metabolism,
Mean DNA Copy No.

0.5
EGFR and transcription factors that regulate diverse genes that promote
the malignant phenotype.1,2,41
(log ratio)

When cell cycle signaling is considered (Fig 1), statistically signif-

icant copy number changes in at least three genes is remarkable. As
0 expected, chromosome 9p, the locus for the gene CDKN2A which
produces the protein product p16, is frequently lost in HPV-negative
TRAF3
CDKN2A versus HPV-positive tumors. Similarly, HPV-negative tumors dem-
−0.5 ATM onstrate prominent amplifications of CCND1 (cyclin D), whereas
1 3 5 7 9 11 13 15 17 19 21 HPV-positive tumors contain a striking focal amplification at 20q11,
Chromosome the location of E2F1. The E2F1 gene, which encodes a key transcrip-
tion factor that promotes genes that mediate cell cycle initiation from
Fig 2. Copy number alterations by chromosome position in human papilloma-
G0 to G1 and cell proliferation, is amplified in approximately 20% of
virus (HPV) –positive and HPV-negative tumors.
HPV-positive HNSCCs but in only 2% of HPV-negative HNSCCs.1
E2F1 together with tumor-suppressor proteins TP53 and RB1, which
are inactivated by the HPV E6 and E7 oncoproteins, critically deregu-
STRUCTURAL ALTERATIONS late cell cycle control and genomic stability in HPV-positive HNSCC.
A fourth gene in the cell cycle pathway, RB1, although only rarely
Several publications have reported the copy number landscape of inactivated, is more often lost in HPV-positive tumors.
HPV-positive versus HPV-negative HNSCC (Fig 2).1,27-35 Across At this time, studies have detected only rare instances of activated
much of the genome, the copy number alterations from HPV-positive fusion oncogenes based on analysis of DNA structural rearrange-
and HPV-negative tumors are highly concordant, and shared ampli- ments. Most prominently, in a minority of HPV-positive samples,
fications are 1q, 3q, 5p, 8q, and others. Common deletions include 3p, several authors have reported FGFR3-TACC3 fusions that seem
5q, 11q, and others. In some instances, comparison of altered regions promising as a therapeutic target.42-44 In addition, there was evidence
across studies is challenging because of the small number of patients in single samples for structural alterations associated with MET
who are HPV positive. Although many regions are clearly shared,
activation. Finally, the most well-described alternative transcript of
others are strikingly divergent. Recently, a region on chromosome
EGFR, the so-called vIII mutant, was only rarely observed, at a rate
14q32 was reported showing deep deletions in HPV-positive tumors
of less than 1%.1
and few, if any, deletions in HPV-negative tumors.1 Focal deletions,
which include homozygous deletions, are relatively rare in a cancer
genome and, when recurrent, are often suggestive of key novel tumor-
MUTATION SIGNATURE
suppressor genes. In combination with sequencing data, homozygous
deletions noted in the TCGA study identified that the tumor-
suppressor gene TRAF3 was inactivated in approximately 20% of It is perhaps somewhat surprising that tumors that presumably
HPV-positive tumors; this inactivation had been previously unrecog- have disparate molecular origins as HPV-positive virally associated
nized in HNSCC, although TRAF3 loss had been seen in nasopharyn- tumors and HPV-negative tobacco-associated tumors share as
geal carcinoma.1,36 TRAF3 is implicated in innate interferon and much homology in their genome landscape as the copy number
acquired antiviral responses, including in Epstein-Barr virus, HPV, data demonstrate. Other similarities between these subtypes in-
and HIV.37-40 TRAF3 also serves as a ubiquitin ligase and negative clude the generally similar mutation rate and number of copy
regulator for nuclear factor-␬B (NF-␬B) –inducing kinase, a criti- number changes in each group. Other aspects of their molecular
cal signal component in activation of transcription factors, which signature are, however, distinctly different.
are implicated in cell survival and other features of the malignant As expected, the transversions associated with smoking in many
phenotype.38 tumors (ie, mutations that change a purine nucleotide to another
A second deletion that presents more prominently in patients purine (A↔G) or a pyrimidine nucleotide to another pyrimidine
with HPV-positive tumors occurs in chromosome 11q. Although the (C↔T) at CpG sites were more frequent in HPV-negative tumors.1 By
deletion occurs in a broad region, which makes a single gene target contrast, a more unique signature has recently been reported in HPV-
difficult to identify, several prominent tumor-suppressor genes are positive tumors and in some other tumors in which a predominance
suggested, including ATM1. In terms of broad and focal amplification, of TpC mutations is noted. These cytosine-to-thymidine C⬎T muta-
perhaps the most prominent difference occurs on chromosome 7, tions have recently been shown most prominently in virally trans-
where HPV-positive tumors exhibit no evidence of amplification. The formed cancers, in which there is increased cytosine deaminase
prominent focal amplification seen at the locus of EGFR is likewise mutagenesis that is characteristic of the apolipoprotein B mRNA ed-
absent in HPV-positive tumors.1 The importance of this distinction iting enzyme catalytic (APOBEC) family of enzymes.45 APOBEC cy-
for anti-EGFR therapy remains obscure, but the lack of genomic tosine deaminase activity is induced and has emerged as a potential
targeting for amplification is evidence that HPV-positive tumors mutagenic factor in virally induced human cancers. Indeed, sequence
do not select for this gene at the clonal level. By contrast, HPV- data from HPV-positive HNSCCs and other human tumors reveal

www.jco.org © 2015 by American Society of Clinical Oncology 3

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Hayes, Van Waes, and Seiwert

genome-wide cytosine mutations consistent with APOBEC deamina- gene is defined as a gene that, when mutated, is responsible for either
tion and edits to the cellular DNA.46,47 The importance of this partic- initiation or progression of a cancer. In the context of cancer sequenc-
ular mutation signature is that it may explain, in part, a striking ing projects in which mutations cannot be individually tested for their
pattern seen in the distribution of mutations of one of the key onco- biologic function, driver mutation status is defined as a mutation that
genes of HNSCC, PIK3CA, between HPV-positive and HPV-negative occurs more often than expected by chance. Many other incidental
tumors. Although PIK3CA is altered in tumors without regard to viral mutations, called passenger mutations, occur in the tumor, but these
association, two specific C⬎T mutations in viral-associated tumors are attributed to the inherent instability of the cancer genome. Because
occur predominantly in two hotspots within the helical domain. These the methods to rigorously define driver mutations continue to evolve,
result in E542K and E545K amino acid substitutions that are impli- additional factors beyond statistical significance are considered when
cated in PIK3CA kinase and oncogene activation.1,27,45 The specificity the most likely candidates for driver gene status are selected; factors
of PIK3CA targeting emphasizes even more strongly the importance include whether the gene is mutated at a specific functional position,
of the 3q locus described in HPV-positive tumors described in the the gene has been reported in other cancers, and mutations occur in
Structural Alterations section. In smoking-related tumors, PIK3CA conjunction with copy number or other structural alterations. Several
mutations, including variants of unknown significance on signaling, reports with significant numbers of patients sequenced suggest that
are seen throughout the gene and are seen much less commonly in the most common driver mutations from the overall population of
hot spots. The signature APOBEC mutation of PIK3CA in HPV- HNSCC (mutations with a frequency ⬎ 5% to 10%) have been de-
positive HNSCC evokes the well-known transversion mutation of tected (Table 1).1,27-31 Data remain limited in subpopulations, includ-
KRAS (G12C and G12V) in smoking-related lung adenocarcinoma.48 ing in HPV-positive tumors, for which the cohorts remain relatively
Strikingly, although squamous cell carcinoma of any site rarely underpowered for the most rigorous statistical approaches.
demonstrates KRAS mutations, they are reported in HPV-positive Many mutations, along with previously described regions of copy
HNSCC.27 Although the data are sparse, the interaction between number and structural alteration, fall into a few cancer pathways.
smoking and KRAS mutation suggests a mechanism through which Most notably, almost every tumor had at least one mutation in the cell
tobacco might augment risk in HPV-positive HNSCC. cycle and survival pathway; mutations included CDKN2A (p16) mu-
tations, deletions, and hypermethylation; RB1 mutations and dele-
tions; E2F1 amplification; CCND1 amplification; MYC amplification;
HPV VIRAL INTEGRATION and TP53. The gene CCND1 bears special mention, because it is
among the most frequently altered oncogenes in HNSCC, although
In addition to the distinct mutational signature (APOBEC v smoking this occurs almost entirely in HPV-negative tumors. By the canonical
pattern), HPV-positive tumors are characterized by a second distinct pathway representation (Fig 1), it is immediately upstream of CDK4/
structural alteration compared with HPV-negative tumors: the inte- CDK6, which is a target of palbociclib, an inhibitor that was recently
gration of viral DNA. Although the true rate of HPV integration into approved in breast cancer.51 Trials in HNSCC are ongoing. Interest-
the human genome is unknown, existing data suggest that at least two ingly, nearly every HPV-negative tumor has inactivation of CDKN2A,
thirds of patients with high expression of E6 and E7 by RNA quantifi- and most have inactivation of TP53. One notable exception was a
cation have detectable integration sites in the genome.49 The exact subset of oral cavity cancers that had p16 loss but wild-type TP53.
nature of the integration is the subject of ongoing investigation. Most Tumors in these patients have a strikingly high rate of CASP8 inacti-
tumors seem to have a single primary integration site, although the vation and concurrent HRAS activating mutations, which suggests a
integration event itself may be complex at sites of gene amplification.50 distinct clinical entity that, in at least one data set, was also associated
Less common are tumors with multiple insertion sites on different with favorable clinical outcomes.1 Other investigators have reported
chromosomes. Most viral integrations occur in or near genes, al- similar favorable outcomes associated with wild-type TP53 in combi-
though the data do not yet support a recurrent site of integration to nation with other genomic alterations.52 The weight of the evidence,
identify selective pressure on a specific location in the genome. As with including the low rate of TP53 in patients with HPV-positive tumors
most structural rearrangements in cancer, the predicted impact who also benefit from favorable outcomes, suggests that wild-type
of integration is generally to silence the gene.1 Several tumor- TP53 status might serve as a universal marker of favorable outcome
suppressor genes have documented viral integration sites, such as without regard to HPV status.
RAD51 and ETS2. Whether the integration event resulted in tumor A second pathway clearly targeted by multiple driver mutations is
initiation or progression has not been definitively shown versus the the receptor tyrosine kinase pathway, which includes the RAS,
alternative possibility that the gene disruption may be a passenger PIK3CA, and PTEN branches. Numerous targets are altered, includ-
event. In the case of a passenger event, the targeting of a gene may ing activating mutations of PIK3CA, HRAS, FGFR3, KRAS, FGFR2,
be nonspecific and may result from a stochastic event that occurred EGFR, and KRAS (Fig 3). The EGFR mutations are worthy of special
at any open chromatin location. The lack of recurrent events gives note, because these are generally not the canonical activating muta-
more weight to the lower relevance of the integration site to tumor tions seen in lung cancer in the kinase domain of the gene. Mutations
initiation and progression. are more often in the juxtamembrane region of the gene and include a
number that have been documented to be functional in other cancer
types, including glioblastoma. Additional kinase pathway mutations
MUTATIONS AND PATHWAYS include loss of function from repressor elements in the pathway;
examples include PTEN, PIK3R1, and NF1.
Among the most anticipated results from recent tumor-profiling proj- Relevant to the recent enthusiasm in immune-based therapy
ects was the accounting of driver gene somatic mutations. A driver in solid tumors, which includes the recent approval of nivolumab

4 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Genetic Landscape of Head and Neck Cancer

dependence on the NF-␬B pathway. Mutations in HLA-A/HLA-B and

Table 1. Genes Commonly Mutated in Head and Neck Squamous
Cell Carcinoma
B2M are also seen in squamous lung cancer and, although these have
not clearly been functionally validated, represent important candidate
Mutated Genes
driver mutations. The immune system remains an important area of
Tumor Group Gene Symbol Mutation Rate (%)ⴱ investigation through integrated data analysis as well. Investigators
All tumors CDKN2A 15-22 with TCGA demonstrated dramatic differences in immune signa-
FAT1 5-23 tures across the molecular subtypes of HNSCC identified by gene
TP53 41-72 expression patterns, most notably with a complete absence of
CASP8 5-9
immune infiltration in the classic subtype and an activated T-cell
AJUBA
PIK3CA 8-21
signature in the mesenchymal subtype.1,54,55
NOTCH1 10-21 Perhaps more difficult to target with current therapeutics, but
KMT2D (ie, MLL2) 5-18 nonetheless frequently altered, are three additional pathways: squa-
NSD1 4-10 mous differentiation, oxidative stress, and WNT signaling. Many tu-
HLA-A mors demonstrated loss-of-function mutations in NOTCH genes,
TGFBR2
almost always with amplifications of chromosome 3q, where the well-
HRAS
TPRX1
known squamous transcription factor TP63 is located. Perhaps less
FLG well appreciated is the increasingly recognized interaction between
DDX3X this differentiation pathway and two frequently lost genes of the WNT
RPIK4 pathway, FAT1 and AJUBA.56,57 Several direct interactors with TP63
CUL3 and NOTCH seem affected by loss of function, including ZNF75058
PTEN and, most important as the master regulator of oxidative stress,
SYNE1
NFE2L2.59 It may be fair to say, if early reports are validated, that the
MED1
FGFR3†
NFE2L2 axis, with its repressors KEAP1 and CUL3, may represent one
NFE2L2† of the early potential therapeutic biomarkers for clinical use. Altera-
PIK3R1† tions of oxidative stress are seen across the spectrum of smoking-
RB1† related cancers and are tightly associated with the classic gene
FBXW7† expression subtype, as has been validated in numerous studies.1,60
SCN9A† Early translational data and preclinical models suggest that activation
CHEK2†
of this pathway may be associated with resistance to radiation, which is
PTCH1†
KRAS†
a cornerstone of therapy in HNSCC.61,62
MLL3† Although there are many other mutations, one in particular bears
FGFR2† special attention: the gene for nuclear receptor binding SET domain
ZNF217† protein 1 (NSD1), which is altered in approximately 10% of tumors.
RIMS2† The NSD1 gene is a histone 3 Lys 36 (H3K36) methyltransferase,
EGFR†
similar to SETD2, a gene most associated with alterations in the clear
NOTCH2†
NOTCH3†
cell subtype of renal cell carcinoma. Just as with SETD2 in clear cell
HPV-positive subgroup TRAF3† carcinoma, mutations of NSD1 were associated with a dramatic,
PIK3CA 22-56 genome-wide hypomethylation phenotype in HNSCC.1,63 Inter-
B2M† estingly, NSD1 seems relevant in relation to head and neck devel-
ZNF750† opment in nonmalignant disease, as do germline inactivating
IFNGR1† mutations, which are associated with craniofacial abnormalities
MLL3†
(Sotos syndrome).
DDX3X†
FGFR2/3† 11-14
NOTCH1†
NF1† INTEGRATED ANALYSIS
KRAS†
FBXW7† In addition to gene sequencing and copy number studies, there are
ⴱ
Data are presented for selected genes with mutation rates ⬎ 10% in at least mounting data from additional genomics platforms relevant to the
one series. understanding of HNSCC. The most mature of these data are gene
†Indicates that the gene did not meet strict statistical criteria for significance
of q score ⬍ 0.1 in at least one study. The gene is reported because the sum expression studies that include microarrays and RNA sequencing
of evidence supports its status as a driver. from TCGA. It is now clear that reproducible subtypes of HNSCC can
be defined by expression data, including the classic subtype associated
with activated oxidative stress, the atypical subtype associated with
lack of chromosome 7 amplification (including most HPV-positive
in squamous lung cancer, several known or suspected driver mu- cases), the mesenchymal subtype characterized by an activated T-cell
tations in the immune and cell death pathways are demonstrated in signature, and the basal subtype associated with wild-type TP53. Mo-
HNSCC.53 Amplifications of FADD and BIRC2, along with loss-of- lecular subtypes are an attractive way to classify tumors in some
function mutations and deletions of CASP8 and TRAF3, highlight a instances for the purposes of consideration of model systems and

www.jco.org © 2015 by American Society of Clinical Oncology 5

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Hayes, Van Waes, and Seiwert

A HLA/ NOTCH/
B HLA/ NOTCH/
RTKs inflammation differentiation RTKs inflammation differentiation

FGFR2/3 HLA A/B, B2M NOTCH1/2/3 EGFR FGFR1/2/3 ERBB2 HLA A/B, B2M NOTCH1/2/3
mut/fus mut mut mut/amp mut/amp mut mut

mut/amp
IGF1R
Other RTK alt. Presumed effects EPHA2 Presumed effects TP63/SOX2
TP63/SOX2
are uncommon on immunogenicity amp DDR2 on immunogenicity amp
MET FAT1/AJUBA
Cell death mut
PI3K signaling RAS signaling NF-kB
FADD
PI3K signaling RAS signaling amp
PIK3CA KRAS TRAF3
mut/amp mut mut/del PIK3CA HRAS CASP8
mut/amp mut mut
CYLD
PTEN NF1/2 mut
Oxidative stress
mut/del mut PTEN NF1/2
mut/del mut NFE2L2
TSC1/2 Inflammation/effects
mut TSC1/2 mut
on immune function
mut CUL
INPP4B HPV E6/E7 TP53
mut mut
exp wt INPP4B
TP53 MYC
mut KEAP1
mut amp Inflammation/
Cell cycle mut
effects on
Cell cycle immune function
E2F1 CDKN2A RB1 Proliferation Impaired
amp wt p16 mut and survival differentiation CCND1 CDKN2A* RB1 Proliferation and Impaired
amp mut/del mut survival differentiation

100 50 0 50 100 % altered

inactivating event activating

rate

Fig 3. Frequent genetic aberrations in selected pathways altered in HPV-positive (A) and HPV-negative (B) head and neck squamous cell carcinoma. Overview of key
genetic aberrations for HPV-positive and HPV-negative head and neck cancers. Shades of gold indicate frequency of activating changes in presumed oncogenes, and
shades of blue indicate frequency of inactivating changes in presumed tumor suppressor genes. amp, amplification; del, deletion; HLA, human leukocyte antigen; mut,
mutation; fus, fusion/translocation; wt, wild type.

potentially of therapeutic groups, although the clinical utility of this CCND1 amplifications and the availability of existing therapeutics.
remains unproved.1 Others have reported predictive gene expression Beyond targeted therapies, preclinical data already suggest that
signatures associated with prognosis, response to therapy, and HPV adherent signaling in the NFE2L2 pathway may predict differential
status; other phenotypes have been reported but have not generally phenotypes with respect to standard chemotherapy and radiation
been validated in clinical trials. The Chicago HNC Genomics cohort in ways that have clinical relevance, such as the selection of primary
has recently reported a striking finding within HPV-positive HNSCC treatment modality. Other broad signatures, such as molecular
of two subgroups defined as prominent inflamed (inflamed mesen- subtypes predicted by expression profiling, might similarly influ-
chymal HPV-positive subtype) or noninflamed (classic HPV-positive ence the selection of therapies, such as immune-based treatment,
subtype) with potential implications for immunotherapy.1,10 Similar and new technologies will allow reliable and reproducible determi-
work has been reported with other biomolecule platforms that target nation of expression signatures as part of clinical trials. The discov-
epigenetics, miRNA, and proteomics. ery of molecular signatures and shared mutations across tumor
types, such as squamous lung cancer and esophageal cancer, offers
OUTLOOK AND FUTURE CLINICAL IMPLICATIONS the potential to unify treatment paradigms across cancers that have
historically been isolated by anatomic boundaries. Among the
As a first step, accurate HPV diagnostics are needed beyond the barriers that face clinicians and researchers in achievement of the
interim proxy biomarker of p16. With confidence in the diagnosis promise of molecular characterization of HNSCC is uncertainty in
of virally associated tumors will come a definitive answer about the the regulatory and reimbursement environment for molecular di-
effectiveness of cetuximab as a function of HPV status and confi- agnostics. The pace of progress may be determined, in large part, by
dence in making HPV-related treatment decisions in the future the ability of researchers and clinicians to order (and justify to
(eg, de-escalation). Furthermore, the increased breadth of payers or trial sponsors) the relevant clinical grade diagnostic tests
genomic data on EGFR and related pathway alterations invites a to fuel molecularly targeted clinical trials and population studies.
re-evaluation of predictive biomarkers of EGFR-directed thera- In conclusion, head and neck cancer is a common and often
pies. Extension of the molecular profiling beyond HPV status to fatal cancer that is typically treated with curative intent at the time
target the most common molecular alterations for both therapeu- of diagnosis. Many therapeutic questions, including the choice of
tic and prognostic indications will become possible if profiling is optimal treatment modality and intensity and implications of HPV
done routinely (eg, associated with clinical studies). Two immedi- status on therapy, remain. By using the data provided by the first
ately obvious candidates include the PIK3CA and CDK4/CDK6 wave of comprehensive genome characterization, studies should
pathways because of the high prevalence of PIK3CA mutations and focus and accelerate progress in this disease by highlighting the

6 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Genetic Landscape of Head and Neck Cancer

frequent gene alterations in the disease. Researchers will be able to

AUTHOR CONTRIBUTIONS
leverage this knowledge in the design of clinical trials, the develop-
ment of model systems, and the assessment of patient outcomes. Conception and design: All authors
Collection and assembly of data: D. Neil Hayes, Tanguy
Y. Seiwert
AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST Data analysis and interpretation: D. Neil Hayes, Tanguy
Y. Seiwert
Disclosures provided by the authors are available with this article at Manuscript writing: All authors
www.jco.org. Final approval of manuscript: All authors

papillomavirus detection methods in head and neck complete response (cCR) to induction chemother-
REFERENCES cancer. Modern Pathol 24:1295-1305, 2011 apy (IC). J Clin Oncol 32:385s, 2014 (abstr LBA6006)
15. Boyer SN, Wazer DE, Band V: E7 protein of 27. Seiwert TY, Zuo Z, Keck MK, et al: Integrative
1. Cancer Genome Atlas Network: Comprehen- human papilloma virus-16 induces degradation of and comparative genomic analysis of HPV-positive
sive genomic characterization of head and neck retinoblastoma protein through the ubiquitin- and HPV- negative head and neck squamous cell
squamous cell carcinomas. Nature 517:576-582, proteasome pathway. Cancer Res 56:4620-4624, carcinomas. Clinical Clin Cancer Res 21:632-641,
2015 1996 2015
2. Gymnopoulos M, Elsliger MA, Vogt PK: Rare 16. Liang C, Marsit CJ, McClean MD, et al: Bio- 28. Agrawal N, Frederick MJ, Pickering CR, et al:
cancer-specific mutations in PIK3CA show gain of markers of HPV in head and neck squamous cell Exome sequencing of head and neck squamous cell
function. Proc Natl Acad Sci U S A 104:5569-5574, carcinoma. Cancer Res 72:5004-5013, 2012 carcinoma reveals inactivating mutations in NOTCH1.
2007 17. Chai RC, Lambie D, Verma M, et al: Current Science 333:1154-1157, 2011
3. Ang KK, Harris J, Wheeler R, et al: Human trends in the etiology and diagnosis of HPV-related 29. Chung CH, Guthrie VB, Masica DL, et al:
papillomavirus and survival of patients with oropha- head and neck cancers. Cancer Med 4:596-607, Genetic alterations in head and neck squamous cell
ryngeal cancer. N Engl J Med 363:24-35, 2010 2015 carcinoma determined by cancer gene-targeted se-
4. Abogunrin S, Di Tanna GL, Keeping S, et al: 18. Bonner JA, Harari PM, Giralt J, et al: Radio- quencing. Ann Oncol 26:1216-1223, 2015
Prevalence of human papillomavirus in head and therapy plus cetuximab for squamous-cell carci- 30. Pickering CR, Zhang J, Yoo SY, et al: Integra-
neck cancers in European populations: A meta- noma of the head and neck. New Engl J Med tive genomic characterization of oral squamous cell
analysis. BMC Cancer 14:968, 2014 354:567-578, 2006 carcinoma identifies frequent somatic drivers. Can-
5. Jiron J, Sethi S, Ali-Fehmi R, et al: Racial 19. Curran D, Giralt J, Harari PM, et al: Quality of cer Disc 3:770-781, 2013
disparities in human papillomavirus (HPV) associ- life in head and neck cancer patients after treatment 31. Stransky N, Egloff AM, Tward AD, et al: The
ated head and neck cancer. Am J Otolaryngol 35: with high-dose radiotherapy alone or in combination mutational landscape of head and neck squamous
147-153, 2014 with cetuximab. J Clin Oncol 25:2191-2197, 2007 cell carcinoma. Science 333:1157-1160, 2011
6. López RV, Levi JE, Eluf-Neto J, et al: Human 20. Bonner JA, Harari PM, Giralt J, et al: Radio- 32. Hayes DN, Kim WY: The next steps in next-
papillomavirus (HPV) 16 and the prognosis of head therapy plus cetuximab for locoregionally advanced gen sequencing of cancer genomes investigation.
and neck cancer in a geographical region with a low Clin Invest 125:462-468, 2015
head and neck cancer: 5-year survival data from a
prevalence of HPV infection. Cancer Causes Control 33. Lamlertthon W, Hayward MC, Hayes DN:
phase 3 randomised trial, and relation between
Emerging technologies for improved stratification of
25:461-471, 2014 cetuximab-induced rash and survival. Lancet Oncol
cancer patients: A review of opportunities, chal-
7. Deschler DG, Richmon JD, Khariwala SS, et 11:21-28, 2010
lenges, and tools. Cancer J 17:451-464, 2011
al: The “new” head and neck cancer patient-young, 21. Machiels J, Licitra LF, Haddad RI, et al: Afa-
34. Meyerson M, Gabriel S, Getz G: Advances in
nonsmoker, nondrinker, and HPV positive: Evalua- tinib versus methotrexate (MTX) as second-line
understanding cancer genomes through second-
tion. Otolaryngol Head Neck Surg 151:375-380, treatment for patients with recurrent and/or meta-
generation sequencing. Nat Rev Genet 11:685-696,
2014 static (R/M) head and neck squamous cell carcinoma
2010
8. Gillison ML, Zhang Q, Jordan R, et al: Tobacco (HNSCC) who progressed after platinum-based ther-
35. Smeets SJ, Braakhuis BJ, Abbas S, et al:
smoking and increased risk of death and progres- apy: Primary efficacy results of LUX-Head & Neck.
Genome-wide DNA copy number alterations in head
sion for patients with p16-positive and p16-negative Ann Oncol 25:1-41, 2014
and neck squamous cell carcinomas with or without
oropharyngeal cancer. J Clin Oncol 30:2102-2111, 22. Vermorken JB, Stöhlmacher-Williams J, Da-
oncogene-expressing human papillomavirus. Onco-
2012 videnko I, et al: Cisplatin and fluorouracil with or
gene 25:2558-2564, 2006
9. Huang SH, Xu W, Waldron J, et al: Refining without panitumumab in patients with recurrent or 36. Chung GT, Lou WP, Chow C, et al: Constitu-
American Joint Committee on Cancer/Union for metastatic squamous-cell carcinoma of the head tive activation of distinct NF-kappaB signals in EBV-
International Cancer Control TNM stage and prog- and neck (SPECTRUM): An open- label phase 3 associated nasopharyngeal carcinoma. J Pathol 231:
nostic groups for human papillomavirus-related oro- randomised trial. Lancet Oncol 14:697-710, 2013 311-322, 2013
pharyngeal carcinomas. J Clin Oncol 33:836-845, 23. Fayette J, Wirth LJ, Oprean C, et al: Random- 37. Eliopoulos AG, Dawson CW, Mosialos G, et
2015 ized phase II study of MEHD7945A (MEHD) vs al: CD40-induced growth inhibition in epithelial cells
10. Keck MK, Zuo Z, Khattri A, et al: Integrative cetuximab (Cet) in ⱖ 2nd-line recurrent/metastatic is mimicked by Epstein-Barr Virus– encoded LMP1:
analysis of head and neck cancer identifies two squamous cell carcinoma of the head & neck. Ann iInvolvement of TRAF3 as a common mediator.
biologically distinct HPV and three non-HPV sub- Oncol 25:iv340-iv356, 2014 Oncogene 13:2243-2254, 1996
types. Clin Cancer Res 21:870-881, 2015 24. Vermorken JB, Mesia R, Rivera F, et al: 38. Hacker H,, Tseng PH, Karin M: Expanding
11. Jordan RC, Lingen MW, Perez-Ordonez B, et Platinum-based chemotherapy plus cetuximab in TRAF function: TRAF3 as a tri-faced immune regu-
al: Validation of methods for oropharyngeal cancer head and neck cancer. New Engl J Med 359:1116- lator. Nat Rev Immunol 11:457-468, 2011
HPV status determination in US cooperative group 1127, 2008 39. Karim R, Tummers B, Meyers C, et al: Human
trials. Am J Surg Pathol 36:945-954, 2012 25. Bonner JA, Harari PM, Giralt J, et al: Associ- papillomavirus (HPV) upregulates the cellular deu-
12. Seiwert T. Accurate HPV testing: A require- ation of human papillomavirus (HPV) and p16 status biquitinase UCHL1 to suppress the keratinocyte’s
ment for precision medicine for head and neck with efficacy and safety data in the phase III radio- innate immune response. PLoS Path 9:e1003384,
cancer. Ann Oncol 24:2711-2713, 2013 therapy (RT)/cetuximab (cet) registration trial for 2013
13. Dreyer JH, Hauck F, Oliveira-Silva M, et al: locoregionally advanced squamous cell carcinoma of 40. Oganesyan G, Saha SK, Guo B, et al: Critical
Detection of HPV infection in head and neck squa- the head and neck (LASCCHN). Ann Oncol 25:iv340- role of TRAF3 in the Toll-like receptor-dependent
mous cell carcinoma: A practical proposal. Virchows iv356, 2014 and -independent antiviral response. Nature 439:
Archiv 462:381-389, 2013 26. Cmelak A: Reduced-dose IMRT in human 208-211, 2006
14. Schlecht NF, Brandwein-Gensler M, Nuovo papilloma virus (HPV)-associated resectable oropha- 41. Engelman JA, Luo J, Cantley LC: The evolu-
GJ, et al: A comparison of clinically utilized human ryngeal squamous carcinomas (OPSCC) after clinical tion of phosphatidylinositol 3-kinases as regulators

www.jco.org © 2015 by American Society of Clinical Oncology 7

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Hayes, Van Waes, and Seiwert

of growth and metabolism. Nat Rev Genet 7:606- actions in primary head and neck cancers. Proc Natl 56. Morris LG, Kaufman AM, Gong Y, et al: Re-
619, 2006 Acad Sci U S A 111:15544-15549, 2014 current somatic mutation of FAT1 in multiple human
42. Capelletti M, Dodge ME, Ercan D, et al: Iden- 50. Akagi K, Li J, Broutian TR, et al: Genome-wide cancers leads to aberrant Wnt activation. Nat Genet
tification of recurrent FGFR3-TACC3 fusion onco- analysis of HPV integration in human cancers re- 45:253-261, 2013
genes from lung adenocarcinoma. Clin Cancer Res veals recurrent, focal genomic instability. Genome 57. Nola S, Daigaku R, Smolarczyk K, et al:
20:6551-6558, 2014 Res 24:185-199, 2014 Ajuba is required for Rac activation and mainte-
43. Yuan L, Liu ZH, Lin ZR, et al: Recurrent 51. Finn RS, Crown JP, Lang I, et al: The cyclin- nance of E-cadherin adhesion. J Cell Biol 195:855-
FGFR3-TACC3 fusion gene in nasopharyngeal carci- dependent kinase 4/6 inhibitor palbociclib in combina- 871, 2011
noma. Cancer Biol Ther 15:1613-1621, 2014 tion with letrozole versus letrozole alone as first-line 58. Sen GL, Boxer LD, Webster DE, et al: ZNF750
44. Di Stefano AL, Fucci A, Frattini V, et al: treatment of oestrogen receptor-positive, HER2- is a p63 target gene that induces KLF4 to drive
Detection, characterization and inhibition of FGFR- negative, advanced breast cancer (PALOMA-1/ terminal epidermal differentiation. Dev cell 22:669-
TACC fusions in IDH wild type glioma. Clin Cancer TRIO-18): A randomised phase 2 study. Lancet 677, 2012
Res doi: 10.1158/1078-0432.CCR-14-2199 [epub
Oncol 16:25-35, 2015 59. Wakabayashi N, Shin S, Slocum SL, et al:
ahead of print on January 21, 2015]
52. Gross AM, Orosco RK, Shen JP, et al: Multi- Regulation of notch1 signaling by nrf2: Implications
45. Henderson S, Chakravarthy A, Su X, et al:
tiered genomic analysis of head and neck cancer for tissue regeneration. Sci Signal 3:ra52, 2010
APOBEC-mediated cytosine deamination links PIK3CA
ties TP53 mutation to 3p loss. Nat Genet 46:939- 60. Cancer Genome Atlas Research Network:
helical domain mutations to human papillomavirus-driven
943, 2014 Comprehensive genomic characterization of
tumor development. Cell Rep 7:1833-1841, 2014
53. Rizvi NA, Mazières J, Planchard D, et al: squamous cell lung cancers. Nature 489:519-525,
46. Burns MB, Temiz NA, Harris RS: Evidence for
Activity and safety of nivolumab, an anti-PD-1 im- 2012
APOBEC3B mutagenesis in multiple human can-
cers. Nat Genet 45:977-983, 2013 mune checkpoint inhibitor, for patients with ad- 61. Kawasaki Y, Okumura H, Uchikado Y, et al:
47. Roberts SA, Lawrence MS, Klimczak LJ, et al: vanced, refractory squamous non-small-cell lung Nrf2 is useful for predicting the effect of chemora-
An APOBEC cytidine deaminase mutagenesis pat- cancer (CheckMate 063): A phase 2, single-arm trial. diation therapy on esophageal squamous cell carci-
tern is widespread in human cancers. Nat Genet Lancet Oncol 16:257-265, 2015 noma. Annals Surg Oncol 21:2347-2352, 2014
45:970-976, 2013 54. Walter V, Yin X, Wilkerson MD, et al: Molec- 62. Shibata T, Kokubu A, Gotoh M, et al: Ge-
48. Dogan S, Shen R, Ang DC, et al: Molecular ular subtypes in head and neck cancer exhibit dis- netic alteration of Keap1 confers constitutive Nrf2
epidemiology of EGFR and KRAS mutations in 3,026 tinct patterns of chromosomal gain and loss of activation and resistance to chemotherapy in gall-
lung adenocarcinomas: Higher susceptibility of canonical cancer genes. PloS One 8:e56823, 2013 bladder cancer. Gastroenterology 135:1358-1368,
women to smoking-related KRAS-mutant cancers. 55. Chung CH, Parker JS, Karaca G, et al: Molec- 2008
Clin Cancer Res 18:6169-6177, 2012 ular classification of head and neck squamous cell 63. Cancer Genome Atlas Research Network:
49. Parfenov M, Pedamallu CS, Gehlenborg N, et carcinomas using patterns of gene expression. Can- Comprehensive molecular characterization of clear
al: Characterization of HPV and host genome inter- cer Cell 5:489-500, 2004 cell renal cell carcinoma. Nature 499:43-49, 2013

■ ■ ■

8 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Genetic Landscape of Head and Neck Cancer

AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Genetic Landscape of Human Papillomavirus–Associated Head and Neck Cancer and Comparison to Tobacco-Related Tumors
The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are
self-held unless noted. I ⫽ Immediate Family Member, Inst ⫽ My Institution. Relationships may not relate to the subject matter of this manuscript. For more
information about ASCO’s conflict of interest policy, please refer to www.asco.org/rwc or jco.ascopubs.org/site/ifc.
D. Neil Hayes Carter Van Waes
Leadership: GeneCentric No relationship to disclose
Stock or Other Ownership: GeneCentric
Consulting or Advisory Role: GeneCentric Tanguy Y. Seiwert
Research Funding: GeneCentric Honoraria: Novartis, Bayer AG, Onyx Pharmaceuticals, Merck, Amgen
Patents, Royalties, Other Intellectual Property: GeneCentric Research Funding: Genentech (Inst), Boehringer Ingelheim (Inst)

www.jco.org © 2015 by American Society of Clinical Oncology

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128
 Hayes, Van Waes, and Seiwert

Acknowledgment
We thank Katie Hoadley, Vonn Walter, Ashley Salazar, and Michele Hayward for assistance in figure and manuscript preparation.

© 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at CAMBRIDGE UNIV MEDICAL LIBRARY on September
Copyright © 2015 American Society
9, 2015 of Clinical Oncology. All rights reserved.
from 131.111.164.128